Establishment of wMel Wolbachia in Aedes aegypti mosquitoes and reduction of local dengue transmission in Cairns and surrounding locations in northern Queensland, Australia [version 2; peer review: 2 approved]

Background: The wMel strain of Wolbachia has been successfully introduced into Aedes aegypti mosquitoes and subsequently shown in laboratory studies to reduce transmission of a range of viruses including dengue, Zika, chikungunya, yellow fever, and Mayaro viruses that cause human disease. Here we report the entomological and epidemiological outcomes of staged deployment of Wolbachia across nearly all significant dengue transmission risk areas in Australia. Methods: The wMel strain of Wolbachia was backcrossed into the local Aedes aegypti genotype (Cairns and Townsville backgrounds) and mosquitoes were released in the field by staff or via community assisted methods. Mosquito monitoring was undertaken and mosquitoes were screened for the presence of Wolbachia. Dengue case notifications were used to track dengue incidence in each location before and after releases. Results: Empirical analyses of the Wolbachia mosquito releases, including data on the density, frequency and duration of Wolbachia Open Peer Review

mosquito releases, indicate that Wolbachia can be readily established in local mosquito populations, using a variety of deployment options and over short release durations (mean release period 11 weeks, range 2-22 weeks). Importantly, Wolbachia frequencies have remained stable in mosquito populations since releases for up to 8 years. Analysis of dengue case notifications data demonstrates nearelimination of local dengue transmission for the past five years in locations where Wolbachia has been established. The regression model estimate of Wolbachia intervention effect from interrupted time series analyses of case notifications data prior to and after releases, indicated a 96% reduction in dengue incidence in Wolbachia treated populations (95% confidence interval: 84 -99%). Conclusion: Deployment of the wMel strain of Wolbachia into local Ae. aegypti populations across the Australian regional cities of Cairns and most smaller regional communities with a past history of dengue has resulted in the reduction of local dengue transmission across all deployment areas.

Introduction
The wMel strain of Wolbachia has been successfully introduced into Aedes aegypti mosquitoes and subsequently shown in laboratory studies to reduce transmission of a range of viruses including dengue, Zika, chikungunya, yellow fever, and Mayaro viruses that cause human disease (Aliota et al., 2016a;Aliota et al., 2016b;Amuzu et al., 2015;Caragata et al., 2016;Caragata et al., 2019;Carrington et al., 2018;Dutra et al., 2016;Ferguson et al., 2015;Frentiu et al., 2014;Kho et al., 2016;Moreira et al., 2009;Pereira et al., 2018;Rocha et al., 2019;Tan et al., 2017;van den Hurk et al., 2012;Walker et al., 2011;Ye et al., 2013;Ye et al., 2015;Ye et al., 2016) Early field trials involving releases of Wolbachia infected Ae. aegypti mosquitoes into two isolated communities in northern Australia showed that the wMel strain of Wolbachia could be deployed and establish in the local mosquito populations with full community support (Hoffmann et al., 2011) and persist (Hoffmann et al., 2014). Further it was shown that the dengue blocking properties of these mosquitoes remained stable several years after establishment (Frentiu et al., 2014). Additional releases into a number of urban settings in Cairns in northern Australia subsequently assessed the effects of release area size and landscape features on Wolbachia establishment and spread into the mosquito populations (Schmidt et al., 2017). Subsequent city-wide Wolbachia mosquito releases were undertaken across the medium-sized city of Townsville in northern Queensland, resulting in successful establishment of Wolbachia in local mosquito populations and complete elimination of local dengue transmission (O'Neill et al., 2018).
All prior field studies resulted in a patchwork of deployments of various sizes across areas of north Queensland, Australia (Hoffmann et al., 2011;Hoffmann et al., 2014;O'Neill et al., 2018;Schmidt et al., 2017). From 2013 to 2017 the World Mosquito Program (formerly known as the Eliminate Dengue Program) undertook a series of additional deployments in nearly all significant dengue transmission risk areas in Australia where wMel Wolbachia had not yet been deployed. These releases involved a variety of release methods, including releases of eggs and adult stages directly by project staff and through community assisted methods involving school children, businesses, community groups, and individual householders. Here we describe the entomological and epidemiological outcomes of this work.

Intervention area
Wolbachia mosquito releases were undertaken across four local government administrative areas in northern Queensland, Australia: the Cairns Region, the Cassowary Coast Region, the Douglas Shire and the Charters Towers Region (Table 1, Figure 1-Figure 3). Within each region, the locations for Wolbachia mosquito releases were selected based on historical dengue case reports, human population density, reported presence of Ae. aegypti and logistical considerations. Depending on the specific objective of each Wolbachia release, for example small scale releases to test different release methods (e.g. egg release trials in Stratford 1-3, Cairns North, Bungalow 1-3, Table 1, Figure 2), or an area-wide release across the entire location (e.g. adult releases across 23 Cairns suburbs between Nov 2016 and Jun 2017, Table 1, Figure 1B, Figure 2), each area was mapped and the target release areas (generally the residential areas and some business areas) were identified. Areas deemed unsuitable for Ae. aegypti, such as uninhabited forested and vegetated areas, open or vacant areas, sporting fields, large industrial and commercial areas, agricultural and farming areas, and major transport infrastructure (major roads, highways, railways and airports) were excluded from releases. The bounded size (km 2 ) of each release area was calculated, along with the residential population and the number of households (Table 1).

