Rapid hippocampal plasticity supports motor sequence learning

Abstract

Recent evidence suggests that gains in performance observed while humans learn a novel motor sequence occur during the quiet rest periods interleaved with practice (micro-offline gains, MOGs). This phenomenon is reminiscent of memory replay observed in the hippocampus during spatial learning in rodents. Whether the hippocampus is also involved in the production of MOGs remains currently unknown. Using a multimodal approach in humans, here we show that activity in the hippocampus and the precuneus increases during the quiet rest periods and predicts the level of MOGs before asymptotic performance is achieved. These functional changes were followed by rapid alterations in brain microstructure in the order of minutes, suggesting that the same network that reactivates during the quiet periods of training undergoes structural plasticity. Our work points to the involvement of the hippocampal system in the reactivation of procedural memories.

Keywords: functional MRI; hippocampus; motor sequence learning; reactivation; structural plasticity.

Fig. 1.

Experimental design and behavior. (A) Experimental design. Twenty participants trained on two sensorimotor tasks: MSL (7) and an active CTL. Functional images (fMRI) were obtained during training on the MSL task. DWIs were acquired for both tasks before and 30 min and 24 h after practice. Long-term memory was assessed at Test, 24 h after practice. (B) Motor sequence learning. During motor practice, MSL was quantified using the intertap interval (depicted in black), i.e., the time elapsed between successive key presses from correctly executed sequences. Green dots represent the mean intertap interval ± SE for each block. Rest periods between task blocks are indicated with vertical gray bars. Inset illustrates the different learning metrics quantified for each block: MOnGs, MOGs, and total learning. MOGs were calculated as the difference (delta) in the mean intertap interval between the last correct sequence of a block and the first correct sequence of the next block, whereas MOnGs were quantified as the difference between the first and last correct sequence within a block. Total learning reflects the sum of MOnGs and MOGs. (C) MOGs and MOnGs. Data points in the violin plots depict the sum of deltas for each learning metric over all blocks across participants (***P < 0.0001; *P < 0.05, NS: not significant, t test against zero, corrected by Bonferroni). A linear regression analysis indicated that total learning was explained by MOGs (MOGs: F(1,18) = 4.5371, P = 0.047; MOnGs:F(1,18) = 7e-04, P = 0.98). Furthermore, MOGs differed from MOnGs (paired t test, t₁₉ = 1.99, P = 0.03). (D) Cumulative sum of changes in performance as a function of block number. This measure was computed based on the cumulative sum of deltas for all blocks. Note that MOGs take place during the first half of learning, before reaching asymptotic performance.

Fig. 2.

Tracking functional and microstructural changes induced by MSL. (A) Functional changes observed during motor learning. Shown are the whole-brain, voxelwise statistical parametric maps (SPMs) for the Task (Task > Rest) and Rest (Rest > Task) periods (P < 0.05, FWE-corrected). Note that cortico-cerebellar and cortico-striatal systems were activated during task execution (in red), whereas the hippocampus and the precuneus increased their activity during the rest periods (in green). (B) Longitudinal changes in microstructure induced by MSL. To identify longitudinal changes in microstructure that differed across scanning sessions and tasks, we conducted a whole-brain, voxelwise task (MSL vs. CTL) by scanning session (baseline, 30 min, 24 h) interaction analysis (FWE-corrected P value <0.05). Barplots show the mean and the 95% CIs corresponding to the time course of MD for each cluster identified in this analysis. Note that MSL induced changes in MD in the hippocampus and the precuneus. (C) Overlap between functional and structural changes induced by MSL. Shown are the SPMs resulting from the Rest fMRI analysis (in green) overlaid on the MD analysis (in magenta).