Neuronal enhancers fine-tune adaptive circuit plasticity

Summary:

Neuronal activity-regulated gene expression plays a crucial role in sculpting neural circuits that underpin adaptive brain function. Transcriptional enhancers are now recognized as key components of gene regulation that orchestrate spatiotemporally precise patterns of gene transcription. We propose that the dynamics of enhancer activation uniquely position these genomic elements to finely tune activity-dependent cellular plasticity. Enhancer specificity and modularity can be exploited to gain selective genetic access to specific cell states, and the precise modulation of target gene expression within restricted cellular contexts enabled by targeted enhancer manipulation allows for fine-grained evaluation of gene function. Mounting evidence also suggests that enduring stimulus-induced changes in enhancer states can modify target gene activation upon restimulation, thereby contributing to a form of cell-wide metaplasticity. We advocate for focused exploration of activity-dependent enhancer function to gain new insight into the mechanisms underlying brain plasticity and cognitive dysfunction.

Neural circuits in the brain are continually reshaped throughout life in response to ongoing experience. The profound role of sensory experience in establishing the pattern of neural connectivity in the developing postnatal brain was first elucidated by the pioneering studies of Hubel and Wiesel, who showed that visual experience sculpts ocular dominance columns in the visual cortex.¹ Subsequent studies have confirmed the importance of early life experience in the development of circuits well beyond just sensory regions of the cortex.^2,3 Even in the fully wired adult brain, activity-dependent plasticity of neural circuits plays an essential role in behavioral adaptation. Neural networks rapidly encode information about the history of environmental experiences via the selective strengthening and weakening of specific synaptic connections within activated circuits.^4,5 However long-lasting activity-dependent synaptic and circuit-level changes engage additional mechanisms, and notably, an extensive and growing body of evidence indicates that the lasting effects of sensory input are mediated at least in part by the concerted induction of a program of gene expression that controls key aspects of circuit refinement.⁶

Neuronal activity-induced transcription has been widely observed in vivo in neurons undergoing experience-dependent plasticity.^7–10 In fact, this association is so pervasive that activity-regulated genes like Fos and Arc have become de facto markers of neural ensembles and memory engrams.^11–13 Regulators and effectors of the activity-induced transcriptional program can directly control aspects of synapse structure and function, including dendritic outgrowth, spine and synapse maturation, synaptic pruning, and excitatory/inhibitory balance, thus providing mechanisms to transduce synaptic signaling into long-lasting changes in circuit function.¹⁴ Importantly, perturbation of activity-regulated transcription factors (TFs) significantly impairs diverse forms of neural plasticity and adaptive behavior, indicating that the transcriptional consequences of activity are causative for plasticity and not merely byproducts of cell signaling.^15–20 Moreover, a number of neurological disorders have been linked to mutations in genes encoding activity-dependent transcriptional regulators, underscoring the critical role of this signaling network in cognitive development and function.²¹

The cellular mechanisms that couple neuronal excitation to regulated gene transcription have been extensively characterized and are recently reviewed elsewhere.²² Briefly, synaptic activity and subsequent membrane depolarization activate intracellular signaling pathways that engage a set of activity-responsive TFs, most notably from the CREB, Elk/SRF, and MEF2 families. These TFs drive the calcium-dependent transcription of a largely common set of immediate-early genes (IEGs) that predominantly encode additional TFs such as Fos. IEG TFs then regulate the expression of diverse cell-type-specific, late-response gene (LRG) programs, including secreted factors and neuronal modulators tailored to effect specific circuitry changes.²³

What has changed in recent years, following the widespread application of genome-wide sequencing technologies, is our understanding of the complex ways in which TFs interact with the genome to produce activity-dependent changes in gene transcription. Historically, research focused almost exclusively on the binding of neuronal activity-regulated TFs at promoters, gene-proximal regions that serve as the ultimate functional site for RNA Polymerase II (Pol II) recruitment and transcriptional initiation. However, over the last 15 years new genome-wide approaches have revealed the involvement of an extensive system of distal activity-responsive cis-regulatory enhancer elements in the transcriptional response to neural activity, dramatically expanding the known cis-regulatory network dedicated to the control of this gene expression program. For example, one study found that >95% of the binding sites for the activity-induced TF FOS lie outside of gene proximal promoter regions ²⁴. Indeed, large-scale profiling studies have documented rapid, widespread changes in the activity of many hundreds of enhancers accompanying both neuronal activation in vitro and various learning paradigms in vivo.^24–28

Decoding the regulatory logic of the genome is an exciting area of inquiry, and no doubt there will be new discoveries made that will advance our understanding of neuronal activity-neuronal gene transcription. Yet, given the complexity of enhancer regulation and the massive proliferation of genomic datasets, one might question whether the study of such mechanisms should remain within the province of gene regulation aficionados. Put simply, beyond an understanding of target gene expression patterns, how will knowledge concerning activity-responsive enhancers impact the study of activity-dependent circuit plasticity?