Community engagement
For Wolbachia mosquito releases between 2011-2014, communication and community engagement activities followed the approach described in Hoffmann et al. (2011). This included consultation with key stakeholders and community groups, one-on-one meetings, displays at community events and centers, and door-knocking and mail-outs to householders to assess support for and participation in Wolbachia mosquito releases. Prior to releases residents were asked to provide permission for release of Wolbachia mosquitoes, either as adult stages from the footpath near their house, or as eggs that were placed into containers by program staff in outdoor shaded areas at their properties. During the release period, we continued to undertake close engagement with the local communities to ensure that residents were informed about the study and were comfortable with continuation of release activities. This included random household surveys, meetings with reference groups of residents and community leaders, and ongoing promotion of a free phone number and accessible city project office in Cairns. Periodic result updates were provided to the communities through letterbox leaflet drops, attendance at community events and meetings, paid advertisements in local newspapers, community newsletters and radio and television media outlets. Residences of a small number of people that did not wish to participate were excluded from release activities.
For Wolbachia mosquito releases between 2015-2017, community engagement activities followed the Public Acceptance Model (PAM) as described in O'Neill et al. (2018). The PAM was comprised of the following four components: 1. Raising broad community and stakeholder awareness. Information was provided to residents and key stakeholders about Wolbachia and mosquito release and monitoring activities via various channels, including mass communication (newspapers, media events), school outreach programs and social media, and direct engagement through face-to-face meetings, stalls at community events and presentations at existing community networks and meetings, information kiosks    and traditional electronic mails outs of information letters and updates.
2. Quantitative surveys to assess community awareness and support. Pre-release surveys were conducted by an external market research company (Compass Research) (see methods in O'Neill et al., 2018), and were undertaken from July 2016 (Table 2). In Charters Towers, Douglas and Cassowary Coast shires additional baseline surveys were undertaken prior to commencement of communication and engagement campaigns, along with an initial survey in Cairns in Nov 2013. Each survey involved 100-300 participants ( Table 2).
3. Establishment of an issues management system. The system enabled community members to easily contact the program with any questions and concerns and have them quickly addressed by program staff typically within 24 hours of receipt. The system also allowed residents to opt in or out of direct participation in release and monitoring activities.
4. Community reference group. Community reference groups were established in each location, with respected community members from key stakeholder groups ( Table 2). The reference group's function was to independently review activities to ensure that engagement was carried out in accordance with our stated Public Participation Principles (O'Neill et al., 2018).
In Cairns, the implementation of the PAM commenced in 2015 and included engagement of traditional mass media, establishment of the Cairns sign-up website (where residents registered their interest in participating in the project), leveraging of existing networks to spread information (particularly through educational institutions), direct-to-premise informative mailouts, information kiosks at community events and public locations, engagement of high-level stakeholders, such as government representatives, indigenous interest groups and environmental advocacy groups, establishment and maintenance of independent community reference groups, maintenance of community feedback channels (telephone survey and online forms), and distribution of quarterly field trial updates to participants and stakeholders. In Charters Towers, Douglas Shire and the Cassowary Coast the PAM was implemented prior to releases.   (Hoffmann et al., 2011). Cages contained up to 10,000 wMel-infected Ae. aegypti and were provided access to human volunteer blood-feeders almost daily. To minimize laboratory adaptation, male Ae. aegypti pupae/adults from field collected Ae. aegypti (F1-F2 eggs) were introduced into the colony each generation, so that they constituted around 10-20% of the new male population. Eggs were collected on flannel cloth in plastic buckets acting as oviposition sites that were spread throughout the cage, incubated for 3 days and then stored in the laboratory.

Release colony maintenance. Two
From 2016 the wMel-infected mosquito colony was maintained at Monash University, Melbourne. The Townsville Ae. aegypti wMel-infected line was established by backcrossing the Cairns Ae. aegypti wMel-infected line to the offspring of Townsville collected wildtype mosquitoes for three generations (O'Neill et al., 2018). Both wMel-infected lines were maintained in controlled laboratory conditions, in 30 cm 2 mesh-sided rearing cages (see description of methods in O'Neill et al., 2018). Each cage contained ~600 adults, and was fed using human volunteers once per week for two gonotrophic cycles. These colonies comprised of a broodstock, and a release-production colony. Male Ae. aegypti adults (from F1-F3 field collected material) were introduced into the broodstock cages at a rate of 10% every generation. Material from the broodstock colony was then transferred to the release-production colony where material was amplified through one generation without the addition of field collected males. Eggs were collected on red flannel cotton strips, and were matured for four days before being dried. Once the drying process was complete (O'Neill et al., 2018), eggs were packed and shipped to field sites.

