Scaled deployment of Wolbachia to protect the community from dengue and other Aedes transmitted arboviruses

Background: A number of new technologies are under development for the control of mosquito transmitted viruses, such as dengue, chikungunya and Zika that all require the release of modified mosquitoes into the environment. None of these technologies has been able to demonstrate evidence that they can be implemented at a scale beyond small pilots. Here we report the first successful citywide scaled deployment of Wolbachia in the northern Australian city of Townsville. Methods: The wMel strain of Wolbachia was backcrossed into a local Aedes aegypti genotype and mass reared mosquitoes were deployed as eggs using mosquito release containers (MRCs). In initial stages these releases were undertaken by program staff but in later stages this was replaced by direct community release including the development of a school program that saw children undertake releases. Mosquito monitoring was undertaken with Biogents Sentinel (BGS) traps and individual mosquitoes were screened for the presence of Wolbachia with a Taqman qPCR or LAMP diagnostic assay. Dengue case notifications from Queensland Health Communicable Disease Branch were used to track dengue cases in the city before and after release. Results: Wolbachia was successfully established into local Ae. aegypti mosquitoes across 66 km 2 in four stages over 28 months with full community support. A feature of the program was the development of a scaled approach to community engagement. Wolbachia frequencies have remained stable since deployment and to date no local dengue transmission has been confirmed in any area of Townsville after Wolbachia has established, despite local transmission events every year for the prior 13 years and an epidemiological context of increasing imported cases. Conclusion: Deployment of Wolbachia into Ae. aegypti populations can be readily scaled to areas of ~60km 2 quickly and cost effectively and appears in this context to be effective at stopping local dengue transmission


Gates Open Research
Any reports and responses or comments on the article can be found at the end of the article. scaled approach to community engagement.
frequencies have Wolbachia remained stable since deployment and to date no local dengue transmission has been confirmed in any area of Townsville after Wolbachia has established, despite local transmission events every year for the prior 13 years and an epidemiological context of increasing imported cases.
: Deployment of into populations can be Conclusion Wolbachia Ae. aegypti readily scaled to areas of ~60km quickly and cost effectively and appears in this context to be effective at stopping local dengue transmission

Introduction
A growing body of evidence shows that the wMel strain of Wolbachia, when introduced into Aedes aegypti, reduces the mosquito's ability to transmit key human viruses such as dengue 1 , Zika 2,3 and chikungunya 4,5 , and this reduction is estimated to have the potential to significantly reduce disease transmission in affected communities 6 . The World Mosquito Program (formerly known as the Eliminate Dengue Program), a not-for-profit consortium, has demonstrated previously that, after small-scale releases, the wMel strain of Wolbachia can be established and maintain itself within isolated Ae. aegypti populations around the city of Cairns in Australia 7,8 . Subsequent pilot releases have also shown that Wolbachia can be established in contiguous urban habitats 9 . In this report, we present the results of the first large-scale deployment of Wolbachia across Townsville, a medium-sized city in northern Australia with a population of ∼187,000 residents.
Our goals for this work were to demonstrate that large scale deployment of Wolbachia was possible 10 , that it could be done quickly and efficiently at low cost, and that it was acceptable to communities. In addition, while not designed as a clinical trial, it also provided an opportunity to examine a time series of observational data on dengue transmission, for 13 years before deployment and four consecutive dengue transmission seasons since deployment began.

Community engagement
One of the key objectives of the Townsville project was establishing a community engagement framework that could be suitably scaled for a citywide deployment and could be used cross-culturally for future deployments. Previous deployments in Cairns had relied on obtaining individual consent from community members for the release activities, an approach that was unsuitable for the required scaling. Instead we developed a Public Acceptance Model (PAM) for our engagement that formed the basis for obtaining community support for the research activities. The PAM was based on a set of Public Participation Principles described in Table 1.
The PAM consisted of four key components 1. Raising awareness by providing information to residents and key stakeholders about the program. These activities included face to face meetings, media events, stalls at community markets, community presentations utilising existing community networks such as community associations, information kiosks in public spaces, traditional and electronic mail outs of information letters and deployment coverage updates, a public billboard and newspaper advertising, a school outreach program and social media incentive program.

Updates from Version 2
This new version adds new data on both Wolbachia monitoring in the local mosquito population as well as dengue case notification and extends the data to cover an additional dengue transmission season in Townsville (new Figure 3 and Figure 4). In addition, a new Interrupted Time Series analysis of the data has been added. Nicholas Jewel and Stephanie Tanamas contributed to the new Interrupted Time Series analysis, and so they have been added to the author list of this version.

UPDATE
2. Quantitative surveys that measured community awareness and acceptance conducted by an external market research company, Compass Research. Each telephone survey was undertaken at roughly six monthly intervals, the first survey being undertaken in March 2014 prior to our community engagement activities starting in the city and each involved 200-600 participants (Table 2).
3. An issues management system that allowed community members to easily contact the program with questions or concerns and have them addressed by program staff typically within 24 hours of receipt. This also allowed residents to opt out of direct participation if they had concerns.

