Transcriptome and epigenome analysis of engram cells: Next-generation sequencing technologies in memory research

Neuroscience and Biobehavioral Reviews 127 (2021) 865–875 Available online 5 June 2021 0149-7634/© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Transcriptome and epigenome analysis of engram cells: Next-generation sequencing technologies in memory research Miguel Fuentes-Ramos, Marta Alaiz-Noya, Angel Barco * Instituto de Neurociencias, Universidad Miguel Hern´ andez – Consejo Superior de Investigaciones Científicas, Av. Santiago Ram´ on y Cajal s/n, Sant Joan d’Alacant, 03550, Alicante, Spain ARTICLE INFO Keywords: Memory Epigenetics Transcriptomics Engram cells Next-generation sequencing Labeling of active neurons Multiomic analysis Chromatin modifications ABSTRACT Transcription and epigenetic changes are integral components of the neuronal response to stimulation and have been postulated to be drivers or substrates for enduring changes in animal behavior, including learning and memory. Memories are thought to be deposited in neuronal assemblies called engrams, i.e., groups of cells that undergo persistent physical or chemical changes during learning and are selectively reactivated to retrieve the memory. Despite the research progress made in recent years, the identity of specific epigenetic changes, if any, that occur in these cells and subsequently contribute to the persistence of memory traces remains unknown. The analysis of these changes is challenging due to the difficulty of exploring molecular alterations that only occur in a relatively small percentage of cells embedded in a complex tissue. In this review, we discuss the recent advances in this field and the promise of next-generation sequencing (NGS) and epigenome editing methods for overcoming these challenges and address long-standing questions concerning the role of epigenetic mechanisms in memory encoding, maintenance and expression. 1. On memory traces, engram cells and epigenetic memory Memory trace refers to the representation of a memory in the brain due to a series of biological changes produced by a learning experience. In 1904, the German zoologist and evolutionary biologist Richard Semon introduced the term “engram” as "… the enduring though primarily latent modification in the irritable substance produced by a stimulus…" (Semon, 1921) to describe memory representations with a more precise scientific nomenclature (Schacter et al., 1978). Years later, Donald O. Hebb hypothesized that groups of interconnected neurons are activated together during learning and that their recurrent and simultaneous activation produces a strengthening of the synapse between them, which leads to the establishment of long-term memory (Hebb, 1949; Josselyn et al., 2017). More recently, experiments aimed at identifying and manipulating cells activated by a specific experience supported the notion that memories are deposited in discrete neuronal assemblies called engrams (Han et al., 2009; Liu et al., 2012; Roy et al., 2016). The cells in these assemblies are characterized by being activated during a certain learning experience, which produces a series of physical and/or chemical changes that contribute to their reactivation when some aspects of the previous experience are presented, thereby evoking memory retrieval. The current models propose that the cellular assemblies that participate in the encoding of a given experience (or memory acquisition) undergo a molecular process of consolidation that leads to long-term memory storage (Fig. 1) (Kitamura et al., 2017; Vetere et al., 2011). The memory trace subsequently remains in a latent state until it is reactivated by a natural stimulus that evokes that experience, re-entering the active state (Reijmers et al., 2007). This framework, based on decades of investigation, has been vital for designing experiments to prove the existence of engrams and undertake the characterization of engram cells at different levels. However, the exact molecular changes that occur in engram cells and underlie memory encoding, consolidation and long-term maintenance have not been fully elucidated. Epigenetic mechanisms regulate gene expression through chemical or conformational changes in the chromatin that do not alter the underlying DNA sequence. These mechanisms include covalent changes in the DNA and histone proteins, and longand short-range chromatin interactions and remodeling (Kouzarides, 2007). Epigenetic changes are regulated by a plethora of proteins that include "writers" that add specific epigenetic modifications, "erasers" that remove them and "readers" that recognize them and bring other activities to these locations. The * Corresponding author. E-mail address: [email protected] (A. Barco). Contents lists available at ScienceDirect Neuroscience and Biobehavioral Reviews journal homepage: www.elsevier.com/locate/neubiorev https://doi.org/10.1016/j.neubiorev.2021.06.010 Received 26 March 2021; Received in revised form 2 June 2021; Accepted 3 June 2021 Neuroscience and Biobehavioral Reviews 127 (2021) 865–875 866 combinations of epigenetic marks constitutes an “epigenetic language” (Gardner et al., 2011) according to which distinct gene states, such as transcriptionally active genes, silenced genes or primed genes (i.e., silenced but poised for expression), display specific patterns that we are only starting to elucidate (Allis and Jenuwein, 2016). Although epigenetic regulation is dynamic, it enables the stable propagation of changes in gene expression so that the change in the transcriptional status of the gene can persist long after the stimulus that triggered it disappeared, which is often described as epigenetic memory. More recently, the concept of epigenetic memory has been extended to neuroscience, where it has been suggested that the formation of memory, which requires the activation of tightly regulated transcriptional programs by learning, may be driven in part by epigenetic mechanisms (Fischer, 2014; Lopez-Atalaya and Barco, 2014). However, the existence of an epigenetic mnemonic code remains a matter of speculation (Ripoli, 2017). 2. Tackling the engram: from experience to engram cells The contribution of epigenetic alterations to learning and memory has been intensively investigated during the last two decades (Campbell and Wood, 2019; Fagiolini et al., 2009; Kyrke-Smith and Williams, 2018; Lopez-Atalaya and Barco, 2014; Sweatt, 2013). Although it is clear at present that epigenetic mechanisms, such as DNA methylation and various forms of histone posttranslational modification (hPTM), importantly contribute to mnemonic processes (Alarc´ on et al., 2004; Guan et al., 2009; Levenson et al., 2004; Miller et al., 2010, 2008; Miller and Sweatt, 2007), there is still no firm evidence linking specific epigenetic modifications to the creation, maintenance or decline of a memory trace. A recurrent explanation for this lack of evidence is that the changes would only occur in a relatively small percentage of cells embedded in a tissue of high complexity and cellular heterogeneity (i.e., the engram cells) and may therefore go undetected. In this review, we summarize the more prominent contributions of novel next-generation sequencing (NGS)-based technologies to the description of the molecular events that occur during memory encoding, consolidation and recall. We classify these studies into four categories according to the specificity of the genetic material analyzed and the stringency in the definition of “engram cells” (Fig. 2A): (i) Investigations of transcriptional and epigenomic signatures in bulk tissue of a brain region involved in memory, typically the hippocampus, amygdala or prefrontal