Adult mosquito rearing for releases.
Between 2011-2013, adult mosquitoes for releases were produced using previously described methods (Hoffmann et al., 2011). Briefly, immature stages were reared in 3 L buckets with 2 L of water and fed a diet of Tetramin For the 2016 Charters Towers releases, wMel-infected mosquito eggs were produced at Monash University and were then shipped to Townsville where the eggs were hatched and the immatures stages were reared to adults in cups in a laboratory maintained at 25-28 °C. Egg strips were either hatched in tap water containing Aqua One Vege Wafers (Aqua Pacific UK Ltd, Southampton, UK, Product number 26050), and two days later approximately 100 larvae were aliquoted into individual release cups (paper drinking cups, 550 mL volume, 100 mm width × 180 mm height, C-DC9787, FPA, Australia) each containing 360 mL of tap water, or egg strips were cut into sections based on the required number of eggs to produce a target hatch of 100 larvae and the eggs were added directly to the individual rearing cups. Immatures were fed Aqua One Vege Wafers (1.5 wafers upon setup and 1.5 wafers on day 4).
A mesh cover was placed on each cup and adults were maintained for 4-6 days on a 50% honey solution. Release cups were transferred to plastic tubs and transported to Charters Towers for release the following day between 0800-1000 hours.
For the 2016-2017 Cairns releases, adult mosquitoes were reared as above, with initial rearing being undertaken at the James Cook University insectary and this transitioned to a laboratory in Cairns, maintained at 25-28 °C, in early 2017.
For the 2017 Douglas Shire releases, wMel-infected mosquito eggs were produced at Monash University and were then shipped to Port Douglas where the eggs were hatched and the immatures stages were reared to adults in a laboratory maintained at 25-28 °C. Rearing followed above procedures except that all eggs were hatched in water containing 1 Tetramin Tropical Tablet, and two days later approximately 100 larvae were aliquoted into individual release cups (Plastic [PET] drinking cup, 425 mL volume, Detpak, Australia) each containing 300 mL of tap water. Plastic (PET) lids (Detpak, Australia) were placed on top of a mesh cover on each cup and adults were maintained for 4-6 days on 50% honey solution. On the morning of release, release cups were transferred to plastic tubs and transported to the field for release between 0800-1000 hours.

Adult mosquito releases
Between 2011 and 2015, weekly releases were undertaken on a per household basis, at a density of 1:3 to 1:10 houses. Between 2016 and 2017, releases were undertaken on an area basis, where the target release area was divided into a series of 100 m × 100 m grids, with a single release point located inside each grid. Adult mosquitoes were transported to the field in release cups in vehicles and released on a single day each week, normally at the property line or front yard by removing the container lids and gently shaking. Releases were generally undertaken between 0800-1600 hours each day.
Between 2011 and 2013 the duration of releases was fixed to between 9 and 16 weeks, except for an initial trial of egg releases in Stratford (SF3) where the methodology was being optimized over a longer period (23 weeks). Shorter duration release periods (7-or 8-weeks) were trialed in Charters Towers and Douglas Shire in 2016, and these were extended in Tully and Innisfail (Cassowary Coast) to 12-week releases and 14-to 16-week releases, respectively, in 2017. The 2016-2017 Cairns releases were staged, with releases continuing in a release area until the frequency of Wolbachia in samples of field-caught mosquitoes was above 50% for two consecutive weeks. The duration of releases in these areas varied from 10-12 weeks, except for Cairns North Ext (CNX) where only two weeks of releases were undertaken. Releases in this area were discontinued after this time because the Wolbachia infection frequency in mosquitoes already exceeded 90% following the spread of Wolbachia from nearby release areas. Details of the releases in each area are summarized in Table 1.

Direct egg releases
Small-scale field trials to develop and optimize the egg release methods were undertaken in parts of Babinda and Machans Beach in 2013, and in Bungalow (BU1-3), Cairns North (CN1) and Stratford (SF1-3) in 2014. For the 2013 trials, 3 L white polypropylene buckets with lids (Piber Plastics, Australia) were used ( Figure 4A). Each bucket had four 6 mm holes drilled 20 mm apart in a square pattern in the side. Tangle-Trap Sticky Coating (Tanglefoot, USA, Product number 300000588) was applied around the perimeter of the emergence holes (1-2 cm away from holes) to prevent the entry of ants into the buckets. The buckets interiors were roughened with sandpaper to provide perching for mosquitoes post-emergence. Containers were filled with 2 L of tap water and 1 lucerne pellet (0.5 g) (LLP, Carole Park, Australia) and 1 TetraMin Tropical Tablet (broken into four pieces) were added, along with an egg strip containing approximately 150 eggs (target emergence rate of 100 adults per container). Containers were placed in a shaded position near the front boundary of selected houses. Containers were serviced every 2 weeks, at which time live/dead larval/ pupal counts were made and the general condition of the container (tipped over, dry/empty, fouled water) was recorded for quality assurance. During the cooler months, immature development in some containers was slowed and necessitated leaving containers in place for 3 weeks prior to servicing. Once immature development was complete, the remaining contents of the containers was discarded and the inside of the container was cleaned with a sponge to remove any residual eggs and larvae/pupae. The container was then refilled with clean tap water, and new food and eggs were added as described above.
