Motor learning- and consolidation-related resting state fast and slow brain dynamics across wake and sleep

Abstract

Motor skills dynamically evolve during practice and after training. Using magnetoencephalography, we investigated the neural dynamics underpinning motor learning and its consolidation in relation to sleep during resting-state periods after the end of learning (boost window, within 30 min) and at delayed time scales (silent 4 h and next day 24 h windows) with intermediate daytime sleep or wakefulness. Resting-state neural dynamics were investigated at fast (sub-second) and slower (supra-second) timescales using Hidden Markov modelling (HMM) and functional connectivity (FC), respectively, and their relationship to motor performance. HMM results show that fast dynamic activities in a Temporal/Sensorimotor state network predict individual motor performance, suggesting a trait-like association between rapidly recurrent neural patterns and motor behaviour. Short, post-training task re-exposure modulated neural network characteristics during the boost but not the silent window. Re-exposure-related induction effects were observed on the next day, to a lesser extent than during the boost window. Daytime naps did not modulate memory consolidation at the behavioural and neural levels. These results emphasise the critical role of the transient boost window in motor learning and memory consolidation and provide further insights into the relationship between the multiscale neural dynamics of brain networks, motor learning, and consolidation.

Figure 1

Study design. (A) One block of the Finger tapping task (FTT): participants perform 20 blocks (30 s/block) separated by 20-s rest intervals on the finger tapping task (FTT) using the four fingers of the left hand (repeated sequence: little/4/, index /1/, ring /3/, middle /2/, little /4/ fingers). (B) Full experimental design. After a first resting state (RS) recording and a learning session (20 FTT blocks), there were 3 testing windows in the MEG scanner (T1: Boost, T2: Silent, T3: Next day), with each time a RS immediately before and after a short (2 FTT blocks) behavioural testing.

Figure 2

Global Performance Index (GPI) evolution. (A) Mean normalised GPI scores across 26 blocks of Finger Tapping Task (FTT); blocks 1–20 correspond to the Learning session; blocks 21–22 to the boost (T1); blocks 23–24 to the silent (T2) and blocks 25–26 to the next day (T3) windows. (B) Normalised GPI score evolution from Best Motor Performance (BMP) at the end of the Learning session to each testing session. Circles: Wake condition; Squares: Nap condition. Error bars represent standard errors.

Figure 3

Spatial topographies of HMM transient states computed over the 7 rest sessions (RS 1–7). Red/blue scales indicate positive/negative correlation values between the envelope and the state activation/inactivation time course (i.e., increased/decreased power during one state visit). For visualisation purposes, the maps are thresholded to 60% of the maximum absolute of the partial correlation values.

Figure 4

HMM, temporal parameters for Sensorimotor/Visual State 4. White violin plots represent non-induced RS sessions, and grey violin plots—induced RS. Medians: solid lines; quartiles: dotted lines. (***p < 0.001 and **p < 0.002, corrected by factor 21).

Figure 5

Increased connectivity (NBS analysis) as a result of re-introduction to the task (induction effect) during the boost period. Nodes are scaled according to their weight (the sum of all edges connected to the node). Significant edges are represented as interconnecting lines between 126 connectome seed regions. TMG Temporal middle gyrus, STG Superior temporal gyrus, pMFC posterior Medial frontal cortex, INS Insula, SMA Supplementary motor area, L left, R right.

Figure 6

Neural network positively correlated with the Best Motor Performance (BMP) Index during RS 3 (induced, boost window). Nodes are scaled according to their weight (the sum of all edges connected to the node). Significant edges are represented as interconnecting lines between 126 connectome seed regions. ORBmid Middle frontal gyrus, orbital part; ORBsupmed Superior frontal gyrus, medial orbital; SFG Superior frontal gurus; ANG Angular gyrus; STG Superior temporal gyrus; L left; R Right.

Figure 7

Neural network positively correlated with Learning index during RS 5 (induced, silent window). Nodes are scaled according to their weight (the sum of all edges connected to the node). Significant edges are represented as interconnecting lines between 126 connectome seed regions. MFG Middle frontal gyrus, CAU Caudate nucleus, INS Insula, THA Thalamus, AMYG Amygdala, TPOsup Temporal pole, superior temporal gyrus, TPOmid Temporal pole, middle temporal gyrus, HIP Hippocampus, FFG Fusiform gyrus, Crus I Crus, Cerebellum, VER Vermis, Cerebellum, L left, R Right.