Aedes aegypti is an important vector of several pathogenic arboviruses including dengue, chikungunya and Zika. Innovative approaches to control Aedes populations, involving synthetic transgenic modifications as well as Wolbachia bacteria, appear promising. For the various techniques requiring offspring inheritance of a trait, released males must successfully compete for mating partners against wildtype males. However, very little is known about mechanisms of mate selection in mosquitoes in general and in particular about potential correlations between mating success and offspring immune performance. Harmonic convergence signals have been proposed as a cue for females to predict male quality. We investigated whether offspring of converging parental pairs showed differences in immune competence compared to offspring derived from non-converging parental pairs using three different types of immune assays. We found that offspring immune responses (melanization response and response to a bacterial challenge) differed between offspring from converging and non-converging parents. However, immune responses were shaped by several interacting factors such as sex, age, reproductive status, and parental mating behavior. Parental mating behavior had a stronger effect on the immune response of male offspring than on female offspring. Further, a population of female offspring derived from converging parental pairs reached their peak dengue virus dissemination rate earlier compared to a population of offspring derived from non-converging parental pairs. Our results provide insight into a wide range of selective pressures shaping mosquito immune function. Evolutionary trade-offs between naturally and sexually selected traits can have important implications for disease transmission and control and should be considered in the development of reproductive control strategies.

Mosquito rearing

Eggs (F2-F4, originating from Kamphaeng Phet Province, Thailand  (16°27′ 48”N, 99°31′ 47”E) were hatched under a reduced pressure vacuum desiccator for 30min in ddH₂O and provided with 0.3 mg of crushed Cichlid pellets (Hikari Cichlid Gold Baby Pellets). After 24h, larvae were dispensed into larval trays at a density of 200 larvae per tray, with each tray containing 1L of distilled water and four Cichlid pellets (Hikari Cichlid Gold Large Pellets). After 7-8d, pupae were collected, and adults allowed to eclose into large mating cages. We maintained adults on 10% sucrose solution ad libitum. To generate eggs, sugar water was removed for 48h and water removed for 24h. Females were then offered a whole human blood meal (Interstate Blood Bank, Memphis, TN, USA) in an artificial glass feeder with a bovine intestine membrane. Eggs were collected on paper towels in a damp oviposition site 3-4d after the blood meal. Eggs were allowed to embryonate for a minimum of 1 week and stored dry on paper in humidified chambers. Mosquito colonies were maintained at a constant 27°C ± 0.5 ^oC and 85% + 5% relative humidity in environmental insect chambers (Percival Scientific) set to a 12 h light : 12 h dark photoregime.

Experimental mosquitoes were reared as above except for two differences: application of moderate stress and lower larval density. Previously, we found that in order to detect differences between offspring of converging and non-converging males, we needed to apply a moderate level of stress (Cator and Harrington 2011). Fourth instar larvae were exposed to 20°C + 0.5^oC for 24h, as moderate stress condition (Cator and Harrington, 2011). Secondly, larval rearing density was lower (usually between 80-150 larvae per tray) as eggs from each female were hatched separately and the total amount of food was adapted accordingly. Pupae were separated according to sex and allowed to eclose in sex specific cages.

Harmonic convergence assays

We recorded acoustic interactions between opposite sex pairs of not mated 3-5 day old Ae. aegypti. Females were anesthetized on wet ice and tethered to a 2cm strand of human hair using Nailene glue (Pacific World Corp., San Diego, CA, USA). The hair was attached to a piece of metal wire as described in Cator et al. (2009). One female per trial was positioned in the center of a 15 × 15 × 15cm Plexiglas mating arena, 2cm above the sensitive face of a particle velocity microphone (NR-21358; Knowles, Itasca, IL). The arena was placed on a heat plate set to 30 °C, and a tray with moist cotton wool was provided to increase humidity in the arena. Female flight was initiated via removal of tarsal inhibition and gentle puffs of air. One male per trial was released into the mating arena and allowed to mate with the female.