The 2014 egg release trials were similar to the above, except these involved 2.3 L plastic buckets containing 1 L of tap water. An alternative feeding mixture involving ground red kidney beans (1.25 g per container) was used in these trials. An assessment of egg hatch rate (hatch rate quality assurance) was undertaken in the laboratory prior to each release round and this information was used to calculate the required numbers of eggs to be added to each container. The target emergence rate for these trials was 75 adults per container.
For the 2015 egg releases, the conditions were the same as the 2014 releases except Aqua One Vege Wafers were used as a food source (4 wafers per containers in spring/summer/autumn months, and 5 wafers per containers in winter months), and each container received 100 viable eggs).  The Townsville Public Health Unit (PHU) provided line-listed data on addresses of all dengue cases notified from THHS and CHHHS, from the operational databases of the Cairns Tropical Public Health Service and the Townsville PHU. This data included one or more addresses per case, with an indication of the primary residential address, and of any address that had been classified as the probable location of dengue acquisition or a possible site of exposure, during the course of public health follow up. We linked the PHU dataset to the NoCS dataset by unique case notification identifier. All NoCS records were retained, with or without a linked PHU record. PHU records without a linked NoCS record were excluded, as NoCS is considered the master source of case notifications data, and holds the travel history variable to distinguish locally-acquired from imported cases.
Approval to access anonymized spatially identifiable dengue case notification data, collected as part of routine disease surveillance, was obtained from the Townsville Hospital and Health Service human research ethics committee (HREC/16/ QTHS/108) and research governance office (SSA/16/QTHS/238), and from the Office of the Director-General, Queensland Department of Health, under the Public Health Act 2005.
Classification of Wolbachia exposure status A case's location, for the purpose of classifying Wolbachia exposure status, was determined using geolocatable address information from the PHU operational database. The address indicated in the PHU dataset as the probable location of dengue acquisition was used where available (53%); if unavailable then the primary residential address was used (22%). For 20% of cases a geolocatable address was available in the operational database but was not designated as either 'acquired' or 'residential', and for the remaining 4% no geolocatable address was available in the PHU database, and the suburb of residence from the NOCS dataset was used to define the case's designated location. Case geolocations were overlaid with Wolbachia release area boundaries in ArcMap (version 10.5, ESRI, Redlands, USA) (QGIS is an open-access alternative) to classify the Wolbachia exposure status of each case at the time of case onset. An area was considered exposed (post-intervention) from the date of last release. Five 'non-release areas' in central Cairns, where BGS monitoring indicated that Wolbachia had established by dispersal from neighboring release areas, were considered exposed from the inferred date that the Ae. aegypti infection curve crossed 80%. The local government area (LGA) of each case was determined from its location, as classified above, and any cases located outside of Cairns, the Cassowary Coast, Charters Towers or Douglas LGAs were excluded from the analysis.

Population data
The populations of each release area and Wolbachia-exposed non-release area ( Figure 1-Figure 3) were estimated by aggregating from mesh blocks (ABS, 2016) to the boundaries of each intervention area (Table 1), in ArcMap. In all but three release areas, the boundaries of the release area and monitored area were aligned, and defined the Wolbachia-exposed population for the purpose of epidemiological analyses. The exceptions were Gordonvale, Yorkeys Knob, and Babinda, where monitoring extended beyond the boundaries of the release area and demonstrated Wolbachia establishment throughout this extended area. The population denominator for these three areas was therefore calculated for the larger monitored area (Table 1 footnote). A further exception was Stratford, where concurrent releases were conducted in three small noncontiguous areas, which represented ~75% of the area and 90% of the population of the suburb of Stratford. For the purpose of epidemiological analysis, the whole suburb population was considered Wolbachia-exposed from the completion of the last releases in the three release zones. For the interrupted time series analysis, monthly aggregate treated and untreated areas (and their resident populations) were calculated, dynamic over time, with the treated area in any given month defined as the total area where Wolbachia deployments had been completed to date (or where, for the five central Cairns non-release areas where Wolbachia established, the inferred local Wolbachia frequency had reached 80%, as above).

