State-Dependent Regulation of Cortical Processing Speed via Gain Modulation

Abstract

To thrive in dynamic environments, animals must be capable of rapidly and flexibly adapting behavioral responses to a changing context and internal state. Examples of behavioral flexibility include faster stimulus responses when attentive and slower responses when distracted. Contextual or state-dependent modulations may occur early in the cortical hierarchy and may be implemented via top-down projections from corticocortical or neuromodulatory pathways. However, the computational mechanisms mediating the effects of such projections are not known. Here, we introduce a theoretical framework to classify the effects of cell type-specific top-down perturbations on the information processing speed of cortical circuits. Our theory demonstrates that perturbation effects on stimulus processing can be predicted by intrinsic gain modulation, which controls the timescale of the circuit dynamics. Our theory leads to counterintuitive effects, such as improved performance with increased input variance. We tested the model predictions using large-scale electrophysiological recordings from the visual hierarchy in freely running mice, where we found that a decrease in single-cell intrinsic gain during locomotion led to an acceleration of visual processing. Our results establish a novel theory of cell type-specific perturbations, applicable to topdown modulation as well as optogenetic and pharmacological manipulations. Our theory links connectivity, dynamics, and information processing via gain modulation.

Neural mechanisms underlying the temporal organization of naturalistic animal behavior

“…These baseline changes may be driven by top-down projections from higher cortical areas or by neuromodulators. In a recurrent circuit exhibiting attractor dynamics, changes in baseline levels modulate the average transition times between metastable attractors ( Mazzucato et al, 2019 ; Wyrick and Mazzucato, 2021 ). In these models, attractors are represented by potential wells in the network’s energy landscape, and the height of the barrier separating two nearby wells determines the probability of transition between the two corresponding attractors (lower barriers are easier to cross and lead to faster transitions, Figure 4B ).…”

“…In particular, a decrease (increase) in pyramidal cells gain can lead to faster (slower) average action timing. This biologically plausible computational mechanism was proposed to explain the acceleration of sensory coding observed when gustatory stimuli are expected compared to when they are delivered as a surprise ( Samuelsen et al, 2012 ; Mazzucato et al, 2019 ); and the faster encoding of visual stimuli observed in V1 populations during locomotion periods compared to when the mouse sits still ( Wyrick and Mazzucato, 2021 ). Transition rates between attractors may also be modulated by varying the amplitude and color of the fluctuations in the synaptic inputs (see Appendix 3), while keeping the barrier heights fixed.…”

“…An alternative robust and biologically plausible way to generate a vast hierarchy of timescales was proposed based on the ingredients of recurrent functional clusters and heterogeneity in synaptic couplings ( Figure 8E ; Stern et al, 2021 ), two common features observed across cortical circuits ( Marshel et al, 2019 ; Kiani et al, 2015 ; Miller et al, 2014 ; Song et al, 2005 ; Perin et al, 2011 ; Lee et al, 2016 ), supported by theoretical evidence ( Litwin-Kumar and Doiron, 2012 ; Wyrick and Mazzucato, 2021 ). In this model, excitatory and inhibitory neurons are arranged in clusters of heterogeneous sizes, generating metastable activity whose typical timescale is measured by the on–off cluster switching time (see Appendix 3).…”

“…The stochasticity in behavioral unit duration can be achieved by generating transitions between attractors at each level via low-dimensional variability, arising from mesoscopic feedback loops involving cortex and subcortical areas ( Recanatesi et al, 2022 ; Figure 3D ). The average transition time between behavioral units at each level will then depend on the barrier height separating the corresponding attractors ( Cao et al, 2016 ; Mazzucato et al, 2019 ; Wyrick and Mazzucato, 2021 ). As one moves up the hierarchy from actions (fast transitions = lower barriers) to sequences and activities (slow transitions = higher barriers), the potential wells separating the corresponding attractors become deeper and the basins of attraction wider.…”

“…Phenomenological birth–death processes fit to spiking network simulations can give a qualitative understanding of the on–off cluster dynamics ( Huang and Doiron, 2017 ; Shi et al, 2022 ), and it would be interesting to derive these models from first principles. Mean field methods for leaky-integrate-and-fire networks can give a qualitative prediction of the effects of external perturbations on metastable dynamics, explaining how changes in an animal’s internal state can affect circuit dynamics ( Mazzucato et al, 2019 ; Wyrick and Mazzucato, 2021 ). These qualitative approaches should be extended to fully quantitative ones.…”