Ethics statement

The COPA study was reviewed and approved by Ponce Medical School Foundation, Inc. Institutional Review Board (protocol number 171110-VR).

Overview of project phases

The project included two major phases. After pre-release activities including mosquito population monitoring and community outreach, Phase I began in September of 2020. The overall goal of the project was to test whether suppressing the wild population of Ae. aegypti in 19 communities of Ponce, Puerto Rico, would lead to a statistically significant reduction in dengue case rates in people who live in these communities compared to 19 other communities where no mosquito control was undertaken. In Phase I, we released Ae. aegypti Wolbachia male mosquitoes into all 19 treatment clusters with the goal of suppressing all 19 treatment clusters at the same time. In December of 2020, mosquito monitoring data showed that the wild population was not sufficiently reduced after three months of releases. In response, we adjusted the release strategy in Phase II (January 2021) and focused releases in four clusters to start, resulting in substantially increased densities of released males relative to wild Ae. aegypti male densities. We continued releases in the first four clusters until the end of the project in December of 2021. As in Phase I, the goal of Phase II was to suppress the wild population of Ae. aegypti in all 19 treatment clusters. However, after one year of releases in Phase II with only moderate impact on the wild mosquito population and insufficient dengue cases for statistical measurement, the project was ended in December 2021.

COPA setting and methods

The COPA project was initiated in 2018, and categorized areas of Ponce, Puerto Rico into distinct study clusters organized by previous arboviral incidence data in Ponce, social and geographic neighborhood delimitation and size, engagement with community leaders, safety issues, and type of housing [31]. The implementation and completion of the Wolbachia male mosquito releases were preceded by a broad educational campaign that continued during the releases. Public perception of the COPA study and the releases were very positive during and after the project period, with very little opposition voiced. To minimize “contamination” (mosquitoes flying from treatment clusters to control clusters) the 38 clusters were separated from each other by a distance of at least 150 meters, the typical flight distance of Ae. aegypti males [32] or by a geographical barrier such as a river, forest, or major roadway, except when adjacent clusters were assigned the same treatment or control status (Fig 1). The clusters ranged in size from 15 to 126 hectares (ha) (Table 1) and spanned the city from the coast to the hills with a mixture of residential and non-residential properties (Fig 1). COPA study clusters and methods have been described in previous publications [28–30]. Briefly, ~3,800 COPA participants ages 1–50 years, residing in 38 distinct clusters have annual interviews including standardized questionnaires on socio-demographics, health status, history of febrile illness, and other health-related questions, as well as providing blood samples for arboviral testing. Through text messaging, participants are also asked every week about fever and other symptoms and offered arboviral and respiratory viruses testing. The COPA study design was powered to detect a 50% reduction in arboviral disease incidence among 3,800 participants over a three-year intervention period [30].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Location and project assignment of COPA clusters, Ponce, Puerto Rico, September 2020–December 2021.

A) 38 polygons showing clusters defined for the COPA study. Phase I releases were conducted in 19 treatment clusters (blue) in 2020. B) Phase II treatment clusters are highlighted in purple and control clusters highlighted in blue. Remaining clusters received no extra activity, but monitoring traps remained active during Phase II. Map source: USDA Farm Service Agency, National Agriculture Imagery Program. Map created using ArcGIS Pro 2.6.0. USDA:NRCS:Geospatial Data Gateway:Home:Direct DownLoad.

https://doi.org/10.1371/journal.pntd.0012839.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of COPA clusters (N = 38), Ponce, Puerto Rico, 2020–2021.

https://doi.org/10.1371/journal.pntd.0012839.t001

Approvals/Permits

The open release of male Ae. aegypti mosquitoes with Wolbachia was approved in 2016 under an Experimental Use Permit (EUP; 89668-EUP-3 2020-12-04) by the United States Environmental Protection Agency (EPA), which expired at the end of 2021 [33]. In 2020, the Puerto Rico Department of Agriculture issued the required permits for the PRVCU to release male mosquitoes with Wolbachia in Ponce. Commercial use of the WB1 isolate of Wolbachia in Aedes aegypti male mosquitoes was granted in the United States on April 16, 2024 by the EPA [34].

Cluster assignment to treatment and control

Based on previously available prevalence data for CHIKV infection, clusters were assigned to one of three categories: 1 = low prevalence (< 28%), 2 = medium prevalence (>28–43%), 3 = high prevalence (>44%). Randomization was then conducted for clusters in each category using the sample R function, which is based on a random number generator without replacement [35]. The location of the 38 clusters and their classification as treatment or control clusters is shown in Fig 1A and cluster characteristics are described in Table 1.

