Skip to main content
  • Loading metrics

From Lab to Field: The Influence of Urban Landscapes on the Invasive Potential of Wolbachia in Brazilian Aedes aegypti Mosquitoes

  • Heverton Leandro Carneiro Dutra ,

    Contributed equally to this work with: Heverton Leandro Carneiro Dutra, Lilha Maria Barbosa dos Santos

    Affiliation Mosquitos Vetores: Endossimbiontes e Interação Patógeno Vetor, Centro de Pesquisas René Rachou—Fiocruz, Belo Horizonte, Minas Gerais, Brazil

  • Lilha Maria Barbosa dos Santos ,

    Contributed equally to this work with: Heverton Leandro Carneiro Dutra, Lilha Maria Barbosa dos Santos

    Affiliation Laboratório de Transmissores de Hematozoários, Instituto Oswaldo Cruz, Fiocruz, Rio de Janeiro, Rio de Janeiro, Brazil

  • Eric Pearce Caragata,

    Affiliation Mosquitos Vetores: Endossimbiontes e Interação Patógeno Vetor, Centro de Pesquisas René Rachou—Fiocruz, Belo Horizonte, Minas Gerais, Brazil

  • Jéssica Barreto Lopes Silva,

    Affiliation Mosquitos Vetores: Endossimbiontes e Interação Patógeno Vetor, Centro de Pesquisas René Rachou—Fiocruz, Belo Horizonte, Minas Gerais, Brazil

  • Daniel Antunes Maciel Villela,

    Affiliation Programa de Computação Científica, Instituto Oswaldo Cruz, Fiocruz, Rio de Janeiro, Rio de Janeiro, Brazil

  • Rafael Maciel-de-Freitas,

    Affiliation Laboratório de Transmissores de Hematozoários, Instituto Oswaldo Cruz, Fiocruz, Rio de Janeiro, Rio de Janeiro, Brazil

  • Luciano Andrade Moreira

    Affiliation Mosquitos Vetores: Endossimbiontes e Interação Patógeno Vetor, Centro de Pesquisas René Rachou—Fiocruz, Belo Horizonte, Minas Gerais, Brazil



The symbiotic bacterium Wolbachia is currently being trialled as a biocontrol agent in several countries to reduce dengue transmission. Wolbachia can invade and spread to infect all individuals within wild mosquito populations, but requires a high rate of maternal transmission, strong cytoplasmic incompatibility and low fitness costs in the host in order to do so. Additionally, extensive differences in climate, field-release protocols, urbanization level and human density amongst the sites where this bacterium has been deployed have limited comparison and analysis of Wolbachia’s invasive potential.

Methodology/Principal Findings

We examined key phenotypic effects of the wMel Wolbachia strain in laboratory Aedes aegypti mosquitoes with a Brazilian genetic background to characterize its invasive potential. We show that the wMel strain causes strong cytoplasmic incompatibility, a high rate of maternal transmission and has no evident detrimental effect on host fecundity or fertility. Next, to understand the effects of different urban landscapes on the likelihood of mosquito survival, we performed mark-release-recapture experiments using Wolbachia-uninfected Brazilian mosquitoes in two areas of Rio de Janeiro where Wolbachia will be deployed in the future. We characterized the mosquito populations in relation to the socio-demographic conditions at these sites, and at three other future release areas. We then constructed mathematical models using both the laboratory and field data, and used these to describe the influence of urban environmental conditions on the likelihood that the Wolbachia infection frequency could reach 100% following mosquito release. We predict successful invasion at all five field sites, however the conditions by which this occurs vary greatly between sites, and are strongly influenced by the size of the local mosquito population.


Through analysis of laboratory, field and mathematical data, we show that the wMel strain of Wolbachia possesses the characteristics required to spread effectively in different urban socio-demographic environments in Rio de Janeiro, including those where mosquito releases from the Eliminate Dengue Program will take place.

Author Summary

Wolbachia are maternally inherited bacterial endosymbionts that induce cytoplasmic incompatibility in mosquitoes, and use this pattern of sterility to facilitate their spread into naïve, uninfected host populations. These bacteria were recently introduced into Aedes aegypti, a species naturally free of Wolbachia, and the primary vector of dengue virus. Wolbachia are currently being used as a biological tool to control dengue transmission in many countries, including Brazil. We performed a series of laboratory and field assays and then created mathematical models in order to understand the invasive ability of the wMel Wolbachia strain in Brazilian Ae. aegypti mosquito populations. We show that wMel infection had no detrimental fitness effects on its new Brazilian host, and would theoretically be able to successfully invade the mosquito populations in the five distinct urban landscapes that we studied. We found that Wolbachia spread more easily where the local mosquito population was smaller. The size of the local human population did not overtly influence the likelihood of spread. These findings provide insight into how the nature of different urban landscapes will affect the likelihood of Wolbachia spreading effectively, which will greatly assist future releases of Wolbachia-infected mosquitoes in Brazil as part of the Eliminate Dengue Program.


Dengue is widely considered the most important vector-borne virus, with an estimated 2.5 billion people living at risk of infection in more than 100 countries [1]. The virus is transmitted by mosquitoes from the genus Aedes, with Aedes aegypti being the primary vector. There are an estimated 100 million new cases of the disease each year, and this imposes a heavy burden on global health and economics, with the average annual cost of the disease estimated to be US$2.1 billion for the Americas alone [2]. Control efforts are complicated by the current lack of an effective dengue vaccine, the emergence of insecticide resistance in natural vector populations, and the difficulties and costliness associated with maintaining effective suppression of mosquito breeding sites. Consequently, alternative approaches are required in order to restrict dengue transmission [3].

A novel and promising form of biocontrol designed to reduce dengue transmission utilizes the endosymbiotic bacterium Wolbachia pipientis, which is naturally present in up to 40% of all insect species [4]. Different strains of this bacterium can cause a variety of phenotypic effects. Of these, the phenotypes of cytoplasmic incompatibility (CI), high bacterial density, and as a result, a high maternal transmission rate are the two main determinants of Wolbachia’s “drive” mechanism, which facilitates the spread of the bacterium into field mosquito populations [5]. CI is a reproductive incompatibility that prevents females without Wolbachia from producing viable offspring after copulation with Wolbachia-infected males. In contrast, Wolbachia-infected females can successfully reproduce after mating with any male. This increases the frequency of Wolbachia infection in a given population with each subsequent generation. Ae. aegypti is not naturally infected with Wolbachia, but has been transinfected with several different Wolbachia strains that cause strong CI [68]. A further, significant discovery was the fact that some Wolbachia strains can prevent or severely restrict infection by key pathogens, including dengue virus, in the host mosquito [7,9]. Together these phenotypes would allow the replacement of wild, dengue-susceptible mosquitoes with dengue-resistant, Wolbachia-infected mosquitoes potentially ameliorating the great burden of dengue [10].

The use of Wolbachia as a biocontrol tool to reduce dengue transmission is currently being tested in several countries ( Here Wolbachia-infected mosquitoes are released in large numbers over a period of weeks, and eventually replace the wild Ae. aegypti population. So far, field trials have already been conducted in four sites in Australia, one in Vietnam, and two in Indonesia, while trials in Brazil have been underway since September 2014 using Ae. aegypti infected with the wMel Wolbachia strain. The Brazilian trials are set to take place at four different field sites over the next two years. Critically, there is a great deal of variability in the physical, environmental and socioeconomic conditions between these four sites, and between those from the other countries involved in the project. These differences include factors such as climate, housing and population density, sanitation conditions and the use of piped or stored drinking water. All of these factors can affect the size and characteristics of the already-established wild mosquito populations [11], which in turn determines the number of Wolbachia-infected mosquitoes that must be released in order for the bacterium to spread to high frequencies in a given area [12]. As such, it is important to identify these factors and determine if any could potentially limit the invasive potential of Wolbachia, as this could necessitate that mosquito release characteristics, including release duration and mosquito numbers, be tailored for individual release sites within a city.