4.
A community reference group that consisted of respected community members from key stakeholder groups and included representation from Townsville City Council, Queensland Health, the local indigenous community, the Defence Force, local business, community development and environmental groups, the tourism sector and the education sector. The reference group's primary function was to independently review our activities to ensure that we had carried out our engagement in accordance with our commitments and stated Public Participation Principles (Table 1). The reference group was tasked to evaluate our activities and make a recommendation to the program management that community engagement had been sufficient for releases of mosquitoes to commence. Before releases began this group met monthly; after releases started they continued to meet every 6-8 weeks. The secondary functions of this group were to test and comment on the suitability of engagement materials and approaches, and to provide the program with feedback on community sentiment towards the program and identify potential issues that might require a proactive response. The reference group was also kept regularly updated on the latest results of the program.

Rearing
In order to establish the colony for release, wild mosquito eggs were collected from ovitraps set at 49 sites across Townsville and used to produce a wildtype colony. Material from this colony was stored as dried eggs and amplified only as required. Amplification of material from this colony was limited to F3 for use in outcrossing during colony maintenance. For stage 1 of the Townsville releases, eggs were produced from insectaries at Monash University, Melbourne or James Cook University, Cairns and shipped to the Townsville field office. For stages 2-4 all mosquito material was produced at Monash University.
The wildtype colony was backcrossed for three generations to a laboratory line infected with the wMel strain of Wolbachia 11 . This new colony, TSV wMel.f was continuously maintained in order to produce ∼800,000 eggs per week. To maintain the material during mass production, the TSV wMel.f line was divided into two distinct colonies: 'broodstock' and 'release material'. The 'broodstock' colony was reared under the more relaxed conditions described in 12 but kept at 26°C. Its purpose was to produce eggs for amplification and production of the 'release material colony'. In order to prevent inbreeding, 10% wildtype males (from the same wildtype material as was used for backcrossing) were added to each generation of the 'broodstock'. The purpose of the 'release material' colony was to produce eggs for release; it did not provide any material for the next generation in the laboratory. In order to facilitate mass production, the 'release material' colony was maintained as described for the broodstock with the following modifications. No wild material was added to the 'release material' colony. Once eggs were hatched, first instar larvae were aliquoted into 500 ml plastic cups For both colonies, females (5-7 days old) were fed with human blood (Monash University Human Ethics approval CF11/0766 -2011000387). They were provided the bloodmeal by introducing the arm of a volunteer into the selected cage. Females were fed until repletion (usually 10-15 minutes). Females were fed once per week, for one or two weeks depending on requirements.
For safety, only one bloodfeeder was used per cage and bloodfeeders who showed any signs of fever or who were taking antibiotics were excluded.
Three 22 cm oviposition strips of red cotton duck cloth were placed in each cage three to five days after bloodfeeding. Oviposition strips were removed from cages four days later, and sandwiched between two double layers of 3mm thick kitchen sponge that had been covered with a single layer of paper towel, covered with a 3mm thick Perspex sheet and placed on a rack. Eggs were allowed to dry this way in an 80%RH controlledtemperature room for up to 24 hours before being placed in humidified containers. The humidity in these containers was maintained at ∼80%RH by providing a saturated KCl solution inside the containers.
After the oviposition strips had been dried, the density of eggs/cm on each strip was estimated to determine the length of egg strip to be cut for subsequent use in Mosquito Release Containers (MRCs). Eggs were then shipped to the Townsville field lab.
Hatch rate was tested for every batch of eggs produced. Matched sets of eggs were taken from a number of strips and photographed to assess desiccation and overall quality of the eggs. One portion of each matched set was shipped to the release site, and one set kept at the rearing facility. Once the eggs reached the release site, both sets of eggs were counted, hatched, and hatch rate determined by counting larvae. Hatch rate of 70% or above was considered acceptable. If hatch rate fell below 70%, the cause of this drop was investigated. In most cases, the cause was determined to be due to fluctuating environmental conditions or to slight changes made to the drying procedure, which was altered slightly throughout releases.
Wolbachia infection frequency was also tested each week of production. 80 females and 80 males were screened from each broodstock cohort using diagnostic qPCR as described below.
If Wolbachia frequency fell below 97% in any broodstock cohort, the eggs from their resultant 'release material colony' would not be used for release, however this issue never arose.
The James Cook University rearing strategy differed slightly from the Monash rearing strategy. A single colony of ∼10,000 Townsville wMel-infected Ae. aegypti sourced from Monash was created in a semi-field flight cages 13 in the Tropical Medicine Mosquito Research Facility located at James Cook University in Cairns. Based upon experience with earlier releases, we assumed that there is a loss of ∼50% of the colony per week. The colony was therefore refreshed with 2500 males and 2500 females each week. We also conducted backcrossing to maintain genetic diversity by adding males (10% of cage male population) sourced from an uninfected wildtype Townsville colony (< F4). To prevent introduction of wild females and potential loss of Wolbachia infection into the colony, we only added males. This was achieved by placing suspected male pupae based on size into cups of 10; any cups containing emerged females were discarded.
Females (5-7 days old) were fed with human blood on volunteers (JCU Human Ethics H4907). They were provided the bloodmeal by introducing 5 volunteer blood feeders into the field cage 3-5 times/week who let mosquitoes feed for 10 minutes. For safety bloodfeeders were screened at every feed for possible exposure to dengue infected mosquitoes using a questionnaire to access travel history, and their temperature was taken to detect fever. Any volunteers with fever, a possible exposure to dengue infected mosquitoes or who were taking antibiotics were excluded for a minimum of 2 weeks.
Eggs were harvested from partially flooded 10 L buckets containing 26 × 30 cm strips of red felt cloth placed in the semi-field cage. A perspex template 31cm in length with 12 1-cm holes drilled into it was placed over the cloth to limit oviposition to the exposed 1 cm area of the ovistrip. The ovistrips were collected 3 times/week, embryonated and dried three days later. Once removed from the cages, oviposition strips were placed on moist paper towel in a sealed plastic container, after 3 days the lid of the sealed container was removed and the eggs were allowed to dry this way in an 80%RH controlled temperature room for up to 24 hours before being placed in humidified containers. The humidity in these containers was maintained at ∼80%RH by providing a saturated KCl solution inside the containers. The cloth was then cut into individual eggstrips containing a single egg clump that could be deployed into egg release containers in the field. The number of eggs on each eggstrip was estimated by using reference photographs of eggstrips with known egg numbers as visual guides for fast estimation.

