Engram mechanisms of memory linking and identity.

Memories are thought to be stored in neuronal ensembles referred to as engrams. Studies have suggested that when two memories occur in quick succession, a proportion of their engrams overlap and the memories become linked (in a process known as prospective linking) while maintaining their individual identities. In this Review, we summarize the key principles of memory linking through engram overlap, as revealed by experimental and modelling studies. We describe evidence of the involvement of synaptic memory substrates, spine clustering and non-linear neuronal capacities in prospective linking, and suggest a dynamic somato-synaptic model, in which memories are shared between neurons yet remain separable through distinct dendritic and synaptic allocation patterns. We also bring into focus retrospective linking, in which memories become associated after encoding via offline reactivation, and discuss key temporal and mechanistic differences between prospective and retrospective linking, as well as the potential differences in their cognitive outcomes.

Selective engram coreactivation in idling brain inspires implicit learning.

Passive priming of prior knowledge to assimilate ongoing experiences underlies advanced cognitive processing. However, the necessary neural dynamics of memory assimilation remains elusive. Uninstructed brain could also show boosted creativity, particularly after idling states, yet it remains unclear whether the idling brain can spontaneously spark relevant knowledge assimilations. We established a paradigm that links/separates context-dependent memories according to geometrical similarities. Mice exploring one of four contexts 1 d before undergoing contextual fear conditioning in a square context showed a gradual fear transfer to preexposed geometrically relevant contexts the next day, but not after 15 min. Anterior cingulate cortex neurons representing relevant, rather than distinct, memories were significantly coreactivated during postconditioning sleep only, before their selective integration the next day during testing. Disrupting sleep coreactivations prevented assimilation while preserving recent memory consolidation. Thus, assimilating pertinent memories during sleep through coreactivation of their respective engrams represents the neural underpinnings of sleep-triggered implicit cortical learning.

Adult neurogenesis modulates the hippocampus-dependent period of associative fear memory.

Acquired memory initially depends on the hippocampus (HPC) for the process of cortical permanent memory formation. The mechanisms through which memory becomes progressively independent from the HPC remain unknown. In the HPC, adult neurogenesis has been described in many mammalian species, even at old ages. Using two mouse models in which hippocampal neurogenesis is physically or genetically suppressed, we show that decreased neurogenesis is accompanied by a prolonged HPC-dependent period of associative fear memory. Inversely, enhanced neurogenesis by voluntary exercise sped up the decay rate of HPC dependency of memory, without loss of memory. Consistently, decreased neurogenesis facilitated the long-lasting maintenance of rat hippocampal long-term potentiation in vivo. These independent lines of evidence strongly suggest that the level of hippocampal neurogenesis play a role in determination of the HPC-dependent period of memory in adult rodents. These observations provide a framework for understanding the mechanisms of the hippocampal-cortical complementary learning systems.

Autophagy Enhances Memory Erasure through Synaptic Destabilization.

There is substantial interest in memory reconsolidation as a target for the treatment of anxiety disorders, such as post-traumatic stress disorder. However, its applicability is restricted by reconsolidation-resistant boundary conditions that constrain the initial memory destabilization. In this study, we investigated whether the induction of synaptic protein degradation through autophagy modulation, a major protein degradation pathway, can enhance memory destabilization upon retrieval and whether it can be used to overcome these conditions. Here, using male mice in an auditory fear reconsolidation model, we showed that autophagy contributes to memory destabilization and its induction can be used to enhance erasure of a reconsolidation-resistant auditory fear memory that depended on AMPAR endocytosis. Using male mice in a contextual fear reconsolidation model, autophagy induction in the amygdala or in the hippocampus enhanced fear or contextual memory destabilization, respectively. The latter correlated with AMPAR degradation in the spines of the contextual memory-ensemble cells. Using male rats in an in vivo LTP reconsolidation model, autophagy induction enhanced synaptic destabilization in an NMDAR-dependent manner. These data indicate that induction of synaptic protein degradation can enhance both synaptic and memory destabilization upon reactivation and that autophagy inducers have the potential to be used as a therapeutic tool in the treatment of anxiety disorders.SIGNIFICANCE STATEMENT It has been reported that inhibiting synaptic protein degradation prevents memory destabilization. However, whether the reverse relation is true and whether it can be used to enhance memory destabilization are still unknown. Here we addressed this question on the behavioral, molecular, and synaptic levels, and showed that induction of autophagy, a major protein degradation pathway, can enhance memory and synaptic destabilization upon reactivation. We also show that autophagy induction can be used to overcome a reconsolidation-resistant memory, suggesting autophagy inducers as a potential therapeutic tool in the treatment of anxiety disorders.

