Altered proportions of retinal cell types and distinct visual codes in rodents occupying divergent ecological niches.

Vertebrate retinas share a basic blueprint comprising 5 neuronal classes arranged according to a common wiring diagram. Yet, vision is aligned with species differences in behavior and ecology, raising the question of how evolution acts on this circuit to adjust its computational characteristics. We address that problem by comparing the thalamic visual code and retinal cell composition in closely related species occupying different niches: Rhabdomys pumilio, which are day-active murid rodents, and nocturnal laboratory mice (Mus musculus). Using high-density electrophysiological recordings, we compare visual responses at both single-unit and population levels in the thalamus of these two species. We find that Rhabdomys achieves a higher spatiotemporal resolution visual code through the selective expansion of information channels characterized by non-linear spatiotemporal summation. Comparative analysis of single-cell transcriptomic atlases reveals that this difference originates with the increased relative abundance of retinal bipolar and ganglion cell types supporting OFF and ON-OFF responses. These findings demonstrate that evolution may drive changes in neural computation by adjusting the proportions of shared cell types rather than inventing new types and show the power of matching high-density physiological recordings with transcriptomic cell atlases to study evolution in the brain.