We argue that the spatiotemporal features of enhancer activation—their highly selective activation in defined neural cell-types, developmental stages, and/or in response to specific extracellular stimuli—make these cis-regulatory elements uniquely well suited to control the physiologically relevant tuning of activity-dependent cellular plasticity. For example, a typical gene is regulated by multiple enhancers that can function separately and exhibit highly context- and cell-type-specific activity, which collectively give rise to that gene’s composite expression profile. The specificity and functional modularity of enhancers enables the use of individual enhancer disruption—either via naturally occurring genetic variation or direct experimental manipulation—to change gene expression within a restricted subset of functional contexts. Indeed, enhancer variation has emerged as an important driver of evolutionary change, phenotypic diversity, and human disease. Moreover, evidence from various studies suggests that persistent stimulus-induced chromatin changes can give rise to sustained primed/refractory enhancer states that result in altered target gene activation upon restimulation. By thus altering the stimulus sensitivity of activity-regulated gene expression, such changes in activity-dependent enhancer function may serve to shift the threshold for activity-dependent cellular plasticity, a process known as metaplasticity.

Several comprehensive reviews of enhancer function in the nervous system have appeared in recent years.^23,29–32 Here, we focus on the specific role of enhancers as regulators of experience-dependent neural circuit refinement and plasticity. After briefly reviewing the current understanding of enhancer mediated-gene regulation, we discuss the key aspects of enhancer function that we argue make them especially relevant to the study of cellular plasticity. Many of the efforts to test hypotheses of how enhancers shape neuronal plasticity are in their early stages. Thus, throughout this discussion, we leverage insights from the extraordinary breadth of work on enhancer regulation and function conducted in non-neural systems, highlighting areas of particular relevance and opportunities for future progress.

Key properties of enhancers

Enhancers are gene-distal cis-regulatory elements that play a pervasive role in gene regulation (Box 1) (Figure 1). One of the first observations to emerge from the large-scale identification of enhancers was support for the idea that enhancer activity is often cell-type-specific. Beginning with the very first genome-scale efforts to profile TF binding, it has been appreciated that most TFs bind only a small fraction of their genomic binding motifs in a given cell type.³³ For a TF to bind, the nucleosomal DNA needs to have been made accessible during development either by the action of cell lineage-determining “pioneer” factors³⁴ or via close, cooperative binding of multiple cell-type-specific TFs.^35,36 As a result, only ~10,000–50,000 of the estimated one million identified genomic enhancers are active in any given cell type.^37,38

Box 1. Enhancer function, characteristics, and means of identification.

Enhancers are gene-distal TF-binding sequences that can potentiate the activity of a core promoter in a distance- and orientation-independent manner.¹⁸⁵ These elements consist of a short (typically 200–500 bp) core region depleted of nucleosomes – the basic structural units of chromatin, consisting of DNA wrapped around the core histone octamer. As such, enhancer DNA is locally accessible for binding by TFs. Enhancers can be found within introns as well as intergenic regions up to hundreds of kilobases away from the target promoter. They are thought to exert their effects on target gene expression through enhancer-target promoter interactions mediated by the looping out of the intervening chromatin (Figure 1).^186,187 Activated enhancers then increase expression of the target gene by increasing the likelihood of RNA Pol II capture at the gene promoter¹⁸⁸ and/or by acting at the level of transcriptional elongation to release Pol II pausing.^189,190

The ability to identify distinctive enhancer-associated chromatin signatures by high-throughput DNA-sequencing methods has revolutionized the identification of putative enhancers by enabling their large-scale genome-wide characterization based on epigenomic criteria. The use of chromatin marks to annotate putative enhancers has been extensively reviewed elsewhere.^140,191,192 Common criteria for defining enhancers include: 1) the presence of nucleosome-free, accessible chromatin, as revealed by DNaseI hypersensitivity or preferential transposon integration into accessible chromatin (aka ATAC-seq);^193,194 2) relative enrichment of the mono- or dimethylated form of histone H3 lysine 4 (H3K4me1/2); 3) high levels of acetylation of histone H3 at lysine 27 (H3K27ac);^195,196 and 4) the recruitment of transcriptional coactivators such as P300, CBP, or Mediator^25,197,198. Active enhancer elements also directly recruit RNA Pol II and are themselves actively transcribed,^25,199 with the resulting enhancer-derived RNAs (eRNAs) constituting another widespread sequencing-based means to identify to active enhancers genome-wide.^200,201

Figure 1. General schema of enhancer-mediated gene regulation.