Statistical methods
Locally-acquired and imported dengue case notifications were cross-tabulated by month of illness onset and LGA. Scaled 'time-since-release' (TSR) was calculated for each case located within a Wolbachia exposed area, as (date of case onset -date of last release in the local release area) rounded to the nearest month. TSR is equal to zero for cases with onset in the same month as releases were completed, and is positive for cases with onset post-intervention. Cases located in Wolbachia non-exposed areas were excluded from the TSR analysis. The staggered deployment of Wolbachia across release areas from January 2011 to May 2017 means the distribution of pre-intervention and post-intervention time within the dengue case time series (Jan 2000 -Mar 2019) is variable between release areas. For each release area, the pre-intervention period was calculated as months from January 2000 to end of releases, and post-intervention period as months from end of releases to March 2019. Locally-acquired and imported dengue case notifications were tabulated by month of TSR, and summary statistics for the distribution of pre-and post-intervention time were calculated.
To better visualize the temporal distribution of locally-acquired and imported dengue case notifications with respect to Wolbachia deployments, release areas were grouped by calendar quarter of last release, and cases were plotted by release area group against date of illness onset, stratified by locally-acquired vs imported cases.
Negative binomial regression was used to model monthly counts of locally-acquired dengue cases (January 2000 -March 2019) in aggregate Wolbachia-treated and not-yet-treated areas. Cases located in Wolbachia non-exposed areas were excluded from the analysis. The regression model was fitted in Stata (SE version 14.2, StataCorp, TX) using generalized estimating equations, with epidemic year (September -August) as the cluster variable to account for temporal autocorrelation in the monthly case counts, adjusting for monthly imported dengue cases (any vs none) and season (dry: June -November vs wet: December -May), with a population size offset. A binary intervention variable was included in the regression model to distinguish the pre-or post-intervention status of each area in any given month, the coefficient of which provided the estimate of intervention effect (incidence rate ratio).
Ethics approval for human blood feeding mosquito colonies in Melbourne was issued from Monash University (CF11/0766 a 2011000387). All volunteers (no children involved) provided written consent.
In Cairns, Human Ethics approval for bloodfeeding was provided by Human Research Ethics Committee, James Cook University (H4907). All adult subjects provided informed oral consent (no children were involved). Names of subjects providing oral consent were recorded in writing.
Verbal and/or written consent from participants was obtained by phone, online or face-to-face to set BG traps, set MRCs, or participate in Community Mosquito Releases.
Surveys were undertaken under Monash ethics: CF13/2407 -2013001272 Community knowledge of dengue and Wolbachia based dengue control and CF13/2805 -2013001515 Community knowledge of dengue and Wolbachia based dengue control -Townsville.
Wolbachia from nearby release areas). By week 12 of releases, the median Wolbachia frequency in mosquitoes was 82.4% and this increased to 92.3% by week 22. After completion of releases (>23 weeks), median weekly Wolbachia frequencies ranged between 66.0-95.0% through until week 52, and were above 80% thereafter.
Longitudinal monitoring of the Wolbachia infection frequency in mosquitoes collected from the initial Yorkeys Knob and Gordonvale release sites (Hoffmann et al., 2011;Hoffmann et al., 2014) indicated that Wolbachia has been maintained in the mosquito populations at high levels with mean Wolbachia frequencies of 94.7% and 95.4%, respectively, for over 8 years ( Figure 6). Similar results from releases undertaken in 2013 in inner Cairns suburbs (Schmidt et al., 2017)

Results and discussion
Overall, there was a predictable and consistent trajectory of Wolbachia establishment in Ae. aegypti populations as a result of relatively short term (median release duration 12 weeks, range 5-23 weeks), low density releases of Wolbachia infected mosquitoes (generally 20% of houses), either as adults or eggs, across a variety of communities in Cairns and surrounding areas in northern Australia ( Figure 5). Wolbachia frequency data from mosquitoes collected during release and post-release monitoring periods were compiled by week (week 1 = first release) for each release area (non-release areas as per Table 1 were excluded from the analyses, along with the following release areas KB, CNX, EHX, MRAX, BUX3, AER and FW as the frequency of Wolbachia in mosquitoes in week 1 was already high as a result of likely spread of   releases ( Figure 7A). Aedes aegypti counts and Wolbachia screening results, aggregated by release area and collection period are available as Underlying data (Ryan, 2019).
Early deployments of Wolbachia involving egg releases, either in combination with adult mosquito releases ( Figure 7B) or on their own (Figure 8), allowed testing and development of the egg release methods, and also calibration of release rates both in terms of the density and the duration of releases.