Adult Neurogenesis Conserves Hippocampal Memory Capacity.

The hippocampus is crucial for declarative memories in humans and encodes episodic and spatial memories in animals. Memory coding strengthens synaptic efficacy via an LTP-like mechanism. Given that animals store memories of everyday experiences, the hippocampal circuit must have a mechanism that prevents saturation of overall synaptic weight for the preservation of learning capacity. LTD works to balance plasticity and prevent saturation. In addition, adult neurogenesis in the hippocampus is proposed to be involved in the down-scaling of synaptic efficacy. Here, we show that adult neurogenesis in male rats plays a crucial role in the maintenance of hippocampal capacity for memory (learning and/or memory formation). Neurogenesis regulated the maintenance of LTP, with decreases and increases in neurogenesis prolonging or shortening LTP persistence, respectively. Artificial saturation of hippocampal LTP impaired memory capacity in contextual fear conditioning, which completely recovered after 14 d, which was the time required for LTP to decay to the basal level. Memory capacity gradually recovered in parallel with neurogenesis-mediated gradual decay of LTP. Ablation of neurogenesis by x-ray irradiation delayed the recovery of memory capacity, whereas enhancement of neurogenesis using a running wheel sped up recovery. Therefore, one benefit of ongoing adult neurogenesis is the maintenance of hippocampal memory capacity through homeostatic renewing of hippocampal memory circuits. Decreased neurogenesis in aged animals may be responsible for the decline in cognitive function with age.SIGNIFICANCE STATEMENT Learning many events each day increases synaptic efficacy via LTP, which can prevent the storage of new memories in the hippocampal circuit. In this study, we demonstrate that hippocampal capacity for the storage of new memories is maintained by ongoing adult neurogenesis through homoeostatic renewing of hippocampal circuits in rats. A decrease or an increase in neurogenesis, respectively, delayed or sped up the recovery of memory capacity, suggesting that hippocampal adult neurogenesis plays a critical role in reducing LTP saturation and keeps the gate open for new memories by clearing out the old memories from the hippocampal memory circuit.

Behavioral, cellular, and synaptic tagging frameworks.

Behavioral tagging is the transformation of a short-term memory induced by a weak experience into a long-term memory through temporal association with a novel experience. This phenomenon was discovered to recapitulate synaptic tagging and capture at the behavioral level. Significant progress has been made in determining the molecular machinery associated with synaptic tagging and capture and behavioral tagging theories. However, the tag setting and recruitment of plasticity-related proteins that occur within the spatiotemporally constrained cell ensemble at the network level (cellular tagging) in the brain where multimodal sensory information is input are just beginning to be understood. Here, we review the evidence for behavioral tagging and the mechanism underlying memory allocation at the network level leading to the overlap of cell ensembles. We also discuss the functional significance of overlapping cell ensembles in association of standard Pavlovian conditioning and distinct memories. Finally, we describe the role of neuronal ensemble overlap in behavioral tagging.

Cofilin1 controls transcolumnar plasticity in dendritic spines in adult barrel cortex.

