The role of the genome in experience-dependent plasticity: Extending the analogy of the genomic action potential

Abstract

Our past experiences shape our current and future behavior. These experiences must leave some enduring imprint on our brains, altering neural circuits that mediate behavior and contributing to our individual differences. As a framework for understanding how experiences might produce lasting changes in neural circuits, Clayton [D. F. Clayton, Neurobiol. Learn. Mem. 74, 185–216 (2000)] introduced the concept of the genomic action potential (gAP)—a structured genomic response in the brain to acute experience. Similar to the familiar electrophysiological action potential (eAP), the gAP also provides a means for integrating afferent patterns of activity but on a slower timescale and with longer-lasting effects. We revisit this concept in light of contemporary work on experience-dependent modification of neural circuits. We review the “Immediate Early Gene” (IEG) response, the starting point for understanding the gAP. We discuss evidence for its involvement in the encoding of experience to long-term memory across time and biological levels of organization ranging from individual cells to cell ensembles and whole organisms. We explore distinctions between memory encoding and homeostatic functions and consider the potential for perpetuation of the imprint of experience through epigenetic mechanisms. We describe a specific example of a gAP in humans linked to individual differences in the response to stress. Finally, we identify key objectives and new tools for continuing research in this area.

Accounts of information processing and integration in the brain often focus on membrane potentials, ion channels, and synaptic signaling. As captured in the classical observation of the electrophysiological action potential (eAP), neurons can integrate signals from many synaptic inputs to generate discrete functional outputs that propagate forward in time. However, the activity of every brain cell is ultimately governed by the genome in the cell’s nucleus, and since the 1980s, it has been clear that the genome responds dynamically to signals important for information processing (1, 2). Across many organismal models of behavior, contexts associated with learning and/or stress have been observed to trigger discrete time-limited pulses of neural gene expression, which analogously, have been termed a genomic action potential (gAP) (3). Mechanistically, the gAP, like the eAP, has cascading consequences that spread across both time and space in the nervous system (Table 1). By altering the landscape of proteins, RNA, and chromatin structure in the cell, a gAP can influence how a cell responds to a subsequent activation event and how it communicates with other cells.

Table 1.

Comparison of the eAP with the gAP

eAP vs. gAP	eAP	gAP
Intracellular integration of inputs	Yes	Yes
Site of integration	Axon initial segment	Cell nucleus
Time course	Milliseconds	Minutes
Refractory phase	Yes	(Yes?)
Intracellular phenotype	Cascading changes in membrane potential	Cascading changes in specific protein amounts
Effects	Increased probability of transmitter release	Synaptic and cellular remodeling, neurotrophin release
Consequence	Propagation of information	Information filtering for long-term encoding

Dynamic Genome in Every Brain Cell

Since the original analogy was formulated in 2000 (3), a wealth of new information has accumulated about epigenetic mechanisms and the ways in which brains adapt to experience. With this new information, we revisit the gAP analogy and extend it further by considering several frames of reference or levels of biological organization. These include (i) the level of individual cells; (ii) the level of neural circuits, where we propose that different gAPs may have distinctive roles in homeostasis vs. engram encoding (Hebbian plasticity); (iii) potential mechanisms propagating the impact of the gAP; and (iv) the potential for gAP mechanisms to contribute to adaptive whole-organism physiology specifically in the context of stress responses. Across all levels, we observe how the gAP may serve effectively as a salience filter, determining which patterns of cellular, neural, or behavioral activity drive lasting functional changes in the organism.