In Phase I of the project from September 3, 2020 to December 29, 2020, Wolbachia males were released into all 19 treatment clusters. In January 2021, to increase suppression levels, Phase II releases were concentrated into four clusters as part of a phased release strategy similar to a rolling carpet or stepped-wedge design [27], but organized into isolated clusters rather than contiguous release areas. We started with four clusters based on preliminary data suggesting that splitting shipments among the number of households in four clusters would lead to male densities necessary to reach the targeted overflooding ratio of 10 released males to 1 wild male, based on successful suppression in previous studies with ratios ranging from 6 to 18 released males for every one wild male [21,26].

For Phase II releases, we ranked the 19 treatment clusters based on disease transmission risk, which we calculated using historical Ae. aegypti female densities as well as an estimate of the proportion of residents under the age of 20 that are, therefore, likely to be susceptible to dengue infection. We selected the top ranked four treatment clusters that could be treated at one time within logistical and release capacity constraints to improve the probability of detecting an epidemiological effect under this phased strategy. This selection process led us to select PL01, PL02, ST02, and VG02 (Fig 1) based on their combined size as well as their high transmission risk. We chose four control clusters that approximately matched the four treatment clusters in size and pre-release female densities such that the starting pre-release suppression level was approximately zero.

Monitoring

Adult Ae. aegypti mosquitoes were monitored using Autocidal Gravid Ovitraps (AGOs) [36,37] deployed in treatment and control clusters. Trap deployment began in some clusters in April 2018 and were finalized in September 2020 before the start of Wolbachia male releases. Trap densities are listed in Table 1, with an average of one trap per 3.2 ha. Traps were visited every week to obtain adult mosquito counts, and full replacement of water, hay, and glue paper was conducted every eight weeks according to standard protocols for AGOs. Monitoring with AGO traps continued until traps were removed during the last week of March 2022, three months after the completion of the releases. To maximize limited resources, AGOs were the primary measure of impact in Phase I. Ovitraps and other traps were not used in Phase I.

We confirmed the presence of an established Ae. aegypti population in all clusters with a range of adult mosquito densities. Baseline counts before Wolbachia mosquito releases are shown in S1 Fig. We hypothesized that we could achieve sufficiently high overflooding ratios to have an impact on the population based on approximate estimates of the wild population stemming from the mean number of females per trap night in the treatment clusters.

In Phase II, additional monitoring was needed to increase male mosquito capture rates in order to measure the ratio of Wolbachia males to wild males. We added 15 BG Sentinel v2 traps (Biogents, Regensburg, Germany) in March 2021 to each of the four Phase II treatment clusters (average of one BG per 3.3 ha). Traps were deployed with BG lures and powered by 12v 35Ah batteries. Traps were active for three trap nights per collection and collected one time per week. Trap fan speed was recorded using a Kestrel 2000 (Nielsen-Kellerman Company, Boothwyn, PA, USA). Collections with a fan speed less than 4 mph were excluded from analysis. Collected male mosquitoes were shipped to Verily in South San Francisco, tested for Wolbachia DNA using a Loop-mediated isothermal amplification (LAMP) assay, and analyzed to calculate overflooding ratio as described previously [21]. In addition to AGO and BG traps, we added ovitraps in Phase II in order to obtain better measurements of Wolbachia male mating rates in the release sites. We deployed 30 ovitraps in each of the four treatment and four control clusters (average of one trap per 1.6 ha) to collect eggs for hatch rate assessment as described previously [21]. Briefly, ovitraps were standard black cups with seed germination paper (38# Anchor Paper Inc., St. Paul, MN, USA) and fermented hay infusion [21]. Traps were deployed as pairs with traps on opposite sides of a property for 3 trap nights. Positive egg papers were stored in an insectary (28°C, ~70–80% relative humidity) for >72hrs [38], flooded using a solution of bovine liver powder and yeast, and hatched larvae counted. Induced sterility was calculated using Abbott’s formula [39] and confidence intervals were calculated using 10,000 bootstrap samples.

To confirm that a population of Ae. aegypti with wAlbB Wolbachia had not established in the treatment areas via accidental release of females with Wolbachia, we sampled larvae from ovitraps in the treatment clusters and test for Wolbachia using the LAMP assay as described above. We tested a total of 135 collections spanning four months in 2021 in all four Phase II treatment clusters. No larvae were positive for Wolbachia.