During sensory deprivation, the barrel cortex undergoes expansion of a functional column representing spared inputs (spared column), into the neighboring deprived columns (representing deprived inputs) which are in turn shrunk. As a result, the neurons in a deprived column simultaneously increase and decrease their responses to spared and deprived inputs, respectively. Previous studies revealed that dendritic spines are remodeled during this barrel map plasticity. Because cofilin1, a predominant regulator of actin filament turnover, governs both the expansion and shrinkage of the dendritic spine structure in vitro, it hypothetically regulates both responses in barrel map plasticity. However, this hypothesis remains untested. Using lentiviral vectors, we knocked down cofilin1 locally within layer 2/3 neurons in a deprived column. Cofilin1-knocked-down neurons were optogenetically labeled using channelrhodopsin-2, and electrophysiological recordings were targeted to these knocked-down neurons. We showed that cofilin1 knockdown impaired response increases to spared inputs but preserved response decreases to deprived inputs, indicating that cofilin1 dependency is dissociated in these two types of barrel map plasticity. To explore the structural basis of this dissociation, we then analyzed spine densities on deprived column dendritic branches, which were supposed to receive dense horizontal transcolumnar projections from the spared column. We found that spine number increased in a cofilin1-dependent manner selectively in the distal part of the supragranular layer, where most of the transcolumnar projections existed. Our findings suggest that cofilin1-mediated actin dynamics regulate functional map plasticity in an input-specific manner through the dendritic spine remodeling that occurs in the horizontal transcolumnar circuits. These new mechanistic insights into transcolumnar plasticity in adult rats may have a general significance for understanding reorganization of neocortical circuits that have more sophisticated columnar organization than the rodent neocortex, such as the primate neocortex.

A Lognormal Recurrent Network Model for Burst Generation during Hippocampal Sharp Waves.

The strength of cortical synapses distributes lognormally, with a long tail of strong synapses. Various properties of neuronal activity, such as the average firing rates of neurons, the rate and magnitude of spike bursts, the magnitude of population synchrony, and the correlations between presynaptic and postsynaptic spikes, also obey lognormal-like distributions reported in the rodent hippocampal CA1 and CA3 areas. Theoretical models have demonstrated how such a firing rate distribution emerges from neural network dynamics. However, how the other properties also display lognormal patterns remain unknown. Because these features are likely to originate from neural dynamics in CA3, we model a recurrent neural network with the weights of recurrent excitatory connections distributed lognormally to explore the underlying mechanisms and their functional implications. Using multi-timescale adaptive threshold neurons, we construct a low-frequency spontaneous firing state of bursty neurons. This state well replicates the observed statistical properties of population synchrony in hippocampal pyramidal cells. Our results show that the lognormal distribution of synaptic weights consistently accounts for the observed long-tailed features of hippocampal activity. Furthermore, our model demonstrates that bursts spread over the lognormal network much more effectively than single spikes, implying an advantage of spike bursts in information transfer. This efficiency in burst propagation is not found in neural network models with Gaussian-weighted recurrent excitatory synapses. Our model proposes a potential network mechanism to generate sharp waves in CA3 and associated ripples in CA1 because bursts occur in CA3 pyramidal neurons most frequently during sharp waves.

Does autophagy work in synaptic plasticity and memory?

Many studies have reported the roles played by regulated proteolysis in neural plasticity and memory. Within this context, most of the research focused on the ubiquitin-proteasome system and the endosome-lysosome system while giving lesser consideration to another major protein degradation system, namely, autophagy. Although autophagy intersects with many of the pathways known to underlie synaptic plasticity and memory, only few reports related autophagy to synaptic remodeling. These pathways include PI3K-mTOR pathway and endosome-dependent proteolysis. In this review, we will discuss several lines of evidence supporting a physiological role of autophagy in memory processes, and the possible mechanistic scenarios for how autophagy could fulfill this function.

Sleep and quiet wakefulness signify an idling brain hub for creative insights.