gAP in the Cell: Molecular Elements

The primary genomic response in the brain to immediate experience (the gAP) involves acute time-limited changes in expression of a core set of Immediate Early Genes (IEGs) within individual cells (3, 4). Activation of intracellular signal transduction—in particular, the MAPK (mitogen-activated protein kinase) pathway and CREB (cAMP-responsive element binding protein)—leads to increased transcription of IEGs, such as cFOS, EGR1, and ARC, with their RNA levels peaking after ∼30–60 min followed by a peak of their proteins after ∼90 min (5). IEGs encode a variety of proteins, including transcription factors (e.g., c-Fos, Egr1), components of intracellular signal transduction pathways (e.g., MAPK phosphatases), and epigenetic readers/writers [e.g., DNA methyltransferase 3 (DNMT3)]. In addition to protein coding messenger RNAs (mRNAs), microRNAs and other noncoding RNAs may also participate in the initial genomic response, and the response may include decreases as opposed to increases in some RNAs (6–8). Underlying or correlated with these transcriptional changes are chromatin-based mechanisms of targeted gene regulation, including histone acetylation (9–11), DNA methylation (12), and even introduction of DNA breaks (13). Initial functional consequences may be directed toward modulation of synaptic transmission as suggested for ARC, an IEG-encoded protein that has been shown to depress glutamatergic transmission through endocytosis of synaptic AMPARs (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors) (14, 15) or possibly even by direct intercellular transfer of RNAs as suggested by the recent discovery that ARC can form virus-like capsids that traffic across synapses (16, 17).

Just as neuronal subtypes exhibit characteristic differences in their eAPs, they may also vary in the set of genes that is induced by acute activity. Moreover, even for a single cell type, a given physiological or behavioral context may result in different patterns of spatiotemporal input to the cell, and there is some evidence that this could affect different subsets of genes. Direct evidence for sensitivity to stimulus structure comes from recent studies of cultured neurons, where different temporal patterns of electrical or chemical depolarization were found to trigger different gene response patterns (18, 19). Within the brain, where neurons are synaptically connected, gAPs induced by somatic calcium influx (i.e., neuronal firing) may differ from ones driven by localized dendritic activity (see below). Interestingly, a recent study of long-term potentiation in the rat hippocampus demonstrated that some memories can be formed in the absence of somatic cell firing altogether (20); in this case, the gAP and associated memory processes may proceed independently from the eAP. Thus, the gAP can be viewed as a “salience filter,” integrating a variety of different cellular inputs, which on exceeding a threshold level (of salience), trigger the induction of the gAP.

Although the gAP has been described as an “all or none” response within an individual cell (3), there are several factors that confound a simple interpretation of this analogy, especially when considering populations of cells. IEG responses are typically measured as aggregate activity in extracts or sections of complex multicellular brain tissue (e.g., analogous to electrophysiological field potential or functional MRI signals). Different response magnitudes could reflect either differential recruitment of cell subpopulations or differential activation levels within individual cells (ref. 21 is an early example where differential recruitment was documented). A recent application of single-cell transcriptomics shows that brain tissue may also contain cells at different temporal stages in the response trajectory to stimulation (22). How sequential gAPs may interact within a single cell is also unclear: do they summate, or might there be refractory phases? In some contexts, sustained repeated behaviors may sum to increase or prolong IEG expression, whereas in other contexts, sequential exposures to a stimulus can result in habituation of responses (23, 24). Temporal interactions could emerge within individual cells (e.g., through epigenetic mechanisms; see below) or through systems-level modulations as originally outlined in 2000 (1). Approaches, such as catFISH (cellular compartment analysis of temporal activity by fluorescent in situ hybridization) and TAI-FISH (tyramide-amplified immunohistochemisty fluorescent in situ hybridization) (25, 26), may provide a means to follow the longitudinal integration of gAP induced by sequential stimuli at single-cell resolution. Other tests of these questions may be developed in mice using increasingly sophisticated optogenetic and pharmacogenetic manipulations to control both timing and location of specific IEG expression.

Attempts to relate IEG expression to the intensity or magnitude of a behavior may be further confounded if the brain region under study does not directly and exclusively determine the magnitude of the behavioral output being measured. If the behavioral readout is modulated by additional circuit elements, salience at the level of individual neurons (i.e., the gAP) may not necessarily correspond to the absolute salience of the experience of the organism. Where control of a measured behavior has been closely mapped to a discrete brain area, such as, for example, in the dedicated avian song control circuit, linearities have been reported between magnitude of behavior (singing) and magnitude of gene expression (of egr1/ZENK) (27).

