Thermonavigation of Convective Plumes in the Mosquito Aedes aegypti

Abstract

The yellow fever mosquito, Aedes aegypti, is a vector for numerous diseases, including Dengue fever, Chikungunya, and yellow fever. Efforts to minimize transmission of this disease focus on minimizing vector-host interactions, with a centralized effort to elucidate strategies employed to effectively locate hosts via windborne cues. Efforts to understand how host-emitted plumes are navigated focus on olfaction. While olfaction allows for easily controlled stimuli and robust behavioral response, it is not the only sensory modality employed during host localization. Indeed, deterrents that inhibit olfactory detection do not fully inhibit host localization. Mosquitoes are well endowed with many sensory receptors for several sensory modalities, including thermosensation. I present a series of experiments that examine and describe plume navigation in response to convective thermal plumes. Female mosquitoes were flown in a wind tunnel (1 m long x 0.26 m wide x 0.26 m tall) with a constant wind speed of 0.5 ± 0.05 m s-1, and either no thermal signal, or a lateralized (left or right) thermal plume present. The heating elements of these plumes were modified such that radiative signal was minimized, resulting in stimulus comprised of only convective thermal signal at distances where behaviors were measured. These thermal plumes were turbulent in structure, and produced temperature differences of up to 6 °C. We defined the location and boundaries of these plumes using a mathematical model based on temperature differences, flow speed, and rates at which the plume mixed in external air. I show, for the first time, that mosquitoes navigate toward thermal sources using only convective signals. The plume navigation behavior is described by a bias to the crosswind (lateral) position of the plume source (p < 0.01), and is achieved by augmenting the ratio of left and right turns during flight. A trend of slowly shifting, or sometimes oscillating, the average heading of the trajectory was also found, and may suggest another thermonavigation strategy. Large scale effects such as these were seen, however no change in distributions or features of kinematics (e.g. velocity, acceleration) were seen. To determine whether such slight changes or biases in kinematics can produce these large scale behaviors by creating an agent based model of mosquito flight. Features of mosquito navigation were approximated by modeling flight as a biased random walk in a driven damped mass system. The model was trained on control (plume absent) behavior, and tested with the presence of the plume, which was represented as a binary signal at the position of the empirical plumes. We defined a stimulus bias as a slight bias toward the last crosswind position a plume was detected, and faster upwind velocities while within the plume bounds. The agent model did not produce the control crosswind distributions, but did produce crosswind position distributions similar to those observed in mosquito behavior, and had similar kinematics in and outside of the plume boundaries. These experiments demonstrate that mosquitoes, when presented a turbulent, convective thermal plume, navigate in a biased random walk, and augment the bias of the walk toward the crosswind plume location.