Mosquito releases
The municipal area of Townsville is ∼190km 2 . However, within this area there were many areas where releases did not take place due to the lack of suitable Ae. aegypti habitat. Releases were restricted to residential and business areas within the city where Ae. aegypti breeding was likely to occur. This resulted in the actual area for release being reduced to approximately 66km 2 to effectively cover the city. The release program was divided into four stages ( Figure 1).
Stage 1 covered a release area of 20km 2 and included the suburbs with known highest dengue transmission risk: South Townsville, Railway Estate, North Ward, Townsville City, Belgian Gardens, Castle Hill, West End, Garbutt, Currajong, Vincent,  Gulliver, Aitkenvale, Mundingburra, Rosslea, Hyde Park, Pimlico, Mysterton and Hermit Park. In this stage, all releases were undertaken using bucket mosquito release containers (MRCs). These were 2.3L white polypropylene pails with lid (Peopleinplastic, Australia), with top 164mm diameter, base 145mm diameter, and height 147mm. Each bucket had four 6mm holes drilled 20mm apart in a square pattern in the side ( Figure 2A). The inside of each bucket was roughened with sandpaper to allow mosquitoes to rest upon emergence. Into each bucket MRC was placed an egg strip containing approximately 100 viable eggs (estimated from hatch rate QA), 5 (summer) or 6 (winter) wafers of Aqua One vege wafer fish food (Aqua Pacific, UK) and 1L water. More food was provided in winter and the servicing cycle for these buckets was extended from 2 to 3 weeks to allow for longer emergence times.
Bucket MRCs for stage 1 were placed by program staff in outdoor shaded areas at approximately 20% of all residential properties in a roughly evenly spaced arrangement in each suburb. They were serviced every two weeks by tipping out the water, cleaning the bucket and adding new food, water and eggs. An average of 88 adult mosquitoes were released from each bucket MRC in stage 1. Releases continued in each suburb until the frequency of Wolbachia in samples of field-caught mosquitoes from that suburb was above 50% for two consecutive weeks.
For stage 1, it required between 7 and 19 weeks of releases for each suburb to reach that target Stages 2, 3 & 4 covered release areas of 18, 18 and 10 km 2 respectively, and included the following suburbs. Stage 2: Cranbrook, Heatley, Kirwan/Thuringowa Central and Mount Louisa; stage 3: Condon, Pallarenda, Rowes Bay, Rasmussen and Kelso; stage 4: Idalia, Oonoonba, Wulguru/Stuart, Annandale and Douglas. Releases for these later stages did not rely on program team members to place all release containers. Instead, they utilised strategies that directly involved the community, such as the use of school students, direct community release, or through collaboration with local businesses. Releases for these stages also used Mozzie Box MRCs ( Figure 2C) which consisted of a 775ml Food Pail (Detpak, Australia) without handle, and with measurements top 104×92mm, base 79×61mm, height 104mm. Four 5mm holes were punched into each MRC -one hole approximately 1cm from the top right and top left corners of each long-side face of the box. Each Mozzie Box MRC received 100 viable eggs (estimated from hatch rate QA), 4 (summer) or 5 (winter) wafers of Aqua One vege waters, and 400ml tap water. Mozzie Box MRCs were not re-used.
In stages 2-4 the goal was again to place MRCs at 20% of residences in the release area. This was done by using community engagement activities to identify participants who would agree to host an MRC. In areas where there were large spatial gaps in participation, the program team would then supplement coverage by visiting additional houses in these areas and obtaining consent to leave MRCs with residents at these locations. Finally, in the last two suburbs of stage 3 (Kelso & Rasmussen) and across stage 4, releases of adult mosquitoes 7 were used to fill in gaps in MRC coverage.
During the 28 months of the release phase (stages 1-4), a total of approximately 4 million mosquitoes were released. Releases were undertaken with regulatory approval from the Australian Pesticides and Veterinary Medicines Authority (APVMA permit numbers PER14797 and PER82947).