gAP Within Different Neural Contexts

To understand the precise functional significance of a particular instance of the gAP, it is important to consider the specific neural context in which it occurs. In behaving organisms, different experiences clearly trigger different patterns of genomic activation throughout the brain. This was demonstrated with formal rigor in the study of Mukherjee et al. (24, 28), where patterns of IEG expression in the mouse brain were compared across 13 different experiences, ranging from reinstatement of feeding and mild foot shock to different regiments of cocaine exposure and volitional sucrose consumption. Each experience was found to be represented by a unique transcriptional signature to the extent that a minimal expression profile of 4 IEGs across 7 brain regions was sufficient to decode an individual mouse’s recent experience with nearly 100% accuracy. Expression changes of the same gene can also support opposite behavioral changes depending on where in the brain they occur. For example, in cichlid fish, both ascent and descent in social hierarchy are associated with expression changes of a subset of IEGs in the brain, but these are found in different anatomical patterns (29).

Within a single brain region or circuit, there may be variations in gAP function depending on the organization of the circuit and the roles of the different cell types involved. One obvious function is the encoding of learned events into memory. Putatively, an “encoding gAP” could be induced by stimuli that drive Hebbian plasticity of synaptic transmission. In this case, the gene expression programs induced by the initiating experience will culminate in transcription of the building blocks for specific synaptic modifications aimed at supporting the long-term maintenance of information storage at synapses within the network while maintaining synapse specificity. A second and possibly distinct function for the gAP would be in mediating homeostatic rebalancing, whereby the induced gene expression programs support the adaptation of neural networks to substantial shifts in activity (a “homeostatic gAP”). These functionally distinct experience-dependent gAPs may comprise different transcription programs, supporting different consequences for the synapses, neurons, and circuits in which they are encoded. Homeostatic mechanisms may be recruited more robustly in situations in which networks are establishing the appropriate gain on incoming stimuli and network activity, as observed during development, or after large shifts in neuronal activity (convulsions, seizures, large shifts in ionic balance, and large transitions in sensory input) as well as during sleep (19, 30–36). A “homeostatic gAP” is, therefore, anticipated to recruit large proportions of principal neurons as well as inhibitory interneurons, which robustly control network gain (37, 38). In contrast, an “encoding gAP,” induced within a neuronal ensemble recruited to encode a specific memory, is expected to recruit a fraction of the principal neurons in a given brain region (39). While most well-studied IEGs (ARC, FOS, EGR1, NPAS4, etc.) have been found to be induced by gAPs that could be classified as either putatively homeostatic or putatively encoding (40–42), there are indications that different transcriptional programs are induced by transient vs. prolonged stimuli, with transient stimuli potentially mimicking single-trial encoding of experience (19). Furthermore, NPAS4 has been associated with gAPs that may be classified as either homeostatic (37, 43) or encoding (41) and has been found to activate distinct programs of late response genes in inhibitory and excitatory neurons (37). A further speculative distinction between homeostatic and encoding gAPs is the expectation that neuropeptides (such as BDNF) may be recruited by homeostatic gAPs to propagate a signal to a large network of neurons, compromising synapse specificity, while encoding gAPs may retain synapse specificity in communication between the neurons in the network (44).

Interestingly, multiple waves of IEG transcription induced by experience have been reported (45–47). Potentially, the first wave of transcription may be encoding an individual experience, while the delayed wave of transcription could be mediating a global homeostatic shift in network function induced by the accumulation of experiences over time, possibly corresponding to the proposed role of sleep in memory consolidation (48). During “real-life” induction of gAPs, homeostatic and encoding gAPs may both be recruited, with the relative contribution of either type of gAP defined by the circumstances: everyday events (49) may be represented primarily by encoding gAPs, while unusual or especially salient events, which evoke substantial release of neuromodulators or stress hormones, may evoke a larger ratio of homeostatic gAPs. Speculatively, the recruitment of a homeostatic gAP may serve to “increase the gain” of the representation of events occurring within the same time window (of a few hours), thus creating a temporal window for the coupling of events in time. This idea may provide a molecular framework for “behavioral tagging,” whereby experiences of significant novelty (putatively inducing homeostatic gAPs) prolong the retention of everyday experiences (putatively inducing encoding gAPs) (49). “Behavioral tagging” is an extension of the “synaptic tagging” concept, demonstrating the essential role of inducible transcription and translation in the formation of proteins that are captured by recently active synapses to support long-term plasticity and long-term memory (50).