Acoustic recordings were taken from time of male release to the end of any type of first physical interaction between male and female or the termination of a successful copulation. Trials in which males took longer than 5 min to initiate an interaction with the female were discarded. The acoustic interactions of mosquito flight tone were recorded and analyzed using Raven 1.0 software (Cornell Laboratory of Ornithology, Ithaca, NY). Harmonic convergence was defined as when the harmonic frequencies of both the male and female during an encounter matched. Frequencies were considered to be matching if they were within less than 4.95Hz of each other and lasted in this state for a minimum of 1s (30, 41).

Regardless of whether or not harmonic convergence occurred within a mating interaction, male and female pairs were maintained together in 500ml paper cups (supplied with 10% sucrose and kept at 27°C ± 0.5 ^oC and 85% + 5% relative humidity) to achieve a high number of successfully mated pairs. The male was then removed after 3 days. Females were blood fed as described above, and eggs were collected from each female from up to three separate gonotrophic cycles and stored as described above.

Immune Assays

To stimulate humoral melanization and bacterial killing, immune challenges were administered to mosquitoes anesthetized on wet ice via an intrathoracic injection into the anepisternal cleft (110) using a microcapillary glass needle (P-97, Sutter Instruments, Novato, CA, USA) attached to a mouth pipette (humoral melanization) or a Nanoject injector (Drummond Scientific, Broomall, PA, USA), for bacterial growth, (70).

Humoral melanization assay

To stimulate the melanization response, mosquitoes in each experimental group were injected with one neutrally charged G-25 Sephadex bead (Sigma, St. Louis, MO, USA) (111). As Sephadex beads range in size from 40 to 120μm diameter, we visually selected the smallest beads for inoculation. Beads were injected using a minimal amount of Schneider’s Drosophila medium (less than 0.5 μl). After the injection, mosquitoes were housed individually in 50ml tubes and maintained as described above. After 24h, mosquitoes that were able to walk were anesthetized on ice and beads were dissected out in a phosphate-buffered saline (PBS) solution. Recovered beads were scored according to their degree of melanization using three categorical classes: unmelanized, partially melanized, or fully melanized (70, 112-114).

Bacterial growth and mosquito mortality

Tetracycline-resistant GFP-expressing Escherichia coli (E. coli), dh5 alpha strain, were grown overnight in Luria-Bertani's (LB) rich nutrient medium in a shaking incubator at 37°C. We inoculated ice anesthetized mosquitoes in each experimental group with 200,000 live E. coli bacteria suspended in Schneider’s Drosophila medium (Gibco) at a final volume of 69nl using intra-thoracic injections.

After injection, mosquitoes were housed individually in 50 mL Falcon tubes and maintained as described above. After 24h, the number of dead mosquitoes was recorded to estimate the effects of experimental treatment on bacterial virulence. Mosquitoes that survived the 24h post-challenge, were homogenized in 1ml of LB medium and the suspension transferred into culture tubes containing 1ml of LB medium. The bacteria solution was incubated for 5h in a shaking incubator (shaking speed 180rpm) at 37°C. Afterwards, 200μl of each sample were pipetted into black 96-well plates in duplicate, and GFP fluorescent signal was measured using a multimode microplate reader (Varioskan, Thermo Scientific, Waltham, MA, USA). Fluorescence values for the initial 200,000 E. coli bacteria injection dose were obtained and used as a baseline to normalize experimental values. Methods associated with validating this protocol are provided in the supplementary information (SI Methods).

Dengue infections

Experimental design

Eggs from parental pairs that had either harmonically converged or did not converge were collected as described above. Eggs from four recording events (from a total of 55 parental pairs) were separated according to parental convergence status (not converged vs. converged) and divided into three replicates. Eggs from each treatment and replicate were hatched separately, as described above. For each experimental treatment and replicate, pupae were transferred into mixed sex cages to allow mating, and adults were maintained as described above. Four to five days after emergence, males were removed, and sucrose was replaced with dH₂O. The next day, 5-6d old females were provided a DENV infectious blood meal.