cortex. (ii) The analysis of memory-related brain regions has sometimes been refined with protocols that enabled the specific isolation of the most relevant cell type (or their chromatin) in that tissue, separating them from other cell types in the tissue (e.g., neurons or a broad category of neurons, such as excitatory neurons). (iii) Description of the transcriptional and epigenomic signatures in the cells activated in a certain brain region during a specific behavioral experience. We refer to these cells as “experience” cells. Arguably, this group of cells includes engram cells but not exclusively those cells. Within the cells activated by a given experience, we will have groups of cells involved in encoding specific components of the memory for that experience, as well as other cells whose functional state may be unrelated or related to a different component of the same experience (Sweis et al., 2021; Tanaka et al., 2018). (iv) Investigations of the transcriptomic and epigenetic changes in engram cells – i.e., the cells activated during a given experience/ context and reactivated when the animal is re-exposed to the same experience/context – suggesting that the cells are part of the neuronal encoding for that experience or context. Hundreds of studies fall in the first category, but far fewer in the second one. Moreover, to date, only a small number of studies have investigated the transcriptional and epigenetics changes occurring in “experience” cells, and only two of them considered the “reactivation” criterion to obtain an enrichment in engram cells (Table 1). Despite recent and noteworthy attempts to describe transcriptomic and epigenetic changes in memory traces, researchers have not yet linked specific epigenetic changes to memory persistence (Asok et al., 2019). However, NGS-based technologies are reaching the required level of sensitivity and resolution to address this challenge and elucidate the epigenomic changes, if any, associated with different memory processes (Fig. 2B). In the next section, we introduce the various methods available to label Fig. 1. Memory phases: In psychology, memory is often broken into three stages: encoding, storage and retrieval, although the duration of these phases may vary depending on the memory paradigm or the context of discussion. Memory encoding or acquisition (left) is the first phase in which new information is transformed into neural codes. Related to encoding, the term memory allocation refers to the process by which neurons (grey triangles) are chosen to be part of the engram (dashed lines). Memory storage or consolidation (center) is the post-acquisition phase in which memory stabilization occurs at different levels of brain organization. Different biological processes take place in the brain that can last from hours to days. On some occasions, the long duration of this phase has led to the distinction between early and late consolidation. Finally, memory recall or retrieval (right) is the process of remembering or retrieving information that has been previously stored. M. Fuentes-Ramos et al. Neuroscience and Biobehavioral Reviews 127 (2021) 865–875 867 active neurons. Next, we review how these methods were used in increasingly sophisticated attempts to achieve the bold objective outlined above. Finally, we will discuss the prospects of NGS methodologies and novel strategies for epigenetic engineering to further characterize the biology of engram cells and the composition of memory traces. 3. Identifying and tagging active neurons Behavioral experiences, such as the exploration of a novel environment, learning a new task or forming an associative aversive memory, are known to trigger the induction of immediate early genes (IEGs), such as Fos and Arc, in sparse neuronal assemblies (Guzowski et al., 2005; Yap and Greenberg, 2018). As a result, the detection of IEG products, mRNAs and proteins by fluorescence in situ hybridization (FISH) and immunohistochemistry (IHC) has been used as the earliest and most approachable means of labeling and studying neurons activated by experience and thus potentially implicated in memory encoding (DeNardo and Luo, 2017). The expression of IEGs is transient by nature and therefore offers limited possibilities to assess later memory processes such as testing whether the cells that expressed those markers in response to an experience are later reactivated during its retrieval. However, a variation of these methods, referred to as cellular compartment analysis of temporal Fig. 2. Different approaches aimed to explore transcriptional and epigenetic changes evoked by experience. A. Most of the studies conducted so far to retrieve transcriptome and epigenome changes involved in learning and memory have used whole brain regions or, in few cases, isolated neurons (left). However, the refinement of NGS technologies and the development of more efficient genetic tools for tagging activated cells have made possible in recent years the investigation of transcriptional and epigenomic signatures in cells activated by an experience (centre panel, first experience cells in red) and even in neurons tagged during learning and reactivated during retrieval (right panel, reactivated cells in yellow), which are likely part of the engram for that experience (red cells connected by a dashed line). B. Representative methods for transcriptome and epigenome analysis. The catalogue of methodologies includes (i) chromatin interactions profiling through high-throughput chromosome conformation capture (Hi-C), promoter capture Hi-C (pcHi-C) and chromatin interaction analysis by paired-end tag sequencing (ChIAPET); (ii) chromatin accessibility through assay for transposase accessible chromatin with high-throughput sequencing (ATAC-seq) and transposome hypersensitive site sequencing (THS-seq); (iii) protein-DNA interaction analysis through chromatin immunoprecipitation (ChIP)-seq, chromatin endogenous cleavage with highthroughput sequencing (ChEC-seq), cleavage under targets and release using nuclease (CUT&RUN) and cleavage under targets and tagmentation (CUT&Tag); (iv) transcriptome analysis by RNA sequencing (RNA-seq), nuclear (nu)RNA-seq, ribosomal RNA immunoprecipitation sequencing (riboRNA-seq), single-nuclear (sn) RNA-seq; and (v) DNA methylation profiling through bisulfite sequencing (BS-seq), whole-genome bisulfite sequencing (WGBS) and single-nucleus reduced representation bisulfite sequencing (snRRBS). Those methods already used in memory research are indicated in red. M. Fuentes-Ramos et al. Neuroscience and Biobehavioral Reviews 127 (2021) 865–875 868 activity by FISH (catFISH), made it possible for the first time to study the induction of IEGs by two behavioral events that occur at different times in a relatively narrow interval (minutes to few hours), such as memory acquisition and early retrieval. This method enables the simultaneous detection of cytoplasmic and nuclear transcripts for specific IEGs (e.g., Arc) in individual cells: while the more stable cytoplasmic mRNA reflects the activity produced by the first behavioral event, the short-life nuclear RNA reflects the response to the second (Guzowski et al., 1999; Sauvage et al., 2019). Of note, for this method to work, the samples need to be collected immediately after the second experience (i. e., retrieval). Neither IEG FISH, IHC, nor even catFISH enable longer-lasting labeling of the cells activated during learning to identify long-term memory engram cells. Such a process would require genetic strategies that enable a delayed or permanent labeling of the cells that expressed the IEGs in a given time window, thereby providing tools for longitudinal analyses and for studying