While the field is developing thanks to contributions of groups studying both large shifts in homeostasis or sensory environment (19, 37, 38, 43, 51, 52) and more subtle, everyday experiences (3, 4, 24, 36, 49, 53–57), it is possible that these two categories of gAP may differ in their mechanisms and consequences. Awareness of the potential distinctions between these 2 forms of experience-dependent plasticity may promote the resolution of fundamental questions regarding the computational significance of the gAP and its role in encoding information as discussed next.

gAP and the Encoding of Engrams

Encoding of a particular memory is likely to involve only a fraction of neurons that otherwise share similar broad afferent and efferent connectivity (58, 59). Coincident activity in such a subset of neurons defines a “neuronal ensemble,” which is thought to represent part of the “engram” or physiological representation of a particular memory. The idea of encoding in neuronal ensembles was first articulated by Hebb (60) almost 70 years ago and has been substantiated by a large number of studies that used IEGs to identify collections of neurons that are coactive during the encoding of an experience (Fig. 1). Recent studies have provided direct evidence linking gAPs in individual neurons and membership in putative engrams. These studies have used IEG-based tagging systems to express optogenetic and chemogenetic actuators in neuronal ensembles that were active at the time of encoding. Subsequent manipulation of these ensembles provides direct evidence that they constitute a component of the engram. Activation of tagged ensembles can induce memory retrieval in the absence of external sensory cues, whereas inhibition of tagged ensembles can prevent memory retrieval in the presence of external sensory cues (61, 62). These studies further show that only fractions of principal neurons are recruited to ensembles encoding learned experiences (39, 63), with interneurons participating more broadly in defining the allocation of principal neurons to ensembles (64, 65).

Fig. 1.

Assumptions regarding the relationship of the gAP to neuronal ensembles encoding an experience. (A) A salient experience is anticipated to induce a gAP in only a fraction of the principal neurons within a defined brain structure. (B) Within an ensemble, IEG expression is assumed to be bistable—maintained at low levels in the absence of stimuli and induced to its full extent after experience, rapidly returning to baseline. (C) Ensembles are presumed to be coherent—such that the ensemble defined by the expression of a single IEG will overlap to a large extent with the ensemble defined by other induced IEGs.

New research suggests that the recent activation history of a neuron, represented by the status of gene activity in the cell (i.e., the gAP), determines whether it becomes allocated to an engram (56). These studies have used a fear-conditioning paradigm in which rodents learn an association between a conditioned stimulus (CS; e.g., a tone) and an unconditioned stimulus (e.g., a foot shock). Using viral vectors to increase excitability in a random, sparse population of excitatory neurons in the lateral amygdala showed that neurons with increased excitability at the time of an event were more likely to be allocated to the engram supporting the memory of that event. Conversely, expressing constructs that decrease neuronal excitability decreased this probability (66–68).

More recent studies have also examined how excitability changes influence the interaction between engrams (55). In experiments where mice were conditioned to fear 2 different-toned CSs, memories were encoded in overlapping neuronal ensembles when the 2 training events occurred close together in time (<6 h). Moreover, they were linked behaviorally—extinguishing one CS led to extinction of fear to the other. In contrast, when the 2 training events were separated in time (24 h apart), there was no evidence for memory linking, and the engrams supporting these distinct memories did not overlap.

These results suggest that memories acquired in close temporal proximity are coallocated to overlapping engrams (thus linking the 2 memories), whereas memories acquired at more distal times are disallocated to distinct engrams (and the memories remain distinct). Similar to initial allocation, coallocation is also mediated by a winner-take-all mechanism in which more excitable neurons win the competition for recruitment to the second engram. Importantly, after they are allocated, neurons remain more excitable for a time period that overlaps with the “coallocation window” (typically 6 h). Broadly, these studies raise the possibility that dynamic gene activity governs how information is organized in neural circuits, with a neuron’s recent activation history (captured in the temporal dynamics of a gAP) defining the coupling of memories and their coallocation to neuronal ensembles.