School releases
The Wolbachia Warriors Program was developed both as a tool to engage children and their parents and make them aware of the program, and as an alternative channel to release mosquitoes. Five different primary schools were selected to run the program over the duration of the Townsville project. One school participated in each stage except for stage 2 where two schools participated. In total 943 students aged 6-12 participated in these programs.
School children were provided with a bucket MRC in stage 1 as used in operational releases in stage 1 but made of clear plastic to encourage student observation ( Figure 2B) and Mozzie Box MRCs in stages 2-4 ( Figure 2C), complete with mosquito eggs, food, instructions, a calendar to track progress, a magnifying glass, a badge for participation, and an educational booklet tailored for either lower (grade P-2) or upper primary (grade 3-6) students ( Figure 2D). Each student was expected to undertake three consecutive releases with their MRC over a six-week period.
Materials were distributed at the schools by program communication and engagement staff, who gave presentations encouraging participation prior to each of the three mosquito release cycles. Students were asked to use their calendar to record the progress of the mosquito life cycle in their MRC, and to return it to program staff at the end of the release.

Direct community release
In these releases, a Mozzie Box MRC was provided directly to residents who set it up and reared the mosquitoes themselves at their place of residence. In stages 2-4, more than 6,000 households directly participated in establishing Wolbachia by managing their own release container. Almost half of these participants contacted the program team to receive an MRC, which was subsequently delivered to their house. The remaining participants were recruited through doorknocking, or through other recruitment methods such as community groups. Additional Mozzie Box MRCs were distributed through large local employers including the City Council, Telstra, The Townsville Hospital, James Cook University and Queensland Nickel. More than 200 people participated in these programs.

Quality assurance procedures
In stage 1, program staff checked 5-10% of all bucket MRCs to determine whether the bucket had failed or not, and if not to count pupal skins to obtain an estimate of adult emergence from which they could estimate release rates. In stage 2 -4, a random selection of 5-15% of all MRCs were checked to determine if they were set correctly. Larvae, pupae and pupal skins were counted to estimate emergence rates in these stages (accounting for potential delayed development of mosquitoes at time of QA due to community members setting up MRCs later than day of delivery). This approach was supplemented in stage 3 with additional sentinel buckets that were set and checked by staff to determine average emergence rates. These data were then used to adjust numbers of eggs placed in MRCs.   Ethical approval was not required to access non-identifiable dengue case notification data collected as part of routine disease surveillance.

