Empirically derived model of the role of sleep in associative learning and recuperative processes

Abstract

Neuronal firing patterns, defined as intraburst temporal relations between sequential spike intervals, were monitored extracellularly from the rostral part of the nucleus reticularis (NR) and from the centrum medianum—parafascicular complex (CM-Pf) of the thalamus in adult freely moving cats during acquisition of bar pressing behavior (BP1) for milk reward, subsequent slow wave sleep (SWS1), and contiguous rapid eye movement (REM1) sleep, and then, 16 to 48 h later, after the animal fully learned the conditioned BP behavior (BP2). The patterns were defined on the basis of relative relations between sequential spike intervals. The intervals were compared in sequential pairs, moving one interval at a time. This method, if applied to random and/or independent values, generates a novel and well-defined theoretical distribution of inequality patterns which is uniquely sensitive to the history of spike train and plastic changes at neuronal membrane. The “information” encoded in single neuronal firing patterns was quantified using two criteria: (a) the magnitude of pattern emissions with reference to the random model; and (b) the difference between pattern distribution during BP1-training/acquisition and after the animal fully acquired the BP behavior. The latter patterns, termed BP2-template, were assumed to be composed only of patterns relevant for learned behavior. Thus, other patterns observed during BP-1, SWS, and REM were classified as spurious. During BP-1 training/acquisition of conditioned behavior, in 28 out of 265 neurons studied, the firing patterns linked to and relevant for BP behavior emerged in background of spurious patterns. Subsequently, during SWS-1, the relevant patterns were selectively suppressed below chance level. Conversely, during REM-1 sleep, contiguous with SWS-1, patterns linked to acquisition of BP behavior were strongly amplified above the level observed during BP1-training episode, while the ‘spurious’ patterns were eliminated. Thus, REM-1 patterns may be regarded as amplified precursors of the BP2-template which, 16 to 48 h later, and often after additional training, emerged with the acquired conditioned BP behavior. With reference to the random model, the emission magnitudes of individual BP-1 patterns were correlated with suppressions of same patterns during subsequent SWS-1. The magnitudes of these deficits were, in turn, correlated with subsequent REM1-induced amplified emission of same patterns. These observations indicate that: (a) acquisition of operantly conditioned behavior is linked to inversions in statistical distribution of patterns during BP1, SWS1, and REM1 episodes; and (b) the inversions are not random but graded, and therefore are likely to reflect the well-known graded nature of receptor desensitization which may result from repetitive exposure to neurotransmitters and/or modulators during a motor task that requires repetitive use of the same set of neuronal assemblies and pathways over a significant period of time. The neuronal firing patterns in “control” episodes of milk drinking from a bowl (BOWL) had, on the average, no significant relationships to BP, SWS, nor REM episodes. The results provide: (a) a novel insight into the information processing based on intraburst temporal organization of spike trains, as opposed to the mean firing rate; (b) a novel concept of homeostatic behavior of neurons which is consistent with current views on the physico-chemical dynamics of neuronal membrane and its receptors; (c) the first quantifiable empirical evidence for the relationships between firing patterns, learning, SWS, and REM sleep; and (d) the results suggest a new model for learning linked to recuperative role of sleep which is based on spatio-temporal inversion of neuronal interactions and depolarization-induced recovery (DIR) of synaptic receptors that had been down regulated during waking state. This notion is compatible with receptor changes in vitro and in vivo.