Cellular Perpetuation of the gAP

Although longer lasting than eAPs, gAPs are still transient by nature. The average mRNA transcript has a half-life of 6–10 h, while the mRNAs encoding regulatory transcription factors, including IEGs, are among those with the shortest half-lives (69). Nevertheless, there is growing evidence that molecular mechanisms help sustain the effects of gAPs for extended periods of time (70) at multiple levels—cellular, multicellular, generational, and potentially, even intergenerationally (Fig. 2). For instance, IEGs that form the core of the gAP (like Fos and Jun) exhibit epigenetic features that support their fast activation rates: bivalent promoters containing activating [histone H3 lysine 4 trimethylation (H3K4me3)] and repressing (histone H3 lysine 27 trimetrylation) chromatin marks (71). In addition, these genes are often bound by “poised” RNA polymerases, fully initiated transcriptionally competent complexes that can be quickly released into productive transcription (72). Furthermore, these genes also display high levels of histone acetylation even before stimulation, which is thought to promote open, transcriptionally competent chromatin. Interestingly, these molecular features are also characteristic of genes described as expressing “transcriptional memory” (73), whereby an initial event drives local epigenetic modifications, leading to accelerated and more pronounced expression in response to subsequent activation. Like IEGs, genes expressing transcriptional memory are decorated with H3K4me3. In addition, their promoters are enriched for transcription factors bound to “memory recruitment sequences” as well as the histone variant H2A.Z. Importantly, molecular features of transcriptional memory can survive DNA replication such that, in dividing cells, increased activation potential can be maintained up to 8 cellular generations. Whether transcriptional memory per se is a robust feature of genes recruited by the gAP (such as IEGs) remains unclear, but given the similarities between IEG regulation and genes expressing transcriptional memory, it seems likely that at least some features of transcriptional memory support the rapid transcription of IEGs and may be subject to regulation based on past experience and activity patterns.

Fig. 2.

Epigenetic mechanisms for activity-dependent perpetuation of gene expression states. (A) Poised transcriptional states allow for fast transcription on experience. (B) Transcriptional memory allows for the faster and stronger reactivation of previously active genes. (C) The BDNF positive feedback loop. In its inactive state, BDNF transcription is repressed by methylation of the promoter, HDAC2-dependent nucleosomal suppression, and binding of MECP2 as well as BDNF antisense long noncoding RNA (BDNF-AS). Neuronal activity triggers demethylation of the promoter, release of MECP2, and repression of BDNF-AS. The resulting expression of BDNF triggers the release of HDAC2 from chromatin, perpetuating an active gene transcription state. MRS, memory recruitment sequences; TTS, transcription start site.

Neuronal activity also results in global increases in histone acetylation, which is important for learning and memory (9). Histone acetylation, occurring concurrently with the unfolding of the gAP or following closely behind it, relaxes chromatin, perpetuating the impact of the gAP (74–77). Interestingly, changes in histone acetylation in response to neuronal activity are linked to histone phosphorylation (75) and DNA methylation (74), suggesting the involvement of multiple epigenetic mechanisms in the gAP. Furthermore, specific acetylated histone residues have been differentially associated with the occurrence of salient experiences (i.e., resulting in memory formation) vs. nonsalient experiences (76).

An example illustrating epigenetic mechanisms that may underlie the perpetuation of a gAP is the positive feedback loop involving the histone deacetylase 2 (HDAC2) in the regulation of BDNF expression. BDNF is a neuronal growth factor involved in the activity-dependent decision regarding the elimination or maintenance of synapses (78).* Sustained BDNF up-regulation through histone acetylation in the prefrontal cortex has been associated with both consolidation and extinction of fear conditioning (79) (Fig. 2). Mechanistically, increased neuronal activity triggers the release of the repressive methyl-CpG (cytosine–guanine dinucleotide)-binding protein 2 (MECP2) from the BDNF promoter, resulting in an initial activity-dependent increase in BDNF expression (80). The action of BDNF at its receptors then triggers the release of HDAC2 from chromatin, increasing histone acetylation at neurotrophin-dependent gene promoters, including BDNF itself (81, 82). BDNF expression is also negatively regulated in humans by an overlapping long noncoding antisense RNA (BDNF-AS), which is itself negatively regulated by neuronal activity (83). Other epigenetic mechanisms, such as DNA methylation, may further contribute to the regulation of BDNF after salient experiences (84), indicating that a complex set of molecular mechanisms may function in the gAP to fine tune BDNF expression, potentially broadly impacting the expression of additional genes.