Results and discussion
Prior deployments of Wolbachia in Australia by the World Mosquito Program utilised a traditional individual informedconsent approach to obtaining community authorisation for the releases 7 . While this approach was adequate for small deployments, it was not considered scalable for an entire city. We therefore developed a Public Acceptance Model (PAM) that proved highly effective in ensuring community awareness and acceptance of the mosquito deployment program in Townsville. We believe this model will be suitable for other settings with appropriate local adaptation, and provides a framework for scaled deployment of this type of intervention globally.
Releases of mosquitoes in Townsville began in Oct 2014 with strong community support (Table 2) and lasted for 28 months. The release program was divided into 4 sequential stages. The approach used in Townsville relied on the use of Mosquito Release Containers (MRC) as the preferred method of deployment ( Figure 2). In each suburb of the city MRCs were set at approximately 20% of residences and then refreshed with new food, water and eggs every 2-3 weeks. MRC release cycles continued until 2 consecutive samples of adult mosquitoes taken from the suburb showed a Wolbachia frequency above 50%; Wolbachia frequency in these areas was then monitored without additional releases. While the city occupies a municipal area of 190km 2 , releases were undertaken over a reduced area of ∼66km 2 as not all areas of the greater municipal area were inhabited or provided suitable Ae. aegypti habitat (Figure 1). The targeted release areas covered all of the suburbs where local dengue transmission had occurred during the prior 10 years and known high-risk suburbs for dengue transmission were targeted in stage 1. Wolbachia monitoring was conducted and infection frequency reported aggregate to suburb boundaries, encompassing an area greater than the 66km 2 of actual release areas. The total area considered 'covered' by Wolbachia in Townsville is 128km 2 , with a residential population in 2016 of 140,000.
Wolbachia establishment across the different suburbs of Townsville for the four stages is shown in Figure 3. In general, establishment of Wolbachia occurred reliably after releases stopped once the 50% threshold was met. In some suburbs, Wolbachia frequencies fluctuated for a number of months before eventually rising to above 80%. In five suburbs, a small number of supplementary releases were undertaken to ensure establishment. In all suburbs, the infection frequency has remained stable without any signs of Wolbachia being lost from the mosquito population (Figure 3).
Laboratory experiments have suggested that maternal transmission of wMel can become unstable in Ae. aegypti at high temperatures and plausibly might limit the field usefulness of the wMel strain 17 . The temperatures used in these incubator experiments were meant to mimic larval rearing temperatures in north Queensland. However, our field data shows long-term stability of wMel, presumably because temperatures used in this study were not truly representative of those experienced by mosquitoes in the field. We assume that mosquitoes predictably seek out non-stressful microhabitat when it exists 18 and larval rearing temperatures do not mirror measured ambient temperatures. Empirical data from this study and other sites 9 suggests that wMel is much more robust to deployment than predicted by 17.
A key feature of using MRCs for mosquito releases is the possibility of mobilising the community to undertake the deployment instead of employed program staff. In stage 1 of the release program staff undertook the deployment by setting and maintaining MRC buckets themselves. In stages 2-4 we used a blended approach of community members setting their own MRCs and then program staff members supplementing these deployments by distributing additional MRCs to meet the target of 20% of residences, to ensure adequate coverage without major spatial gaps. Community-based releases were undertaken in three ways; school programs where students were given MRC kits to take home, direct community releases where MRC kits were given to householders who had signed up to participate through community engagement activities, and finally by having large employers within the city distribute MRCs to staff who were willing to participate. Of the three methods, providing MRCs directly to the community was the most cost effective. It also allowed for more targeted deployment and better coordination with field staff, ensuring adequate coverage across a suburb. This blended approach of community-based deployment supplemented with programmatic targeted deployment is considered the most appropriate for future large-scale operations. The schools program -while being less efficient and costlier -proved to be an excellent community engagement vehicle, with the release outcome of secondary importance. Its success was highly dependent on working with an actively engaged teacher who could serve as a champion for the program.
Episodic outbreaks of locally transmitted dengue have occurred annually in Townsville since 2001. Outbreaks occur against a background of regular importations of dengue into Townsville by international travellers (Figure 4). In the period since  In none of the previous 53-month moving windows since 2001 were there fewer than 69 locally-acquired cases notified. Importantly, only one of the four local cases since November 2014 was resident in an area where Wolbachia had been established. However, public health investigation found that this case was highly mobile and therefore the likely place of acquisition was uncertain. The model-based estimate of intervention effect from the interrupted time series analysis suggests a 95% reduction in dengue incidence in Wolbachia treated populations (95% confidence interval: 84-98%), adjusted for season, imported cases, and allowing for temporal autocorrelation of cases (Table 5). These findings, coupled with continuous validation of the impaired vector competence of wMel-infected Ae. aegypti in release areas 19 , represent empirical epidemiological evidence consistent with modelling projections of wMel-mediated elimination of dengue transmission in most settings 6 .
The cost of undertaking the program per person, and per km 2 , varied between stages, and when time to complete each stage was also considered stage 2 was most efficient (Table 3).
Considering the low population density of this city we expect the cost per person, for the same deployment methodology, would be dramatically reduced in many tropical cities with much higher population densities. Furthermore, the costs for the deployment in Townsville were inflated as the work was undertaken as a research activity, with much more monitoring than would be expected in an operational public health intervention. The breakdown of costs by major activity are shown in Table 4. Community engagement activities accounted for a significant part of the cost of deployment, which shows the prioritization of and importance given to these activities by the World Mosquito Program. This, together with the cost of deployment (staff, vehicles etc.), accounted for more than half the cost of the implementation, and represents the areas where significant cost reductions might occur in future operational deployments. Given the costs for this study, and considering that future deployments should utilize less monitoring and occur in settings of higher population density, we estimate that deployment cost should be able to be reduced to less than US$1 per person. Additionally, in contrast to most other interventions, this cost should not be ongoing since once Wolbachia is introduced it is expected to maintain itself in populations. This suggests that the use of Wolbachia for arbovirus control as described in this study has the potential to be an extremely cost effective intervention compared with existing methods and many other proposed interventions that feature the release of modified mosquitoes 10 .
This study demonstrates that: the wMel strain of Wolbachia can be deployed effectively across large geographic areas at low cost; that once the intervention is deployed it is stable and self-sustaining; and that communities are accepting of the release of mosquitoes and are willing to participate in deployments when effectively engaged. From this study, we were able to identify a number of key learnings to take into future studies. These include: the understanding that community engagement approaches can be successfully scaled without compromising their quality, that shipping eggs from a remote production facility is possible but that care is needed with the shipping method to avoid excessive mortality, that managing egg strips for quality and to estimate quantity was laborious and a key step to improve in future scale-up. Finally, a time series analysis of notified dengue cases within the city over a 18-year period is consistent with modelling predictions of a large impact on dengue transmission 6 -and indeed in this city the observational data is consistent with elimination of local transmission.

Data availability
The data underlying Figure 3

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Wolbachia in Aedes populations to suppress dengue transmission. Nature. 1.

