Reactivating an Engram: Context Matters

Memory exerts a powerful influence over behavior. The ability of past experiences to guide future actions suggests there is an internal representation of those experiences stored in the brain. The term “engram” was introduced by Richard Semon to describe the neurobiological basis of memory storage. More concretely, engrams are defined as “the enduring offline physical and/or chemical changes that were elicited by learning and underlie the newly formed memory associations” (Josselyn and Tonegawa, 2020). They are thought to be activated during initial learning experiences and reactivated upon subsequent exposure to stimuli present during encoding to enable memory retrieval.

Fear conditioning is a powerful paradigm for studying in rodent models the behavioral expression of memory, including the existence and influence of engrams. In fear conditioning, a neutral cue becomes associated with an aversive stimulus over repeated pairings, such that the previously neutral cue starts to elicit the expression of stereotyped freezing behavior. This expression of fear via freezing provides a behavioral read-out, or index, of memory retrieval that is easily measurable.

Recent technical developments have enabled researchers to identify engrams and elucidate their mechanistic role in producing freezing responses. The existence of engram cells was shown by using fluorescent proteins to separately tag neurons active during fear conditioning, and those reactivated during freezing, i.e., fear memory retrieval (Reijmers et al., 2007). Engram cells were defined as neurons expressing both fluorescent markers. Using similar techniques, researchers can express light-gated ion channels specifically in engram populations. This enables them to test the causal role of engram cell activity on memory retrieval, by manipulating the activity of these cells while monitoring behavior. After fear conditioning, inactivating engram cells when the animal is in the fear-eliciting context disrupts freezing (Han et al., 2009), whereas optically stimulating these engram cells in a novel neutral context induces freezing responses (Liu et al., 2012). These findings, and many others (Josselyn and Tonegawa, 2020), cast engram cells as both necessary and sufficient for driving freezing following fear memory recall.

Although many studies use freezing as a behavioral readout of fear recall, the reliability of this readout can be questioned. In general, inferring internal states in rodents is more complex than simply measuring freezing in a fear conditioning paradigm (Bouwknecht and Paylor, 2008). For example, it has been shown that optogenetic reactivation of hippocampal fear engram cells can produce active avoidance responses, like place avoidance, instead of freezing (Ramirez et al., 2013). Similarly, Chen et al. (2019) showed engram reactivation can enhance or reduce freezing depending on the context. Thus, relying solely on freezing to determine whether memory retrieval has occurred may lead to false conclusions. The findings also hint at the intriguing possibility that the brain incorporates contextual cues to produce the most appropriate behavioral response during memory reactivation, rather than being limited to a single, stereotyped response such as freezing.

Dorst et al. (2024) directly tested this hypothesis by investigating the flexibility of the behavioral output of engram reactivation in different environmental contexts. Animals underwent classic contextual fear conditioning (placed in a chamber while receiving foot shocks), during which time activity-dependent neuronal labeling was performed. The use of the tetracycline-controlled transactivator and doxycycline for time control of fluorescent protein expression allowed the authors to specifically tag hippocampal fear engrams cells while animals were in a standard conditioning chamber (Dorst et al., their Fig. 1A). Subsequently, labeled engram cells in the dentate gyrus were optically reactivated in three distinct contextual settings: a small box (SB), with the same dimensions as the original conditioning chamber, a medium box (MB), and a large box (LB). All boxes were distinct from the original conditioning chamber in other sensory details. When the fear engram was reactivated in the SB, the animals displayed robust freezing behavior, replicating previous studies (Liu et al., 2012). Freezing behavior was also observed, although slightly reduced, in the MB during engram reactivation. However, fear engram reactivation in the LB failed to elicit any freezing (Dorst et al., their Fig. 1C,D). The lack of freezing in the LB upon engram reactivation suggests that either retrieval failed to occur, or successful retrieval was manifested through defensive responses other than freezing, because of the distinct environmental context.

