Inhibition stabilization is a widespread property of cortical networks

Abstract

Many cortical network models use recurrent coupling strong enough to require inhibition for stabilization. Yet it has been experimentally unclear whether inhibition-stabilized network (ISN) models describe cortical function well across areas and states. Here, we test several ISN predictions, including the counterintuitive (paradoxical) suppression of inhibitory firing in response to optogenetic inhibitory stimulation. We find clear evidence for ISN operation in mouse visual, somatosensory, and motor cortex. Simple two-population ISN models describe the data well and let us quantify coupling strength. Although some models predict a non-ISN to ISN transition with increasingly strong sensory stimuli, we find ISN effects without sensory stimulation and even during light anesthesia. Additionally, average paradoxical effects result only with transgenic, not viral, opsin expression in parvalbumin (PV)-positive neurons; theory and expression data show this is consistent with ISN operation. Taken together, these results show strong coupling and inhibition stabilization are common features of the cortex.

Response nonlinearities in networks of spiking neurons

“…In this work, we have investigated responses in networks of spiking neurons at finite coupling. In this regime, which has recently been suggested to underlie cortical dynamics [ 24 , 25 , 41 , 42 ], we have shown that two types of nonlinearities emerge: response-onset and saturation. The network response transitions between these two nonlinearities as feedforward input increases; for intermediate inputs, the response matches that of the balanced-state model up to corrections of order .…”

VIP interneurons in mouse primary visual cortex selectively enhance responses to weak but specific stimuli

“…Pyr → VIP → SST → Pyr) that depends upon, and contributes to, network dynamics. Stabilized supralinear network (SSN) models have been proposed to account for a variety of contrast-dependent response properties in visual cortex ( Rubin et al, 2015 ; Ahmadian et al, 2013 ), including the transition from a high gain regime at low contrast to a feedback inhibition dominated low gain regime at high contrast ( Adesnik, 2017 ; Sanzeni et al, 2020 ), as well as cortical noise correlations ( Hennequin et al, 2018 ), surround suppression ( Liu et al, 2018 ), and effects of feature and spatial attention on neural activity ( Lindsay et al, 2020 ). In SSNs, high gain is achieved through supralinear single-neuron transfer functions (e.g.…”

Performance in even a simple perceptual task depends on mouse secondary visual areas

“…r vis : visual response period for analysis in C. Black and blue bars: duration of visual stimulus and optogenetic inhibitory stimulus, respectively. Vertical scale bars in panels vary, to more clearly illustrate relative suppression with optogenetic stimulation; quantification in panel C. The rightmost panel shows at the highest power a transient associated with ISN dynamics ( Sanzeni et al, 2020, Figure 5–Figure Supplement 2 ), which has ended by the time the visual response analysis period begins. V1, lateral, medial panels: N = 14, 6, 7 single units.…”