Although a nascent field, evidence is also mounting in support of the intriguing prospect that epigenetic mechanisms are involved in the transmission of the impact of experience across generations (85–89). Furthermore, although certain epigenetic marks correlate tightly with gene expression changes, in many cases the relationship between epigenetic marks and gene expression is still unclear (90).

gAP in the Organism: Modulating Responses to Stress

Responding to stress is a critical organismal challenge in which the gAP plays central roles, recording stressful associations and shaping future responses to similar circumstances. In this section, we extend the analogy of the classical gAP to gene expression changes induced by environmental/stressful stimuli via nuclear hormone receptors. Such a broader gAP may be induced by stress-driven activation of the hypothalamus–pituitary–adrenal (HPA) axis and release of glucocorticoids (GCs): GCs trigger rapid transcriptomic responses via intracellular receptors, which act as transcription factors [e.g., the glucocorticoid receptor (GR)]. GR activation propels a wave of downstream gene regulation (i.e., a GC gAP) starting within 30–60 min after stress exposure and peaking 3–6 h thereafter. A GC gAP can occur across different tissues beyond the brain and may be analog in nature (rather than digital/all or none), as GCs elicit graded transcriptional responses depending on timing and dose. In this context, it is worth mentioning the concept of the “neuroendocrine action potential,” (nAP) which refers to the propagation of a gAP, induced by experience in the brain, to a neuroendocrine response throughout the body, which can be propagated over hours to months (91). The nAP can support stable changes in the behavior of an organism corresponding, for example, to adaptations to changing seasons or a shift in social hierarchy. In this sense, the nAP provides a conceptual framework for considering how some gAPs may propagate to impact behavior over long periods of time.

In addition to inducing a stress transcriptome profile, GR activation also induces local epigenetic changes at its DNA binding sites, the glucocorticoid response elements (GREs). This has been described for GR in the context of transcription factor binding-mediated DNA demethylation at GREs (92), which subsequently facilitates the transcriptional effects of the GR on the target gene (92, 93). In this issue, Provençal et al.^† demonstrate lasting changes in DNA methylation at GREs and bivalent enhancers after GC exposure in hippocampal progenitor cells, suggesting that early life GC exposure could lead to the establishment of poised states that alter the gAP induced by subsequent stress and in turn, the ensuing behavioral responses.

The set point of the magnitude of the gAP induced by GR may also be altered at the level of the GR gene itself. Reception of maternal care by rat pups was shown to be sensed and transduced into lasting changes in (HPA axis) stress-dependent GC release via a mechanism that is initiated by binding of an IEG protein (EGR1) to regulatory sites in the GR gene in the hippocampus. This binding has lasting consequences, as it suppresses DNA methylation at these sites, sustaining higher levels of GR gene expression in the hippocampus of the pups after they mature and tempering the HPA response to subsequent stressors (94). Increased methylation of the GR promoter after reduced maternal care seems to be commonly observed in 7 of 10 studies of this model of early life adversity (95). Recent studies in macaques suggest that social context, such as social status, may also alter the transcriptional response to GC (96): in this case, in peripheral blood cells. These studies bring up interesting future research questions of (i) whether more complex social stimuli can also influence gAPs, (ii) whether environmental factors may influence gAP in a cross-tissue manner, and (iii) how changes in the set point of the gAP induced by the same environmental challenge map between different target tissues.

