Mechanisms of local and global motion detection in Drosophila

Abstract

Visual motion is an essential cue for many sighted animals. This can either be caused by the movement of an object, or the relative movement of the entire world caused by self-motion of the animal. Accordingly, the brain must compute both local motion cues, corresponding to spatiotemporal changes in luminance, and global motion patterns composed of many local motion vectors. In the fly eye, which is composed of hexagonally arranged visual units, the first-direction selective cells, the T4 and T5 neurons, are known as local motion detectors. Global motion was thought to be computed downstream, in large wide-field cells that sample information from many local motion detectors. Despite many years of research, the detailed mechanisms underlying local motion tuning in T4 and T5 cells, as well as the transformation of local into global motion information is not fully understood. In this thesis, I first studied the mechanisms of local motion detection and how local motion information is transferred into a global information about self-motion. First, blocking GABAergic signaling in the whole brain leads to a loss of direction-selectivity in T4 and T5 cells, arguing for a significant role of GABAergic circuits for local motion computation (Fisher et al., 2015a). However, GABAergic cell types and their potential interactions with the neuronal circuit responsible for motion detection had not yet been identified. Based on a behavioral genetic screen, in vivo calcium imaging and genetic analyses, we propose a GABAergic feedback mechanism, implemented by the two columnar C2 and C3 cells, to be required for directional tuning of T4 and T5 cells. Both neurons mainly interact with neurons upstream of T4 and T5 cells, indirectly affecting motion processing. While our data suggest a specific role of C2 for suppressing responses into the neuron’s non-preferred direction in T4 cells, C3 silencing affected the temporal properties of T4 and T5. T4 and T5 cells have been classified anatomically into four subtypes, ostensibly responding to the four cardinal directions of visual motion (Fisher et al., 2015a; Maisak et al., 2013). How these four motion axes arise, given the hexagonal arrangement of visual units in the fly eye, was not clear. Furthermore, it was not known how local cardinal motion from T4/T5 inputs can be transformed into complex optic flow fields encoded downstream. To understand how global motion is represented by the population of T4 and T5 cells, I used in vivo two-photon calcium imaging to characterize the direction tuning of T4 and T5 cells across visual space and the extent of the lobula plate. In contrast to the four anatomically subtypes described previously, we found six functional subtypes of local motion detectors at the population level / across the lobula plate. On average, tuning of these six subtypes matches the hexagonal structure of the eye. Tuning of neighboring motion detectors gradually changes, such that all T4/T5 cells of one subtype encode global motion patterns induced by translational and rotational self-movements of the fly. Together, the T4/T5 population represents six types of self-motion encountered during flight. Thus, downstream LPTCs can simply pool information from the local motion detectors, T4 and T5, to compute diverse complex flow fields. This population code for optic flow is reminiscent of coding of retinal ganglion cells in the vertebrate retina where only four directions of self-motion faced during walking are represented (Sabbah et al., 2017). While the number of motion dimensions encoded by the local motion detectors differ, this suggests a general coding strategy of visual systems to extract self-motion of the animal, adapted to the complexity of maneuvers encountered during locomotion. Taken together, the data presented in this thesis provide new insights about local as well as global mechanisms of visual motion processing in the fly and suggest striking parallels but also highlight differences between the vertebrate and invertebrate visual system. This is critical not only for understanding computational principles of visual systems but also for understanding how evolution adapts neuronal coding strategies to the animal’s behavioral constraints.