Overall comments:
The manuscript "Scaled deployment of to protect the community from transmitted Wolbachia Aedes arboviruses" reports a successful strategy for the long-term establishment of on a citywide Wolbachia scale. Previous pioneer work from this group have shown that Mel strain can establish Aedes w Wolbachia itself in the field after surpassing a certain unstable equilibrium threshold (Hoffmann et al., Nature 2011). However, in the current study, the establishment of is done at a larger scale, across 66 km of Wolbachia Townsville, a city in northern Australia with a population of approximately 187,000 residents. Interestingly enough, the authors also report details on a community engagement approach used to achieve establishment in a cost-effective manner. Wolbachia Remarkably, the results presented here provide the first field-based evidence that the self-sustainable deployment strategy works against dengue. That being said, this study was not designed as a Wolbachia clinical trial experiment, as pointed out by the authors. Although extremely promising and exciting, the results presented here should not be taken, at least at this point, as a definitive evidence that Wolbachia will block dengue transmission in endemic areas. Townsville, has a relatively low number of dengue cases a year, compared to a great number of cities in tropical and sub-tropical areas of the globe. On that regard, this non-profitable consortium has ongoing field-trials in places like South America and Asia, serious contenders to the -based strategy. Wolbachia

Major comments:
The reader is left wondering if non-factors in Townsville (e.g. climate) are contributing to Wolbachia the remarkable decrease of reported locally acquired dengue cases after release Wolbachia (2016-2018). This work would greatly benefit from providing additional epidemiological data. Showing the dengue cases for other cities in Northern Australia with no release over the Wolbachia same period as shown in Figure 4 (2002-2018). This data could be incorporated in Figure 4 as an additional histogram. Ideally this histogram should show an average of locally acquired dengue case notifications from 2002 to 2018 from several cities with epidemiology similar to Townsville from 2002 to 20014 (prior to release). Although not a rigorous control, this data would at Wolbachia least give some indication if non-factors in that region (may be climate?) could be Wolbachia playing a role for the dengue decrease.
(note added in proof: the other reviewer also had a somewhat similar comment).
A small discussion of the challenges faced by their approach as well as lessons learned from Townsville that should be considered in areas were deployment is imminent would be Wolbachia valuable to the field.

Minor comments and suggestions by page
Original text in Main Text: italics : Title: P. 1 Scaled deployment of Wolbachia to protect the community from Aedes transmitted arboviruses Although there are several studies demonstrating efficient pathogen blocking effect of against Wolbachia several arboviruses transmitted by , including many from this research group, this specific Aedes aegypti manuscript only shows epidemiological data on dengue. Therefore, the title would be more precise if changed to reflect the specific arbovirus evaluated.
: Rearing (2 paragraph): " …" P.4 The wildtype colony was backcrossed for three generations Although there is data pointing towards the inexistence of mutations in the local population kdr Ae. aegypti present in Queensland, indicating susceptibility to pyrethroid insecticides, there are several recent reports indicating the sporadic detection of non-native mosquitoes carrying insecticide resistance Ae. aegypti alleles not found in Australia. Areas with intense international flux like seaports and airports are a point-of-entry for these alleles into the local population (Endersby-Harshman ., 2017 ). How is the et al WMP taking the potential risk of insecticide resistance into consideration when rearing their mosquitoes for field releases? Few sentences regarding this aspect would be helpful, given that insecticide resistance alleles could highjacks establishment in areas heavily treated with insecticides by indoor Wolbachia residual spraying (IRS). Are the mosquitoes selected for backcrossing checked for chemical compounds resistance? What is the level of synchrony between the WMP approach and the guidelines established by the Queensland dengue management plan 2015-2020 which indicates the use of IRS as an approach against disease vectors?
: Rearing (2 paragraph): " P.4 In order to prevent inbreeding, 10% wildtype males were added to each ." generation of the 'broodstock' To avoid inbreeding is of great relevance. That being said, is this value of 10% resulted from a pool of males collected across all the same 49 sites in Townsville, the same used to establish the original wildtype colony? Additionally, it is to be expected that at a certain point, given the establishment of in the field, that the males collected would harbor the bacterium. Did the authors screened a Wolbachia portion of the males added to the cages after the copulation period?
: Rearing (7 paragraph): " P.5 Wolbachia infection was also tested each week of production, 80 females …" and 80 males were screened from each broodstock cohort These numbers of female and males tested represents what percentage of the total population? Given the potential for maternal transmission leakage of when considering the Wolbachia transmission dynamics of this bacterium, something speculated to be one of the factors contributing to the difficult establishment of in Cairns, another area where releases by Wolbachia the WMP took place (Schmidt ., 2018 ), it is somewhat unexpected that the field colony was et al not screened prior to release, only the broodstock. What would be the reason for that? May be the consistency by which the mosquitoes were reared under laboratory conditions? : Rearing (8 paragraph): " P.5 The colony was therefore refreshed with 2500 males and 2500 females ." each week The same question previously asked ( : Rearing (2 paragraph). Are these mosquitoes obtained from P.4 a pool across all collection sites or derived from a single site?
: Rearing (8 paragraph): " P.5 This was achieved by placing suspected male pupae based on size into …" cups of 10; Given the difficulty of visually sexing each pupae cup without the aid of software and hardware-based engineering, what was the overall confidence level in this sex by size strategy? Although there is no risk for the community, as the CI and female-based deployment deals with the issues associated with accidental female release in this case, it would be interesting to address this aspect of the method used by the research team. : Mosquito releases (3 paragraph): " P.7 For stage 1, it required between 7 and 19 weeks of releases for ." each suburb to reach that target It is known that the rate of dispersion of correlates with the human density in a given area. As Ae. aegypti such, how long did the authors wait to start screening field collected mosquitoes, as a way to avoid screening the same mosquitoes that were released in a particular area? It is not clear how far the BG traps were set apart from the MRC's.
: Mosquito releases (2 and 4 paragraphs): In summary, MRC's had 100 eggs / 1L of water and 5 P.7 (summer) / 6 (winter) wafers of Aqua One vege wafer fish food, while Mozzie boxes had 100 eggs/ 400mL of water and 4 (summer) / 5 (winter) wafers of Aqua One vege wafer fish food. In terms of quality assessment of the fitness of the mosquitoes being released, how was the comparison between the MRC's and the Mozzie boxes? How was the level of engagement along the three cycles? Given the author's interesting approach, this information could be helpful as a proxy for the predicted efficacy of this strategy in release areas worldwide.
: Results and discussion (3 paragraph): " P.12 In general, establishment of Wolbachia occurred reliably " after releases stopped once the 50% threshold was met.
To date, does the WMP continues to screen these areas where was released roughly 2-3 Wolbachia years ago? If the answer is yes, does the infection frequency still high in these areas? I am particularly interested in the Condo area, where the last data point shows an infection frequency of 51.79%.