The FKBP5 gene is among the transcripts most strongly induced by GR activation in a number of tissues, including the brain (79, 97). Studies in both animals and humans are providing fascinating evidence for complex interactions between stress, genotype, and FKBP5 expression, which may influence susceptibility to psychiatric disease. Overexpression of the FKBP5 gene in several limbic brain regions has been associated with increased anxiety and decreased stress coping behaviors in laboratory animals (98). In humans, functional genetic haplotypes moderate the induction of FKBP5 by GR activation. The haplotype associated with increased transcriptional activation by GCs has been shown, in combination with exposure to early life adversity, to predict risk for a number of psychiatric disorders, including depression, PTSD (post-traumatic stress disorder), and psychosis (97). This gene × environment interaction is likely additionally mediated by allele-specific DNA demethylation at GREs of FKBP5 (96), further derepressing FKBP5 transcription in individuals exposed to adversity. Only the combined genetic and epigenetic transcriptional effects would increase FKBP5 protein levels beyond the threshold for initiating a cascade of events that may lead to the development of psychiatric disorders. The stress-induced epigenetic and transcriptional changes of FKBP5 are illustrated in Fig. 3. An interesting point is that FKBP5, via its effects as a cochaperone of the GR itself, also influences the neuroendocrine response to stress. States of higher FKBP5 expression have been associated with a prolonged cortisol response to stress in laboratory animals and humans (97), suggesting that alterations of the gAP of FKBP5 would have consequences on neuroendocrine set points.

Fig. 3.

Lasting epigenetic effects after GC exposure change subsequent stress gAP. (A) Prestress baseline. GR activation triggers a transcriptional response and also, leaves lasting epigenetic changes, here in the form of reduced DNA methylation within GREs. Whether these changes in DNA methylation status are lasting is likely influenced by the developmental timing of the stress exposure and genetic factors. (B) Poststress/new set point. The epigenetic landscape after a prior priming stress exposure now shows reduced DNA methylation within GRE. A subsequent GR activation now causes an enhanced transcriptional response.

FKBP5 also interacts with a number of other proteins highly relevant for key processes in neuronal function. It could thus represent a molecular hub, initiating multiple cellular processes on stress that would ultimately alter synaptic strength of specific neuronal ensembles. These processes include autophagy, cell proliferation, migration, apoptosis, and DNA methylation (99). Interestingly, FKBP5 also directly interacts with CDK5, a kinase that is known to phosphorylate and activate DNMT1. FKBP5 was found to decrease the interaction of CDK5 with DNMT1, reducing global DNA methylation via reduced phosphorylation and enzymatic activity of DNMT1 (100) as well as specific effects on epigenetic regulation of BDNF. By this, FKBP5 could also contribute to altering the gAP set point for other targets. Alterations of the FKBP5 gAP in response to stress would thus be propagated to other targets both via direct protein–protein interactions as well as possibly by interfering with key epigenetic mechanisms. An additional regulation on FKBP5 has been described via binding of the microRNA miR-15a to the 3′ end of FKBP5, decreasing FKBP5 translation (101). In animal models, the stress-induced up-regulation of this noncoding RNA was shown to buffer behavioral effects of stress.

In summary, exposure to stress or threat may lead to lasting alterations of gAPs by epigenetic effects in promoter and enhancer regions. Prior stressful experiences could alter future stress-induced gAP and by this, change the set point for the saliency filter for subsequent adverse experience. Furthermore, a “neuroendocrine action potential” may also translate into lasting coordinated physiological responses to stress across tissues.

Roadmap for Future Investigation

We have argued that gAPs are at the core of dynamic adaptations to experience, helping to filter, encode, and assimilate information over time. Effects of gAPs can be observed and studied across multiple levels of biological organization—molecular, cellular, systems, and organismal. Here, we consider some of the major questions, challenges, and opportunities for further research, building up from molecular through cellular and organismal to evolutionary perspectives.

Establishing the Causal Role of gAP Components in Encoding Experience.

A major challenge for the field is in defining the functional role of the different molecular components comprising the gAP in encoding experience. This challenge is exacerbated by the complexity of transcriptional networks and the intrinsic redundancies and compensation mechanisms that support their robustness (102). The development of temporally regulated, cell-specific genome editing tools that can target genes (and their regulatory elements) in various combinations is anticipated to assist in resolving the functions mediated by the products of activity-dependent expression with cellular and temporal specificity (103, 104).

Decoding the Activity Transcription Transfer Function.