Therefore, the objective of this work was to determine whether the likelihood of a successful Wolbachia invasion could be influenced by differences between release sites, in preparation for the release of Wolbachia-infected Ae. aegypti in Brazil. To address this objective, we first examined whether the key invasion phenotypes of CI and maternal transmission were present in Ae. aegypti with a Brazilian genetic background, infected with the wMel Wolbachia strain, as these phenotypes had only previously been characterised in mosquitoes from other genetic backgrounds. We then investigated the characteristics of Ae. aegypti populations in five Brazilian suburbs in Rio de Janeiro with different socio-demographic characteristics. Of special interest were those factors that could directly influence or restrict the spread of Wolbachia into natural mosquito populations, including the Ae. aegypti population size, and mosquito daily survival rate. Finally, by amalgamating these data we were able to simulate the likelihood that Wolbachia would disseminate and spread so that all mosquitoes at the five sites became infected. We observed that successful invasion would likely occur for all sites; however the means by which is did so varied between the sites, and invasion occurred more slowly at sites where local mosquitoes were more abundant. Through the results of these models we were then able to develop optimized release strategies for each field site, which will greatly benefit the upcoming biocontrol trials, and likely prove applicable to field sites in other cities and countries that share similar characteristics.


Ethics statement

MRR experimental protocols were submitted to and approved by FIOCRUZ Ethical Committee (CEP/FIOCRUZ protocol no. 22286313.7.0000.5248). Permission was obtained from the Rio de Janeiro Department of Health before releases commenced. Local residents received a full explanation of the project by at least one of the co-authors, and by the same health agent who visited their home 4–6 times per year during routine mosquito surveillance. This explanation highlighted the relevance and objectives of the MRR experiment.

Households were only entered after receiving formal written consent from the householders.

Laboratory assays

Mosquito rearing.

All mosquitoes used in these experiments were maintained in a climate controlled insectary, at 27 ± 1°C and 70 ± 10% relative humidity, with a 12:12 hour light:dark cycle. Larvae were reared in filtered and dechlorinated water, and fed with one Tetramin Tropical Fish Food tablet every two days. Adult mosquitoes received 10% sucrose solution ad libitum, and females were blood-fed weekly with human blood using a Hemotek membrane feeder (Hemotek Ltd) when eggs were required for experimental or colony rearing purposes. Prior to feeding, the blood was tested for the presence of dengue virus using the Dengue NS1 Ag Strip test (BioRad Laboratories).

Backcrossing of wMel-infected Ae. aegypti.

wMel-infected Ae. aegypti were imported to Brazil from Australia (IBAMA license 11BR005873/DF). Prior to field releases in Brazil it was necessary to create a Wolbachia-infected line with a Brazilian genetic background. To that end, the Australian wMel-infected mosquitoes were backcrossed with wildtype Brazilian mosquitoes collected from four different districts in Rio de Janeiro (Paquetá, Jurujuba, Belford Roxo and Vila Valqueire) following a previously described protocol [13]. The parental generation in the backcross consisted of 250 virgin, Wolbachia-infected, Australian females and 200 uninfected, Brazilian males. This process was repeated for nine generations in order to generate the backcrossed Wolbachia-infected line named wMel_Br.

Generation of Ae. aegypti line cured of wMel infection.

A group of wMel_Br mosquitoes were cured of their Wolbachia infection through treatment with antibiotics over three successive generations. Adult mosquitoes were provided with 10% sucrose solution containing 0.1mg/mL of tetracycline hydrochloride for a period of 10 to 14 days, which encompassed the first gonotrophic cycle [14]. Approximately 1000 adult mosquitoes were treated every generation, in order to limit detrimental genetic effects that might occur as a consequence of low population size. In each generation, twenty-four females and twelve males were randomly screened using qPCR (as described below) to confirm the loss of Wolbachia. Antibiotic treatment was discontinued when 100% of samples were negative for Wolbachia. This process took three generations to complete. The resulting line, named wMel_BrTET, was allowed two generations for microbiota reacquisition prior to the start of experiments. During this period, wMel_BrTET larvae were reared in trays dosed with the larval water from wildtype mosquitoes that were not treated with antibiotics.

Monitoring Wolbachia infection status.

Conventional PCR for the Wolbachia surface protein (wsp) gene was used to examine the Ae. aegypti colonies for the presence of Wolbachia [15]. DNA was extracted from individual mosquitoes by homogenization in 80μL of Squash buffer using a Mini-Beadbeater-16 (BioSpec) [16]. PCR reactions were performed using a Veriti Thermal cycler using previously published primers for wsp and the Ae. aegypti Ribosomal S17 (rps17) gene [9], following previously described reaction conditions [14]. In cases where the PCR result was unclear, samples were rerun using real-time PCR on a 7500 Real-Time PCR System (Life Technologies). For real-time PCR, crude DNA was extracted from individual mosquitoes according to existing protocols [16], and then quantified in duplicate for both wsp and rps17, using a two-step reaction (Vol. = 10μL) using 2X SYBR Green PCR Master Mix (Applied Biosystems) [14].

Cytoplasmic Incompatibility and maternal transmission.

To determine if wMel caused CI in the Brazilian genetic background, reciprocal crosses were performed between wMel_Br and wMel_BrTET lines. In two replicate experiments, 50 male and female pupae from each line were sexed by visual analysis of their terminalia using a stereomicroscope, and then placed in a single cage (each cage representing one of the reciprocal crosses depicted in Fig 1) where they were allowed to mate for five days. On the fifth day, females from all groups were blood-fed for one hour using a Hemotek feeder. At 72 hours post-feeding, thirty females from each group were randomly selected and placed individually in inverted petri dishes containing moist filter paper for oviposition [17]. The total number of eggs laid by each female (fecundity) was counted using EggCounter V 1.0 [18]. The number of hatched eggs from each female (fertility) was counted with the aid of a stereomicroscope. Conventional PCR for the wsp gene (as described above) was used to check the infection status of males and females in all reciprocal crosses. A TaqMan-based assay was used for verification when conventional and SYBR Green-based real-time PCR results were unclear. In order to evaluate the rate of Wolbachia maternal transmission, females originally used in the cytoplasmic incompatibility experiment described above, had their eggs hatched and their offspring reared and tested for the presence of Wolbachia. Total DNA from larvae, pupae and adults was extracted and then used in the conventional PCR assay for wsp (as described above).

Fig 1. wMel causes strong cytoplasmic incompatibility in Brazilian Ae. aegypti mosquitoes.

Percentage of hatched eggs per female for all potential crosses between wMel-infected Brazilian Ae. aegypti (green) and Wolbachia-uninfected Brazilian Ae. aegypti (black). Each circle represents a single adult female, while the red horizontal lines indicate the median number of hatched eggs for each cross. The cross highlighted in yellow is the incompatible cross and was statistically different from the other three groups, while the two crosses highlighted in blue had significantly different hatch rates according to Dunn’s test. These data were pooled from two independent biological replicates. The total number of females (n) and the total number of eggs (neggs) are indicated above each cross.

Field assays

Study areas.

For our field release assays, we selected five distinct field sites in the vicinity of Rio de Janeiro, Brazil. Collectively, these sites are representative of the broad range of socio-demographic and economic conditions found across the city, with each providing a distinct environment for mosquito release (Table 1). Demographic data from all five of the sites were collected from the 2010 Census ( The remaining descriptive data were collected via observation during the field site characterization, and during previous experiments [19,20].