Mosquito rearing and shipping

The PRVCU partnered with Verily Life Sciences for the production and shipment of male Ae. aegypti mosquitoes with Wolbachia. All Wolbachia mosquitoes used in this project carried the WB1 isolate of wAlbB [40]. We collected eggs from multiple locations in Ponce, Puerto Rico and created a wild-type colony. Males from this colony were backcrossed with WB1 females for five generations to generate a backcrossed WB1 colony with Puerto Rican background genetics. We sequenced samples of genomes from the parental colonies and the backcross colony and confirmed the background genetics matched the Puerto Rican parental colony at a rate of 95% on average. Mosquitoes were mass reared, and sex separated at Verily facilities in San Francisco, as described previously [21].

Males were packed and shipped up to five times per week. Several versions of the shipping containers were developed and modified during the project period, but the following is a description of the version used throughout all phases of the project. After sex-sorting, males were chilled and placed into small containers using proprietary processes and materials. Small containers were packed into vessels that were packed into larger custom boxes to maintain cool temperature during transit from San Francisco to the PRVCU lab facility in Ponce.

Based on preliminary laboratory testing, we determined that male mosquitoes would be most effective in the field if they could be delivered to the field site in Ponce and unpacked within 24 hours of sedation and packing. After arrival in Puerto Rico, mosquito containers were required to undergo an inspection by the Puerto Rico Department of Agriculture that could be delayed depending on the time and day of arrival. Once cleared, the containers were transported by van for approximately 1.5 hours to Ponce, Puerto Rico, where the mosquitoes were unpacked at the PRVCU laboratory into release tubes with 10% sugar water as described previously [21] to then be released in the selected treatment communities according to a predetermined release schedule. A summary of the transportation logistics from San Francisco to Ponce is presented in S2 Fig.

Mosquito releases

A modified van-based delivery system developed by Verily was used for the male releases, as described previously [21]. The delivery system uses a computer-controlled Google maps-based automated system to deliver the male mosquitoes at select release points along a release map. A predetermined number of males were puffed onto the road edge via holes in the passenger side of the vehicle as the vehicle was moving.

We calibrated the number of males to be released in each of the treatment clusters to be proportionate to the approximate number of households in each cluster. Each shipment was divided among treatment clusters to maintain correct proportions within each week. Van-based releases were adjusted at various points during the project in response to consistently high female collections in some of the AGO traps. Adjustments were made to increase the rate of release from the van on given streets in order to increase the density in certain areas of the cluster with higher female numbers. In Phase I, mosquitoes were released three times per week in each treatment cluster following a rotating schedule to cover the 19 clusters, with releases occurring Monday through Saturday. Preliminary analyses of this phase results showed that suppression levels were suboptimal, and a phased release strategy was implemented. In Phase II, we released Wolbachia males into each of the four treatment clusters five times per week.

Container-based male releases by hand

Throughout the project period, we identified areas within treatment clusters with consistently high female counts in AGOtraps. For some of these areas, vans were not able to drive through given the housing configuration, with only pedestrian access available such as in cluster ST02. In ST02, hand releases were conducted starting in June 2021 in an apartment complex where vans could only access facility edges through parking lots. Hand releases were added to courtyards that were enclosed on four sides by buildings. In PL01, hand releases were added to increase male density inside an apartment complex similar to ST02. In PL02, hand releases were conducted inside a cemetery. Hand releases began in PL01 and PL02 during the second week of August. The last hand releases were completed during the third week of November 2021 due to logistical constraints, approximately one month before project completion.

Integrated vector management

The goal of this project was to measure the efficacy of Wolbachia-male releases as a tool for reducing mosquito populations in Ponce, PR, and ultimately disease transmission, so we chose to minimize integrated control measures. While the community education aspect of the project may have resulted in increased mosquito awareness among residents, it was unlikely to have resulted in significant impacts on the wild population without more dedicated efforts. Other than a single application of larvicide to three subterranean water meter pits in ST02 and unknown levels of private pest control activity, no additional, systematic chemical applications were made in the treatment or control clusters during the project period.

Longevity assay

Life span of male mosquitoes was measured using a cage-based longevity assay as described previously [21]. Briefly, samples of ~100–200 packed and shipped male mosquitoes with Wolbachia and held-back males were released into a cage either in Ponce on the morning when they would have been released in Ponce (“packed and shipped”), or at Verily’s facility (“held-back” or unshipped controls). Males were provided with water in the cage, but no sugar, and dead males were counted every morning until all had died. Average lifespan was calculated as the average number of days it took for males to die after being released into the cage.