Long-term potentiation of synaptic strength is a fundamental aspect of learning and memory. Memories are believed to be stored within specific populations of neurons known as engram cells, which are subsequently reactivated during sleep, facilitating the consolidation of stored information. However, sleep and offline reactivations are associated not only with past experiences but also with anticipation of future events. During periods of offline reactivation, which occur during sleep and quiet wakefulness, the brain exhibits a capability to form novel connections. This process links various past experiences, often leading to the emergence of qualitatively new information that was not initially available. Brain activity during sleep and quiet wakefulness is referred to as the 'idling brain'. Idling brain activity is believed to play a pivotal role in abstracting essential information, comprehending underlying rules, generating creative ideas and fostering insightful thoughts. In this review, we will explore the current state of research and future directions in understanding how sleep and idling brain activity are interconnected with various cognitive functions, especially creative insights. These insights have profound implications for our daily lives, impacting our ability to process information, make decisions and navigate complex situations effectively. This article is part of a discussion meeting issue 'Long-term potentiation: 50 years on'.

Prefrontal coding of learned and inferred knowledge during REM and NREM sleep.

Idling brain activity has been proposed to facilitate inference, insight, and innovative problem-solving. However, it remains unclear how and when the idling brain can create novel ideas. Here, we show that cortical offline activity is both necessary and sufficient for building unlearned inferential knowledge from previously acquired information. In a transitive inference paradigm, male C57BL/6J mice gained the inference 1 day after, but not shortly after, complete training. Inhibiting the neuronal computations in the anterior cingulate cortex (ACC) during post-learning either non-rapid eye movement (NREM) or rapid eye movement (REM) sleep, but not wakefulness, disrupted the inference without affecting the learned knowledge. In vivo Ca2+ imaging suggests that NREM sleep organizes the scattered learned knowledge in a complete hierarchy, while REM sleep computes the inferential information from the organized hierarchy. Furthermore, after insufficient learning, artificial activation of medial entorhinal cortex-ACC dialog during only REM sleep created inferential knowledge. Collectively, our study provides a mechanistic insight on NREM and REM coordination in weaving inferential knowledge, thus highlighting the power of idling brain in cognitive flexibility.

Neuronal stimulation induces autophagy in hippocampal neurons that is involved in AMPA receptor degradation after chemical long-term depression.

Many studies have reported the roles played by regulated proteolysis in synaptic plasticity and memory, but the role of autophagy in neurons remains unclear. In mammalian cells, autophagy functions in the clearance of long-lived proteins and organelles and in adaptation to starvation. In neurons, although autophagy-related proteins (ATGs) are highly expressed, autophagic activity markers, autophagosome (AP) number, and light chain protein 3-II (LC3-II) are low compared with other cell types. In contrast, conditional knock-out of ATG5 or ATG7 in mouse brain causes neurodegeneration and behavioral deficits. Therefore, this study aimed to test whether autophagy is especially regulated in neurons to adapt to brain functions. In cultured rat hippocampal neurons, we found that KCl depolarization transiently increased LC3-II and AP number, which was partially inhibited with APV, an NMDA receptor (NMDAR) inhibitor. Brief low-dose NMDA, a model of chemical long-term depression (chem-LTD), increased LC3-II with a time course coincident with Akt and mammalian target of rapamycin (mTOR) dephosphorylation and degradation of GluR1, an AMPA receptor (AMPAR) subunit. Downstream of NMDAR, the protein phosphatase 1 inhibitor okadaic acid, PTEN inhibitor bpV(HOpic), autophagy inhibitor wortmannin, and short hairpin RNA-mediated knockdown of ATG7 blocked chem-LTD-induced autophagy and partially recovered GluR1 levels. After chem-LTD, GFP-LC3 puncta increased in spines and in dendrites when AP-lysosome fusion was blocked. These results indicate that neuronal stimulation induces NMDAR-dependent autophagy through PI3K-Akt-mTOR pathway inhibition, which may function in AMPAR degradation, thus suggesting autophagy as a contributor to NMDAR-dependent synaptic plasticity and brain functions.