Characteristics of mosquito populations at field sites.

Approximately 30 BG-Sentinel traps (Biogents) were deployed at each of the five field sites in order to collect Ae. aegypti females, as a means to estimate the mosquito population size, as well as the species composition at each field site. The traps were checked weekly between October 2012 and October 2014, for a total of 104 weeks of mosquito collection. Each trap was analysed individually. All captured mosquitoes were transported to the laboratory, mosquito numbers were counted and individual mosquitoes were identified down to the species level using taxonomic keys [21].

Probability of daily survival.

In order to estimate mosquito survival we conducted mark-release-recapture experiments in Jurujuba, a working class area, and Vila Valqueire, a middle class area. Information regarding the probability of daily survival (PDS) from the other three field release was obtained during previous MRR experiments [19,22]. The mosquitoes used in these experiments were the F1 and F2 generations of eggs collected from 40 ovitraps that were deployed at the field sites. These mosquitoes were reared as described above, unless specified below. Wildtype mosquitoes were used in these assays in order to obtain an accurate picture of the fitness of mosquitoes already present at the sites where Wolbachia-infected mosquitoes will be released.

Mark and release.

Before each experiment, eggs were divided into two groups, which were hatched 3 days apart to produce two independent cohorts of adult females. Females from one cohort were blood-fed approximately 18h before release (Cohort A), whereas females from the second cohort were released while unfed (Cohort B). The reason for this disparity was to observe potential effects on the PDS due to blood feeding. A greater PDS amongst Cohort A mosquitoes could promote more effective invasion by Wolbachia as the mosquitoes would not need to spend time searching for a host. We sought to assess whether there were any benefits to releasing using this technique. Each cohort was marked with a different colour fluorescent powder (Day-Glo Color Corp.) in small cylindrical cups (12 x 10cm). Females from both cohorts were released into the test sites on the morning of their fourth day post-eclosion (between 8:00 AM and 9:00 AM), 1h after powder application. Mosquitoes were released outside of the houses at the field sites, with an average of 4.2 mosquitoes/house released in Vila Valqueire and 7.5 in Jurujuba.


Marked females were captured with CDC backpack aspirators (John W. Hock), and sticky traps [20]. Captures started one day after mosquito release. Fifteen houses were randomly selected per day for backpack aspiration. Collections were completed within 15–20 mins for each house. The entire property, inside and outside, was aspirated. If the house in question had a backyard or garden, mosquitoes in that area were also collected, but only within the limits of the property. Sticky traps were installed in 15 randomly selected houses. These were monitored daily for the presence of dust-marked females, throughout the course of the assay. Occasionally, aspiration was conducted in the same house where a sticky trap was installed. Daily captures were conducted by nine consecutive days. Captured mosquitoes were examined and sorted by cohort/powder type under UV light (S1 Table).

Climatic conditions during MRR assays.

The Climate in Rio de Janeiro is characterized by a dry winter (May–August) and a wet summer season (November–March). MRR experiments were performed during September (Vila Valqueire) and November (Jurujuba). During the thirteen days of fieldwork (one day to release and twelve of capturing marked mosquitoes), we registered a mean temperature of 25.7 ± 4.8°C, 6.8 mm rainfall, mean relative humidity of 64.54% and an average wind speed of 1.67m/s in Vila Valqueire. While in Jurujuba there was a mean temperature of 24.7 ± 2.6°C, 55.8mm total rainfall, mean relative humidity of 81.88%, and an average wind speed of 2.31m/s. Air temperature and precipitation data for these periods were recorded every hour and were obtained from meteorological stations located less than 10km away from the two study areas.

Survival rate analysis.

PDS for each of the two cohorts was estimated by fitting two models, the first being exponential [23], and the second, nonlinear [24]. From these models we derived two values: the average life expectancy (ALE), defined as 1/loge PDS, and longevity, defined as PDS10 (where 10 is the estimate of the duration of the extrinsic incubation period for dengue in days given the temperature of the field sites) [25], which gave us the expected proportion of mosquitoes surviving long enough to transmit dengue virus. We fit both models to our data, using linear and nonlinear least squares standard procedures using R version 3.0.1 [26].

Modelling the invasive potential of wMel in the five field sites

Abundance estimation.

We used a logistic regression model based on a Lincoln type index to estimate the size of the Ae. aegypti population at each of the five field sites. This model incorporated the PDS estimations calculated from the MRR experiments and also demographic data from each field site (using two parameters; (1) the number of premises, and (2) number of inhabitants). The number of captures of marked mosquitoes (observations) was modeled as a binomial variable “yi ~ Bin(Ni, pd)”, where “pd” is the probability of capture, and “Ni” is the number of mosquitoes released at each site. We assumed that the probability of capture in the MRR experiments varied depending on each site's conditions, namely the number of people living in each area (hpopsize) and the number of premises in the area (premises), as represented by a logistic model: where “logit” represents a logistic function. The coefficient k0 describes the proportion of the mosquito capture rate that is associated with intrinsic characteristics of the type of trap used. The coefficients k1, and k2 indicate the weighting of each of the two variables (number of premises and human population size) indicating the extent to which each variable influenced the probability of mosquito capture. These coefficients were estimated using the GLM framework in the R statistical platform. These two variables were included in the model because they both had a statistically significant influence on the likelihood of mosquito capture (at the 95% significance level), and because a two-variable model was a better fit for our MRR data, than compared to models that included additional variables (geographical area, human density). These other variables did not have a significant effect on mosquito capture. This type of modeling approach differs to the more traditional use of a Lincoln Peterson estimation, which considers all variables for each site as a single coefficient. Although there were clear landscape heterogeneities between the sites, the recapture protocol we used was identical, which meant that the significance of these parameters could be estimated using logistic regression [27]. The estimate of mosquito abundance at each field site was described by the relationship between the average number of mosquitoes collected each week and their recapture probability. These values were derived from the model and adjusted for weekly differences, and then used to represent the carrying capacity of each site in the mathematical models used to examine wMel invasion (see below).

Invasion analysis.

In order to study the spread of Wolbachia into the five field sites, we used models based on ordinary differential equations (ODE) [28] to mathematically describe the dynamics of the invasive potential of the wMel strain. The models simulated 12 weekly releases of a group of wMel-infected mosquitoes into each of the field sites (the size of this group varied in order to determine the number of mosquitoes required for successful spread), and then described changes to the proportion of Wolbachia-infected individuals in the total mosquito population over time. We assumed that the female-to-male ratio was equal for both populations [28]. These models incorporated parameters derived from our phenotypic characterization of wMel-infected Brazilian mosquitoes, and characteristics of the local mosquito population at each field site (Table 2). In this model Wolbachia infection imposed a fitness cost, increasing the mortality of infected individuals [11].

Table 2. Field site mosquito population characteristics & model parameters.

The two ODE were as follows: 1 2 Where Eq 1 describes the wMel-infected population, and Eq 2 the-uninfected, wildtype population. The variables I and U respectively describe the number of infected and uninfected individuals. The equations contained several parameters describing the release and spread of Wolbachia. v—the vertical transmission rate of wMel. b—the recruitment rate (b=log(1+α.p)), where α is the mean number of eggs laid, and p is the probability that any given egg would result in a live adult mosquito. The value for p was derived from the PDS data from each field site. s—the life shortening effect associated with wMel infection. K—the carrying capacity of the mosquito population at each field site, as calculated using the logistic regression models. f—the cost of CI in the uninfected population, , where c is the rate of CI.

For each site, we considered scenarios involving changes to the maternal transmission rate and decreased longevity resulting from wMel infection in order to determine their effect on the likelihood of Wolbachia spreading to infect all mosquitoes in the area. We then considered changes to the number of Wolbachia-infected mosquitoes released each week, in order to determine the optimal release strategy at each site.