the molecular events related to memory consolidation and long-term storage (Choi et al., 2020). We can classify these approaches based on technical differences, such as the delivery method and the type of promoter used or, more interestingly, based on the duration and features of the label (Fig. 3): Table 1 Studies that used NGS technologies to investigate transcriptional and epigenetic changes in physiologically activated neurons. Region Paradigm NGS methods Strategy to access activated assemblies Highlights Ref. Experience cells DG Novelty exploration snRNA-seq Fos + cells -Transcriptional changes in activated DG cells 1 h after NE. (Lacar et al., 2016) -Pseudotemporal pattern of activation states obtained from gene expression. HPC Novelty exploration snRNA-seq Fos + & Arc + cells -Stronger transcriptional changes in activated DG cells 1 h after NE compared with CA1 pyramidal cells and VIP interneurons. (Jaeger et al., 2018) -Activity-induced transcription persist in DG 5 h after NE. HPC Novelty exploration ATAC-seq Fos + cells -Similar chromatin changes take place in neurons after physiological and chemically-induced activation. (Fernandez-Albert et al., 2019) DG Novelty exploration snRRBS snRNA-seq Fos + cells -Differentially methylation regions locate into gene bodies. (Odell et al., 2020) - Differentially methylated regions present intermediate methylation levels TeA FC (memory encoding) RNA-seq Arc-dVenus transgenic mice -Profiling of transcriptional changes in cortical assemblies 6 h after auditory FC. (Cho et al., 2016) DG FC (memory encoding) RNA-seq Arc-dVenus transgenic mice -Differential pattern of gene expression 24 h after FC in cells activated during training. (Rao-Ruiz et al., 2019) mPFC FC (memory consolidation) RNA-seq cFos-RPL22HA transgenic mice -Transcriptional changes in mPFC cells 30 days after FC. (Dillingham et al., 2019) vCA1 Foot shock or female exposure RNA-seq AAV-cFos-tTA & AAV-TRE-EYFP -Transcriptional differences between cells activated after a positive or a negative experience. (Shpokayte et al., 2020) mPFC FC (remote memory) scRNA-seq FosTRAP2 mice - Persistent transcriptional programs in diverse cell populations related with remote memory. (Chen et al., 2020) HPC FC (memory encoding and consolidation) ATAC-seq, pc-HiC, RNAseq ArcTRAP mice -Memory encoding increases enhancer accessibility. (Marco et al., 2020) -Memory consolidation produces chromatin reorganization and promoter-enhancer interactions. Reactivated cells DG Context re-exposure (memory recall) snRNA-seq Fos + & Arc + colocalization -Transcriptome analysis in reactivated cells. (Jaeger et al., 2018) -Reactivated cells show weaker transcriptional response than newly activated cells. HPC FC (memory recall) ATAC-seq, pc-HiC, RNAseq ArcTRAP mice -Memory recall puts enhancers primed during encoding in contact with their promoters and upregulated genes related to local translation in synaptic compartments. (Marco et al., 2020) Dentate Gyrus (DG), Hippocampus (HPC), Temporal association cortex (TeA), Ventral CA1 (vCA1), Medial prefrontal cortex (mPFC), Vasoactive Intestinal Polypeptide (VIP), Fear Conditioning (FC), Novelty exploration (NE). Fig. 3. Memory phases and duration of different activity tags. The main strategies used for activity-dependent cell marking in studies aimed at identifying experience-responding cells (dashed gray box) vary in the labeling duration (gray bars) and the time window for the labeling (green bars). Note that in the IEG-tTA models the slow elimination of doxycycline (dox) from the animal’s body causes that dox removal (dox-OFF) does not result in the immediate labeling of active cells. In contrast, in IEG-creERT2 models, the labeling triggered by 4-hydroxytamoxifen (4−OHT) injection can be faster and more constrained. These technical differences make each strategy more or less appropriate depending on the phases of memory under study (orange boxes). M. Fuentes-Ramos et al. Neuroscience and Biobehavioral Reviews 127 (2021) 865–875 869 3.1. Methods for delivering the reporter construct These methods include (i) the generation of knock-in strains in which reporter or effector genes are introduced into IEGs, (ii) transgenic strains based on the use of activity-regulated promoters, and (iii) transduction with viruses or pseudoviruses based on activity-regulated promoters. The latter strategy is particularly versatile and may potentially provide a better signal-to-noise ratio than transgenic and knock-in strains, but it is spatially limited (DeNardo and Luo, 2017). 3.2. Type of activity-dependent promoters The options available include (i) naturally activity-regulated promoters (i.e., the promoter of IEGs such as Fos, Egr1 or Arc) and (ii) activity-dependent synthetic promoters, in which researchers have combined minimal promoters with regulatory elements from IEGs to produce activity-responsive constructs that are not found in nature, such as the enhanced synaptic-activity responsive element (E-SARE) (Kawashima et al., 2013), or the Robust Activity Marking module (RAM) (Sørensen et al., 2016). An advantage of synthetic promoters is that they are smaller and therefore can be more reliably delivered into specific brain regions through viral vectors. 3.3. Different types of labels for experience-responsive neurons Regarding the duration and time window for the labeling, we can also distinguish two main strategies: (i) those that use IEG promoters to directly drive a reporter gene; and (ii) those that use IEG promoters to drive an effector protein (e.g., a bacterial recombinase or transcriptional transactivator) whose action relies on a secondary construct localized in a different region of the genome (frequently the Rosa26 locus) or in a viral vector, thereby generating a binary system. The primary limitation of the first strategy is that neuronal labeling is transitory, exhibiting a time window of less than a day, and relies on the kinetics of activation of the IEG promoter (DeNardo and Luo, 2017). In contrast, the second strategy features a wider time window and more versatile labeling. Two types of inducible binary genetic switches have been used to label and investigate engram cells: those based on the tetracycline transactivator (tTA) and those based on the creERT2 recombinase. In the first switch, the promoter of an IEG drives the expression of doxycycline (dox)-repressible tTA. When dox is removed from the medium, tTA binds to the tetracycline response element (TRE) and the reporter gene under that element is expressed, tagging activated cells for several days (Reijmers et al., 2007). In the second type of switch, a variant of the Cre recombinase regulated by tamoxifen (TMX), known as creERT2, is expressed under the control of an IEG promoter and can be easily combined with fluorescent reporters to provide permanent access to the cells (Allen et al., 2017; Guenthner et al., 2013). Furthermore, these inducible binary systems have also been combined with chemoand optogenetic tools to manipulate and investigate the function of activated cells in different behavioral contexts (Choi et al., 2020). Overall, these cell labeling methods have allowed the identification and investigation of ensembles of experience-responding cells in many brain regions, such as the amygdala, hippocampus and cortex (DeNardo et al., 2019; Nonaka et al., 2014; Ramirez et al., 2013), enabling their morphology, excitability, connectivity and synaptic plasticity to be studied (Choi et al., 2018; Erwin et al., 2020; Nonaka et al., 2014; Redondo et al., 2014; Roy et al., 2019; Ryan et al., 2015) [for review, see (Josselyn and Tonegawa, 2020)]. However, to date, only a few studies have investigated the transcriptional and epigenetic changes that occur in engram cells during memory encoding, consolidation or maintenance. 