Final thoughts and suggestions:
The following questions are not within the scope of what is proposed for this manuscript, just a couple suggestions.
Have the authors considered screening areas where deployment did not occur but were Wolbachia rd nd th th nd rd rd 1. 2.

3.
We confirm that we have read this submission and believe that we have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however we have significant reservations, as outlined above. Wolbachia-mediated control is one of the most exciting recent developments in the struggle against dengue and other mosquito borne diseases. We already know from this group's previous work in Cairns that the replacement of wild-type populations with Wolbachia-infected forms is both Aedes aegypti feasible and sustainable in towns of around 150,000 people.
Epidemiologically, it would have been more interesting to see a summary of impacts on dengue transmission in Cairns as, historically, that town experiences far more dengue transmission and more dengue imports than Townsville.
The novelty in the current report is that this is the first "citywide scaled deployment". This scaling refers to the direct involvement of the community and local school children in executing the releases.

Comments:
I would have liked a little more background and explanation on the following: What is the rationale for backcrossing the wMel strain with a local wildtype for three generations prior to release? Is there any underlying empirical basis for this in terms of fitness and/or genetic homogeneity?
It appears from Fig 3 that, by the end of the monitoring period (mid 2016), very few mosquitoes were being captured and that almost all were Wolbachia-infected. Was that very low density a result of a hostile climate? Is it possible that mosquito suppression as well as replacement is having an impact on transmission here? Is declining mosquito density a feature of wMel establishment? Were any Aedes endemic, non-release areas monitored for comparison?
The authors state that the outcomes of releases by schools-based programs were of secondary importance to their value as instruments of engagement. In this paper, it is flagged as a major component of "scaled deployment" so it would have been interesting to report on the operational success of the schools-based program. Was there any evaluation of compliance (observation of the releases made by children) or of the rates of Wolbachia replacement in areas specifically 4.

7.
the releases made by children) or of the rates of Wolbachia replacement in areas specifically targeted by schools?
It seems likely that Wolbachia is reducing transmission risks in Townsville, but other contributing factors may be being ignored. My understanding is that releases of Wolbachia in Townsville have coincided with some of the hottest and driest years on record. Climate has a direct impact on mosquito survival and is strongly correlated with transmission.
There's no real discussion of limitations or challenges here. The authors state that economies of scale, in regions of high population density, will result in successful deployments costing less than $1 per capita. The assumption is that population replacement and dengue-blocking across hyperendemic urban sprawls will be cheap and simple.
Townsville has very limited dengue transmission and adult indices appear lower than for many tropical cities. Does the WMP not see some issues with extrapolating successes in Townsville to the rest of the world?
The authors dismiss work on Wolbachia loss and heat stress (Ulrich et al PLoS NTDs, Ross et al PLoS Pathogens) but Aedes do demonstrate fast and successful development in the field at water temperatures ≥ 30 C and the truth is that we don't know much about the operational impacts of those conditions on Wolbachia stability.
This paper certainly suggests that, in Townsville, Wolbachia-infection is stable. But in Townsville monthly average high temperatures range between 31.5 and 25.1 C. In Bangkok, they range from 35.4 to 31.7 C. A global Wolbachia operation will at least need to consider the potential impacts of high water temperatures in aquatic habitats.
What are the factors that result in a minority of sites hovering around 70% Wolbachia coverage? Is it immigration, or the presence of some protected, wild-type reservoir. Or is it related to mosquito density and a relatively small number of dispersal events?
Overall, this paper represents a keenly awaited progress report on the stability and feasibility of Wolbachia releases in Australia and the ways in which those releases might be made more cost effective. The reviewer's statements around novelty are incorrect in stating that we already know that we can deploy in cities of 150,000 by referring to deployments around Cairns, Australia. The only studies published from Cairns relate to small pilot deployments. Many small pilots have been undertaken in Cairns and surrounding areas since 2011 that have focused on obtaining data on the best ways to deploy mosquitoes. In the last 2 years, we have "filled in" around these pilot areas to provide broad coverage to all of North Queesnland's major dengue risk areas. This study is actually the first published study of a citywide deployment being undertaken as a single project which shows that a "greenfield" city can be engaged and the intervention deployed at scale, cost effectively and quickly. The use of the community to deploy is an additional component of the paper but not the primary aim.

