续接01

Discussion

We investigated physiological, behavioural and neural substrates underlying the effects of overnight consolidation on suppression of aversive memories. Behaviourally, the insertion of a 24-h delay made aversive memories become more resistant to suppression, with less suppression-induced forgetting and enduring emotional reactivity. These results were replicated in an independent cohort of 25 participants. Overnight memories appeared less dependent on the hippocampus and more dependent on the neocortex, a result most likely due to overnight consolidation. Suppression of these consolidated memories was associated with higher prefrontal engagement but less concomitant hippocampal and amygdala disengagement, with distinct hippocampal functional connectivity with prefrontal executive systems when compared with newly acquired memories. These effects were paralleled by a shift away from hippocampal-dependent representational patterns to distributed neocortical representational patterns following overnight consolidation. Our findings suggest that overnight consolidation promotes the assimilation of newly acquired memories into more distributed neocortical regions, making memories more resistant to suppression through prefrontal-hippocampal inhibitory pathways.

Beyond previous studies focusing on newly acquired memories4,11, our behavioural and SCR data converge on the notion that overnight consolidation makes aversive memories more resistant to suppression. For newly acquired memories, we replicated findings from many previous studies by showing prominent suppression-induced forgetting for the NoThink trials compared with the baseline trials3,11,18, which is in conjunction with significant reduction in SCR. These behavioural and physiological effects, however, were less pronounced after overnight consolidation, suggesting less effective voluntary suppression of aversive memories and associated emotional reactivity. Notably, our behavioural and physiological results were reproduced by an independent cohort of 25 participants, indicating the reliability and robustness of the observed effects. In addition, evidence from our behavioural control experiment with neutral stimuli provides similar effect in suppression of neutral memories after overnight consolidation, implying a general role of consolidation process on memory suppression irrespective of emotional specificity. From a broader perspective, our behavioural effects are reminiscent of results from recent studies indicating the reduced efficiency of voluntary suppression of established autobiographical memories30,31. This reduction is thought to result from memory consolidation processes32,33, including rapid eye movement sleep34. It is thus conceivable that overnight sleep may contribute to our observed effects of consolidation on aversive memories.

In conjunction with the above behavioural and SCR results, multiple aspects of our neuroimaging data from brain activation, functional connectivity and neural representation patterns consistently suggest that overnight consolidation promotes the assimilation of newly acquired memories into more distributed neocortical representations, thus making these memories more resistant to suppression. First, as predicted by classical consolidation models1,15, we observed that retrieval of newly acquired memories was associated with relatively greater hippocampal engagement, while recall of memories after consolidation was associated with greater engagement in the neocortical regions such as the LPC and angular gyrus. Similar hippocampal–neocortical functional reorganization involving consolidation of newly learned materials has been demonstrated in previous studies in humans and laboratory animals8,22,35. After consolidation, these memories may become assimilated into the neocortex and develop into more stable and less hippocampus-dependent representation through strengthening cortical-cortical connections7,9,36. In other words, overnight consolidation leads to large-scale functional reorganization of the MTL–neocortical networks involved in the transformation of aversive memories into more stable representations.

Second, suppression of aversive memories after consolidation appears to associate with increased prefrontal engagement but less concomitant hippocampal and amygdala disengagement when compared with newly acquired memories. Numerous studies involving the suppression of newly acquired materials have found increased engagement of the right DLPFC along with disengagement of the MTL regions4,5,11. We found a similar pattern in the suppression of newly acquired aversive memories. Although some regions involved in memory suppression seem to partially overlap with the default mode network (DMN), they differ substantially from the core DMN nodes when compared with the template derived from a large-scale meta-analysis of over 10 000 fMRI studies37. In fact, we observed significantly higher activation in the DLPFC and concomitant lower hippocampal activation during NoThink compared with Think trials. This is contradictory to the classic DMN hypothesis which predicts higher activity in DMN but lower activity in the central executive network such as DLPFC38. Consistent with findings from previous studies5,12, we believe that intentional suppression-induced forgetting involves the active and effortful inhibition of retrieval and access to memories5,12, rather than disengagement from the task.

More importantly, we observed a significant increase in right DLPFC activation during the suppression of aversive memories after overnight consolidation when compared with newly acquired memories. It appears that the right DLPFC is critical for inhibitory control over the hippocampal memory system in voluntary suppression processing4,11,12,18. The increased right DLPFC activation here may reflect more cognitive effort needed to suppress consolidated relative to recently acquired memories. This is in line with Anderson’s executive control hypothesis, suggesting a U-shaped relationship between the levels of activation of the unwanted memory and the engagement of cognitive control39. In the present study, consolidated memories that became distributed in the neocortical network were more difficult to suppress, and thus required higher prefrontal inhibitory control than more recently acquired memories.