Mating competition assays

Male competitiveness was measured using cage competition mating assays. Wild-type males and females from a colony established from locally collected eggs were reared in insectary conditions using standard protocols. Wolbachia-male mosquitoes were either sampled from production batches and held at Verily’s facility in California or sampled from shipments to Ponce. For each cage assay,10 3–5 day old wild-type females, 10 3–5 day old wild-type males, and 10 Wolbachia-males were placed in a medium BugDorm (30cm³, BioQuip) cage and allowed to mate for one hour. We have found that one hour is enough time for 10 females to be mated while minimizing multiply-mated females, providing a reliable competitive measure. Females were then placed individually in fly vials with seed germination paper and given a blood meal to elicit oviposition. All females were included in the analyses except those that did not lay eggs, were found to be uninseminated, or perished. Eggs were given 72 hrs to mature in insectary conditions and then flooded for hatching [38]. Any egg batch that produced at least one larva was considered wild-type mated and male competitiveness was calculated as the proportion of females that were mated to Wolbachia-males in each cage.

Mark-release recapture assessment

We conducted two mark-release recapture assessments to evaluate dispersal distance and survival of packed and shipped Wolbachia males. We placed 34 BG Sentinel traps baited with BG lures in an evenly spaced grid extending out to 350m in a residential neighborhood within the treatment cluster SA01. BG traps were turned on ~24 hours after each release and serviced every 24 hours for five days. Fan speed and any trap issues were noted for each collection. All Ae. aegypti males were returned to Verily and tested for Wolbachia DNA using a LAMP assay as described previously [21]. Corrected recapture rates and daily survival were estimated by fitting a non-linear model, and confidence intervals were calculated using simulations, both based on the method of Buonocorsi et al [41]. Mean distance traveled was calculated by first correcting for variable trap density [42,43] at increasing distances from the release point and taking the mean of recapture estimates in 25m annuli.

Measuring impact on wild populations

Since female mosquitoes take blood meals and are responsible for transmitting disease, wild female densities are the primary endpoint used to assess the impact of the treatment. We expected that if the intervention had no effect, female densities would be equivalent between treatment and control clusters. If the intervention had an effect, female densities would be significantly lower in treatment clusters compared to control clusters. For Phase I, we estimated the preliminary impact of the male releases on the wild population of female Ae. aegypti by comparing all treatment clusters in aggregate to all control clusters in aggregate using Abbott’s formula [31], and the 95% confidence interval was calculated using 10,000 bootstrap samples.

For Phase II, we incorporated the male releases in a model including additional parameters to better estimate the impact of the releases. Trap counts of female Ae. aegypti mosquitoes (per trap-week) were regressed on collection week using a generalized additive model, assuming a Poisson distribution for weekly trap counts and using a log link function[44]. The effect of week was represented by a polynomial spline of order three; treatment, temperature, relative humidity, and days since the last full trap service were included as terms in the linear operator. On the link scale, smooth functions of time (averaged across nested traps and sites) were estimated for (log)counts in the treated group and for (log)counts in the control group. These smooth functions were evaluated at the means of the other continuous variables in the model. To obtain the ratio of counts in the treatment group to the counts in the control group, the differences between the treatment curve and the control curve were computed at each time. Exponentiated, these differences become ratios on the response scale.

Using the estimated covariance matrix of the parameter estimates (including basis function parameters), standard errors for differences at each time point were computed. Residual degrees of freedom were used to find the t-distribution critical value, and confidence intervals for the differences were computed as point estimate ± t*se. The results were exponentiated to obtain confidence intervals for the ratios at each time point.

All data analysis was conducted in R v4.2.1[35], including R packages mgcv[45], RgoogleMaps[46], gplots [47], and RColorBrewer [48].

Overcompensation analysis

To test for ecological compensation and overcompsensation, we re-calculated several statistics from our trapping data. First, we re-calculated summary statistics from both adult and egg collections in rolling one-month windows. To calculate the change in suppression, we calculated suppression by comparing one month windows of female trap counts as described above, then calculating the difference between consecutive windows. Other statistics were calculated as described above. Second, we tested for significant correlations between statistics using the cor.test function (method = ‘pearson’) in R.