An implicit assumption of the approach represented here is that an “encoding gAP” is a transcriptional relay of specific circuit-level events driven by experience. Thus, a major objective should be decoding the “activity transcription transfer function” defining the transformation of specific synaptic activity patterns to specific transcriptional programs. By defining the principles underlying the spatiotemporal association of activity required for induction of a gAP, we expect to also gain insight into the thresholds for neuronal salience filtering and define boundary conditions for transition to a commitment to long-term encoding of information in the brain. The development of an “algorithmic” definition of the rules governing the transformation of activity to transcription will support a major aim of the burgeoning field of “Behavioral Transcriptomics,” which is to associate defined experiences with their corresponding brain-wide gAPs, resolving their activity transcription transfer functions (24).

Associating Neuronal gAP History with Ensemble Selection.

Memories that occur close in time (or are related in content) are expected to be coallocated to overlapping neuronal ensembles, while memories that are separated in time or unrelated in content are expected to be encoded in nonoverlapping ensembles. Evidence is accumulating that the neural space is not an even playing field and that a neuron’s recent history determines whether it participates in information coding (56). While current studies have demonstrated the allocation of temporally adjacent events to overlapping neuronal ensembles, allocation to a memory trace is anticipated to work over multiple timescales, with epigenetic changes possibly having very long-term consequences for where and how information is encoded in neural circuits. Tests of these ideas may be developed in mice using increasingly sophisticated optogenetic and pharmacogenetic manipulations to control both timing and location of specific IEG expression: for example, in hippocampus during fear memory formation and recall (105).

Analysis of the gAP Within Diverse Cell Populations.

The recent development and rapid improvement of approaches for single-cell analysis (scRNAseq, snRNAseq, and smFISH) (22, 30, 106) are supporting the transition from population-based investigation to sensitive analysis with cell-specific resolution, enabling accurate measurement even of low-abundance transcripts, such as transcription factors. One of the contributions of this technical development has been the recognition that the genomic response to stimuli extends beyond neurons and is observed to be distributed across all cell types found in the brain (30, 106, 107). Disambiguating the gAP response in different cell types (and to different patterns of cellular stimulation) is expected to provide novel insight into how information is encoded in the brain.

Individual Differences in the gAP.

Here, we have outlined how the gAP may contribute to the filtering of experience, the formation and persistence of memory, and stress reactivity. Individuals differ greatly in these attributes—could variations in the gAP contribute to these individual differences? If so, what sort of gene × environment × development interactions contribute to variations in the gAP? Animal studies already suggest that the sensitivity or structure of the gAP can vary with developmental stage as seen, for example, in songbirds (108) and rats (109), and human studies have provided evidence for genetic influences on gAP reactivity (97). Might variations in the sensitivity and temporal parameters of the gAP have consequences on how information is filtered and stored either in different individuals or in different physiological or developmental states?

Translational Relevance of the gAP.

How might an understanding of the gAP be used to provide immediate benefits to human society? Direct measurement of the gAP in individuals could easily have clinical applications, but this is impossible with current techniques, as it would require multiple invasive measurements of brain tissue. Ultimately, it may be feasible to develop noninvasive markers of IEG expression and epigenetic mark deposition in the human brain: for example, using MRI or intravital fluorescence for detection (110, 111). Alternatively, there may be a link between responses in the blood and responses in the brain, allowing traces of a brain gAP to be identified in blood (112, 113). Such a connection has been reported in the context of stress, but current research suggests that it may extend more broadly to everyday life events (114). If these approaches can be refined and validated, they may herald a new quantitative psychiatry based on measuring and perturbing components of the gAP that mediate maladaptive memory processes (e.g., PTSD, drug addiction).

Evolutionary Perspectives on the gAP.

It is worth considering how the time course of the gAP may have evolved in response to the temporal demands of natural experience (115). Might the genomic response time course place a limit on behavioral plasticity? How might this align with the requirements of different ecological niches? Is the gAP itself subject to evolutionary plasticity?

These are exciting times to be involved in the investigation of the role of the genome in experience-dependent plasticity. The explosion of technical innovation across neuroscience and biology now supports deep investigation of the hypotheses and unifying principles described herein, with potential to impact health care.