Egg releases into small isolated sites (SF1-3), involving weekly releases at 20% of houses for between 5-23 weeks all resulted in successful establishment of Wolbachia ( Figure 8A). Egg releases into four areas with higher populations of Ae. aegypti (BU1-3 and CN1) at similar or lower release the Cassowary Coast (12.2 km 2 , 7,940 households), Charters Towers (6.9 km 2 , 3,359 households) and Douglas Shire (7.1 km 2 , 2,585 households) between 2015-2017 ( Figure 10-Figure 14). Similar patterns were observed in terms of Wolbachia establishment in Cairns, Charters Towers and Douglas Shire releases, although frequencies were more variable in Palm Cove (PC) where the Wolbachia frequency in mosquitoes didn't reach 80% until after week 72 ( Figure 10D). Releases in the Cassowary Coast -Innisfail area were undertaken between Mar-Jun 2017 and coincided with a period of generally low Ae. aegypti adult mosquito numbers. Wolbachia frequencies in mosquitoes after 14-16 weeks of releases were relatively high (mean 67.8%, range 50.0-100.0%) across the releases areas, although this was followed by a drop in Wolbachia frequencies and high variability between weeks 20-40 (mean 58.1%, range 30.0-80.2%) (Figure 13). In three of these areas (subarea in Innisfail, Mourilyan and South Johnston) the Verily Debug project undertook releases of wAlbB infected male Ae. aegypti mosquitoes which were expected to induce sterility when mated to wMel infected females and well as uninfected females (Axford et al., 2016). Release dates were not stated, however wAlbB infected male Ae. aegypti mosquitoes were detected in BGS trap collections between weeks 39-66 in Mourilyan and weeks 41-66 in Innisfail and South Johnstone ( Figure 13B). The release of wAlbB male densities (5-10% of houses), resulted in the slower establishment of Wolbachia in three of these areas ( Figure 8B, BU1 and BU3, > 1 year to reach 80% frequency). In contrast, in BU2 where the Wolbachia frequency only reached 50% after 9 weeks of releases, the frequency of Wolbachia declined to less than 10% after week 30 ( Figure 8B). The longterm decline in Wolbachia frequencies in mosquitoes in BU2 suggests that releases in this area did not result in Wolbachia exceeding the threshold frequency of infection, above which frequencies systematically increase (Schmidt et al., 2017). Given the decline in Wolbachia frequencies, egg releases re-commenced at week 62 for 16 weeks and resulted in Wolbachia establishment. Larger, operational scale egg releases were undertaken across the remaining inner Cairns areas in 2015 (Figure 9) (11.5 km 2 ; 13,823 households) and involved a fixed density of releases at 20% of households, but with variable durations of releases of between 6-15 weeks (mean 11.5 weeks). Wolbachia frequencies in mosquitoes at the end of releases in each area ranged from 67.4 to 100%. Periodic monitoring of mosquitoes from these areas between weeks 75-171 indicated that the Wolbachia frequency in mosquitoes was high across all sites (mean 98.7%, range 75-100%).
Large scale adult mosquito releases were undertaken across the remaining Cairns suburbs (46.4 km 2 , 35,899 households),  Ae. aegypti mosquitoes coincided with a period when there was a drop in wMel Wolbachia infection frequency in mosquitoes in Mourilyan (mean 61.5% between weeks 39-66), although the wMel Wolbachia frequency increased and was maintained generally above 80% from week 55 onwards. There was no appreciable effect of wAlbB male Ae. aegypti releases on the wMel frequency in either Innisfail or South Johnstone ( Figure 13B). Wolbachia (wMel) infection frequencies in mosquitoes across the three wAlbB release areas were high from weeks 70 onwards (mean 94.0%, range 85.7-97.0%), indicating that wMel Wolbachia Ae. aegypti persisted in these areas, despite the release of relatively large numbers (3 million, Verily Debug project) of incompatible wAlbB male Ae. aegypti mosquitoes.
Overall, short-term releases of between 5-23 weeks involving either egg or adult stages, resulted in the establishment of Wolbachia in mosquito populations across all release areas. There were no clear differences between egg or adult releases in terms of Wolbachia establishment. The overall duration of egg and adult release periods were similar (average 12 weeks duration for both egg and adult releases), although egg releases were generally undertaken every 2 weeks compared with weekly adult mosquito releases. Although Ae. aegypti populations varied seasonally across the study areas, there were no clear seasonal effects on Wolbachia establishment, indicating that under the north Queensland conditions releases can be undertaken year round. Operationally, egg releases provided advantages over adult mosquito releases in that there was no need to rear immatures stages to adults in a local insectary. For egg releases, eggs were produced centrally in an insectary, and then transferred to the field and placed into mosquito release containers, either by staff or via community members themselves. These simple, low-cost egg release methods may represent a more scalable approach for future large-scale implementations, particularly in low resource settings where infrastructure for mass rearing of adult mosquitoes is limited.