Data analysis

The CI data were analysed using Kruskal-Wallis one-way analysis of variance, followed by pairwise comparison using Dunn’s tests, and fecundity data were compared using Mann-Whitney U-tests. Differences in survival rates between the field sites were evaluated using exponential and non-linear models. Exponential models have traditionally been used to estimate Ae. aegypti survival but these possesses two fundamental drawbacks; they assume a priori that mosquito mortality does not vary with increasing age, and do not account for the removal of mosquitoes from the population during the recapture process. Consequently, we utilised a non-linear survival model, which incorporated a correction for the removal of individuals [24]. All results were analysed through GraphPad Prism version 5.03 for Windows (GraphPad Software, CA, (, or using R 3.0.1 software [26].


CI and fecundity

Reciprocal crosses between the wMel_Br and wMel_BrTET mosquito lines revealed strong cytoplasmic incompatibility due to the presence of wMel in Ae. aegypti mosquitoes with a Brazilian genetic background (Fig 1). A CI rate of 99.51% was observed when uninfected females were crossed with infected males (the “incompatible” cross), where only 15 of 3035 eggs hatched, a significantly lower proportion than what was observed in the other crosses (Kruskal-Wallis, H = 124.1, df = 3, P < 0.0001). Conversely, the median percentage egg hatch per female in the other three “compatible” crosses ranged between 90.3 and 96.9%. Additionally, the cross between infected females and uninfected males resulted in a higher number of hatched eggs compared to the cross between uninfected females and uninfected males (Mann-Whitney U test, U = 603.5, df = 1, P < 0.0001). All females involved in the three compatible crosses laid at least one viable egg, indicating that they all had mated successfully. Although there was slightly higher fecundity in crosses involving wMel-infected females, there were no significant statistical differences between the four experimental crosses (Fig 2; Kruskal-Wallis, H = 4.141, df = 3, P = 0.2466).

Fig 2. wMel does not affect fecundity in Brazilian Ae. aegypti.

Number of eggs laid by females from the reciprocal crosses between wMel-infected (green) and antibiotic-treated (black) Brazilian Ae. aegypti mosquitoes. Each circle represents a single adult female, while the horizontal red lines indicate the median number of eggs laid in each cross. The presented data are pooled from two independent biological replicates. No significant differences were detected between the four crosses. The total number of females (n) and the total number of eggs (neggs) are indicated above each cross.

Maternal transmission rate

All adult wMel_Br females that were involved in the first replicate of the CI experiments were screened for Wolbachia infection using qPCR, with a 100% infection rate observed. The resulting progeny from these mosquitoes were also screened for Wolbachia to provide an estimate of the maternal transmission rate in the Brazilian genetic background. Overall, 757 larvae, pupae and adult mosquitoes were screened. Of these, 725 were infected and 32 were uninfected, which implied a 96% maternal transmission rate of wMel, in Brazilian Ae. aegypti mosquitoes. This experiment was then repeated using the progeny from the second CI experiment. This time only adult mosquitoes were screened for Wolbachia, and 239 of 239 samples were identified as positive, indicating a 100% maternal transmission rate.

Mosquito population characteristics at field sites

Over the course of the two-year collection period we collected a total of 181,937 wild mosquitoes across the five field sites; 82,316 of these were Ae. aegypti, 2,867 were Aedes albopictus and 96,754 were Culex quinquefasciatus. We observed differences in the mean weekly number of mosquitoes captured between the sites (Table 3). The highest numbers of collected mosquitoes were obtained from Vila Valqueire, the middle class area, where 6043 more mosquitoes were collected than in Tubiacanga, the lower middle class area, which had the next highest number collected. Aside from the slum, Amorim, where the collection period was shorter than for the other sites, total collections were the lowest in Urca, the upper-middle class area. Mean (SD) weekly Ae. aegypti collections were the highest in Vila Valqueire (256 ± 74.3), almost double that of Urca (144.1 ± 40.6), which had the lowest collection rate.

MRR rate

During the MRR assay, a total of 1,730 Ae. aegypti females were released in Vila Valqueire. Daily captures occurred for 9 consecutive days after the initial release. A total of 98 marked females were collected during this period for a recapture rate of 5.66%. Of these, 67 (68.4%) were captured with backpack aspirators, while the remainder were recovered from sticky traps. In Jurujuba, 3,000 Ae. aegypti females were released, with collections again occurring daily for 9 days post-release. We collected 95 marked mosquitoes, corresponding to a recapture rate of 3.17%. Of these, 44 (46.3%) were collected using backpack aspirators, and 51 (53.7%) in sticky traps. The overall recapture rate was much higher for Vila Valqueire than for Jurujuba (Table 4). In Vila Valqueire we observed very similar recapture rates for both Cohorts, however for Jurujuba the recapture rate of Cohort A, where mosquitoes were blood fed prior to release, was almost twice that of Cohort B (Table 4).

Table 4. Recapture rate and longevity of Ae. aegypti females released in Vila Valqueire and Jurujuba.

Daily survival rates and longevity estimates

We used the MRR data to create two models, a non-linear model and an exponential model, in order to estimate the PDS of the two mosquito cohorts that were released at each of the two field sites. The non-linear model provided higher estimates of PDS than the exponential model. This was expected because that model corrects for the removal of mosquitoes collected over previous days (Table 4). Mosquitoes released in the Jurujuba had a higher overall survival rate than those released in Vila Valqueire (Student’s t test, t = 6.75, df = 1, P < 0.001). At the latter site, mosquitoes released soon after blood-feeding (Cohort A) had a better survival rate, with the PDS 10.27% higher than for those released under conditions of starvation (Cohort B) (Table 4). However, we observed the opposite pattern in Jurujuba, where the PDS of Cohort B was 13.76% higher than that of Cohort A. We observed variation in estimates of average life expectancy (ALE) between the two field sites, and between the different cohorts released at the same sites. Regardless of the pre-release feeding regime, the expected lifespan of mosquitoes released in Jurujuba was higher than those released in Vila Valqueire, as evidenced by a two-day difference in the ALE upper limit estimate.

Likelihood of successful wMel invasion at field sites

We used statistical models based on logistic regression that incorporated data from the MRR experiments to estimate the average Ae. aegypti population size at each of our five field sites (Table 2). The highest estimate of abundance was obtained in Vila Valqueire, with Jurujuba having the second largest population of mosquitoes. The estimated Ae. aegypti population size at the other three sites was three to four times smaller than either Jurujuba or Vila Valqueire.

We then used mathematical models based on ordinary differential equations in order to simulate the spread of wMel-infected Ae. aegypti at each of the five field sites. These models incorporated parameters that were determined from the laboratory experiments and the abundance values derived from the MRR experiments (Table 2). In these models, only the variables of human population size and number of houses were significant predictors. The geographical area and density of inhabitants had no effect. We then compared the effects of different maternal transmission rates and life-shortening levels on the ability of the mosquitoes to spread. The initial scenario assumed a 96% rate of maternal transmission of wMel, and life shortening of 10% associated with infection [7]. We considered scenarios where the numbers of wMel-infected female mosquitoes released each week varied between 200 and 400 (Fig 3). For all five sites, a release number of 200 mosquitoes led to Wolbachia becoming established in Amorim, Tubiacanga and Urca inside of 50 days. A release size of 300 was sufficient for the Wolbachia infection frequency to reach 100% within 100 days in Jurujuba. For Vila Valqueire the release of 400 female mosquitoes per week was required for complete invasion, and this occurred inside of 100 days.

Fig 3. Modeling the invasive potential of wMel at different field sites.

