Transition between Functional Regimes in an Integrate-And-Fire Network Model of the Thalamus

Abstract

The thalamus is a key brain element in the processing of sensory information. During the sleep and awake states, this brain area is characterized by the presence of two distinct dynamical regimes: in the sleep state activity is dominated by spindle oscillations (7 − 15 Hz) weakly affected by external stimuli, while in the awake state the activity is primarily driven by external stimuli. Here we develop a simple and computationally efficient model of the thalamus that exhibits two dynamical regimes with different information-processing capabilities, and study the transition between them. The network model includes glutamatergic thalamocortical (TC) relay neurons and GABAergic reticular (RE) neurons described by adaptive integrate-and-fire models in which spikes are induced by either depolarization or hyperpolarization rebound. We found a range of connectivity conditions under which the thalamic network composed by these neurons displays the two aforementioned dynamical regimes. Our results show that TC-RE loops generate spindle-like oscillations and that a minimum level of clustering (i.e. local connectivity density) in the RE-RE connections is necessary for the coexistence of the two regimes. We also observe that the transition between the two regimes occurs when the external excitatory input on TC neurons (mimicking sensory stimulation) is large enough to cause a significant fraction of them to switch from hyperpolarization-rebound-driven firing to depolarization-driven firing. Overall, our model gives a novel and clear description of the role that the two types of neurons and their connectivity play in the dynamical regimes observed in the thalamus, and in the transition between them. These results pave the way for the development of efficient models of the transmission of sensory information from periphery to cortex.

Firing-rate based network modeling of the dLGN circuit: Effects of cortical feedback on spatiotemporal response properties of relay cells

“…The eDOG model, on which is based, assumes the response properties of dLGN cells to be linear, mimicking the firing behavior in the tonic mode [ 109 ]. To investigate network behavior when dLGN cells are in burst mode, spiking neuron models including calcium-like slow conductances [ 35 , 110 ] or biophysically detailed models with Hodgkin-Huxley style conductances [ 38 , 97 ] are likely necessary.…”

“…While network simulations based on integrate-and-fire type neuron models (e.g., [ 96 , 110 ]) and biophysically-detailed neuron models (e.g., [ 38 , 97 ]) for entire dLGN nuclei are becoming computationally feasible with modern computers, there will still be a need for conceptually and mathematically simpler network simulation tools such as the present tool based on the eDOG model. Such models are important to gain intuition about how the different circuit components may affect the overall circuit behavior, and will also be important for guiding the choice of the numerous, typically unknown, parameters in more comprehensive dLGN network simulations.…”

Thalamocortical Spectral Transmission Relies on Balanced Input Strengths

“…Here we summarize the main features of the model: the network is an extension of a thalamic network model introduced in Barardi et al. ( 2016 ) connected to a previously developed cortical network model Mazzoni et al. ( 2008 , 2011 ).…”

“…This is necessary for the onset of sustained thalamic oscillations (Barardi et al. 2016 ). The model includes thalamocortical afferents by considering synaptic connections from TC neurons to cortical excitatory and inhibitory populations.…”

“… Recurrent connections between RE neurons are defined by random and sparse connectivity scheme superimposed on a small-network arrangement (Barardi et al. 2016 ). We will use this parameter set throughout this manuscript, unless otherwise stated.…”