SIMONE PAPA

The mammalian retina is a sophisticated neural network that processes visual information in parallel, performing complex computations such as motion detection and direction selectivity (DS).

Starburst amacrine cells (SACs) play a key role in this process by providing asymmetric inhibition to direction-selective ganglion cells (DSGCs). While circuit-based mechanisms of DS have been extensively studied, the role of SAC dendritic morphology in shaping directional responses remains an open question.

In this project, by running simulations with a morphologically-faithful model of starburst amacrine cells, we aim to confirm that branching patterns of the neuronal processes can themselves serve as one of the underlying mechanisms of direction selectivity in the retina. This study investigates the computational principles underlying DS by refining and optimizing the Y-Branch SAC model, which simulates SAC dendritic architecture and its influence on motion processing.

Using a multicompartmental SAC model, revised from the RSME format of Ezra-Tsur E, et. al. (2021) through preliminary results from the host lab (Hiroki Asari, EMBL Rome), this research explored how dendritic branching patterns contribute to DS. High-performance computing (HPC) resources were exploited to systematically test branching angles, improving the biophysical realism of the SAC simulations.

A new optimization framework was developed in NEURON, allowing for efficient modifications to dendritic parameters while maintaining computational accuracy. Directional selectivity indices were computed for various stimulus speeds, revealing that centripetal selectivity emerges preferentially at specific dendritic angles. The results highlight the critical role of dendritic architecture in DS computations, supporting the hypothesis that intrinsic SAC morphology contributes to motion selectivity independently of synaptic mechanisms.

These findings provide a more nuanced understanding of DS, reinforcing the interplay between morphology, synaptic organization, and retinal computations. Although animal models have historically contributed to understanding DS in SACs through pharmacological and electrophysiological studies, computational models offer a promising non-invasive approach. However, animal validation will be ultimately required for confirming computational findings.

Future research integrating computational simulations with experimental models will be crucial in uncovering those mechanisms. Beyond fundamental neuroscience, this research could have broader implications for artificial vision, neuroprosthetics, and clinical interventions. The Y-Branch SAC model, refined in this work, serves as a powerful tool for investigating neuronal computations in the retina.