Models of the invasive potential of wMel at the five different field sites based on ordinary differential equations assuming a maternal transmission rate of 96% and a longevity cost of 10%. Each plot shows the changes to the total number of Wolbachia-infected (solid lines) and—uninfected (dashed lines) mosquitoes (y-axis) in the population over time (x-axis), during the course of a Wolbachia invasion involving 12 weekly releases of different sized cohorts of infected mosquitoes. When Wolbachia becomes fixed the size of the uninfected population decreases to zero. Red lines depict a weekly release of 200 Wolbachia-infected females, which was sufficient for wMel to spread to infect all mosquitoes in Amorim, Tubiacanga, and Urca. Blue lines depict a release of 300 females, which led to successful invasion at all sites except Vila Valqueire, which required 400 females per week (black lines). The differences in the number of mosquitoes at each site reflect the Ae. aegypti abundance derived from experimental data.

Next, we examined a scenario where the life shortening associated with wMel was greater than expected—increasing from 10% to 33%. Under these conditions the number of mosquitoes required for weekly release increased dramatically (Fig 4). In both Urca and Tubiacanga, the required number was 900 females, with a 100% infection rate being reached inside 50 and 100 days respectively at the two sites. For Jurujuba and Vila Valqueire, which had a greater natural abundance of Ae. aegypti, the release number required was 1500 and 2000 females respectively, and in both cases, Wolbachia was predicted to spread to all mosquitoes inside 100 days. In Amorim, the increased mortality rate had no effect on the number of mosquitoes required, with a 100% infection rate occurring within 50 days under a release regime of 200 or 300 mosquitoes.

Fig 4. The effect of a stronger life shortening on the invasive potential of wMel.

Models depict the invasive potential of wMel with a higher than expected life shortening effect (33% rather than 10%). Each plot shows the changes to the total number of Wolbachia-infected (solid lines) and—uninfected (dashed lines) mosquitoes (y-axis) in the population over time (x-axis), during the course of a Wolbachia invasion involving 12 weekly releases of different sized cohorts of infected mosquitoes. In this scenario far greater numbers of released mosquitoes were required for Wolbachia to reach 100% infection (than for a 10% life shortening effect), and these numbers differed greatly between sites. In Jurujuba, stable infection occurred after weekly releases of 1500 mosquitoes (blue lines), but not 900 (red lines). In Urca, invasion occurred with 900 mosquitoes (blue), but not 200 (red). For Tubiacanga, a release number of 900 mosquitoes was also sufficient to lead to a 100% infection (blue), but 300 was not (red). In Vila Valqueire, releases of 900 mosquitoes was insufficient (red), and 2000 mosquitoes per week were required (blue). In Amorim, releases of both 200 (red) and 300 (blue) mosquitoes led to all mosquitoes becoming infected.

We then considered the effect of a higher maternal transmission rate (99%) for both of the above scenarios. When life shortening was 10%, the increased maternal transmission rate had no effect on the spread of Wolbachia at Tubiacanga or Urca (S1 Fig). For Amorim, the decreased fitness cost led to equivalent invasion profiles with either 200 or 300 females per week. However, at both Jurujuba and Vila Valqueire, the number of released mosquitoes required to achieve stable infection was decreased, and the time to achieve a stable infection was shorter. When life shortening was increased to 33%, the increased maternal transmission did not affect the release dynamics at any of the field sites, with cohorts of similar size to those at the lower maternal transmission rate required to increase the frequency of Wolbachia infection (S2 Fig).


Characterisation of wMel infection in Brazilian Ae. aegypti

The use of Wolbachia to control dengue transmission represents a novel and promising method of mosquito biocontrol. Although the technique is still in early-stage field trials and restricted to small or isolated areas in a handful of countries, the scope and scale of these trials will inevitably expand. Releases of Ae. aegypti infected with the wMel Wolbachia strain have recently commenced in Brazil, where regulatory agencies provided approval for releases at four isolated sites through 2016 (these being four of the field sites examined in this article, with the exception of Amorim). Given that there could potentially be great variability in the conditions present at these sites, it is important to understand how differences in physical characteristics, human population demographic factors, and local mosquito population characteristics between these sites might affect the likelihood of successful invasion by Wolbachia. If there was such an effect, it would necessitate that key components of the release process (e.g. mosquito release numbers, number of release points, release period length) be tailored to individual release sites.

For Wolbachia to successfully invade in a given area, the infected mosquitoes must display a high level of cytoplasmic incompatibility [5]. This would compensate for imperfect maternal transmission of Wolbachia and any detrimental effects on host fitness resulting from infection, if either of those things were to occur [29]. In the case of the Brazilian releases, it was first necessary to transfer the wMel strain to mosquitoes with a Brazilian genetic background. Our results indicated that near-complete cytoplasmic incompatibility occurred when infected males mated with uninfected females. This would provide infected mosquitoes with a large reproductive advantage over uninfected mosquitoes in a mixed population, as might occur during a field release. Interestingly, we observed that egg viability was higher when uninfected males mated with uninfected females than with Wolbachia-infected females. This suggests that there may be a small cost associated with wMel infection that lead to reduced egg viability. When mated with uninfected males, the majority of Wolbachia-infected females displayed an egg hatch level that was comparable to that produced by matings of uninfected males and females, while a small group of these infected females laid eggs with greatly reduced viability. It is possible that this group of females all had higher than normal levels of Wolbachia, as this can lead to more extreme phenotypes in the host [30].

Fecundity is an important life-history trait in invertebrates, and is also important for Wolbachia-based control strategies, as higher fecundity will promote a more rapid spread of the bacterium [31]. In our experiments, we observed no significant difference in fecundity levels due to the presence of the wMel infection, which suggests that there will be no fitness impact on this trait due to infection. This was similar to what was observed in wMel-infected Ae. aegypti with an Australian genetic background [7].

In one of two experiments, we detected a lower rate of wMel maternal transmission in the Brazilian genetic background than what has previously been described (96% as opposed to 100% in the other experiment) [6,7]. Loss of transmission efficiency can be associated with environmental factors like high temperature [32], or with the host genetic background [33]. Given the consistency of laboratory temperatures, a difference due to the change in host genetics represents a more plausible scenario. Another potential explanation could be due to the Ae. aegypti-Wolbachia relationship, which is still rather novel from an evolutionary point of view. Consequently, we would expect a stronger host immune response to Wolbachia [34,35], which could potentially promote the loss of infection in some individuals [36]. This could occur if the immune response of Brazilian mosquitoes to Wolbachia is different to that of other clades. It is also possible that the maternal transmission rate could have been underestimated, and some of the samples were falsely identified as negative, perhaps due to insufficient dilution prior to qPCR at that time. Given the high levels of maternal transmission that have previously been observed for wMel [7], we were of the opinion that the results of the second experiment were more likely to be accurate. Nevertheless we also included the lower rate in the subsequent models.

Characterization of local mosquito populations

The ability of Wolbachia to become fixed in wild mosquito populations has previously been examined using theoretical models [37]. However, none of these have considered how conducting the experiments in large, heterogeneous cities such as Rio de Janeiro would affect the likelihood of a Wolbachia strain spreading effectively. Ae. aegypti life history traits and population size may vary dramatically between districts, depending on the nature and influence of urban landscapes and human demographic factors. This in turn would influence how Wolbachia might spread. To test these effects, we sought to characterise five distinct field sites in Rio de Janeiro, and the mosquito populations dwelling within, and then used mathematical models to predict how the resulting differences might affect invasion by Wolbachia.