DENV in vitro culturing and mosquito infections

We propagated DENV-2 virus stocks (strain PRS225488), originally isolated from human serum in Thailand in 1974 and acquired from the World Reference Center for Emerging Viruses and Arboviruses at the University of Texas Medical Branch, using previously described methods (79). The blood meal consisted of 50% (vol/vol) human red blood cells (washed three times with RPMI and cleared from white blood cells and residual plasma), 33% (vol/vol) Dulbecco’s Modification of Eagle’s Medium (DMEM, Corning) containing DENV at a final concentration of 2.8 x 10⁵ (replicate 1), 1.5 x 10⁵ (replicate 2), and 2.3 x 10⁵ (replicate 3) PFU/ml (titrated from a sample of the prepared blood meal and determined using the Spearman-Karber TCID₅₀ method (115)), 20% (vol/vol) FBS, 1% (wt/vol) sucrose and 5mM ATP. The blood meal was provided through a glass feeder as described above.  Afterwards, blood engorged females were randomly distributed into 500ml paper cups with 23-28 females each. Adults were maintained in cups as described above.

Determination of infection and dissemination

Dengue infection in mosquitoes can be assessed through three key stages: infection, dissemination, and infectiousness. Per parental convergence group and replicate, 20 mosquitoes were processed on days 3, 6, 9, 12, 15 and 18 post infection (n = 120 females per time point, n = 760 females in total). To do this, females were immobilized on ice and then transferred to a chill table (Bioquip, Rancho Dominguez, CA). Legs and wings were removed, and the proboscis inserted into a 200µl pipette tip with the end cut-off and containing 30µl of salvation mix (1900 µl FBS, 20 µl 300mM ATP, 80µl red food dye). Legs were kept for analysis of dissemination and transferred into 2ml tubes containing DMEM with 1x antibiotic/antimycotic. Females were allowed to salivate for 40min on a heat plate. Successful salivation was confirmed by the presence / absence of red food dye in female abdomens. Afterwards, the saliva was transferred into Eppendorf tubes containing 700µl of DMEM with 1x antibiotic/antimycotic to test for infectiousness. Heads of females were then cut off and added to the tubes containing the leg samples to test for dissemination. The remaining body was transferred into Eppendorf tubes containing 700µl DMEM with 1x antibiotic/antimycotic to test for infection.

DENV assays

Using plaque assays as described previously (79, 116), we tested for the presence of viable DENV particles in the body, head and legs, and saliva as a proxy for mosquitoes that are infected, have disseminated infection, or are infectious, respectively (117). All body samples taken were analyzed for infection. Those animals that tested positive for infection, subsequently had head and leg samples tested for DENV dissemination. Finally, those that successfully disseminated virus had saliva tested for infectiousness. Body, and head and leg, samples were homogenized using a QIAGEN TissueLyzer at 30 cycles/s for 30s and centrifuged at 17000g for 5min at 4^oC. After centrifugation, the supernatant was used to inoculate Vero cells. After an initial infection period of 2h, media was removed and an overlay 1.5% UltraPure low melting point agarose (Invitrogen)/DMEM with 1x antibiotic/antimycotic was added. Samples were subsequently incubated at 37 ֯C, 5% CO₂ for six days. After incubation cells were fixed with 10% formalin and stained with 1% crystal violet.

Cypopathic effect assays (CPE) were performed like described above for plaque assays, however, cells were scored for whole well cytopathic effect instead of individual plaques. Molecular detection of dengue RNA was performed with the DENV-specific primers D1 (5'-TCAATATGCTGAAACGCGCGAGAAACCG-3') and D2 (5'-TTGCACCAACAGTCAATGTCTTCAGGTTC-3') producing a 511bp fragment (118). Total RNA was extracted using the PureLink RNA extraction kit as per manufacturer’s instructions (Thermo Fisher Scientific, Waltham, MA, USA). Reverse transcription and PCR reactions were performed in a single step procedure with iTaq Universal Probes One-Step Kit as recommended by the manufacturer (Bio-Rad, Hercules, CA). Reaction conditions were first optimized using cell culture supernatants with a known starting PFU of 10⁷/ml. Supernatants were serially diluted 10-fold with the final dilution representing 1 PFU/ml before RNA extraction and single step RT-PCR. PCR product was assessed with 1% agarose gel electrophoresis.