Previous analyses of the spatial spread of Wolbachia from the 2013 inner Cairns release sites (EHW, PP and WC, Figure 2A) indicated that spread of Wolbachia from the release sites were spatially heterogeneous, Wolbachia moved relatively slowly at 100-200m per year (Schmidt et al., 2017).   and this was possibly due to barriers to Ae. aegypti dispersal, higher incidence of long-range Ae. aegypti dispersal, and intergenerational loss of Wolbachia (Schmidt et al., 2018). The current analyses of Wolbachia infection frequencies in mosquitoes from five non-release areas (Figure 2A, 2B, Figure 15) indicated that Wolbachia became established in mosquito populations throughout each area. In the case of the Pyramid Estate non release area (PE NR) which was located west of a main highway which separated it from the initial Gordonvale release site ( Figure 2B), Wolbachia infection frequencies remained low (<20%) for over 100 weeks, despite the high (>80%) Wolbachia frequency in mosquitoes in Gordonvale during the same period ( Figure 15A). Periodic monitoring of mosquitoes in Pyramid Estate at week 228 indicated that the Wolbachia frequency in mosquitoes had reached 50%, and had further increased to above 80% from week 268 onwards. Although monitoring was only undertaken periodically in Pyramid Estate from week 100, the increase in the Wolbachia frequency in this area, despite a relatively low frequency during the first 100 weeks, suggests that the natural introduction of Wolbachia mosquitoes, as either eggs or adults from nearby Gordonvale, was sufficient to result in eventual establishment of Wolbachia. Similar results were found in the four other non-release sites, although in each of these sites the Wolbachia infection frequency was generally correlated with the Wolbachia infection frequency in mosquitoes in nearby release sites ( Figure 15B-E). In these four inner Cairns non-release sites there were only limited boundaries to mosquito movement, and once Wolbachia became established in nearby release areas the infection spread into mosquitoes in nearby nonrelease areas. Overall, these five non-release areas constituted 3.23 km 2 and some 3,261 households, and indicated that releases of Wolbachia mosquitoes do not need to be undertaken in all areas where Ae. aegypti occur (Turelli & Barton, 2017). This opens the way for more efficient deployment strategies in the future where areas are left intentionally with no mosquito releases and instead rely on natural spreading of Wolbachia. This natural spreading may take some time depending on the    Figure 2 and Table 1).
size of deliberate deployment "holes" and local variation in mosquito density and mosquito habitat (Hancock et al., 2019) but has the potential to significantly reduce deployment costs. Moreover, these non-release areas can be considered in the vertical dimension and not just the horizontal dimension given the nature of dispersal of Ae. aegypti within buildings (Liew & Curtis, 2004). In this scenario deployments in upper floors of collections of high-rise apartment buildings may not be needed, instead relying on Wolbachia to naturally invade these areas, simplifying deployment logistics.
In the Cairns releases between 2011-2014, communication and community engagement activities followed earlier approaches described in Hoffmann et al. (2011), and relied heavily on face-to-face consultation with key stakeholders and community groups, including one-on-one meetings, attendance at community events, door-knocking and mail-outs to householders. This proved effective in building awareness of the project and generated support for and participation in releases. From 2015 as release activities scaled up to cover larger areas of Cairns, the PAM was used for community engagement, and this proved effective in building awareness of the project and broad support for activities (Table 2). Community members volunteered to participate in activities and this lead to a pre-registered participant database through which field staff could distribute mosquito release containers and mosquito monitoring traps. In addition to hosting mosquito release containers and mosquito monitoring traps, local ownership of the WMP's Wolbachia method was achieved through a school program conducted at Bentley Park College in 2017. Known as the Wolbachia Warriors Program (O'Neill et al., 2018), the voluntary, appliedscience program was undertaken by 636 students aged from five to 12 years of age. The free program involved the engagement of teachers, parents and students to enable participants to grow and release Wolbachia carrying mosquitoes in their yards at home, three times, over six weeks. Each participant was provided with an instructive project booklet and three Mozzie Boxes (mosquito egg release kits). By participating in an applied science program, students learnt basic natural history that complemented in-class learning, while directly contributing to public health outcomes.
Similar to the Cairns releases above, the PAM model was also implemented in Charters Towers, Douglas Shire and the Cassowary Coast. This was implemented prior to releases and included the same components as in Cairns (Table 2), and generated significant awareness and support, and direct participation of communities in releases: •  (Figure 16). Nearly all locally-acquired cases (94%), and two-thirds of imported cases, were notified during the monsoonal months December -May, and the large majority (87%) were in the populous Cairns regional area. The Department of Health responded in 1998 with emergency vector control activities through a specialized unit (Dengue Action Response Team) to conduct extensive source reduction and chemical intervention activities that included targeted interior residual spray, deployment of lethal ovitraps, and application of larvicides to water holding containers (Queensland Health, 2015). Despite these efforts, and with the increasing numbers of imported dengue cases every year from 2000-2019 ( Figure 16B), local dengue transmission occurred most years in Cairns ( Figure 16A), with large outbreaks in 2003 (450 local cases) and 2009 (776 local cases).