Enhancers and other distal cis-regulatory elements interact with gene promoters via chromosomal looping to driven gene expression. These modular elements bind sequence-specific transcription factors (TFs), which recruit coactivators and core transcriptional machinery, including RNA polymerase II. TF binding must overcome the nucleosomal barrier; poised and active enhancers are characterized by accessible chromatin regions (top).

A second key observation about enhancers is that in any given individual cell type, there are pervasive changes in enhancer activity accompanying neural development, differentiation, and maturation.^39–43 More than 90% of putative enhancer loci in the mouse forebrain exhibit dynamic changes in activity across brain development.³⁷ Moreover, transgenic studies have confirmed the ability of such elements to drive highly spatiotemporally restricted patterns of gene expression in vivo.^44–46

A third observation is that the number of active genomic enhancers greatly exceeds that of their gene targets, such that each gene can be subject to regulation by multiple enhancers. At some loci, the presence of multiple seemingly redundant enhancers can serve to maintain phenotypic robustness, buffering gene expression patterns in the face of environmental and genetic perturbations.⁴⁷ In these cases, gene expression patterns remain largely unaffected by the loss of individual enhancer elements under normal developmental conditions, although defects become apparent in these mutants in the presence of developmental stressors.^48–50 However, for many other genes individual enhancer elements contribute to distinct aspects of a gene’s overall spatiotemporal expression profile. For instance, three different enhancer elements promote expression of the Dlx1/Dlx2 bigene cluster in distinct interneuron subtypes in the developing murine telencephalon.^51,52

This general enhancer architecture has significant implications for how these gene regulatory elements contribute to both cellular physiology and to disease pathology. The fact that different enhancers can be used to regulate the expression of a single gene in different cell types helps to explain why genetic variants that disrupt enhancers may affect gene function only within a specific subset of contexts compared to corresponding gene coding mutations. For example, while coding mutations in the homeobox TF gene TBX5 give rise to Holt-Oram syndrome, characterized by cardiac and forelimb malformations, rare mutations that selectively disrupt cardiac-specific TBX5 enhancer function effectively decouple the heart and limb phenotypes, yielding heart-specific developmental defects.⁵³ Such examples have motivated efforts to comprehensively profile functional cis-regulatory elements across human cell types.^54–59 Researchers have begun to intersect the resulting data with disease- and trait-linked non-coding variants to gain insight into the functional importance of cell-type-specific regulatory elements.^40,60–64

Rare variant mutations provide stark examples, but they also raise the possibility that common variants that alter enhancer function could underlie variable phenotypes in complex disorders and human variation more generally. Indeed, genome-wide association studies (GWAS) have identified thousands of non-coding loci that are associated with disease susceptibility and other complex traits,⁶⁵ with much of the underlying causal variation believed to affect enhancer function.^66,67 Indeed, common sequence variation in putative enhancer elements has been repeatedly linked to susceptibility to various neuropsychiatric disorders^40,68,69 and addiction-associated behavioral traits.^62,63 To date, these data remain largely correlational; however, direct evidence for the importance of enhancer variation at least in the regulation of gene expression has been obtained at specific loci. For example, schizophrenia-associated single-nucleotide polymorphisms (SNPs) that lie within putative enhancer elements present in the first intron of the CACNA1C locus, which encodes a voltage-sensitive calcium channel subunit, exhibited a diminished ability to promote transcription when tested in heterologous cells.⁶⁸

Evidence that variations in enhancer sequence contribute to cognitive function also comes from comparative studies of genes and behavior across evolution. The gain and loss of enhancers is appreciated to be a major driver of phenotypic diversity and evolutionary change insofar as the relatively diminished burden on fitness of individual enhancer elements is thought to allow some enhancer variants to reach high frequency in populations.^70,71 Enhancer modifications have been implicated in major morphological evolutionary transitions^72,73 as well as more subtle adaptive differences between related species.^74–79 Likewise, evidence exists for enhancer involvement in the evolution of complex, innate behaviors,⁸⁰ with naturally occurring genetic variation in putative distal enhancer elements also implicated in intraspecies behavioral differences in several species.^81,82

Enhancer-mediated transcriptional control thus represents a complex, context-dependent, yet pervasive feature of gene regulation. While enhancers have been best understood in the context of their central role in driving developmental and cell-type-specific patterns of gene expression, key features of enhancer function also make them especially suited to mediate aspects of cellular plasticity. This is the topic to which we now turn.