Furthermore, we observed less pronounced hippocampal and amygdala disengagement after overnight consolidation than for newly acquired memories. Previous studies using memory suppression tasks have demonstrated that memory suppression is associated with reduced activity in brain areas critical for episodic recollection (for example, hippocampus)11,18 and emotional memory (for example, amygdala)4. Our prediction analyses based on machine learning algorithms confirmed that the hippocampal disengagement is predictive of suppression-induced forgetting only in the newly acquired condition. Our observed maintenance of hippocampal and amygdala activity during suppression of consolidated aversive memories may reflect less effective prefrontal inhibitory control over the hippocampus and amygdala after overnight consolidation. This interpretation is consistent with less suppression-induced forgetting and less pronounced SCR reduction in the overnight consolidation condition.

It is worth noting that, even when only considering suppression-induced intentional forgetting versus incidental forgetting trials, we observed a very similar pattern of increased prefrontal engagement and less concomitant hippocampal and amygdala disengagement in suppression of overnight memories compared with newly acquired memories. This provides strong evidence that our observed effects truly reflect successful suppression-induced forgetting rather than the attempt to suppress. Although we cannot fully rule out potential confounds related to differences in general memory performance, memory decay on time, and interference between overnight and newly acquired memories (cf. Hulbert and Norman40), we undertook several steps to mitigate these confounding factors. For instance, we opted for an experimental design with specific baseline conditions separately for newly acquired and overnight consolidated memories. We then restricted our analyses for SCR and neuroimaging data by artificially matching the number of trials between these two conditions while taking subsequent memory status (remembered versus forgotten) into account. Together, converging evidence from behavioural, physiological and brain activation suggests that overnight consolidation makes aversive memories more resistant to suppression, which requires more prefrontal inhibitory control over hippocampal memory and amygdala emotional systems.

Third, suppression of overnight memories engages distinct hippocampal–prefrontal functional coupling processes and is associated with a shift away from hippocampal-dependent memory to more distributed neocortical representation patterns. Coordinated functional interactions between prefrontal executive and hippocampal memory systems are known to play a central role in memory suppression. Voluntary suppression of a memory is believed to involve prefrontal inhibitory control of the hippocampus where individual memories are processed18. Indeed, we observed increased hippocampal functional coupling with the right DLPFC during suppression of recently acquired memories. Analysis of forgotten trials further confirmed higher functional coupling between hippocampus and right DLPFC in the successful voluntary forgetting (that is, intentional versus incidental forgetting) of overnight memories. Interestingly, higher hippocampal-prefrontal functional coupling was predictive of greater suppression-induced forgetting of newly acquired memories. The opposite pattern of results was, however, observed in suppression of memories after overnight consolidation. These findings suggest that voluntary suppression through prefrontal inhibitory control over hippocampal memory circuitry is less effective and a distinct mechanism may be engaged in suppression of consolidated memories.

Moreover, results from multivoxel pattern similarity analysis revealed that individual aversive memories became more similar (that is, less discrete) following overnight consolidation in terms of inter-item multivoxel activity patterns (that is, less dissimilar to each other) in the hippocampus. These results provide evidence that individual memories encoded by the hippocampus may develop into less separable representations after overnight consolidation. Pattern separation is considered a hallmark of episodic memory, which allows numerous experiences to be differentiated41. The hippocampus is believed to play a central role in pattern separation by encoding disparate aspects of episodic memories as distinct non-overlapping representations to prevent interference across different memories42,43. Our observed increase in inter-item multivoxel pattern similarity in the hippocampus after overnight consolidation supports the notion that individual aversive memories become less discrete over time44. It is thus conceivable that the less efficient suppression of aversive memories found after overnight consolidation might be the result of less discrete neural representations of these memories, which make them less susceptible (thus more resistant) to prefrontal–hippocampal inhibitory processes as a result of a shift to distributed neocortical representations after overnight consolidation. In support of this idea we observed that higher multivoxel pattern dissimilarity in the hippocampus selectively predicted more effective suppression of newly acquired memories, but higher pattern dissimilarity in the LPC predicted more effective suppression of consolidated memories. In addition, we found that higher multivoxel pattern similarity in the hippocampus was associated with higher DLPFC engagement in memory suppression, suggesting increased prefrontal inhibitory efforts were required to suppress less discrete memory representations. This also explains our observed higher DLPFC activation in suppressing aversive memories after consolidation.