The authors speculated that the observed context-dependent freezing may have been driven by the contextual similarity between the SB and original conditioning chamber. To directly test this, a larger conditioning chamber with dimensions matching the LB was used. Surprisingly, when fear engrams tagged in this larger conditioning chamber were reactivated, the animals no longer displayed freezing in any environment (Dorst et al., their Fig. 1E,F). The absence of freezing in the LB in this experiment potentially rules out contextual similarity as the key driver of freezing. Importantly, the authors observed a significant increase in rearing behavior when engram cells were optically reactivated in the LB context (Fig. 4B). Rearing might reflect an escape-seeking response (Lever et al., 2006), which could be considered a defensive behavior akin to freezing but manifesting differently due to the environmental context. The authors propose that hippocampal engram reactivation successfully retrieved the fear memory but that memory retrieval can elicit flexible, context-dependent defensive behaviors beyond mere freezing. This suggests freezing alone is an imperfect read-out of memory retrieval following optical stimulation.

Previous work has shown that artificially reactivating hippocampal engram cells increases c-Fos expression in downstream brain regions involved in learning and affective states (Roy et al., 2022). Therefore Dorst et al. (2024) posited that differences in c-Fos expression across candidate regions might explain the distinct behavioral outputs observed across contexts. They first compared c-Fos density in the reactivated dentate gyrus engram cells, between animals that underwent engram reactivation in the SB versus the LB. In all these animals, fear engrams were originally tagged during conditioning in the smaller conditioning chamber (Dorst et al., their Fig. 5A). For the SB reactivation group, which displayed light-induced freezing, there were significant increases in c-Fos density compared with control animals (that lack ChR2) in the dentate gyrus, CA3, and CA1 subregions of the hippocampus as well as in the ventromedial hypothalamus—areas known to be involved in fear/defensive behaviors and memory (Dorst et al., their Fig. 5F). The LB reactivation group, which did not display light-induced freezing, did not show significant differences in c-Fos density compared with LB control animals (Dorst et al., their Fig. 5G). No differences in dentate gyrus c-Fos density were found between the SB recall and LB recall groups, indicating engram cell reactivation engaged the dentate gyrus similarly in both contexts. Thus, the authors showed that the dentate gyrus recruited downstream brain areas in a context-dependent manner, suggesting reactivation does not consistently produce the same neural activity patterns, which may explain the differences in observed behavioral outputs.

The authors further probed whether reactivating the engram cells to induce freezing produced similar neural ensemble activity as freezing behavior triggered by natural memory retrieval upon re-exposure to the original conditioning context. Interestingly, despite both conditions evoking identical freezing behavior, the c-Fos activity patterns revealed significant differences between the optogenetic reactivation and natural retrieval conditions. These findings provide a critical insight: different patterns of brain activity can produce apparently identical behavioral outputs.

Several important conclusions can be drawn from the work by Dorst et al. (2024). First, activation of a specific engram does not induce a stereotypic, hard-wired behavioral response. Instead, behaviors appear to be selected flexibly to produce the most appropriate response in the current environmental context. The authors interpret these results from an evolutionary perspective: in the case of threat in a confined environment, the most appropriate response may be freezing; but in a larger area, more active anxiety-like or escape behaviors may be appropriate.

Second, Dorst et al. (2024) showed evidence that context influences the pattern of brain activity elicited by recall of an engram. Moreover, both engram reactivation conditions elicited brain activity patterns different from those elicited by natural recall, suggesting that even if the behavioral output is similar, the brain activity that underpins the behavior can be different. Together, these findings provide important information to consider when investigating the biological substrate of memory storage.

An important takeaway from this work is the need to go beyond simplistic behavioral readouts like freezing, to include more nuanced behavioral phenotypes, or even physiological measures, when inferring internal brain states. For example, while mice in the LB do not overtly freeze following engram reactivation, they show dentate gyrus activation patterns and increases in brainwide correlated activity similar to those of mice in the SB. This difference suggests fear engram reactivation may induce a negative affective state in the LB, but this was not detected because the physical manifestations of this state were not measured. By using more advanced methods of behavioral analysis (e.g., motion sequencing) in conjunction with physiological measurements, such as heart rate variability and blood pressure, we may more precisely understand how engram reactivation affects behavior. More in-depth behavioral/physiological analyses should open the door for future studies using ethologically relevant environments, such as those that include a possible escape route or those containing other animals.