Sleep and Memory Consolidation

Modeling sleep and memory consolidation
With Terrence Sejnowski (Salk Institute, USA), we have elaborated a theory of memory consolidation during sleep. In collaboration with Diego Contreras and Mircea Steriade, we have characterized the spatiotemporal structure of LFP and unit activity during wake and sleep states in the cat [1]. This analysis suggests the deepest phase of slow-wave sleep consists of the alternance between slow-waves and short episodes of “wake-like” activity (“up-states”), which are electrophysiologically almost undistinguishable from the activity during wakefulness. This cyclic structure is consistent with the fact that up-states would represent “replayed” events that have occurred previously during the wake state [1]. Interestingly, during the early phase of slow-wave sleep, another type of oscillation, sleep spindles, seem to be optimal to induce a massive calcium entry in cortical pyramidal neurons [2]. This conclusion was reached by combining computational models with intracellular measurements during spindles in cortex. Such a massive calcium entry, at a frequency around 10 Hz, is an ideal signal to activate molecular gates, such as protein kinase A (PKA). Sleep spindles would therefore provide a physiological signal similar to the repeated tetanus used to induce long-term synaptic changes in slices. However, instead of inducing potentiation directly, spindles may in fact provide a “priming” signal, opening a gate that allows permanent changes to subsequent inputs (the “replayed” events above) following the sleep spindles.

Taking those two observations together leads to the following scenario for how these different mechanisms may contribute to memory consolidation during sleep [3, 4]. During conscious experience, latent memories are formed throughout the cortex, together with links to the hippocampal formation that allow top-down retrieval to occur. During the early stages of sleep, spindle oscillations would mobilize the molecular machinery needed for memory consolidation. In the deeper phases of slow-wave sleep, during the brief periods of wake-like activities (“up-states”), the hippocampal formation would activate latent memories stored in the neocortex (“replay”) and induce permanent changes in intrinsic or synaptic conductances. This hypothetical mechanism of memory consolidation during sleep is consistent with all electrophysiological characteristics of sleep oscillations, and it predicts that special correlations between hippocampal and cortical activities should occur during the up-states of slow waves (see details in [4]). Such correlations have been found recently between cortical slow-waves (up states) and hippocampal sharp waves by Buzsaki’s and McNaughton’s groups. No computational model has been proposed to date to explain these observations, and certainly constitutes one of the most exciting directions to pursue towards exploring the role of sleep oscillations.

The predicted strong impact of inhibition from intracellular recordings in anesthetized cats, combined with modeling [2], was recently confirmed from awake rats [5]. Using multisite neuronal recordings in rat mPFC, it was shown that during sleep spindles, oscillatory responses of cortical cells are different for different cell types and cortical layers. Interneurons were always more modulated than pyramidal cells, both in firing rate and phase, suggesting that the dynamics are dominated by inhibition. In the deep layers, where most of the hippocampal fibers make contacts, pyramidal cells respond phasically to SPWRs, but not during spindles. These results demonstrate that during sleep spindles, the cortex is functionnaly “de-afferented” from its hippocampal inputs, based on processes of cortical origin, and presumably mediated by the strong recruitment of inhibitory interneurons. The interplay between hippocampal and thalamic inputs may underlie a global mechanism involved in the consolidation of recently formed memory traces.

An intriguing hypothesis was proposed more recently [6], namely that the floating threshold for synaptic plasticity may explain why and how slow oscillations could subserve a consolidation of memory traces. It was shown that the floating threshold enables networks of neurons to display spontaneous activity without destruction of changes in synaptic weights (a well-known problem, called the “catstrophic forgetting effect”). However, if applied to a pattern in which the network displays recurrent periods of silence (“Down states”), the effect is fundamentally different. During the silences, the floating threshold diminishes, which leaves the network sensitive to any subsequent pattern of activation (“Up states”). Thus, in this case, the alternation of Up- and Down-states could be a mechanism to allow plasticity to occur during the Up-states. This constitutes the first ever proposed biophysically-plausible mechanism for how slow oscillations with Up & Down states could induce permanent synaptic changes, while sustained activity (as in wakefulness) would not affect synaptic weights (see [6]). No network simulations of this mechanism have been performed yet, but this constitutes a nice challenge for future modeling studies.

In a recent study using multi-electrode recordings in human and monkey [7], we found that fast beta and gamma oscillations of wakefulness are also present in slow-wave sleep, but with higher coherence. This observation may also be relevant to memory concolidation during sleep, because it suggests that fast oscillations may be associated to the replay of information during sleep, but their high coherence may be a substrate to induce long-term changes in cortical circuits during this reactivation. such a possible role for fast oscillations in memory consolidation during sleep is also an interesting venue to explore by future models.

[1] Destexhe, A., Contreras, D. and Steriade, M. Spatiotemporal analysis of local field potentials and unit discharges in cat cerebral cortex during natural wake and sleep states. J. Neurosci. 19:4595-4608, 1999 (see abstract) [2] Contreras, D., Destexhe, A. and Steriade, M. Intracellular and computational characterization of the intracortical inhibitory control of synchronized thalamic inputs in vivo. J. Neurophysiol. 77:335-350, 1997 (see abstract) [3] Sejnowski, T.J. and Destexhe, A. Why do we sleep? Brain Research 886: 208-223, 2000. (see abstract) [4] Destexhe, A. and Sejnowski, T.J. Thalamocortical Assemblies, Oxford University Press, 2001 (see abstract) [5] Peyrache, A., Battaglia, F. and Destexhe, A. Inhibition recruitment in prefrontal cortex during sleep spindles and gating of hippocampal inputs. Proc. Natl. Acad. Sci. USA 108:17207-17212, 2011 (see abstract) [6] El Boustani S, Yger P, Frégnac Y and Destexhe, A. Stable learning in stochastic network states. Journal of Neuroscience 32: 194-214, 2012. (see abstract) [7] Le Van Quyen, M., Muller, L., Telenczuk, B., Cash, S.S., Halgren, E., Hatsopoulos, N.G., Dehghani, N. and Destexhe, A. High-frequency oscillations in human and monkey neocortex during the wake-sleep cycle. Proc. Natl. Acad. Sci. USA 113:9363-9368, 2016 (see abstract)

Alain Destexhe