Causal evidence for retina dependent and independent visual motion computations in mouse cortex

How do computations in the sensory periphery, specifically the retina, contribute to the processing of visual motion in the primary visual cortex (V1)? While direction-selective retinal ganglion cells (DS cells) are known to project to the thalamus and cortex, the causal link between these retinal computations and cortical direction selectivity has remained unclear.

The researchers disrupted retinal direction selectivity in mice using two distinct methods: FRMD7 mutants, which specifically lose horizontal direction selectivity, and the genetic ablation of starburst amacrine cells, which eliminates direction selectivity across all cardinal axes. They then used two-photon calcium imaging to record neuronal activity in layer 2/3 of V1 while presenting visual motion stimuli at various speeds and directions. By comparing these results to control mice, they assessed how the loss of retinal input altered cortical motion processing.

In control mice, V1 exhibits a clear overrepresentation of cells preferring posterior motion, which corresponds to the dominant visual flow experienced during forward locomotion. When retinal direction selectivity was disrupted, this posterior-motion bias disappeared, and cortical responses to high-speed stimuli were significantly reduced. However, the cortex retained a baseline level of direction selectivity that was more evenly distributed across directions and speeds. This indicates that cortical direction selectivity is a composite of two functionally distinct computations: one inherited from the retina and another generated locally within the cortex or thalamus.

This study provides causal evidence that the cortex does not simply receive pre-processed information from the retina but integrates it with locally computed signals. By demonstrating that retinal inputs are specifically tuned to high-speed posterior motion, the findings suggest that the visual system is adapted to the statistics of an animal's natural movement. This dual-pathway architecture allows the cortex to expand the dynamic range of motion processing beyond what is available in the periphery.