Incompatible mosquito shipping and releases

Wolbachia-male Ae. aegypti were mass-reared in South San Francisco, California, chilled, packaged, and shipped to Puerto Rico via commercial airline cargo shipping services using novel packing and shipping technology. Across both years of the project, a total of 107.6 million males were shipped, with weekly totals starting at 0.75M during the ramp-up in 2020 and peaking at 2.35M males per week in 2021 (Fig 2). Various constraints in the shipping and regulatory inspection pipeline limited shipments to five out of seven days per week. Successful SIT or IIT interventions require a consistent presence of released males, so disruptions to release schedules can cause reductions in efficacy, especially when the disruptions are clustered in time. The males were delivered in 345 shipments with an overall on-time success rate of 79%. We experienced many delays and complications related to the ongoing COVID-19 pandemic during the project, resulting in 53% of weeks during the project period experiencing some delay, with 9 weeks seeing on-time rates less than 50%. In some cases, shipments were delayed or canceled resulting in delayed or canceled scheduled releases, and ultimately resulting in the loss of 7.6M males during the project that were never released.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Male mosquitoes with Wolbachia released in COPA clusters, September 2020–December 2021, Ponce, Puerto Rico.

A) Weekly male shipment totals (in Millions) shown on y-axis, with shipment date on the x-axis. Each bar is one week, summarizing five shipments per week. The orange bars show the total number of males that arrived on-time and were successfully released. The green bars show the number of males that arrived between 24–48 hrs that were released. Black bars show the number of males from shipments delayed beyond ~48 hrs were not released. B) The mean number of Ae. aegypti males collected in AGO traps in 2021 calculated across 4 treatment clusters (yellow) and all control clusters (grey). Shaded area shows 95% bootstrap confidence intervals. The vertical dotted lines show when releases began and ended in 2021. C) Overflooding ratio shown on y-axis from BG Sentinel collections according to dates on x-axis, with the cluster name shown for each sub-panel. Each dot is one BG trap collection. The middle horizontal line shows the mean for a weekly collection, with the 95% confidence interval shown with the vertical bar.

https://doi.org/10.1371/journal.pntd.0012839.g002

Phase I: Releases in 19 clusters

During the first phase of the releases from September to December 2020, we shipped 22.31M males, and we successfully released 20.76M males across all 19 treatment clusters. We released an average of 56 (range 36–70) males per household per week (or 1,024 males/ha/week range 665–1281). Using the marginal male capture efficiency of AGOs, we detected a modest but significant increase in male Ae. aegypti in the treatment areas relative to controls (S3 Fig). However, after 10 weeks of releases, we observed only a minor impact on the wild female population with a preliminary reduction of 15% (95% confidence interval, 4–24%) when comparing all treatment clusters in aggregate to control clusters. We concluded that our Wolbachia-male release rate was insufficient to achieve strong reductions in the wild population.

Phase II: Step-wise releases

After a brief three week pause in releases after Phase I, we began Phase II releases in the four treatment clusters in the last week of January 2021. From January until December 2021, we released a total of 79.71M Wolbachia-males (Fig 3). We started releases by applying the same average release rate to each of the 4 clusters at a rate of 375 males/household/week (or 6,338 males/ha/week).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3.

A) Induced sterility calculated as the scaled difference between treatment clusters aggregated hatch rate vs. all control clusters aggregated scaled by the control clusters hatch rate. Shaded area shows the 95% bootstrap confidence interval. Dotted lines show when releases started on left and ended on right. B) Mean number of female Aedes aegypti per trap per week in treatment clusters versus control clusters. December 2020–May 2022. Ponce, Puerto Rico. Dotted lines are the same as panel A. C) Cumulative rain over 21-day windows plotted on the same weekly schedule as in B for three weather stations in Ponce. D) Ratio of female Aedes aegypti mosquitoes in treatment clusters (n = 4) to control clusters (n = 4), Ponce, Puerto Rico, January–December 2021. Shaded area shows point-wise 95% confidence region. Dotted lines same as panel A.

https://doi.org/10.1371/journal.pntd.0012839.g003

During Phase II releases, we monitored male densities in the treatment clusters in several ways. First, we found the mean number of male Ae. aegypti in the AGOs rose to ~5 per 7-day collection period compared to ~0.1 in the control clusters. Second, we added BG sentinel traps that are designed to also capture adult male mosquitoes and found that the mean number of males increased to an average of 39 males per collection in the treatment clusters compared to an average of three in the control clusters (Fig 2C). We also spot-checked the overflooding ratio more directly by testing males from treatment clusters for Wolbachia DNA using a LAMP assay and found that the ratio (Wolbachia:wild males) varied substantially among trap collections, with ratios close to 1:1 in some collections while we detected only Wolbachia-males in others (Fig 2D). The mean overflooding ratio was generally high, with means greater than ~15:1 (Wolbachia: wild males) in all three sampled weeks in PL01 and VG02, while the means were more variable in PL02 and ST02. Taken together, these results suggest that consolidating the releases into four clusters resulted in substantial increases in male densities and overflooding ratios closer to our target levels.

