Fly VS neurons compared to optimized motion detectors

Abstract

Listen

Translate article

Event Abstract Back to Event Fly VS neurons compared to optimized motion detectors A function of dendrites is synaptic-integration. The specific electrophysiological dynamics involved in this process have been a subject of neurobiological research for many decades. However, some questions remain unresolved, such as what determines the morphology of dendrites: the need to perform signal transformations or to connect to incoming axons in an efficient manner, or both? We employ a recently developed inverse approach to study the dendritic morphology-function relationship [2]. Briefly, this approach consist of (1) defining a computational function to be performed by a model neuron, (2) optimizing a model neuron endowed with a dendritic morphology to perform the targeted function, and (3) analyze the emerged contingencies to quantify the morphology-function relationship. By specifying a desired function we evidently focus on the functional characteristics of dendrites. We investigated the hypothesized function of the VS cell from the fly lobular plate which is spatio-temporal integration of small-field direction sensitive inputs. This integration step is also referred to as wide-field motion detection, i.e., detection of motion of over the complete visual field of the fly. We constructed a model in which the small-field motion detection is performed by Reichardt detectors [1] and their output was projected in a realistic way onto the model neuron that has to perform the spatio-temporal integration. The output signal of the integration cell is a smooth signal indicating the direction of the movement. The morphology and the distribution of potassium and sodium conductances is optimized by using the inverse approach. We ran three optimization runs for three different velocities of movement. All neurons could perform the spatio-temporal integration well. Moreover, we found that some (5 out of 9) from a qualitative point of view (visually) resembled the VS cells. Moreover, we found similar distribution of ion-channels in all model neurons. Thus, the dendritic morphology is optimized for performing a computational-electrophysiological function instead of only collecting inputs. It can be argued that due to the biologically realistic constraints imposed by the projection of the small-field sensitive inputs onto the model neuron, the neuron is actually forced to take a shape that corresponds to the shape of the true cells. However, since some model neurons had an entirely different morphology than observed in nature, but still performed the function well, we can rule this option out. Hence, we can conclude that dendritic morphology is not only optimized for wiring, but also to perform specific function in some neuronal substrate.