4. Epigenetic modifications in memory-related brain tissue and specific cell-types It has been known for more than 50 years that de novo transcription is required for the transition from shortto long-term forms of memory (Benito and Barco, 2010; Yap and Greenberg, 2018). This molecular switch is regulated by resident transcription factors (TFs), such as CREB and SRF, whose activity depends on signal transduction cascades that are initiated at synapses and reaching the neuronal nucleus (Alberini and Kandel, 2015). These TFs are upstream of the aforementioned IEGs used to label experience and engram cells. Some of these IEGs (e.g., Fos and Npas4) encode additional TFs that mediate a second wave of activity-dependent transcription, while others encode synaptic effector molecules that are directly responsible for synaptic rearrangements or strengthening in response to experience (e.g., Arc) (Yap and Greenberg, 2018). More recent studies have shown that neuronal activation also induces specific chromatin modifications that co-occur with behavioral experiences, such as the exploration of a novel environment or different forms of associative learning (Fischer et al., 2007; Levenson et al., 2004; Miller et al., 2008; Swank and Sweatt, 2001). Therefore, it has been proposed that the activation and maintenance of new gene programs in neurons in response to experience relies on the interplay between TF activation and the epigenetic enzymes that interact with these TFs to introduce changes in chromatin, such as DNA methylation, hPTMs and changes in higher-order chromatin organization (Alarc´ on et al., 2004; Miller and Sweatt, 2007; Sando et al., 2012; Zovkic et al., 2014). Taken together, these studies strongly support a crucial role of epigenetic mechanisms in memory processes but fail to identify the specific epigenetic events that underlie the formation of memories in specific neuronal assemblies. Possibly, because the genetic material of the cells that do not participate in the encoding of the memory (likely most of the cells in the tissue) hinders the detection of cell typeand memory-specific changes. Various approaches have been undertaken to overcome this limitation. 4.1. Molecular signature of supraphysiological neuronal activation Electrical stimulation and chemical compounds that produce widespread neuronal activation have been used to investigate epigenetic changes induced by activity, and later these findings have been extrapolated or confirmed in neurons activated by more physiological paradigms. This approach was employed in the genome-wide analysis of DNA methylation (Guo et al., 2011) and chromatin accessibility (Su et al., 2017) conducted in the adult dentate gyrus (DG) after electrical stimulation, or in the multiomic analysis of kainate-activated hippocampal neurons conducted by our lab (Fernandez-Albert et al., 2019). 4.2. Molecular signature in specific cell types Another approach has been to refine the preparation of the genetic material for enhanced specificity. For instance, the nuclear marker NeuN was used to separate neuronal (NeuN + ) from nonneuronal (NeuN − ) cells and examine the induced changes in histone modifications and DNA methylation during memory acquisition and maintenance in the two populations (Halder et al., 2016). The global changes detected in that study were weakly related to differential gene expression, which was interpreted by the authors considering that the modifications were involved in priming the cell for future events. Alternatively, it is possible that the analysis did not have the required sensitivity or specificity to detect memory-specific changes because only a small subset of the NeuN + cells respond to the learning experience. An alternative strategy to enrich the preparation in cells specifically implicated in memory is the use of genetic tools that label a specific cell type. This approach was utilized in our previously referenced study, in which the activation of creERT2 driven by the promoter of calcium/ calmodulin-dependent kinase 2 alpha (Camk2a) led to the expression of a fluorescent reporter (Sun1-GFP) specifically in the nuclear envelope of forebrain excitatory neurons. This strategy was combined with fluorescence activated nuclear sorting (FANS) to isolate the nuclei of excitatory hippocampal neurons and perform a multiomic analysis M. Fuentes-Ramos et al. Neuroscience and Biobehavioral Reviews 127 (2021) 865–875 870 including nuclear (nu)RNA-seq, Assay for Transposase Accessible Chromatin with high-throughput Sequencing (ATAC-seq) and Highthroughput Chromosome conformation capture (Hi-C), which revealed chromatin accessibility changes and long-range interactions that remained long after neuronal stimulation (Fernandez-Albert et al., 2019). Although studies using chromatin obtained from bulk tissue or isolated neurons have identified transcriptome and epigenetic changes related to memory and neuronal activation, they are likely to overlook the precise transcriptional and chromatin state dynamics that take place in sparse cell populations. 5. Transcriptional and epigenetic signatures of experience cells Only a small number of cells in a complex structure, such as the hippocampus, are expected to participate in the encoding of a specific experience or context. This phenomenon makes the application of NGSbased technologies very challenging. Due to the intrinsic difficulties of the task, few studies have yet attempted to describe the transcriptional and epigenetic changes that specifically occur in experience-activated cells (Table 1). The first studies on physiologically activated neurons used Fos immunoreactivity to sort recently activated cells in the DG. Lacar and colleagues compared the transcriptional profiles of DG granule neurons that responded to the exploration of a novel environment (Fos + ) with those of nonactivated (Fos − ) DG cells using single nucleus (sn)RNA-seq, reporting the first individual transcriptional signatures of activation (Lacar et al., 2016). Similarly, Jaeger and collaborators used snRNA-seq to describe a stronger transcriptional response in Fos + DG granule neurons than in CA1 pyramidal neurons and vasoactive intestinal polypeptide (VIP) interneurons (Jaeger et al., 2018). Notably, these researchers reported an early induced transcriptional program that included genes related to epigenetic mechanisms, such as DNA or histone demethylation, supporting the notion that neuronal activation triggers epigenetic changes (Belgrad and Fields, 2018). They also reported a late signature response encompassing effector genes that encode proteins related to neurite growth and the extracellular matrix. These temporally segregated patterns of gene expression are likely to underlie the molecular changes responsible for memory encoding and consolidation. In the already discussed study by Fern´ andez-Albert and colleagues, we compared Fos + and Fos − nuclei of excitatory hippocampal neurons 1 h after exposure to a novel environment and reported abundant changes in chromatin accessibility and TF occupancy in genes related to neuronal plasticity and memory function (Fernandez-Albert et al., 2019). More recently, Odell and collaborators investigated DNA methylation heterogeneity in active (Fos + ) and nonactive (Fos − ) DG granule cells (Odell et al., 2020). They observed that most of the regions differentially methylated in the two populations were enriched in gene bodies and characterized by an intermediate state of methylation, that is, the regions were not fully methylated or demethylated. Based on this observation, they