Modelling visual-olfactory integration in free-flying Drosophila

Abstract

Flying fruit flies (Drosophila melanogaster) locate a concealed appetitive odour source most accurately in environments containing vertical visual contrasts (Frye et al, 2003). To investigate how visuomotor and olfactory responses interact to cause this phenomenon, I implement a tracking system capable of recording flies’ flight trajectories in three dimensions. I examine free-flight behaviour in three different visual environments, with and without food odour present. While odour localisation is facilitated by a random chequerboard pattern compared to a horizontally striped one, a single vertical landmark also facilitates odour localisation, but only if the odour source is situated close to the landmark. I implement a closed-loop systems-level model of visuomotor control consisting of three parallel subsystems which use wide-field optic flow cues to control flight behaviour. These are: an optomotor response to stabilise the model fly’s yaw orientation; a collision avoidance system to initiate rapid turns (saccades) away from looming obstacles; and a speed regulation system. This model reproduces in simulation many of the behaviours I observe in flies, including distinctive visually mediated ‘rebound’ turns following saccades. Using recordings of real odour plumes, I simulate the presence of an odorant in the arena, and investigate ways in which the olfactory input could modulate visuomotor control. In accordance with the principle of Occam’s razor, I identify the simplest mechanism of crossmodal integration that reproduces the observed pattern of visual effects on the odour localisation behaviour of flies. The resulting model uses the change in odour intensity to regulate the sensitivity of collision avoidance, resulting in visually mediated chemokinesis. Additionally, it is necessary to amplify the optomotor response whenever odour is present, increasing the model fly’s tendency to steer towards features of the visual environment. This could be viewed as a change in behavioural context brought about by the possibility of feeding. A novel heterogeneous visual environment is used to validate the model. While its predictions are largely borne out by experimental data, it fails to account for a pronounced odour-dependent attraction to regions of exclusively vertical contrast. I conclude that visual and olfactory responses of Drosophila are not independent, but that relatively simple interaction between these modalities can account for the observed visual dependence of odour source localisation.