Statistical analysis

Statistical analyses were performed within the RStudio integrated development environment for R (119, 120). To assess the effects of convergence status (non-converged vs. converged), sex (male vs. female), age (1 vs. 3 vs. 5 days old), and reproductive stage (unmated vs. mated or non bloodfed vs. bloodfed) on mosquito immune performance, we ran three model analyses with the following fixed effects. Model 1 investigated the effects of i) convergence status, ii) sex, and iii) age (1 and 3 day old unmated individuals: melanization assay; or 1, 3, and 5 day old unmated individuals: bacterial survival and mosquito mortality) with the potential for both two and three-way interactions. Model 2 explored the effects of i) convergence status, ii) sex, and iii) mating status in three-day-old individuals only. Finally, Model 3 explored the effects of i) convergence status and ii) bloodfed status and the potential for two-way interactions in five-day old females only. All effects were specified as categorical in nature with the exception of age, which was included as a continuous variable, but after centering and scaling with the means and standard deviations respectively (121). Intercepts were allowed to vary randomly among parental pairs. Results of the melanization experiment were analyzed by ordered logistic regression with mixed effects using the package “ordinal” (122). The dependent variable was specified as non-, partially- or fully-melanized beads in an increasing order. For analyzing the in-vivo bacterial growth assay, statistical models were specified with fluorescence intensity (after subtracting background fluorescence and averaging across duplicates) as the dependent variable. Linear mixed models were initially tested after transformation with [log (y +1)] or without transformation of fluorescent intensities, but neither approach was able to account for the heteroscedasticity. However, rounding fluorescence intensities to the nearest integer allowed modeling the variable as a discrete count instead of using generalized linear mixed models (GLMMs), as described for another dataset (121). All GLMMs were performed in the “glmmTMB” package (121) and specified with a Generalized Poisson distribution and “log” link. Intercepts were allowed to vary randomly among parental pairs across all model analyses. Finally, to analyze mosquito mortality following bacterial challenge, linear mixed models were specified as suggested in the package “lme4” (123). The number of dead mosquitoes at each time point was specified as the dependent variable with the total number of mosquitoes in each cup included as an offset to account for differences between cups.

For analyzing DENV vector competence, binomial GLMMs with a logit link were performed, also in the “glmmTMB” package. The dependent variable was expressed as the proportion of mosquitoes positive for DENV-2 from the total number sampled. Fixed effects consisted of 1) convergence status (non-converged vs. converged), 2) tissue (body, head and legs, or saliva), and 3) time course of infection (days post-infection) modeled up to four-way interactions. While treatment and tissue were specified as categorical predictors, time post-infection was centered and scaled as described above. Further, to estimate how the proportion of DENV positive mosquitoes changed over the post-infection period, time was specified as both linear and quadratic (“hump-shaped”) effects on infection/infectiousness or dissemination. Since multiple tissues were sampled for DENV from the same individual, random intercepts for mosquito number was also nested within each biological replicate. To further investigate whether the overall temporal trends in the proportion of mosquitoes that became infected, disseminated infection, or became infectious over time differed between converged and non-converged groups we ran an additional, separate model analysis for offspring from converged and non-converged parents. These models used the same fixed and random effect structure as described above with the exception that time (days post-infection) was specified as a categorical/group-level predictor.

For all model analyses, model selection and choice of family were based on likelihood-based information criteria assessed using the package “bbmle” (124). Reference groups for all analyses were chosen as follows: non-converged for convergence status, males for sex, not mated for mating status, blood fed for blood feeding status. For all the models, residuals were examined for normality with the “DHARMa” package (125) using a predetermined threshold ratio of squared Pearson residuals over the residual degrees of freedom <1.5 and a Chi-squared distribution of the squared Pearson residuals with p > 0.05. Once overdispersion was accounted for, the marginal means estimated by the model were used to perform pairwise comparisons using Tukey’s contrasts in the “emmeans” package (126). Statistical analysis for mosquito survival after DENV infections was performed using a Kaplan-Meyer survival analysis (log rank test).

Data sets include:

Dengue infection experiment in Aedes aegypti: "aagdenv.csv"

Bacterial survival experiment in Aedes aegypti: "bacterial.survival.csv"

Melanization experiment in Aedes aegypti: "melanization.csv"

Code for the Statistical Analysis in R:

"Rcode.docx"