suggested that pre-existing methylation states can influence neuron excitability and thus contribute to eligibility for a coding cell assembly (i.e., to memory allocation (Josselyn and Frankland, 2018)). However, this study cannot rule out the possibility that the methylation differences between activated and nonactivated neurons are due to de novo methylation or demethylation produced by the activation per se, more so considering that the transcriptional program induced by early activity includes genes related to epigenetic mechanisms (Jaeger et al., 2018). All these studies are based on the direct detection of Fos and thereby explore the initial transcriptional and epigenetic responses that follow neuronal stimulation. The analyses are restricted to the changes that occur at the time of activation or shortly after activation due to the highly transient nature of IEG induction, and they do not provide any information regarding the persistence of the changes at later time points. The aforementioned genetic strategies for delayed or permanent labeling of active neurons (Choi et al., 2020; DeNardo and Luo, 2017) enable the analysis of later time points and the investigation of second-wave effector genes (whose induction is slower than for IEGs) and late processes related to memory consolidation. For instance, several studies have used the Arc-dVenus reporter mouse line, in which the Arc promoter drives the expression of dVenus, a short-life destabilized fluorescent reporter. The kinetics of dVenus expression enable transcription kinetics to be monitored for a longer time than Arc itself, but not permanent cellular labeling (Eguchi and Yamaguchi, 2009). Therefore, this system can be used to retrieve changes associated with memory acquisition/encoding or early consolidation. Using these mice and RNA-seq, Rao-Ruiz and colleagues showed strong transcriptional changes in activated DG cells 24 h after fear conditioning (FC) (Rao-Ruiz et al., 2019). Differences in gene expression were also detected in the temporal association cortex 6 h after FC, which may be related to memory recall (Cho et al., 2016). The investigation of enduring or permanent changes in experience cells requires genetic tags whose expression becomes independent of the IEG promoter. One of the most promising approaches to monitor cells activated during a given experience at later time points is the binary system known as targeted recombination in active populations (TRAP), in which the TMX-dependent recombinase creERT2 is expressed under the control of the Arc or Fos promoter (Guenthner et al., 2013). A recent study using ArcTRAP mice did not obtain large transcriptional changes in activated hippocampal neurons (which became eYFP + ) during memory encoding (1.5–2 h after FC) compared to the basal state (Marco et al., 2020), which contrasts with the prominent transcriptional response reported in the previously discussed studies based on Fos immunoreactivity. This discrepancy could be explained by differences in the brain subregion analysed, the paradigm to which the mice are exposed, differences between Fosand Arc-expressing neurons, or to the background recombination reported in the hippocampus of ArcTRAP mice (Guenthner et al., 2013). Notably, despite weak transcriptional changes, this study reported an epigenetic priming mechanism characterized by increased chromatin accessibility at enhancers (Marco et al., 2020). Overall, all the studies described above support the view that during the first 24 h following learning, changes in transcription and chromatin conformation occur in the nucleus of the cells activated by a learning experience. What happens at later time points is however unclear since later changes have been considerably less investigated. Physiology and cell biology studies have shown that systems memory consolidation requires the strengthening of synapses in regions involved in remote memory and the weakening of synapses in regions involved in recent memory, such as the DG (Gulmez Karaca et al., 2021; Kitamura et al., 2017). In light of these studies, the question arises of whether the transcriptional and epigenetic changes reported during the first 24 h after learning are maintained at later time points or whether new changes appear in different brain regions or genomic locations. In the study with ArcTRAP mice introduced above, Marco and colleagues followed their multiomic analysis of changes during memory encoding with RNA-seq, ATAC-seq and promoter capture Hi-C (pcHi-C) screens intended to identify transcriptional and epigenetic changes in learning-activated hippocampal neurons during memory consolidation (5 days after FC). They reported consolidation-specific transcriptional and chromatin accessibility changes and chromatin rearrangements in the form of novel enhancer-promoter interactions (Marco et al., 2020), reinforcing the notion that memory consolidation is a dynamic process during which different waves of gene transcription occur. The TRAP system developed by the laboratory of Liqun Luo and used by Marco and colleagues has been further improved to reduce leakiness and enhance inducibility with the introduction of a second-generation TRAP strain based on the Fos promoter, referred to as Fos-2AiCreERT2 or simply TRAP2 (Allen et al., 2017; DeNardo et al., 2019). Chen and colleagues used FosTRAP2 mice to label cells activated during memory recall (16 days after FC) in the medial prefrontal cortex (mPFC) M. Fuentes-Ramos et al. Neuroscience and Biobehavioral Reviews 127 (2021) 865–875 871 and to study their transcriptional signature days after this experience using scRNA-seq. This study unveiled heterogeneous transcriptional programs specific for different neuronal and non-neuronal cells involved in remote memory consolidation, including astrocytes and microglia (Chen et al., 2020). Interestingly, the study did not observe a significant overlap between the transcriptional profiles obtained shortly after learning and during remote memory, which suggests that memory consolidation requires transcriptional programs that differ from those activated during memory encoding. Similarly, recent analyses in a transgenic strain that expresses a tagged ribosomal protein in a Cre-inducible manner revealed differences in the translatome (i.e., the collection of transcripts engaged in active translation at a given time) of mPFC neurons that were activity-tagged 30 days earlier during a FC task (Dillingham et al., 2019). This study further supports the role of the mPFC in memory encoding and highlights the existence of a specific transcriptional response in learning-activated neurons that persists long after the experience and that may rely on epigenetic mechanisms. Another novel aspect of the transcriptional regulation of cells activated by experience was recently contributed by Shpokayte and colleagues. These researchers used activity-dependent viral constructs based on the inducible tTA system to label activated cells in the ventral CA1 layer (which become YFP + ) after exposure to a positive (female exposure) or negative (foot shock) emotional experience (Shpokayte et al., 2020). Transcriptional differences were found between these two groups of cells, indicating that different sets of cells may exist that respond preferentially to positive or negative emotional valences. In conclusion, in the last several years, there has been considerable progress in the transcriptomic and epigenomic analysis of experienceresponsive cells. Transcriptional changes have been more extensively studied than epigenome changes, but both types of signatures seem to be dynamically regulated during learning, from memory acquisition to consolidation and maintenance. 