Memory reactivations during sleep.

Neuronal activities that occur during awake periods are often reactivated again during sleep, to consolidate recently encoded memories, a process known as consolidation. In recent years, advanced tools, specially optical techniques and in-vivo live Ca2+ imaging, have revealed a deeper understanding to the offline periods' neuronal activities and their correspondence to later awake behavioral outputs. Recently, there is a growing consensus that sleep is more of an active process. Sleep has been associated with various functions, memory updating, future imaginations of possible familiar scenarios, decision making and planning by replaying past memories. Also, boosting insightful thoughts, creative thinking and problem solving by forming new associations and connections that were not present in awake states. Sleep activities have been directly associated with many "EUREKA" or "AHA" moments. Here, we describe recent views on memory reactivations during sleep and their implications on learning and memory.

Requirement for hippocampal CA3 NMDA receptors in artificial association of memory events stored in CA3 cell ensembles.

The N-methyl-D-aspartate receptors (NRs) in hippocampal CA3 are crucial for the synaptic transmission and plasticity within the CA3 recurrent circuit, which supports the hippocampal functions, such as pattern completion, and reverberatory association of sensory inputs. Previous study showed that synchronous activation of distinct cell populations in CA3, which correspond to distinct events, associated independent events, suggesting that the recurrent circuit expressing NRs in CA3 mediates the artificial association of memory events stored in CA3 ensembles. However, it is still unclear whether CA3 NRs are crucial for the artificial association of memory events stored in the CA3 ensembles. Here we report that the triple transgenic mice (cfos-tTA/KA1-Cre/NR1 flox/flox), which specifically lack NRs in the CA3 cell ensembles, showed impairment in artificial association between two events, which in control mice triggered artificial association. This result indicates that NRs in the hippocampal CA3 are required for the artificial association of memory events stored in the CA3 cell ensembles.

Suppression of a neocortical potassium channel activity by intracellular amyloid-ß and its rescue with Homer1a.

It is proposed that intracellular amyloid-ß (Aß), before extracellular plaque formation, triggers cognitive deficits in Alzheimer disease (AD). Here we report how intracellular Aß affects neuronal properties. This was done by injecting Aß protein into rat and mouse neocortical pyramidal cells through whole-cell patch pipettes and by using 3xTg AD model mice, in which intracellular Aß is accumulated innately. In rats, intracellular application of a mixed Aß(1-42) preparation containing both oligomers and monomers, but not a monomeric preparation of Aß(1-40), broadened spike width and augmented Ca(2+) influx via voltage-dependent Ca(2+) channels in neocortical neurons. Both effects were mimicked and occluded by charybdotoxin, a blocker of large-conductance Ca(2+)-activated K(+) (BK) channels, and blocked by isopimaric acid, a BK channel opener. Surprisingly, augmented Ca(2+) influx was caused by elongated spike duration, but not attributable to direct Ca(2+) channel modulation by Aß(1-42). The Aß(1-42)-induced spike broadening was blocked by electroconvulsive shock (ECS), which we previously showed to facilitate BK channel opening via expression of the scaffold protein Homer1a. In young 3xTg and wild mice, we confirmed spike broadening by Aß(1-42), which was again mimicked and occluded by charybdotoxin and blocked by ECS. In Homer1a knock-out mice, ECS failed to block the Aß(1-42) effect. Single-channel recording on BK channels supported these results. These findings suggest that the suppression of BK channels by intracellular Aß(1-42) is a possible key mechanism for early dysfunction in the AD brain, which may be counteracted by activity-dependent expression of Homer1a.

The locus coeruleus as a regulator of memory linking.

