The logic of recurrent circuits in the primary visual cortex

Although most synapses within the neocortex are recurrent, the functional impact of this recurrence remains unclear. Prior experimental and theoretical work in the primary visual cortex (V1) suggests that recurrent excitation amplifies responses when signals are weak in order to optimize detection, while recurrent inhibition suppresses responses when signals are strong to optimize discrimination. Deepening our understanding of the logic for when recurrent activity facilitates cooperation among cortical ensembles and when it mediates competition beyond these initial results is fundamental to understanding cortical computation.
A major challenge to addressing this logic is parsing the impact of local recurrence within the circuit from that of feedforward and feedback inputs. To overcome this obstacle, we used multiphoton holographic optogenetics to activate ensembles of neurons in Layer 2/3 of V1 in the absence of any external stimulus. Leveraging the spatial specificity of two-photon holographic optogenetics, we find that both the spatial and stimulus preference organization of the ensemble affect the recruited activity. While these two organizing principles interact to produce non-intuitive population responses, with the strongest effects occurring within 30 µm of a stimulated cell, the sign of this interaction for individual responding neurons is dictated by their relative tuning to the ensemble. Specifically, compact (i.e., densely packed) co-tuned ensembles are able to recruit nearby iso-tuned neurons, while suppressing others.
Computational modeling suggests that a combination of highly local recurrent excitatory connectivity and the selective convergence of co-tuned excitatory neurons onto local inhibitory neurons can give rise to these principles of recurrent activity. Further, the model calls attention to the previously undervalued trade-off occurring between the excitatory E→E and suppressive E→I→E recurrent pathways. This combination of in vivo and in silico circuit interrogation explain how recurrent cortical circuits selectively amplify or suppress cortical activity depending on the precise pattern of co-active neurons.