If the released Wolbachia-males were successfully penetrating the landscape and mating with wild females, we would expect eggs collected from these landscapes to hatch at a lower rate than those collected from control clusters. To test for this effect, we added ovitraps to each treatment and control cluster. We compared the proportion of eggs that hatched in treatment clusters to that of the control clusters and found that our male releases induced significant levels of sterility throughout Phase II release period (Fig 3A). Except for several outlier weeks, induced sterility levels remained approximately 0.5 from the start of releases until the end of August, at which point induced sterility levels began trending downwards (Fig 3A), possibly consistent with either a drop in mating pressure from the Wolbachia-males, or an increase in migration from untreated areas nearby.

The goal of the Wolbachia-male releases was to reduce the number of wild female mosquitoes that could contribute to disease transmission. We monitored female densities in the treatment and control clusters using AGOs throughout the project period and using BG sentinels from February through August 2021. Prior to the start of releases, the mean number of females in the treatment and control clusters were well correlated and both fluctuated substantially, sometimes by 2-fold over the span of two weeks (S4 Fig), emphasizing the importance of considering population density changes in the control clusters and other environmental factors when evaluating impact of the Wolbachia releases. After releases began, the mean number of females in both the treatment and control clusters decreased on average for the first 12 weeks of releases (Fig 3B). Although the populations in the treatment clusters decreased at a faster rate than those in the control clusters, the impact of the Wolbachia-male releases is confounded by a drop in mosquito populations in the region, likely due to a marked lack of rain during these months (Fig 3C). Following the first sufficiently consistent rains of the year in mid-April, populations in the control clusters and to a lesser extent the treatment clusters rebounded, suggesting the presence of a substantial egg bank in these communities. From May until October, we observed a repeating saw-tooth pattern in both treatment and control clusters, followed by a large increase in all populations starting in mid-October, consistent with local rain patterns driving large hatching events (Fig 3B).

To best estimate the effect of the Wolbachia-male intervention while accounting for environmental and operational effects, we used a Generalized Additive Modeling approach and found that the weekly counts of female Ae. aegypti mosquitoes in the four treatment clusters were 51% (95 CI 37–71%) of the counts in the control clusters (Fig 3C). Maximum reduction was achieved at the beginning of May 2021 when the rains began (mean of 4.1 and 8.5 females in treatment and control, respectively), and the reductions remained around 50% for most of the Phase II release period. Once releases stopped, the average weekly female mosquito counts in the treatment clusters increased steadily, reaching the same levels as the control clusters in about 2 months. These results suggest incompatible male releases had a clear suppressive effect on the wild population. However, the reduction observed in the four Phase II clusters was lower than the target level required to bring the Ae. aegypti population to a level where viral transmission is reduced [49,50]. Below, we describe various adjustments made to the release strategy as well as attempts to identify opportunities to improve the strength of the intervention.

Adaptive release strategies

As releases were ongoing, we noted that female densities were spatially variable within the treatment clusters, so we made a series of adjustments to focus releases in these areas. In one case, we observed a single AGO trap (ST02-PON-14-6) in ST02 that consistently captured higher female counts than other traps in this cluster (Fig 4). We increased the release rate from the van on February 25, 2020, at all release points within ~50m of the trap location. The female densities in this trap were reduced by more than half three weeks after this adjustment. In the same cluster, ST02, we noted that traps inside or on the edge of a residential apartment complex consistently captured high numbers of females. We had been using the release van to treat this area by treating the adjacent roads and also releasing in parking lots that enter the edge of the property. To improve our monitoring of the wild population in this complex we added two AGO traps inside the complex and confirmed that the wild population was high. After confirming the need for additional treatment, we began releasing inside the complex by hand releasing from small containers in the last week of June at designated locations inside of the courtyards that were blocked from the exterior streets where vans could release. Similar to the impact we observed at ST02-PON-14-6, the number of females in traps in and around the residential complex began to decrease ~3 weeks after the addition of hand releases (Fig 4A, green box). We similarly added hand releases to a large cemetery and a residential complex, both in PL01, and made a number of smaller adjustments to van release rates throughout all four treatment clusters, in an attempt to improve control of problematic breeding locations detected by traps, and thus suppression overall.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4.