Are sufficient details of methods and analysis provided to allow replication by others? Yes
We have a companion paper in preparation that will report the results from deployments in the city of Cairns that will reinforce the findings of this study.

Specific Comments
1. We undertook three generations of backcrossing to make sure our release strain closely approximated the genetic background of the Townsville target population. This was done as a precautionary measure and was not based on specific empirical data that aimed to characterise any differences between the Cairns and Townsville genetic backgrounds of the resident mosquitoes.
2. Figure 3 shows total counts of mosquitoes that were run through diagnostics in a given sampling period. This is presented to help provide an estimate of sample size that underpins a given frequency estimate. There is a correlation between this number and actual mosquitoes caught but it cannot be used to estimate population size because the number of traps contributing the sample was variable. We reduced BG traps in a given area after Wolbachia was considered established (often by more than 60%) and these traps were moved to areas where active releases were being undertaken. As a result, the impression that mosquito populations declined after release is not accurate. Modelling predicts that mosquito populations should reduce slightly after the introduction of Wolbachia but we did not attempt to measure that in this study 3. Because the schools program used volunteer students to undertake the releases and these children were scattered geographically they were supplemented with programmatic releases in children were scattered geographically they were supplemented with programmatic releases in adjoining areas which made it impossible to compare their effectiveness from an entomological perspective with purely programmatic releases. We did undertake QA procedures on a sample of student release containers to evaluate the program. The major consideration for us was that this form of release was quite expensive compared to other forms so from a purely economic perspective it was inefficient. However, its value to us was more from a community engagement perspective as Schools are fundamentally trusted in the community and it was an indirect method to engage parents of school children through undertaking the releases and increase community awareness of the program. 5. The main assumption underlying reduction in costing in based on the fact that major urban tropical cities have much higher population densities than Australian cities but many of the costs in deployment relate to the area of deployment. This means that costs will automatically decline in areas of higher population density and that is what we have seen in other work of the WMP where deployments have already been undertaken successfully as part of clinical trials that are underway in other countries.
6. Our criticism is that the heat stress experiments have been undertaken either in incubators or in semi-field settings and then the results extrapolated to infer a potential problem in stability of Mel w in the field. However, the field is much more complex than an incubator and many breeding sites are cryptic and difficult to assay, so the true water temperature experienced by a population of mosquitoes in the field cannot be accurately measured. What we do know from this and levels of insecticide resistance. These approaches and results will be presented in future papers describing results from those study sites. We feel that it would be better to address those topics in those papers rather than in this paper that did not examine insecticide resistance deeply.
P4. Our collections of wild type material were made from pooling across many ovitraps placed in areas where Wolbachia had not been released as we did for the initial backcrossing. Our goal was to generate genetically diverse material so it was pooled. Given that we could store the eggs we were able to maintain material for outcrossing during the release period as described in methods. As a result, we did not monitor for Wolbachia in the males. In future studies, there is no reason to exclude already treated areas or for the included males not to be Wolbachia infected. We have added text to the revised paper to improve clarity.
P5. Wolbachia infection testing -1. We tested 160 individuals from each cohort. A cohort was on average around 7500 individuals so around 2% of individuals. The sample size of 160 individuals was determined from a prior power analysis. 2. We have seen no evidence for reduced maternal transmission in our mass production.
P5 -Yes these mosquitoes were derived from the same material used to outcross the colony in our Melbourne labs as well as backcross.
P7 -Screening started prior to releases starting and was undertaken weekly. MRCs had variable placement due to the dependence f volunteers to host MRCs. It was not possible with that system to designate fixed differences between MRC's and BGs. However, if you examine the shape of the establishment curves you can see that in most suburbs catching back of release material was not a big issue. If it was then you would expect to see Wolbachia frequencies climb quickly and then potentially decay after releases stopped. We only see those patterns in areas where adult releases took place.
P7 -No fitness comparisons were made between the two but regular QA procedures indicated that emergence rates were similar between the two. P7 -The mixed approach used in the final experiments was undertaken due to operational issues with the completion of the projects and staff contracts completing. This was done to complete the study in time to coincide with staffing contracts finishing and was not set up as a comparison -it can't really be analysed that way as a result. Comparisons between deployment methods are being undertaken in other sites in a more rigorous manner and these will be able to report on this comparison.
P8 -Generally we found difficulty in sustaining interest in repeating the releases by the third round and our engagement officers needed to work harder to maintain interest levels in students for later round releases. As indicated in the paper we felt that school releases were better primarily as an engagement tool and direct public releases more effective from an establishment and cost perspective.
P12 we have updated the figures with the latest monitoring data across the city to provide a more recent view of establishment success and dengue cases. Later monitoring has involved LAMP diagnostics and this protocol has also been included in the methods description. We have also updated the epidemiological figure with the latest data showing an extension of post release monitoring till Oct 2018 and continued impact on locally acquired dengue cases.