| Popis: |
Retinal amacrine cells are a diverse population of inhibitory interneurons, posing a challenge to understand the specific roles of those interneurons in computations of the similarly diverse ganglion cell population. Here we study the predictive computation of motion anticipation, which is thought to compensate for processing delays when encoding moving objects. We recorded the membrane potential of the salamander amacrine cell population optically while recording electrically from ganglion cells with a multielectrode array. We find unexpectedly that ganglion cells with the greatest anticipation for moving stimuli exhibit a new type of predictive motion anticipation that is inconsistent with prior models of delayed inhibition. Based on the spatiotemporal correlations between thousands of amacrine and ganglion cell pairs, we modeled the contribution of the traveling wave of activity for different amacrine cell populations to the encoding of a moving bar. These models indicate that the population responses of slow biphasic amacrine cells create the greatest contribution to both types of ganglion cell motion anticipation, supporting a role for this specific amacrine cell class in the predictive encoding of moving stimuli.Significance StatementThe prediction of moving stimuli is a widespread function occurring in both visual and auditory systems. The diversity of interneuron populations makes it a challenge to understand the mechanisms of these computations. To analyze how the retina anticipates motion, we optically measured inhibitory amacrine cell population activity simultaneously with electrical recording from ganglion cell populations. We then used computational modelling to assess which amacrine cell types have the right spatiotemporal responses to generate motion anticipation. In contrast to previous suggestions of a general role for inhibition, we find that slow biphasic amacrine cells specifically have the greatest contribution to motion anticipation, thus highlighting the need to directly measure and model the effects of interneuron populations in complex sensory computations. |