Hotspot areas in cluster ST02. A) Heatmap showing Ae. aegypti female counts by trap in rows and collection week in columns with colors according to key above. Trap labels to the right of heatmap. Actual counts are labeled in each cell. Specific traps discussed in text shown with yellow and green boxes. Some traps were added after the releases began. Yellow and green bars below indicate weeks used for means shown in panels B (green) and C (yellow). B) Four-week mean number of females in April 2021 (green bar under heatmap) in traps in ST02 according to legend colors showing outlier collection numbers in trap 14–6. Weekly collections for this trap shown in panel A. C) Same as B but for June 2021 (yellow bar under heatmap) showing high female means in the center of the cluster corresponding to an apartment complex. Map source: OpenStreetMap.org (CC BY-SA 2.0). Open Street maps were labeled in R using the Rgooglemaps package. https://www.openstreetmap.org/#map=14/18.01318/-66.60985.

https://doi.org/10.1371/journal.pntd.0012839.g004

Mosquito quality assessments

Reductions in male fitness related to shipping or other factors could limit their efficacy in the field, so we conducted a few assays to assess male quality. To test for large reductions in fitness related to the packing and shipping process, we conducted caged mating competition assays comparing Wolbachia-males, either held back or shipped, to wild-type males. If the two types of males were equally fit, we would expect approximately equal numbers of females mated to each type resulting in a mating rate of 0.5. We conducted a number of mating assays during the development stages of the shipping technology as well as over the duration of the project. We found that, although these assays are highly variable, Wolbachia-males that were held back before shipping competed for mates at an approximately equal rate to wild-type males (Fig 5), but mating rates of shipped males were reduced, with marked differences by shipping vessel type and if shipments were on time or delayed. At the beginning of the project in 2020, we made several adjustments to the shipping vessel and conducted mating assays for each prototype iteration. Mean male mating rates improved from 0.17 to 0.34 over the first three prototypes (v1-v3, S5 Fig) in our preliminary assays with simulated shipping conditions, but further improvements produced a fourth prototype (v3.1) that resulted in a mean mating rate of 0.44 (n = 4, S5 Fig). In the late stages of the project, we tested two additional modified versions of the shipping vessel and saw mixed results with one version (v3.2) resulting in mean mating rate of 0.29 (n = 4), while the other version (v4.0) resulted in a mean mating rate of 0.58 (n = 9). We used v3.1 to ship Wolbachia-males for most production shipments during the project and found a mean mating rate of 0.37 (n = 5) after shipping (Fig 5). We also assayed two samples from a single delayed shipment batch and found reduced mating rates with the Wolbachia-males (mean = 0.31). Cage-based mating assays provide coarse measurement of fitness, but we detected a clear impact of the packing and shipping on male fitness. It remains unclear how this fitness cost translated to field efficacy.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Male quality assessments.

A) Mating competitiveness cage assays. Each dot indicates the proportion of females mated with a Wolbachia-male in competition with wild-type males. Violin plots show median of distribution with white dot and quartiles. B) Reaction norm plot comparing mean lifespan for samples of males from the same batch assayed in San Francisco (held back) or in Ponce (shipped), either on-time or delayed according to legend. Violin plots summarize distribution of held back and shipped samples separately. C) Proportion recaptured (y-axis) in mark-release recapture (MRR) study as a function of distance from release point (x-axis) for two replicates according to legend. Each dot shows the total collected for a trap, and the lines show corrected collection rate (see methods). D) Proportion recaptured (y-axis) in MRR study over days since release (x-axis). Solid lines show empirical data and dotted lines show best-fit model for each replicate. E) MRR summary statistics. Estimated recapture rate and estimated proportion daily survival rate refer to values inferred from model fitting with 95% confidence intervals calculated from simulations (see Methods). Mean distance traveled in meters after correcting for variable trap density (see Methods).

https://doi.org/10.1371/journal.pntd.0012839.g005

Throughout the project, we measured male longevity of samples held back at the Verily production facility as well as samples after both on-time shipping and delayed shipping to Ponce. We compared samples taken from the same rearing batch and found that on-time shipped longevity was significantly lower than longevity for held back samples that were not shipped (Wilcoxon signed rank test P = 0.0379, n = 75 batches, Fig 5), although the difference between means was small (5.06 days for shipped vs. 5.31 days for held back), suggesting only minor impacts overall of packing and shipping on male longevity under laboratory conditions. We only assayed a small number of samples from delayed shipments, but the impact appeared stronger for these samples with a mean longevity of 3.14 (n = 4 batches) compared to 4.54 for held back samples from the same batch, which is consistent with our observations that males from delayed shipments appeared sluggish and unfit. Fig 5B reveals longevity estimates for shipped males that both increase and decrease underscoring the noise associated with this assay, and the overall reduction of the shipped class is small but significant under these conditions. It remains unclear how the differences we measure in the lab translate to the field, but it is clear that packing and shipping had at least some negative impact on male fitness.

