単一細胞摂動が視覚野における適応的で刺激依存的な再発を明らかにする

Cortical processing must adapt dynamically to the strength of sensory inputs, integrating information when signals are weak and refining representations when they are strong. Experimental and theoretical studies suggest this adaptation involves recurrent cortical circuits through cooperative amplification and competitive interactions within functional networks. Recent studies in mouse primary visual cortex (V1) have begun to support this idea, consistent with novel circuit motifs whereby excitatory and inhibitory cells are strongly connected. Despite these advancements, neither models or data provide a clear picture of the behavior of cortical columns within circuits with functional topography. Here, neighboring cells (excitatory and inhibitory) are selective for similar features in visual space, raising the question as to whether cortical columns act as amplifiers or engage in feature competition. Moreover, columnar network models make a strong prediction: cortical circuits should engage in functional amplification during weak sensory drive, transitioning to suppressive interactions with increasing sensory drive-a prediction not yet directly tested. Using single-cell perturbations in ferret V1, we show that cortical networks switch between amplification and suppression, depending on stimulus strength. Combining two-photon optogenetic stimulation with a generalized linear model (GLM) to quantify perturbations, we uncovered spatially broad suppressive influence modulated by stimulus contrast. At low contrast, we observed amplification between functionally-coupled cells. At high contrast, these networks switched to suppression. This reversal emerged in a recurrent network model with strong inhibitory-excitatory connectivity and weaker excitatory-excitatory connections. This class of models predicts markedly stronger suppressive influences from inhibitory cells onto excitatory populations, which we confirmed with cell-specific perturbations. Our results show that cortical recurrence with functional topography toggles between amplification and suppression, providing direct evidence long-predicted by normative and mechanistic circuit models of visual cortex.