The staggered deployment of Wolbachia across Cairns in 2011 -2017, and into the urban centres of the Cassowary Coast, Charters Towers and Douglas regions in 2016 and 2017, led to Wolbachia establishment throughout communities with a total resident population of 165,000 people. When the timing of notified dengue cases is scaled relative to the local completion of Wolbachia deployments (Figure 17), this demonstrates the near elimination of locally-acquired dengue cases from Wolbachia-treated communities. Only four local cases were notified from Wolbachia-treated areas in the eight years since completion of the first releases in March 2011, while dengue case importations continued in these areas. All four of these local cases were notified in January -March 2014, more than five years ago, and three of the four were from early central Cairns release areas (Parramatta Park and Westcourt) which were at that time 8-10 months post-release and surrounded by untreated areas.
The spatial distribution of locally-acquired and imported dengue cases across the intervention areas is illustrated in Figure 18, which highlights that the 2003 and 2009 outbreaks widely affected most parts of Cairns and the other urban centers. In striking contrast, in the years following the first Wolbachia releases in Cairns in 2011, substantial local transmission continued to occur but was concentrated each season within the ever-diminishing area in which Wolbachia had not yet been released. In total, 515 locally-acquired dengue cases were notified   across the four regions since 2011, of which only four have been located in Wolbachia-treated areas.
In an interrupted time series analysis of this case notification data, the regression model estimate of Wolbachia intervention effect indicated a 96% reduction in dengue incidence in Wolbachia treated populations (95% confidence interval: 84 -99%), adjusted for season, imported cases, and allowing for temporal autocorrelation of cases.
The wMel strain of Wolbachia has been deployed across the major regional cites of Cairns and Townsville, as well as nearby smaller regional communities that have historically been affected by dengue transmission. Consistent with modelling projections (Ferguson et al., 2015), Wolbachia deployments have been associated with cessation of local dengue transmission. Alternative explanations for the absence of local dengue transmission are unlikely; there has been no change to local vector control activities and the number of notified imported dengue cases has not diminished with time. Ongoing long-term monitoring is expected to confirm the durability of Wolbachia and its persistence in local Ae. aegypti populations (Ritchie et al., 2018), and the "dengue proofed" status of northern Queensland. Additional public health evidence for Wolbachia's impact on dengue transmission will be delivered by a large cluster randomized trial currently underway in Indonesia (Clinical Trials.gov Identifier: NCT03055585).
This current report demonstrates Wolbachia deployment is scalable, safe, long-lasting, acceptable to communities and is associated with cessation of dengue transmission. With the global burden of dengue clearly not adequately controlled by existing public health tools, the Wolbachia approach should be considered for communities at risk of or endemic for dengue. This project contains all data underlying results presented in Figure 6- Figure 15.

Data availability
Human dengue case notification data was provided to us by Queensland Health. The conditions of the release of the raw dengue case notifications data to us by the Communicable Disease Branch of Queensland Health do not permit further sharing to a third party. This data (locally acquired and imported dengue case notifications) can be acquired by application to Queensland Health: https://www.health.qld.gov.au/clinicalpractice/guidelines-procedures/diseases-infection/surveillance/ reports/notifiable/data-request. Access to data will be considered following completion of the Data request form (PDF, 237 kB). As stated in the manuscript, this work was needed to summarise the methodology and the results from several studies and interventions carried out in Queensland. It describes the steps undertaken to achieve Aedes aegypti population replacement, from the community and stakeholder engagement to the mathematical modelling and the strategies to release Wolbachiainfected mosquitoes.
The conclusions are fully justified and the reduction in dengue incidence achieved by spreading Wolbachia into the local Ae. aegypti populations is unprecedented worldwide. No other tool has shown such a consistent and egalitarian result in fighting dengue virus transmission so far.
In my opinion there will be two crucial tests for the Wolbachia technology. The first one will be in highly urbanised areas of LMIC where arbovirus transmission is high and both Ae. aegypti and Aedes albopictus thrive and live in sympatry. The second one is time related: being Wolbachia-Ae. aegypti is a very recent symbiosis, will we continue observing the same arbovirus blocking effects in the decades ahead? © 2019 Powell J. This is an open access peer review report distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
(CNS_Montoring_Results.xlsx). Overall mean sample size was 32.4 Ae. aegypti per collection period (median 16.0, interquartile range 6.0 -38.0). 5. The monitoring in the five non-release areas did not specifically measure wave speed as was the aim in Schmidt et al., 2017. The non-release areas in the current study were either small areas (0.13 -0.53 km 2 ) surrounded by areas where Wolbachia had been released, or an isolated area (1.99 km 2 ) separated by a main highway from a Wolbachia release site. The spread and establishment of Wolbachia in all of these areas is consistent with previous estimates of Wolbachia spread at 100-200m per year as previously defined by Schmidt et al., 2017. As we point out in the discussion, the main implication of this is the potential for more efficient deployment strategies where "areas are left intentionally with no releases and instead rely on natural spreading of Wolbachia."