6. Transcriptional and epigenetic signatures of engram cells If we restrict the definition of engram cells to cells that are activated both by the experience that generated the memory and during the retrieval of that memory (Tonegawa et al., 2018), their transcriptomic and epigenomic analysis become even more challenging than the previously discussed screens in experience cells. It is a particularly arduous task to separate cells related to memory from the rest of the cells detected as active in a given moment whose functionality may be different. This has been attempted by looking at the reactivation of experience cells during the retrieval of that experience. To track cell reactivation, it is necessary to combine two tags for activated cells that either have different kinetics or can be set independently. To date, only two studies have analyzed transcriptional and epigenetic changes specifically in reactivated cells (Table 1). Jaeger and colleagues exploited the faster decay of Fos expression compared to Arc when measured by flow cytometry (Jaeger et al., 2018), following a similar reasoning that in the catFISH method described earlier in this review. The authors exposed the mice to a given context and after 4 h either re-exposed the animals to the same context or to a different one, sacrificing the mice 1 h later. The expression of Arc induced by the first experience was still detected after more than 4 h, whereas the Fos signal induced by the first experience had already decayed. As a result, only those cells activated during the second experience expressed Fos. Although this method strongly constrains the interval between encoding and retrieval, the authors could separate “reactivated” and “newly activated” cells based on the coexpression of Fos and Arc using flow cytometry. As expected, reactivated cells were more abundant in mice that were re-exposed to the same context than those exposed to a new context, pinpointing the existence of an engram for the original context. Furthermore, the authors compared the transcriptome of both populations, describing a specific transcriptional signature for reactivated cells and elaborating a timeline for some IEGs in the different groups of neurons (“reactivated”, “newly activated” and “not reactivated”) within the 5 h period after the first contextual exposure. Notably, the burst of IEG expression was weaker in the reactivated cells than in newly activated cells. To identify reactivated neurons more than a few hours after the learning experience, it is necessary to combine the detection of an IEG, such as Fos, during retrieval with genetic strategies for activitydependent neuronal labeling during memory encoding. Colocalization of both labels enables the identification of reactivated cells that are likely to be part of the engram (Josselyn and Tonegawa, 2020). Consistent with this notion, the aforementioned study by Marco and colleagues using ArcTRAP mice investigated the transcriptome and epigenome of reactivated cells after retrieval of a contextual FC memory formed 5 days earlier (Marco et al., 2020). The multiomic analysis in these neurons showed that some enhancers primed during memory encoding by increased frequency of contacts with their promoters were associated with the positive regulation of genes related to local translation in the synaptic compartments at the later time point, thereby providing a likely mechanism for memory consolidation. These two seminal studies have provided the first screens for changes encompassing “experience” and “reactivation” neurons. However, the studies did not explore whether the reactivated cells were functionally involved in the engram. As a further refinement in the study of “reactivated” cells towards the analysis of bona fide engram cells, one could include the additional criteria of demonstrating in parallel that the reactivation of those cells is sufficient to retrieve memory, whereas their inhibition interferes with memory retrieval. Binary genetic strategies used to label engram cells can also allow deferred access (e.g., for activation or inhibition) to those cells. In particular, optogenetic and chemogenetic manipulation of specific neuronal assemblies has contributed crucially to consolidating engram theory (Roy et al., 2019). Some studies inhibited the cells that are part of the engram during memory retrieval causing amnesia (Denny et al., 2014; Lacagnina et al., 2019), while others were able to retrieve a fear response in a neutral context by optogenetic reactivation of neurons that were activated during fear conditioning (Liu et al., 2012) or even to create a false or artificial associative memory (Lau et al., 2020; Ramirez et al., 2013; Vetere et al., 2019). Therefore, further research should combine omic screens with functional assays to confirm that the reactivated cells analyzed throughout NGS methods play actual roles in memory storage or retrieval by coupling omic analyses with chemoor optogenetic tools. 7. The promise of NGS technologies in engram’s research The development and continuous optimization of methods for transcriptome and epigenome analysis provide the opportunity to address questions that were considered inaccessible until recently. Progress in two key areas is particularly likely to have an impact on engram’s research: 7.1. Low-input methods for epigenome analyses While some NGS-based methods, such as RNA-seq and ATAC-seq, have been successfully scaled down to be used with a low number of cells and are widely used, other methods may require further improvement before they can be applied to the investigation of epigenetic changes in very few or single cells. For instance, improvements are required for chromatin immunoprecipitation (ChIP)-seq, whole-genome bisulfite sequencing (WGBS), and diverse methods to investigate longrange chromatin interactions, such as Hi-C and chromatin interaction analysis by paired-end tag sequencing (ChIA-PET), which still require a relatively large amount of chromatin (obtained from several million cells) for optimal results (Policastro and Zentner, 2018). However, this situation is changing rapidly. For instance, different alternatives to ChIP-seq in which cell fixation is not necessary and/or the amount of input required is reduced have been recently published, such as M. Fuentes-Ramos et al. Neuroscience and Biobehavioral Reviews 127 (2021) 865–875 872 Chromatin Endogenous Cleavage with high-throughput sequencing (ChEC-seq) (Zentner et al., 2015), Cleavage Under Targets and Release Using Nuclease (CUT&RUN) (Skene and Henikoff, 2017) and Cleavage Under Targets and Tagmentation (CUT&Tag) (Kaya-Okur et al., 2019). The CUT&Tag protocol is particularly promising; in this method, a hyperactive Tn5 transposase is fused with protein A (Tn5-pA) and preloaded with sequencing adapters, allowing the mapping of chromatin components with high resolution. This innovation also reduces background levels, time and the cost of the protocol (Kaya-Okur et al., 2019). There are also new methods to investigate long-range chromatin interactions that have a lower input requirement and improved signal-to-noise ratios (Mumbach et al., 2016), which may open the possibility of performing more refined longitudinal analyses investigating chromatin interactions in engram cells. Moreover, improvements in library preparation methods that maximize input recovery after bisulfite conversion (Zhou et al., 2019) render the identification of DNA methylation changes suitable for engram studies. 