Mark-Release-Recapture (MRR) assessments can provide more realistic estimates of longevity and flight performance in the field, so we conducted a series of two MRRs in a community not involved in active suppression releases. We recaptured a very small proportion of the released males for both replicates (rep 1 = 0.004, rep 2 = 0.006). We placed BG Sentinel 2 traps out to 350m from the release point and estimated mean distance traveled of 155m and 183m for replicate 1 and 2, respectively, after correcting for variable trapping effort (Fig 5C and 5E). To jointly estimate a corrected recapture rate and daily survival, we fit a non-linear model (Fig 5D) to the recapture data [41] and obtained very low daily survival rates (0.37 and 0.40, Fig 5E) and corrected recapture rates around 2% and 3% for replicates 1 and 2, respectively. Daily survival rates of ~0.4 suggests that only 40% of males survive each day resulting in very low male densities after only two days post release. Although the recapture rates were very low, data from two MRR replicates are consistent with significant negative impacts of the shipping on the daily survival of Wolbachia-males, likely limiting efficacy of the intervention.

Ecological compensation of the wild population

Another possible explanation for limited suppression is that our intervention produced a compensatory or overcompensatory response, where survival to adulthood is either unchanged or increased, respectively, after extrinsic mortality, such as predation or IIT interventions, is applied to the wild Ae. aegypti population[51,52]. Recent studies suggest that removal of Ae. container breeding mosquitoes at the embryonic stage, as is the case during IIT and SIT interventions, can relieve resource competition in the larval habitat resulting in higher survival to adulthood than without the intervention [53,54]. To test for evidence of compensation or overcompensation, we summarized induced sterility, mean female numbers, and suppression on a monthly basis and made several comparisons to ask whether our data fit expectations under these models. First, we compared the change in suppression (after three-week lag, calculated using Abbot’s formula, see Methods) to induced sterility using a Before-After Control-Impact (BACI) analysis at the cluster level and saw a large proportion of months consistent with compensation in all clusters where induced sterility was at least moderate with little to no impact on suppression three weeks later (Fig 6). A small number of months were in line with an additive effect, but a much larger proportion of months showed negative change in suppression despite moderate levels of induced sterility (Fig 6) consistent with over-compensation. Importantly, we used a GMM to infer suppression while accounting for other variables such as rain that are clearly important drivers of female densities, and comparison of BACI analysis using GMM model output with induced sterility levels provides support for a compensatory response throughout the project (Fig 6).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Evidence for ecological compensation and overcompensation.

Each dot represents the mean for a four-week window for each Phase II treatment cluster according to the legend. A) Change in suppression calculated using before-after-control-impact comparing windows separated by three weeks. Analysis of fitted Generalized Mixed Model shown in black. Ovals show regions where values under Additive, Compensation, and Over-compensation models would be expected. B) Impact on suppression as a function of mean female densities at the time of application. C) Impact on suppression as a function of mean egg densities at the time of application. D) Induced sterility as a function of female densities in the same window. Dotted lines show best-fit line and results of Pearson’s correlation test shown in bottom right.

https://doi.org/10.1371/journal.pntd.0012839.g006

One of the recommendations for avoiding compensatory and over-compensatory effects during vector control interventions is to remove larvae, or induce sterility, when the population density is lower than normal [51]. To test whether this dynamic may have impacted our project, we compared the change in suppression to both female densities (ρ = -0.19; P = 0.0067, Fig 6) and egg densities (ρ = -0.29; P = 2.0e-5, Fig 6) and saw a significant correlation in both cases, suggesting less suppression at higher female counts and egg counts. Interestingly, we also compared the level of induced sterility to the mean number of females at the time of application and saw significant and negative correlation (ρ = -0.39; P = 6.1e-9, Fig 6). Together these data are consistent with a scenario where the impact of male releases was directly related to the wild population density as would be expected if release rates or male fitness were insufficient to compete with high density wild populations.