By conducting mosquito collections over a period of two years, we were able to make estimates of existing mosquito population sizes, which varied greatly amongst our five field sites. The average weekly mosquito collection did not correlate with the development index of the sites. Of the four sites where collections ran the full 104 weeks, Vila Valqueire had the highest collection rate, and Urca the lowest. Interestingly, these two sites had the highest developmental index of our five sites. Both sites contain large houses/duplexes, have piped water, and organized garbage collection and have good mosquito control programs. This suggests that the developmental index of these areas is not a major determinant of the number of mosquitoes at these sites. In Urca there was no regular use of external water tanks, which may also have contributed to the lower prevalence of Ae. aegypti in the area.

In our MRR experiments we observed that the mosquito recapture rate was approximately twice as high for Vila Valqueire as for Jurujuba. This could have occurred because of increased access to properties in Vila Valqueire as a result of greater assistance from local residents. Alternatively, it is possible that insecticide usage may have been higher in Jurujuba during the experiment. During this experiment we released two different cohorts of mosquitoes, one blood fed prior to release, and the other unfed. The bloodfed mosquito cohort (A) was recaptured at a far greater rate than the non-bloodfed cohort (B) in Jurujuba, but there was no difference in the recapture rate between the two cohorts at Vila Valqueire. It is possible that this difference in recapture rate occurred due to a specific micro-environmental effect in Jurujuba that differentially affected the two cohorts, however it is not currently clear what that might be. Alternatively, this observation could have been an artefact of small collection numbers, rather than a real difference. We also observed a differential effect of pre-release blood feeding on survival, with the PDS of Cohort A mosquitoes higher than Cohort B in Vila Valqueire, but lower in Jurujuba. This suggests that there may certain conditions present at some field sites where it would be advantageous to release blood fed individuals. Similar experiments should be conducted with wMel-infected mosquitoes to determine if this effect is consistent in the presence of Wolbachia.

Comparable MRR experiments for the other three field sites; Amorim, Urca and Tubiacanga had been performed previously [19,22]. In these experiments only non-blood fed females were released, so the results were comparable to our data for Cohort B. Overall, we observed a much lower recapture rate for Jurujuba and Vila Valqueire than what was previously described at the other sites. A comparison of all sites revealed the highest recapture rates were in Amorim (13.10%) and Tubiacanga (12.82%) during the wet season. These values decreased slightly during the dry season but were still higher than either of the two new sites, as was the recapture rate in Urca (9.77%). The ALE of mosquitoes at Vila Valqueire was similar to those from Urca and Tubiacanga (during the wet season). While ALE values at Jurujuba were similar to those from Amorim (during both the wet and dry seasons) and Tubiacanga (during the dry season).

We observed great variation in the daily survival rate of mosquitoes between our five field sites. The highest PDS (0.93) was observed in the less affluent district of Amorim [22], while the lowest PDS of 0.60 was observed in Urca, which had the highest level of social development. The PDS of the other three sites ranged between 0.72 and 0.81. Although we obtained data for only five sites, we saw a very strong negative correlation between field site affluence and mosquito PDS. This suggests that the developmental area of a particular area could potentially have some effect on the survival of local mosquitoes, which is likely reflective of a disparity in mosquito control measures. We also saw a strong positive correlation between human density and PDS, which could be explained by the fact that higher density makes it easier for mosquitoes to find hosts and feed successfully. The fact that there were such considerable differences in PDS between the sites indicates that this is a factor that must be considered when planning releases. A lower PDS coupled with a large wild population could make it more difficult for Wolbachia-infected mosquitoes to become established, especially taking into account potential life shortening effects due to infection.

Invasive potential of wMel-infected Brazilian Ae. aegypti

We utilised the data obtained through the laboratory and field experiments described above to model the spread of wMel through the five field sites. Our models indicated that two field sites, Jurujuba and Vila Valqueire, had far greater abundance of Ae. aegypti than the other three sites. There are several potential explanations for the differences we observed in mosquito abundance. There could have been differences in the numbers, location, and favorability in breeding sites, differences in the likelihood of finding a host, differences in mosquito control measures, or in environmental conditions. All of these factors could not be directly considered in the statistical models. Interestingly, our data did not indicate that there was a strong relationship between PDS and abundance. Vila Valqueire and Jurujuba had high abundance but far lower numbers of houses and inhabitants that the other three field sites. This is somewhat counterintuitive, as a positive correlation between abundance and these factors makes more sense when considering the biological requirements of mosquito populations for hosts and breeding sites. Given that there are a large number of factors that affect the size of a mosquito population, some of which have not been considered in our models, it is unclear if this effect is something real or an artifact of considering only five locations.

Estimates of abundance at each of the five sites were directly incorporated in our models, and used to represent the initial size of Wolbachia-uninfected mosquitoes naturally present at each site prior to the time of release. The release protocols used in these models were simulated based on the general format of Wolbachia-infected mosquito releases as utilised as part of the Eliminate Dengue Project [11]. This involves the weekly release of a number of Wolbachia-infected female mosquitoes for a period of several months, during which time the frequency of Wolbachia-infected individuals in the population typically rises and then eventually reaches 100%.

Our models indicated that the wMel infection frequency would likely reach 100% at each of the five field sites, with only a small number of mosquitoes required for release each week. Interestingly though, the requirements for a successful release differed between the sites in terms of the numbers of mosquitoes required, and the expected period of time required for Wolbachia to become fixed. Unsurprisingly, these differences proved to be strongly related to the predicted mosquito abundance at the sites. Under the initial modelled conditions of a low cost of infection and slightly decreased maternal transmission rate, we determined that a weekly release of 200–300 females was typically sufficient for a successful invasion for all sites with the exception of Vila Valqueire, which had the greatest abundance and mosquito carrying capacity of all the sites. This represented a release cohort far smaller than what was released in Australia [11]. For Amorim, Tubiacanga and Urca, which all had smaller wild mosquito populations, the models predicted that the infection would spread to all mosquitoes in the area within the first 50 days after the initial release occurred. These results were expected, given the high importance of mosquito abundance on the ability of Wolbachia to spread quickly [12]. Even so, a period of less than 100 days was predicted for the two sites with higher abundance. The models indicated that a lower maternal transmission rate was a large determining factor in the dynamics of invasion, as an increase in the modelled rate from 96% to 99%, led to a decrease in the number of mosquitoes and time required for the Wolbachia infection to spread to 100%.

The success of Wolbachia-infected mosquitoes in all of these locations is strongly linked to fitness costs resulting from the infection [11,38]. While the cost associated with wMel infection is fairly low [11] we also considered a scenario where a stronger than expected life shortening effect occurred amongst released mosquitoes. This resulted in less effective invasion. At the three sites with lower abundance, the number of mosquitoes required for weekly release tripled, while at the sites with higher abundance, five times more mosquitoes were required. These numbers may appear high, but they are still significantly lower than the what was released each week at field sites in Australia [11]. Our models also indicated that the detrimental effect of this fitness cost on invasive potential was far greater than the benefit obtained through a higher maternal transmission rate, as a combination of these two parameters led to invasion patterns that were largely similar to those seen with the lower maternal transmission rate. This scenario could be considered a worst-case scenario for invasion with wMel, and is also rather unlikely to eventuate given that such extreme fitness effects have not been observed in wMel-infected mosquitoes after more than two years in the field [38]. What is encouraging is that, although the invasion was less effective under these conditions, our data suggest that it is still possible for Wolbachia to spread and completely infect the mosquito population at any of the five sites.

