Google meet virtual room: https://meet.google.com/feh-gbqi-eua
Abstract:
Nikola Jajcay,
Klaus ObermayerThe hierarchical nesting of sleep spindles and cortical slow oscillations is considered a precursor of successful episodic memory consolidation where it, presumably, sets the stage for memory traces migration from short-term hippocampal storage to longer-lasting neocortical sites[1]. Spindles are thought to be generated in the thalamus and projected onto cortical sites, while slow oscillations originate in the cortex and migrate in wave-like patterns to the thalamus, hence the interplay between the two rhythms is orchestrated within the thalamocortical circuitry. The state-of-the-art mass models of the thalamocortical loop however only consider relevant motifs and patterns in sleep but discard network effects. Here we take the first steps towards integrating thalamocortical projections into a large-scale brain model and modelling whole-brain cortical slow- wave activity and thalamic spindles as seen in non-REM sleep.
We model the thalamus as a network node containing one excitatory and one inhibitory mass representing thalamocortical relay neurons and thalamic reticular nuclei, respectively. With little deviations, our model follows the thalamic component developed in [2]. In the thalamic submodule, we investigated its spindling behaviour upon changing, firstly, conductances of rectifying and T-type calcium current, by which the thalamus can be parametrised in three oscillatory regimes: fast oscillations, dominated by Ca current; spindle regimes with a balanced interplay of Ca and rectifying currents, and slow delta oscillations for strong hyperpolarisation. Next, by the application of external excitatory firing rate drive, which simulates excitatory source connected to the thalamus, we found dynamically interesting spindle-promoting regimes in interaction with thalamic conductances (see Fig. 1 for estimated number of spindles), and by changing the parameters or external drive, we were also able to control the inter-spindle interval, spindle duration, and shape of spindle envelope.
Secondly, we connected the thalamic node to one cortical node, modelled as interconnected excitatory and inhibitory adaptive exponential integrate-and- fire neuronal masses[3]. The excitatory mass contains a spike-triggered adaptation mechanism by which the node is parametrised to sit in the limit cycle generating slow oscillations by the means of excitation-adaptation feedback loop. We investigated the dynamical repertoire of the connected model concerning the connection strength and network delays. Our preliminary results indeed show that, upon connection, the thalamic spindles are imprinted into cortical node activity
where they modulate slow oscillation envelope, and that slow oscillation activity of the cortex, in turn, shapes spindling behaviour of thalamocortical relay mass by affecting spindle duration and inter-spindle interval. Our connected model also conserves the phase-phase and phase-amplitude couplings reported in the literature on observed EEG data or other thalamocortical models.
Our results suggest that thalamic mass model of spindle activity can be connected to various mass or mean-field models of cortical nodes and, after careful treatment of network connections and delays, we believe that our conclusions would carry over to the large-scale network model.
[1] Ji D, Wilson M A (2007). Nat Neurosci 10(1).
[2] Schellenberger Costa M, et al. (2016). PLoS Comput Biol 12(9).
[3] Cakan C, Obermayer K (2020). PLoS Comput Biol 16(4).
