Correlated electrophysiological and connectomic analyses of ganglion cell circuits in macaque retina

A comprehensive understanding of a neural circuit requires correlating its structure and function. This study investigates the extent to which the stimulus-driven electrical activity of retinal ganglion cells can be explained by the structure of synaptic connections in the retinal circuitry. Electrophysiological and connectomic analyses on a single piece of retina tissue were combined to evaluate the relationship between the functional response of ganglion cells and the structure of their synaptic connections. Single-cone resolution white noise stimuli were presented to a macaque retina, and the electrical activity of ganglion cells was simultaneously recorded using a 512-electrode array. Stimulus-response mapping revealed individual cone locations within the receptive field of each ganglion cell and the weights of their individual contributions to the ganglion cell response. The functionally characterized retinal sample was imaged under an electron microscope to collect a large volume spanning 750×1000×150 μm³. I then correlated the physiological and anatomical data by registering the receptive fields of ganglion cells with their dendritic arbors and by registering the cone locations within ganglion cell receptive fields with the imaged cone inner segments in the photoreceptor layer. The complete feedforward wiring diagrams of six selected ganglion cells were reconstructed: two OFF midgets, two ON midgets, and two small bistratified cells. This involved detailed reconstructions of synaptically connected bipolar cells and their ribbon synapses with the cones, which in turn allowed the synaptic weights of the cones on the ganglion cells to be estimated. Comparing the functionally estimated cone weights against the synaptically measured weights revealed a strong positive correlation between them. This suggests that the functionally estimated cone connectivity of ganglion cells accurately reflects the structure of their anatomical circuits. To investigate the extent to which the structure of a ganglion cell wiring diagram can predict its stimulus-driven electrical activity, I developed a simple neural network model whose structure and weights were based on the wiring diagram of the ganglion cell. The prediction accuracy of this anatomically constrained model was similar to that of the statistical linear-nonlinear model fitted only on the functional data without anatomical constraints, suggesting that the structure of the circuit is indeed informative of its computation.