Measuring the precision of spot localization in the population code of retinal ganglion cells.

Abstract

Purpose : What is the precision with which the retina transmits information about the spatial location of small objects? This is a fundamental question about the retinal code, yet the answer remains unclear.

Methods : We used a variety of techniques to measure both the functional output and anatomical dendrite configurations of identified retinal ganglion cells (RGCs). The measurements were made in ex vivo preparations of the mouse retina during the presentation of small (20 – 30 µm) spots of light presented at multiple nearby locations. Our data include (1) spikes and excitatory input currents from up to 4 simultaneously recorded, neighboring RGCs of the same type, (2) spike data from full RGC mosaics recorded on multi-electrode-arrays, (3) calcium responses of the nearly full populations of RGCs (>150 cells) encoding a target region of space, and (4) 3D reconstructions of the dendritic arbors of sets of 4 neighboring RGCs of the same type.

Results : Data from these different modalities and recorded in different labs all support the conclusion that populations of RGCs can represent spatial information down to a precision 10 – 20 times finer than that of the receptive-field centers of the individual RGCs. Different homogeneous and heterogenous populations are compared. The anatomical data suggests that dendrites are coordinated within at least some homogeneous RGC populations to decorrelate responses to spots and nearby locations and thus increase spatial precision in the output.

Conclusions : Simultaneous recordings of populations of identified RGCs using different modalities offer powerful datasets in which to study the output code of the retina. These datasets also require new mathematical tools for their analysis. Precision in object localization within populations of RGCs can be more than an order of magnitude smaller than the size of the receptive-field centers of individual RGCs.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.