A core function of our memory system is the ability to link appropriate events. In this issue of Neuron, Chowdhury et al. (2022) pinpoint a specific neuromodulatory circuit for contextual memory linking, but not formation, in the hippocampus.

A short-term memory trace persists for days in the mouse hippocampus.

Active recall of short-term memory (STM) is known to last for a few hours, but whether STM has long-term functions is unknown. Here we show that STM can be optogenetically retrieved at a time point during which natural recall is not possible, uncovering the long-term existence of an STM engram. Moreover, re-training within 3 days led to natural long-term recall, indicating facilitated consolidation. Inhibiting offline CA1 activity during non-rapid eye movement (NREM) sleep, N-methyl-D-aspartate receptor (NMDAR) activity, or protein synthesis after first exposure to the STM-forming event impaired the future re-exposure-facilitated consolidation, which highlights a role of protein synthesis, NMDAR and NREM sleep in the long-term storage of an STM trace. These results provide evidence that STM is not completely lost within hours and demonstrates a possible two-step STM consolidation, first long-term storage as a behaviorally inactive engram, then transformation into an active state by recurrence within 3 days.

Hippocampus as a sorter and reverberatory integrator of sensory inputs.

The hippocampus must be capable of sorting and integrating multiple sensory inputs separately but simultaneously. However, it remains to be elucidated how the hippocampus executes these processes simultaneously during learning. Here we found that synchrony between conditioned stimulus (CS)-, unconditioned stimulus (US)- and future retrieval-responsible cells occurs in the CA1 during the reverberatory phase that emerges after sensory inputs have ceased, but not during CS and US inputs. Mutant mice lacking N-methyl-D-aspartate receptors (NRs) in CA3 showed a cued-fear memory impairment and a decrease in synchronized reverberatory activities between CS- and US-responsive CA1 cells. Optogenetic CA3 silencing at the reverberatory phase during learning impaired cued-fear memory. Thus, the hippocampus uses reverberatory activity to link CS and US inputs, and avoid crosstalk during sensory inputs.

A cortical cell ensemble in the posterior parietal cortex controls past experience-dependent memory updating.

When processing current sensory inputs, animals refer to related past experiences. Current information is then incorporated into the related neural network to update previously stored memories. However, the neuronal mechanism underlying the impact of memories of prior experiences on current learning is not well understood. Here, we found that a cellular ensemble in the posterior parietal cortex (PPC) that is activated during past experience mediates an interaction between past and current information to update memory through a PPC-anterior cingulate cortex circuit in mice. Moreover, optogenetic silencing of the PPC ensemble immediately after retrieval dissociated the interaction without affecting individual memories stored in the hippocampus and amygdala. Thus, a specific subpopulation of PPC cells represents past information and instructs downstream brain regions to update previous memories.

Sevoflurane-induced amnesia is associated with inhibition of hippocampal cell ensemble activity after learning.

General anesthesia could induce amnesia, however the mechanism remains unclear. We hypothesized that suppression of neuronal ensemble activity in the hippocampus by anesthesia during the post-learning period causes retrograde amnesia. To test this hypothesis, two experiments were conducted with sevoflurane anesthesia (2.5%, 30 min): a hippocampus-dependent memory task, the context pre-exposure facilitation effect (CPFE) procedure to measure memory function and in vivo calcium imaging to observe neural activity in hippocampal CA1 during context exploration and sevoflurane/home cage session. Sevoflurane treatment just after context pre-exposure session impaired the CPFE memory, suggesting sevoflurane induced retrograde amnesia. Calcium imaging showed sevoflurane treatment prevented neuronal activity in CA1. Further analysis of neuronal activity with non-negative matrix factorization, which extracts neural ensemble activity based on synchronous activity, showed that sevoflurane treatment reduced the reactivation of neuronal ensembles between during context exploration just before and one day after sevoflurane inhalation. These results suggest that sevoflurane treatment immediately after learning induces amnesia, resulting from suppression of reactivation of neuronal ensembles.