Felix Franke^1,2, Marc Büttner^1,2, Matej Znidaric^2,3, Roland Diggelmann^2,3, Federica B. Rosselli², Annalisa Bucci^2,3, Andreas Hierlemann³; ¹University of Basel, ²Institute of Molecular and Clinical Ophthalmology Basel (IOB), ³Eidgenössische Technische Hochschule Zürich (ETH)

Along the processing hierarchy of sensory systems, neurons become progressively more functionally specialized, i.e., their responses become increasingly selective to specific stimulus features. In the visual system, early stages encode changes in local contrast, whereas later stages selectively respond to complex stimulus features such as textures and object identities. Traditional approaches to study stimulus encoding in sensory systems are effective in revealing localized receptive fields for many neurons in early stages of the processing hierarchy. However, as functional specialization increases, and using traditional methods, the fraction of neurons showing strong and predictive receptive fields decreases. We demonstrate that this effect is already evident in the mouse retina where a significant fraction of retinal ganglion cells exhibits only weak and unpredictive receptive fields. The situation worsens one synapse downstream, in the nucleus of the optic tract (NOT), preventing receptive field identification for most neurons. We developed a method using structured, parameterized stimuli that are white in a nonlinear reparameterization of the stimulus space. Our approach efficiently identifies receptive fields, for nearly all recorded neurons, in both mouse and primate retinae, as well as in the mouse NOT. Most of the newly identified receptive fields are localized but nonlinear, i.e., the cells respond strongest to moving stimuli and not to unstructured contrast changes. Consequently, we describe the receptive fields in a movement space, rather than classical light intensity space. We use the approach to highlight differences in functional specialization between the primate and the mouse retina, as well as between the mouse retina and its direct projection target, the NOT.

Acknowledgements: Funded by SNSF (CRSII5_173728, CRSII5_216632; PCEFP3_187001; CRSK-3_220987, CRSK-3_221257; 310030_220209); ERC Advanced Grant (neuroXscales, 694829); Sedinum Foundation.