Sleep modulates the neural substrates of both spatial and contextual memory consolidation

Abstract

It is known that sleep reshapes the neural representations that subtend the memories acquired while navigating in a virtual environment. However, navigation is not process-pure, as manifold learning components contribute to performance, notably the spatial and contextual memory constituents. In this context, it remains unclear whether post-training sleep globally promotes consolidation of all of the memory components embedded in virtual navigation, or rather favors the development of specific representations. Here, we investigated the effect of post-training sleep on the neural substrates of the consolidation of spatial and contextual memories acquired while navigating in a complex 3D, naturalistic virtual town. Using fMRI, we mapped regional cerebral activity during various tasks designed to tap either the spatial or the contextual memory component, or both, 72 h after encoding with or without sleep deprivation during the first post-training night. Behavioral performance was not dependent upon post-training sleep deprivation, neither in a natural setting that engages both spatial and contextual memory processes nor when looking more specifically at each of these memory representations. At the neuronal level however, analyses that focused on contextual memory revealed distinct correlations between performance and neuronal activity in frontal areas associated with recollection processes after post-training sleep, and in the parahippocampal gyrus associated with familiarity processes in sleep-deprived participants. Likewise, efficient spatial memory was associated with posterior cortical activity after sleep whereas it correlated with parahippocampal/medial temporal activity after sleep deprivation. Finally, variations in place-finding efficiency in a natural setting encompassing spatial and contextual elements were associated with caudate activity after post-training sleep, suggesting the automation of navigation. These data indicate that post-training sleep modulates the neural substrates of the consolidation of both the spatial and contextual memories acquired during virtual navigation.

Figure 1. Virtual environment and navigation tasks.

The map depicts an aerial view of the colour 3D virtual town in which subjects navigated at the ground level using a keypad. The 10 possible starting points are represented by letters (from A to J) with associated symbols and colours indicating the location to reach, out of 3 possible targets (Bin, Lee and Station). The 4 snapshots display samples of the environment as seen by the participant in the Natural, Recognition, Alternate or Impoverished conditions. For the Recognition task, subjects first navigated in the environment following colour dots on the ground (left panel). They were instructed to determine whether and where environmental changes had been made as compared to the town explored during the training period. At the end of each walk, a four-choice panel composed of 3 pictures taken from the path (one of them representing a change made in the environment), and a white square was presented. Subjects had to respond by selecting the modified image or the white square if they thought that no modification had been made (right panel). Also displayed in the insets is the navigation-related activity in each condition. Each image thus shows areas where BOLD response is greater than the mean (baseline) activity, at the population level, during navigation blocks. Contrasts are displayed at p^corr <0.05 (FWE, corrected for multiple comparisons in the whole brain volume) superimposed on a representative subject's MRI, in RS (blue blobs) and TSD participants (red blobs). Purple blobs represent activations observed in both groups. For the Alternate condition, the image represents activations during the whole blocks of navigation, that is not distinguishing between Detour or Routine behaviours.

Figure 2. Within-subjects correlation analyses between brain activity and trial to trial performance in the Natural condition.

Contrasts are displayed at p<0.001 (uncorrected) superimposed on the average T1-weighted MR scan. a) Higher correlations with performance in the RS than TSD group in the left (−10, 4, 12; Z = 3.72; top panel) and right (14, 14, 18; Z = 3.39; bottom panel) caudate nuclei. b) Parameter estimates (arbitrary units±standard deviations) of correlation between performance and brain activity in the left caudate nucleus (−10 14 12), indicating a positive association in the RS group (white plot), and a negative association in the TSD group (grey plot).

Figure 3. Sleep-dependent modulation of correlation between brain activity and performance in the Recognition condition.

Contrasts are displayed at p<0.001 (uncorrected) superimposed on the average T1-weighted MR scan. Correlations were computed at the within-subject level (i.e., between brain activity and individual variations in trial-to trial performance). Left panel: higher correlations in the RS than in the TSD group in the left frontal gyrus (−14 52 40, Z = 3.19). Right panel: higher correlations in the TSD than in the RS group in the right parahippocampal gyrus (18 −18 −26, Z = 3.38).