A Retinal Circuit Generating a Dynamic Predictive Code for Oriented Features

Jamie Johnston,Sofie-Helene Seibel,Léa Simone Adele Darnet,Sabine Renninger,Michael Orger,Leon Lagnado

doi:10.1016/j.neuron.2019.04.002

Jamie Johnston, Sofie-Helene Seibel + Show 4 more

Open Access

https://doi.org/10.1016/j.neuron.2019.04.002

Copy DOI

Abstract

SummarySensory systems must reduce the transmission of redundant information to function efficiently. One strategy is to continuously adjust the sensitivity of neurons to suppress responses to common features of the input while enhancing responses to new ones. Here we image the excitatory synaptic inputs and outputs of retinal ganglion cells to understand how such dynamic predictive coding is implemented in the analysis of spatial patterns. Synapses of bipolar cells become tuned to orientation through presynaptic inhibition, generating lateral antagonism in the orientation domain. Individual ganglion cells receive excitatory synapses tuned to different orientations, but feedforward inhibition generates a high-pass filter that only transmits the initial activation of these inputs, removing redundancy. These results demonstrate how a dynamic predictive code can be implemented by circuit motifs common to many parts of the brain.

Highlights

A general principle in understanding the design of sensory systems is the need to encode information efficiently which, in turn, requires removal of redundancies in the signal received from the outside world (Barlow, 1961; Fairhall et al, 2001; Srinivasan et al, 1982; Sterling and Laughlin, 2015)
Imaging the Input and Output from Individual Retinal Ganglion Cells The signals retinal ganglion cells (RGCs) deliver to the brain depend on the integration of a variety of synaptic inputs, often with mixed properties
Understanding how the activity of these inputs determines the output of the neuron has been difficult because individual synapses on a dendrite cannot be isolated using electrophysiology (Branco and Ha€usser, 2011; Schmidt-Hieber et al, 2017)

Summary

Introduction

A general principle in understanding the design of sensory systems is the need to encode information efficiently which, in turn, requires removal of redundancies in the signal received from the outside world (Barlow, 1961; Fairhall et al, 2001; Srinivasan et al, 1982; Sterling and Laughlin, 2015). This principle helps us understand why the retina does not operate like a camera conveying a stream of intensity values for each pixel. A number of models have been proposed for the neural implementation of predictive codes (Bastos et al, 2012; Hosoya et al, 2005; Rao and Ballard, 1999), the circuit mechanisms have not been clearly identified

Results

Discussion

Conclusion