A trade-off between dry season survival longevity and wet season high net reproduction can explain the persistence of Anopheles mosquitoes

Abstract

Background: Plasmodium falciparum malaria remains a leading cause of death in tropical regions of the world. Despite efforts to reduce transmission, rebounds associated with the persistence of malaria vectors have remained a major impediment to local elimination. One area that remains poorly understood is how Anopheles populations survive long dry seasons to re-emerge following the onset of the rains.

Methods: We developed a suite of mathematical models to explore the impact of different dry-season mosquito survival strategies on the dynamics of vector populations. We fitted these models to an Anopheles population data set from Mali to estimate the model parameters and evaluate whether incorporating aestivation improved the fit of the model to the observed seasonal dynamics. We used the fitted models to explore the impact of intervention strategies that target aestivating mosquitoes in addition to targeting active mosquitoes and larvae.

Results: Including aestivation in the model significantly improved our ability to reproduce the observed seasonal dynamics of vector populations as judged by the deviance information criterion (DIC). Furthermore, such a model resulted in more biologically plausible active mosquito survival times (for A. coluzzii median wet season survival time of 10.9 days, 95% credible interval (CrI): 10.0-14.5 days in a model with aestivation versus 38.1 days, 95% CrI: 35.8-42.5 days in a model without aestivation; similar patterns were observed for A. arabiensis). Aestivation also generated enhanced persistence of the vector population over a wider range of both survival times and fecundity levels. Adding vector control interventions that target the aestivating mosquito population is shown to have the potential to enhance the impact of existing vector control.

Conclusions: Dry season survival attributes appear to drive vector population persistence and therefore have implications for vector control. Further research is therefore needed to better understand these mechanisms and to evaluate the additional benefit of vector control strategies that specifically target dormant mosquitoes.

Keywords: Aestivation; Anopheles mosquitoes; Mathematical modelling; Persistence; Plasmodium falciparum; Vector ecology.

Fig. 1

Schematic of the model structure. a The modelled life stages. The aquatic stages (eggs, larvae and pupae) are grouped together as immature (I) vectors and the non-aquatic vectors as mature (M) adult mosquitoes. The parameter, p, represents the development of immature vectors to adult vectors. The parameters μ_I and μ_M are the mortality rates for the immature and mature vectors, respectively. In adverse weather, the adult vectors (M) are assumed to adapt and move into the dormant state M_d. The movement (plasticity) of mosquitoes between the active and dormant compartments is modelled by the parameters d (adaptation) and w (reactivation). b The curves used to model vector rainfall-based fitness, adaptation and reactivation between the dry and the wet seasons

Fig. 2

Model fits to the dry and wet season mosquito population data. Panels a-d show results for A. coluzzii and panels e-h for A. arabiensis. The null model H:0/M₀ assumes no aestivation and no reduced fecundity in the dry season. Model H:1/M₁ assumes no aestivation but reduced reproduction fitness in the dry season. Model H:2/M₂ includes aestivation but that only the active vectors are observed. Model H:3/M₂ assumes that both active and aestivating vectors are observed

Fig. 3

Predicted adult mosquito lifespans. H:0 and H:1 predict non-aestivating mosquitoes to be long-lived while the aestivation models (H:2 and H:3) predict only aestivating mosquitoes to be long-lived. The error bars represent the credible intervals and the dots the median of the estimated parameters. Panels a and b show the predicted lifespan for A. coluzzii and A. arabiensis adult mosquitoes

Fig. 4

The trade-off between fecundity and survival longevity drives population persistence. A. coluzzii mosquito populations are shown to go extinct if fecundity and survival times are reduced. Panel a illustrates extinction (in model M₁), survival time is increased while keeping fecundity low (F = 5 eggs). Panel b shows that persistence is achieved be increasing F to 15 eggs. Panels c and d demonstrate that persistence is easily sustained in the aestivation model, M₂, at low active adult mosquito survival times and minimal reproduction. Panels e-g show the trade-off between fecundity and survival longevity in the no-aestivation model. Increasing fecundity (eggs laid), allows relatively short-lived mosquitoes to persist. Panels h-j, demonstrate more robust persistence under similar conditions in the aestivation model. Parameters used are as given in Tables 1 and 2; however, the parameters F (fecundity) and μ_I or μ_d (survival) are varied as shown in the panels

Fig. 5

Evaluation of vector control strategies. Simulations showing the modelled changes in the abundance of active adult vectors when interventions with different mechanisms of action are used to control the mosquito population. Panels a, d and g show the effects of interventions that increase the larval death rate (e.g. larvicides) whilst panels b, e and h show the effects of interventions that kill adult mosquitoes (e.g. IRS, LLINs or other insecticides). Panels c, f and i show the effects of combining these two targets. Panels a-c are predicted from a model with no aestivation; d-f from a model including aestivation in which only the active vectors are targeted and g-i from the aestivation model if dormant vectors are also targeted. Parameter values used are given in Tables 1 and 2. Intervention efficacies of 0, 20 and 80% were used, and these represent percentage increase in mortality of vectors/larvae induced. The dotted line (0%) means no intervention; the red line and the black line represent interventions that increase vector/larvae mortality by 20 and 80%, respectively. See Additional file 1: Text 1 for further details on how interventions were simulated