Cell-type specific inhibitory dynamics shape binocular response normalization in visual cortex

Abstract

Response normalization during sensory processing is a canonical cortical computation thought to emerge from inhibition-stabilized networks (ISNs). A key prediction from ISNs is that inhibition tracks and scales with excitation, leading to sublinear response summation, but which interneuron classes mediate this operation is unclear. Here, we investigate temporal dynamics and response summation in parvalbumin (PV) and somatostatin (SST) interneurons in mouse binocular visual cortex (bV1). We find that both PV and SST interneurons are binocular but exhibit distinct temporal dynamics and summation profiles. Binocular responses in PV interneurons tracked pyramidal neuron responses, exhibiting sublinear summation consistent with ISN models. In contrast, binocular responses SST interneurons exhibited rapid temporal dynamics and linear or supralinear summation, deviating from ISN predictions. While both PV and SST neurons received excitatory drive via interhemispheric callosal input, they exhibited distinct cellular properties, synaptic dynamics and input-output transformations. Biophysically constrained simulations indicated that cellular and synaptic differences alone were insufficient to reproduce the observed response dynamics in vivo during binocular integration, which were better explained by class-specific differences in local inhibitory circuit motifs. Together, our results indicate that heterogeneous, cell-type-specific inhibitory dynamics shape response normalization in bV1, with PV but not SST interneurons having ISN-like properties.