The most important factor influencing the spread of Wolbachia at any of the field sites was local mosquito abundance, and our data reinforce the need to release Wolbachia-infected mosquitoes in higher numbers at sites where there are larger populations of wildtype mosquitoes [28,29,37]. This factor is clearly more of an influence than any of the urban landscape conditions that we analysed. However there is little doubt that there are socio-demographic factors that are important to determining abundance. For instance a larger human population and greater human density would likely provide a higher chance of locating a blood meal host, and potentially an increase in the number of favourable breeding site containers. Better sanitation conditions at the field site would be expected to have the opposite effect, decreasing breeding site availability. We observed the greatest abundance of Ae. aegypti at Vila Valqueire and Jurujuba. These sites had by far the smallest number of houses and total human population of the five sites. They differed from each other in terms of sanitation (no garbage collection in Jurujuba), and human density (high in Jurujuba, but low in Vila Valqueire). It is difficult to establish relationships between these variables and abundance, and part of the reason may be that they are only indirect or partial determinants. A similar study in Australia, focusing more specifically on micro-environmental conditions at the residential block level, indicated that the type of house, level of shade on the property, and availability of breeding sites all affected the proportion of mosquitoes in the area that were infected with Wolbachia [39]. Our study was not designed to assess differences between sites at this level, however there is likely a strong link between these parameters, and those from our models.

There were some factors that were not considered in our models that could potentially contribute to differences in wildtype mosquito abundance between the sites. The first of these is that seasonal differences were not taken into account. Estimations of mosquito population size at each site were an average taken across all collection points, and normalized the effects of time and spatial homogeneity across the area. In addition, the MRR data, which was utilised to calculate abundance, were obtained from experiments conducted over a comparatively shorter period of time, and could have been influenced by seasonal effects. As the MRR experiments at Jurujuba and Vila Valqueire were conducted separately to those for the other sites, this could also have contributed to the higher abundance and lower recapture rates observed there. The data obtained from Amorim were also interesting. There we observed the highest PDS of all the sites, yet the carrying capacity was the lowest, and the site was the easiest of the five for Wolbachia to invade. It is possible that given the shorter mosquito collection period at this site, the carrying capacity was underestimated, or perhaps the mosquito population is just naturally low at that location in spite of the high human density. Given the complexity and heterogeneity of the landscapes encompassing our field sites it is entirely possible that there are other unmeasured or potentially cryptic factors that could contribute to local mosquito abundance or to the successful spread and invasion of Wolbachia. It is also possible that, given the developing nature of the field sites, and other future release areas in different countries there may be changes in the landscape over time that may have a similar effect. To that end, it would be useful to consider future modelling at these or other sites in the context of high resolution, real-world landscape data, as such models would be both more rigorous and more sensitive [40,41]. At the time of this study such data were not available, and while they would have no doubt proven to be valuable, they were too costly to obtain.

General conclusions

Our results suggest that the wMel strain of Wolbachia will likely be able to successfully invade five distinct field sites around the city of Rio de Janeiro, Brazil. We observed both strong CI, and no decrease to fecundity associated with infection, both of which will promote invasion. Our survival data indicated that there were differences in the daily probability of survival for mosquitoes at the different sites, with the better-developed sites being less hospitable. We also observed that wMel was able to spread more effectively when the local mosquito population was smaller. This suggests that Wolbachia could potentially struggle to spread at better-developed sites that have more abundant mosquito populations, as we observed with Vila Valqueire. Critically, although we saw a difference in the invasive potential of wMel between the field sites, the maximum numbers of mosquitoes required for weekly releases was only 2000, and under most conditions, the number was far lower. These data support the hypothesis that releases would be more effective during the dry season, when the mosquito population is naturally smaller [11].

We show that field sites in Rio de Janeiro are complex landscapes that support mosquito populations of greatly different sizes, and this can affect the dynamics of Wolbachia. However, inter-site differences in key aspects of these landscapes do not serve as an impediment to the overall success of invasion by wMel. In spite of this it can be beneficial to adequately characterise field sites prior to the release of Wolbachia-infected mosquitoes as this will help determine the optimal release strategy. Our results provide a useful framework for the ongoing field-releases of wMel-infected Ae. aegypti in Brazil, and will likely hold similar utility for future releases of these mosquitoes in other similarly heterogeneous urban landscapes.

Supporting Information

S1 Table. Raw capture data from the Mark-Release-Recapture experiments in Jurujuba and Vila Valqueire.

Marked mosquitoes were released into the field sites and then re-captured over the course of 9 days using both backpack aspirators and Sticky traps. Marked mosquito numbers represent both Cohorts A and B (see methods section for details). Unmarked mosquitoes were not released as part of the experiments.


S1 Fig. Modeling the invasive potential of wMel with a 99% vertical transmission rate.

Under this scenario the maternal transmission rate of wMel increased to 99%, while the life shortening effect remained at 10%. All modeled releases sizes were as per Fig 3 (red; 200 females, blue; 300, black; 400). In this case a release cohort size of 300 led to a Wolbachia infection frequency of 100% at all sites. In general, invasion occurred more quickly at all sites after the initial release than with a maternal transmission rate of 96%.


S2 Fig. Modeling the invasive potential of wMel with a 99% vertical transmission rate, and increased life shortening.

The models here depict the invasive potential of wMel with the maternal transmission rate increased to 99% and the life shortening effect increased to 33%. Mosquito release numbers were the same as those depicted in the models from Fig 4. The models here indicate that the adverse effects of increased life shortening, which necessitate the release of additional mosquitoes each week, are not ameliorated by the increase in vertical transmission. As such the dynamics of wMel invasion were not different from the data presented in Fig 4.



We are grateful to all members of the Laboratório de Malária (LAMAL—CPqRR/FIOCRUZ), Laboratório de Transmissores de Hematozoários (LATHEMA—FIOCRUZ), members of Programa de Computação Científica (PROCC—FIOCRUZ), to the Brazilian and Australian coordinators of the Eliminate Dengue program. The Australian wMel-infected Aedes aegypti line and TaqMan probes were kindly provided by Prof Scott O’Neill. We are also indebted to the Municipal Health Secretariats of Rio de Janeiro and Niterói for providing logistics support and health agents.

Author Contributions

Conceived and designed the experiments: HLCD LAM LMBdS RMdF. Performed the experiments: HLCD JBLS LMBdS DAMV. Analyzed the data: EPC HLCD DAMV. Wrote the paper: HLCD EPC DAMV RMdF LAM.