7.2. Single-cell and single-nucleus analyses Engrams can include other cell types than neurons and are likely to integrate neurons with different functionalities (Sun et al., 2020) in which the epigenetic changes may occur in different or even opposite directions. This possibility makes single-cell (sc) and single-nucleus (sn) RNA-seq particularly important in this field of research. While scRNA-seq enables us to observe cell heterogeneity within the cell assembly, snRNA-seq has emerged as a powerful alternative with important advantages because the analysis of nuclear RNA (nuRNA) provides better temporal resolution than traditional mRNA-seq if the objective is to examine a dynamic transcriptional response. NuRNA-seq analysis also reduces the inherent variability associated with tissue disruption in which different cells can stochastically conserve larger or smaller fractions of the cytoplasm, axon and dendritic processes. Furthermore, nuclei are more resistant to perturbation than dissociated cells, preventing the stress response triggered by tissue disruption from occluding neuronal plasticity-related gene induction (Fernandez-Albert et al., 2019; Stroud et al., 2020). Novel scor sn-NGS methods to explore the epigenome can be combined with scRNA-seq or snRNA-seq to characterize other aspects of gene regulation (Packer and Trapnell, 2018). For instance, chromatin accessibility and occupancy have been studied at the single-cell level using scATAC-seq (Buenrostro et al., 2015; Cusanovich et al., 2015) and sc Transposome Hypersensitive Site sequencing (scTHS-seq) (Lake et al., 2018); scChIP-seq has been used to track the heterogeneity of a cell population through histone marks (Rotem et al., 2015); sc Bisulfite Sequencing (scBS-seq) (Smallwood et al., 2014) has made it possible to obtain DNA methylation profiles from single cells; and sn or scHi-C (Flyamer et al., 2017; Ramani et al., 2017), respectively, has allowed the study of three-dimensional structure of the chromatin in individual cells. However, these techniques require further optimization to work well in conditions of low input, and their applicability remains limited due to the poor coverage compared to techniques based on pooled cells or nuclei. To date, these techniques have not been applied for the study of memory engram studies, but in the coming years, they will become an indispensable tool for the study of the epigenetic changes involved in the formation and persistence of memory. 8. Mimicking and erasing epi-signatures in engram cells The optimization of genetic activity-dependent neuronal labeling tools and the development of novel NGS-based technologies should soon clarify long-standing questions concerning the contribution of epigenetic mechanisms in memory engrams. Genomic screens and multiomic analysis should provide promising candidate regions undergoing remodeling during acquisition and consolidation of memory. Next, candidate regions and mechanisms should be pursued to evaluate whether changes are necessary and/or sufficient to cause transcriptional and functional changes within the cellular assembly. As a first attempt in that direction, a recent study examined the epigenetic manipulation of DNA methylation levels by the overexpression of DNA methyltransferase 3a2 (Dnmt3a2) specifically within DG-allocated cell assemblies, which led to enhanced memory and improved engram reactivation (Gulmez Karaca et al., 2020). However, such a global approach would have a genome-wide impact and would not address the causal role of epigenetic modifications at specific loci. This level of specificity could be achieved through epigenome engineering tools based on the chimeric fusion of epigenetic enzymatic activities with artificial zinc finger proteins (ZFPs), transcription activation-like effectors (TALEs) or the nuclease-deficient (dCas9) clustered regularly interspaced palindromic repeats (CRISPR) system, which can enable locus-specific rewriting of epigenetic information (Yim et al., 2020). The combination of these tools with synthetic transcriptional activator or repressor domains, and with epigenetic effector domains, have been employed to bidirectionally regulate the expression of specific genes in vivo and to explore the behavioral consequences of the manipulation (Bustos et al., 2017; Carpenter et al., 2020; Heller et al., 2014; Kwapis et al., 2018; Zheng et al., 2018). Most Fig. 4. Evaluation of the functional relevance of epigenetic signatures. A learning experience activates a sparse number of cells that become part of an engram (yellow cells). The application of epigenome editing tools – zinc finger proteins (ZFPs), transcription activation-like effectors (TALEs) and the nuclease-deficient Cas9 protein (CRISPR-dCas9) – to engram cells might favour the maintenance or stability of the engram, thereby enhancing memory retrieval (upper right scheme with the solid yellow neurons), or interfere with their reactivation, thereby weakening and even erasing the memory trace (bottom right scheme with the dashed yellow neurons). Epigenome editing tools contains a DNA binding domain (light blue shape) and an effector module (dark blue oval). The connections between engram cells (in yellow) are represented by the dashed lines. M. Fuentes-Ramos et al. Neuroscience and Biobehavioral Reviews 127 (2021) 865–875 873 current efforts have focused on CRISPR-dCas9 due to its versatility, ease of engineering, and cost-effectiveness (Brocken et al., 2018; Xu and Qi, 2019). The protein has been tethered to a number of epigenetic effector proteins to mimic or erase epigenetic marks, emulating the physiological regulation of gene expression (Yim et al., 2020). For example, studies have addressed the manipulation of DNA methylation status, both to erase hypermethylation marks associated with Fragile X syndrome (Liu et al., 2018) and to mimic locus-specific methylation dysregulation related to autism pathogenesis (Lu et al., 2020). The role of histone acetylation in activity-driven transcription (Chen et al., 2019) or neuronal identity (Lipinski et al., 2020) has also been explored, as well as the manipulation of three-dimensional genome folding (Kim et al., 2019; Morgan et al., 2017). However, there are still important barriers to applying these tools in the study of engram cells. First, the combination of CRISPR-dCas9 with epigenetic effectors results in large constructs that exceed the limited packaging capacity of the viral vectors most commonly used in neurosciences (Hamilton et al., 2018). Continued innovations expanding the CRISPR-dCas9 toolbox, such as developing alternative dCas9 proteins with smaller sizes (Xu et al., 2020) or splitting the dCas9 protein into different modules (Truong et al., 2015), reduce the size problems, thereby opening new avenues for neuroepigenome editing in vivo. Second, precise spatial and temporal control is necessary to elucidate the causal function of experience-dependent gene regulation in the sparse population of engram cells. Although epigenome-editing tools have been recently engineered to be expressed in a cre-dependent manner and have been combined with cre-driver transgenic strains for cell specificity (Yim et al., 2020), technical refinement is still necessary to apply this approach to the study of specific neuron assemblies. Further development of these techniques may open the possibility of engineering engrams through manipulation of the neuronal epigenome (Fig. 4). Possible future studies may aim at erasing specific epigenetic marks associated with memory encoding or consolidation, or mimic the changes triggered by the experience. Evaluating the impact of these manipulations on memory expression would definitively confirm the link between experience-triggered epigenetic changes and long-term memory processes. Acknowledgements We thank Jose V. Sanchez-Mut and Felix Leroy for the critical reading of the manuscript and useful comments. M.F.-R. is the recipient of a “Formaci´ on de Profesorado Universitario (FPU)” fellowship from the Spanish Ministry of Education and M.A.-N. is the recipient of an ACIF fellowship from Generalitat Valenciana. A.B. research is supported by grants SAF2017-87928-R and SEV-2013-0317 from MICINN co-financed by ERDF, grant PROMETEO/2020/007 from the Generalitat Valenciana, and grant RGP0039/2017 from the Human Frontiers Science Program Organization (HFSPO). The Instituto de Neurociencias is a “Centre of Excellence Severo Ochoa”. References Alarc´ on, J.M., Malleret, G., Touzani, K., Vronskaya, S., Ishii, S., Kandel, E.R., Barco, A., 2004. Chromatin acetylation, memory, and LTP are impaired in CBP+/- mice: a model for the cognitive deficit in Rubinstein-Taybi syndrome and its amelioration. Neuron 42, 947–959. https://doi.org/10.1016/j.neuron.2004.05.021. Alberini, C.M., Kandel, E.R., 2015. The regulation of transcription in memory consolidation. Cold Spring Harb. Perspect. Biol. 7, a021741 https://doi.org/ 10.1101/cshperspect.a021741. Allen, W.E., DeNardo, L.A., Chen, M.Z., Liu, C.D., Loh, K.M., Fenno, L.E., Ramakrishnan, C., Deisseroth, K., Luo, L., 2017. Thirst-associated preoptic neurons encode an aversive motivational drive. Science 357, 1149–1155. https://doi.org/ 10.1126/science.aan6747. Allis, C.D., Jenuwein, T., 2016. The molecular hallmarks of epigenetic control. Nat. Rev. Genet. 17 (8), 487–500. https://doi.org/10.1038/nrg.2016.59. Asok, A., Leroy, F., Rayman, J.B., Kandel, E.R., 2019. Molecular mechanisms of the memory trace. Trends Neurosci. 42 (1), 14–22. https://doi.org/10.1016/j. tins.2018.10.005. Belgrad, J., Fields, R.D., 2018. Epigenome interactions with patterned neuronal activity. Neuroscientist 24, 471–485. https://doi.org/10.1177/1073858418760744. Benito, E., Barco, A., 2010. CREB’s control of intrinsic and synaptic plasticity: implications for CREB-dependent memory models. Trends Neurosci. 33, 230–240. https://doi.org/10.1016/j.tins.2010.02.001. Brocken, D.J.W., Tark-Dame, M., Dame, R.T., 2018. dCas9: a versatile tool for epigenome editing. Curr. Issues Mol. Biol. 26, 15–32. https://doi.org/10.21775/cimb.026.01. Buenrostro, J.D., Wu, B., Litzenburger, U.M., Ruff, D., Gonzales, M.L., Snyder, M.P., Chang, H.Y., Greenleaf, W.J., 2015. Single-cell chromatin accessibility reveals principles of regulatory variation. Nature 523, 486–490. https://doi.org/10.1038/ nature14590. Bustos, F.J., Ampuero, E., Jury, N., Aguilar, R., Falahi, F., Toledo, J., Ahumada, J., Lata, J., Cubillos, P., Henríquez, B., Guerra, M.V., Stehberg, J., Neve, R.L., Inestrosa, N.C., Wyneken, U., Fuenzalida, M., H¨ artel, S., Sena-Esteves, M., VarelaNallar, L., Rots, M.G., Montecino, M., van Zundert, B., 2017. Epigenetic editing of the Dlg4/PSD95 gene improves cognition in aged and Alzheimer’s disease mice. Brain 140, 3252–3268. https://doi.org/10.1093/brain/awx272. Campbell, R.R., Wood, M.A., 2019. How the epigenome integrates information and reshapes the synapse. Nat. Rev. Neurosci. 20, 133–147. https://doi.org/10.1038/ s41583-019-0121-9. Carpenter, M.D., Hu, Q., Bond, A.M., Lombroso, S.I., Czarnecki, K.S., Lim, C.J., Song, H., Wimmer, M.E., Pierce, R.C., Heller, E.A., 2020. Nr4a1 suppresses cocaine-induced behavior via epigenetic regulation of homeostatic target genes. Nat. Commun. 11, 504. https://doi.org/10.1038/s41467-020-14331-y. Chen, L.-F., Lin, Y.T., Gallegos, D.A., Hazlett, M.F., G´ omez-Schiavon, M., Yang, M.G., Kalmeta, B., Zhou, A.S., Holtzman, L., Gersbach, C.A., Grandl, J., Buchler, N.E., West, A.E., 2019. Enhancer histone acetylation modulates transcriptional bursting dynamics of neuronal activity-inducible genes. Cell Rep. 26, 1174–1188. https://doi. org/10.1016/j.celrep.2019.01.032 e5. Chen, M.B., Jiang, X., Quake, S.R., Südhof, T.C., 2020. Persistent transcriptional programmes are associated with remote memory. Nature 587, 437–442. https://doi. org/10.1038/s41586-020-2905-5. Cho, J.-H., Huang, B.S., Gray, J.M., 2016. RNA sequencing from neural ensembles activated during fear conditioning in the mouse temporal association cortex. Sci. Rep. 6, 31753. https://doi.org/10.1038/srep31753. Choi, J.-H., Sim, S.-E., Kim, J., Choi, D.Il, Oh, J., Ye, S., Lee, J., Kim, T., Ko, H.-G., Lim, C.-S., Kaang, B.-K., 2018. Interregional synaptic maps among engram cells underlie memory formation. Science 360, 430–435. https://doi.org/10.1126/ science.aas9204. Choi, J.-E., Kim, Jiwon, Kim, Jinhyun, 2020. Capturing activated neurons and synapses. Neurosci. Res. 152, 25–34. https://doi.org/10.1016/j.neures.2019.12.020. Cusanovich, D.A., Daza, R., Adey, A., Pliner, H.A., Christiansen, L., Gunderson, K.L., Steemers, F.J., Trapnell, C., Shendure, J., 2015. Multiplex single-cell profiling of chromatin accessibility by combinatorial cellular indexing. Science 348, 910–914. https://doi.org/10.1126/science.aab1601. DeNardo, L., Luo, L., 2017. Genetic strategies to access activated neurons. Curr. Opin. Neurobiol. 45, 121–129. https://doi.org/10.1016/j.conb.2017.05.014. DeNardo, L.A., Liu, C.D., Allen, W.E., Adams, E.L., Friedmann, D., Fu, L., Guenthner, C.J., Tessier-Lavigne, M., Luo, L., 2019. Temporal evolution of cortical ensembles promoting remote memory retrieval. Nat. Neurosci. 22, 460–469. https://doi.org/ 10.1038/s41593-018-0318-7. Denny, C.A., Kheirbek, M.A., Alba, E.L., Tanaka, K.F., Brachman, R.A., Laughman, K.B., Tomm, N.K., Turi, G.F., Losonczy, A., Hen, R., 2014. Hippocampal memory traces are differentially modulated by experience, time, and adult neurogenesis. Neuron 83, 189–201. https://doi.org/10.1016/j.neuron.2014.05.018. Dillingham, B.C., Cameron, P., Pieraut, S., Cardozo, L.M., Yoo, E.J., Maximov, A., Stowers, L., Mayford, M., 2019. Fear learning induces long-lasting changes in gene expression and pathway specific presynaptic growth. bioRxiv. https://doi.org/ 10.1101/571331. Eguchi, M., Yamaguchi, S., 2009. In vivo and in vitro visualization of gene expression dynamics over extensive areas of the brain. Neuroimage 44, 1274–1283. https://doi. org/10.1016/j.neuroimage.2008.10.046. Erwin, S.R., Sun, W., Copeland, M., Lindo, S., Spruston, N., Cembrowski, M.S., 2020. A sparse, spatially biased subtype of mature granule cell dominates recruitment in hippocampal-associated behaviors. Cell Rep. 31, 107551 https://doi.org/10.1016/j. celrep.2020.107551. Fagiolini, M., Jensen, C.L., Champagne, F.A., 2009. Epigenetic influences on brain development and plasticity. Curr. Opin. Neurobiol. 19 (2), 207–212. https://doi.org/ 10.1016/j.conb.2009.05.009. Fernandez-Albert, J., Lipinski, M., Lopez-Cascales, M.T., Rowley, M.J., MartinGonzalez, A.M., del Blanco, B., Corces, V.G., Barco, A., 2019. Immediate and deferred epigenomic signatures of in vivo neuronal activation in mouse hippocampus. Nat. Neurosci. 22, 1718–1730. https://doi.org/10.1038/s41593-0190476-2. Fischer, A., 2014. Epigenetic memory: the Lamarckian brain. EMBO J. 33, 945–967. https://doi.org/10.1002/embj.201387637. Fischer, A., Sananbenesi, F., Wang, X., Dobbin, M., Tsai, L.H., 2007. Recovery of learning and memory is associated with chromatin remodelling. Nature 447 (7141), 178–182. https://doi.org/10.1038/nature05772. Flyamer, I.M., Gassler, J., Imakaev, M., Brand˜ ao, H.B., Ulianov, S.V., Abdennur, N., Razin, S.V., Mirny, L.A., Tachibana-Konwalski, K., 2017. Single-nucleus Hi-C reveals unique chromatin reorganization at oocyte-to-zygote transition. Nature 544, 110–114. https://doi.org/10.1038/nature21711. Gardner, K.E., Allis, C.D., Strahl, B.D., 2011. Operating on chromatin, a colorful language where context matters. J. Mol. Biol. 409, 36–46. https://doi.org/10.1016/j. jmb.2011.01.040. M. Fuentes-Ramos et al.