  1. 1. Bhatt S, Gething PW, Brady OJ, Messina JP, Farlow AW, et al. (2013) The global distribution and burden of dengue. Nature 496: 504–507. pmid:23563266
  2. 2. Shepard DS, Coudeville L, Halasa YA, Zambrano B, Dayan GH (2011) Economic impact of dengue illness in the Americas. Am J Trop Med Hyg 84: 200–207. pmid:21292885
  3. 3. Maciel-de-Freitas R, Avendanho FC, Santos R, Sylvestre G, Araújo SC, et al. (2014) Undesirable consequences of insecticide resistance following Aedes aegypti control activities due to a dengue outbreak. PLoS One 9: e92424. pmid:24676277
  4. 4. Zug R, Hammerstein P (2012) Still a host of hosts for Wolbachia: analysis of recent data suggests that 40% of terrestrial arthropod species are infected. PLoS One 7: e38544. pmid:22685581
  5. 5. Sinkins SP (2004) Wolbachia and cytoplasmic incompatibility in mosquitoes. Insect Biochem Mol Biol 34: 723–729. pmid:15242714
  6. 6. McMeniman CJ, Lane R V, Cass BN, Fong AWC, Sidhu M, et al. (2009) Stable introduction of a life-shortening Wolbachia infection into the mosquito Aedes aegypti. Science 323: 141–144. pmid:19119237
  7. 7. Walker T, Johnson PH, Moreira LA, Iturbe-Ormaetxe I, Frentiu FD, et al. (2011) The wMel Wolbachia strain blocks dengue and invades caged Aedes aegypti populations. Nature 476: 450–453. pmid:21866159
  8. 8. Xi Z, Khoo CCH, Dobson SL (2005) Wolbachia establishment and invasion in an Aedes aegypti laboratory population. Science 310: 326–328. pmid:16224027
  9. 9. Moreira LA, Iturbe-Ormaetxe I, Jeffery JA, Lu G, Pyke AT, et al. (2009) A Wolbachia symbiont in Aedes aegypti limits infection with dengue, Chikungunya, and Plasmodium. Cell 139: 1268–1278. pmid:20064373
  10. 10. McGraw EA, O’Neill SL (2013) Beyond insecticides: new thinking on an ancient problem. Nat Rev Microbiol 11: 181–193. pmid:23411863
  11. 11. Hoffmann AA, Montgomery BL, Popovici J, Iturbe-Ormaetxe I, Johnson PH, et al. (2011) Successful establishment of Wolbachia in Aedes populations to suppress dengue transmission. Nature 476: 454–457. pmid:21866160
  12. 12. Barton NH, Turelli M (2011) Spatial waves of advance with bistable dynamics: cytoplasmic and genetic analogues of Allee effects. Am Nat 178: E48–E75. pmid:21828986
  13. 13. Yeap HL, Mee P, Walker T, Weeks AR, O’Neill SL, et al. (2011) Dynamics of the “popcorn” Wolbachia infection in outbred Aedes aegypti informs prospects for mosquito vector control. Genetics 187: 583–595. pmid:21135075
  14. 14. Baton LA, Pacidônio EC, Gonçalves DDS, Moreira LA (2013) wFlu: characterization and evaluation of a native Wolbachia from the mosquito Aedes fluviatilis as a potential vector control agent. PLoS One 8: e59619. pmid:23555728
  15. 15. Braig HR, Zhou W, Dobson SL, O’Neill SL (1998) Cloning and characterization of a gene encoding the major surface protein of the bacterial endosymbiont Wolbachia pipientis. J Bacteriol 180: 2373–2378. pmid:9573188
  16. 16. Fu Y, Gavotte L, Mercer DR, Dobson SL (2010) Artificial triple Wolbachia infection in Aedes albopictus yields a new pattern of unidirectional cytoplasmic incompatibility. Appl Environ Microbiol 76: 5887–5891. pmid:20601501
  17. 17. Valencia MD, Miller LH, Mazur P (1996) Permeability of intact and dechorionated eggs of the Anopheles mosquito to water vapor and liquid water: a comparison with Drosophila. Cryobiology 33: 142–148. pmid:8812093
  18. 18. Mollahosseini A, Rossignol M, Pennetier C, Cohuet A, Anjos A Dos, et al. (2012) A user-friendly software to easily count Anopheles egg batches. Parasit Vectors 5: 122. pmid:22713553
  19. 19. David MR, Lourenço-de-Oliveira R, Freitas RM de (2009) Container productivity, daily survival rates and dispersal of Aedes aegypti mosquitoes in a high income dengue epidemic neighbourhood of Rio de Janeiro: presumed influence of differential urban structure on mosquito biology. Mem Inst Oswaldo Cruz 104: 927–932. pmid:19876569
  20. 20. Maciel-de-freitas R, Neto RB, Gonçalves JM, Codeço CT, Lourenço-de-oliveira R (2006) Movement of Dengue Vectors Between the Human Modified Environment and an Urban Forest in Rio de Janeiro. J Med Entomol 43: 1112–1120. pmid:17162942
  21. 21. Consoli RAGB, Oliveira RL de (1994) Principais mosquitos de importância sanitária no Brasil: 228.
  22. 22. Maciel-de-Freitas R, Codeço CT, Lourenço-de-Oliveira R (2007) Daily survival rates and dispersal of Aedes aegypti females in Rio de Janeiro, Brazil. Am J Trop Med Hyg 76: 659–665. pmid:17426166
  23. 23. Gillies MT (1961) Studies on the dispersion and survival of Anopheles gambiae Giles in East Africa, by means of marking and release experiments. Bull Entomol Res 52: 99–127.
  24. 24. Buonaccorsi JP, Harrington LC, Edman JD (2003) Estimation and comparison of mosquito survival rates with release-recapture-removal data. J Med Entomol 40: 6–17. pmid:12597647
  25. 25. Salazar MI, Richardson JH, Sánchez-Vargas I, Olson KE, Beaty BJ (2007) Dengue virus type 2: replication and tropisms in orally infected Aedes aegypti mosquitoes. BMC Microbiol 7: 9. pmid:17263893
  26. 26. R Core Team (2014) R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria.
  27. 27. Royle JA, Chandler RB, Sollmann R, Gardner B (2013) Spatial Capture-Recapture. Waltham MA, USA: Academic Press.
  28. 28. Keeling MJ, Jiggins FM, Read JM (2003) The invasion and coexistence of competing Wolbachia strains. Heredity (Edinb) 91: 382–388. pmid:14512953
  29. 29. Turelli M (2010) Cytoplasmic incompatibility in populations with overlapping generations. Evolution 64: 232–241. pmid:19686264
  30. 30. Min KT, Benzer S (1997) Wolbachia, normally a symbiont of Drosophila, can be virulent, causing degeneration and early death. Proc Natl Acad Sci U S A 94: 10792–10796. pmid:9380712
  31. 31. Rasgon JL, Scott TW (2003) Wolbachia and cytoplasmic incompatibility in the California Culex pipiens mosquito species complex: parameter estimates and infection dynamics in natural populations. Genetics 165: 2029–2038. pmid:14704183
  32. 32. Stevens L (1989) Environmental factors affecting reproductive incompatibility in flour beetles, genus Tribolium. J Invertebr Pathol 53: 78–84. pmid:2915149
  33. 33. Stouthamer R, Breeuwer J, Hurst G (1999) Wolbachia pipientis: microbial manipulator of arthropod reproduction. Microbiology: 71–102.
  34. 34. Kambris Z, Cook P, Phuc H, Sinkins S (2009) Immune activation by life-shortening Wolbachia and reduced filarial competence in mosquitoes. Science (80-) 134.
  35. 35. Rancès E, Ye YH, Woolfit M, McGraw EA, O’Neill SL (2012) The relative importance of innate immune priming in Wolbachia-mediated dengue interference. PLoS Pathog 8: e1002548. pmid:22383881
  36. 36. Zug R, Hammerstein P (2014) Bad guys turned nice? A critical assessment of Wolbachia mutualisms in arthropod hosts. Biol Rev Camb Philos Soc 49.
  37. 37. Hancock PA, Godfray HCJ (2012) Modelling the spread of Wolbachia in spatially heterogeneous environments. J R Soc Interface 9: 3045–3054. pmid:22675165
  38. 38. Hoffmann AA, Iturbe-Ormaetxe I, Callahan AG, Phillips BL, Billington K, et al. (2014) Stability of the wMel Wolbachia Infection following invasion into Aedes aegypti populations. PLoS Negl Trop Dis 8: e3115. pmid:25211492
  39. 39. Hoffmann AA, Goundar AA, Long SA, Johnson PH, Ritchie SA (2014) Invasion of Wolbachia at the residential block level is associated with local abundance of Stegomyia aegypti, yellow fever mosquito, populations and property attributes. Med Vet Entomol 28 Suppl 1: 90–97. pmid:25171611
  40. 40. Keeling MJ, Woolhouse ME, Shaw DJ, Matthews L, Chase-Topping M, et al. (2001) Dynamics of the 2001 UK foot and mouth epidemic: stochastic dispersal in a heterogeneous landscape. Science 294: 813–817. pmid:11679661
  41. 41. Meentemeyer RK, Cunniffe NJ, Cook AR, Filipe JAN, Hunter RD, et al. (2011) Epidemiological modeling of invasion in heterogeneous landscapes: spread of sudden oak death in California (1990–2030). Ecosphere 2: art17.