Enhancers as highly tuned sensors of distinct types of neuronal activity

Emerging evidence indicates that the fundamental principles of enhancer-gene regulatory logic gleaned from studying enhancers in development also extend to the control of activity-responsive gene programs in postmitotic neurons. For example, comparative epigenomic and transcriptomic profiling reveals an average of ~4–6 activity-responsive enhancer elements per activity-regulated gene.^25–27 Moreover, activity-regulated genes that are induced in a cell-type-specific manner are associated with the differential regulation of cell-type-specific enhancers.^83–86 Indeed, this cell-type-specificity is underscored by the recent appreciation that while TF-encoding IEGs are largely (but not entirely, see below) similar across neuronal cell types, downstream late-response effector gene programs differ substantially across neuronal and non-neuronal cell types.²³ These cell-type-specific differences appear to reflect the selective recruitment of widely induced IEG TFs to enhancers by cell-type-restricted stimulus-unresponsive transcriptional regulators and other chromatin differences.^87,88

What is special about applying the rules of enhancer function to activity-regulated genes is that the specificity of enhancer-dependent transcription can now extend to the response of a gene to distinct upstream stimuli. For example, membrane depolarization and brain-derived neurotrophic factor (BDNF) treatment activate markedly different enhancer programs within the same culture paradigm.⁸⁹ Moreover, this specificity of enhancer activation by different stimuli holds at the level of individual genes. For instance, stimulus-specific activation of distinct sets of enhancers has been shown to underlie the broad transcriptional induction of the Fos gene in response to a wide range of stimuli.^90,91

The response of different enhancers to neuronal stimuli depends on the specific combination of enhancer-bound TFs, which can differ in their sensitivity to activation by intracellular signaling pathways. In the case of Fos activation, the enhancers bound by the TFs CREB and SRF are highly responsive to upstream stimuli that activate intracellular calcium and cAMP signaling, whereas a single Fos enhancer bound by members of the MEF2 TF family selectively mediates activation of Fos transcription downstream of BDNF-mediated TrkB-coupled signaling.⁹⁰ This finding that enhancers can respond differentially to different signaling cascades also raises the question as to whether, depending on the complement and organization of TFs that bind a particular cis-regulatory region, different enhancers may preferentially drive gene expression in response to distinct patterns of synaptic input. While yet to be addressed directly, this outcome seems highly likely: different activity patterns have been shown to drive distinct neuronal gene expression responses.^92–96 Moreover, various signaling pathways that mediate excitation-transcription coupling have been shown to temporally integrate intracellular calcium oscillations in different ways,^97–99 resulting in the differential amplitude and frequency tuning of specific calcium-responsive TFs.^94,100–102

A practical application of this role of enhancers as differential mediators of excitation-transcription coupling relates to the essential contribution of activity-induced transcriptional responses to functional circuit mapping. IEGs such as Fos, have long been used to mark activated circuits following sensory and behavioral stimuli. More recently, a suite of IEG-dependent viral and transgenic tools has been developed to monitor and manipulate activated neural ensembles such that the modern study of memory engrams has been closely intertwined with the use of such IEG-based reporters.¹⁰³ Existing tools are almost exclusively based on IEG promoters and/or synthetic binding arrays for activity-induced TFs.¹⁰⁴ Yet, the precise relationship between IEG-tagged cells and the engram itself has yet to be fully defined. For example, while overlap between cells tagged during training and testing exceeds chance levels, the overall correspondence between these two cell populations remains relatively low (typically ~10–40%).¹⁰⁵ The extent to which this limited overlap reflects overbroad IEG-based labeling that obscures the functionally relevant subpopulation (perhaps defined by specific activity patterns), other technical considerations, or the dynamic nature of the engram itself as it is refined and consolidated, remains unclear. Moreover, while the field has largely employed various ensemble-tagging reagents interchangeably, evidence not only suggests that different IEGs can mark functionally distinct neuronal ensembles – consistent with possible differential sensitivity of IEGs to neuronal activity patterns – but also that different tagging methods based on the same IEG can yield limited overlap.^106,107

Future development and use of reporter reagents driven by curated stimulus-responsive enhancer elements should allow more precise fine-grained targeting of specific cell types and/or cells responding to distinctive neural activity patterns. Importantly, while calcium- and membrane voltage-based sensors already permit high temporal resolution imaging of neuronal activity dynamics, enhancer-based reporters will also provide genetic access to selectively manipulate specific cell types within activated circuits, potentially without the need for recombinase-dependent approaches. In this regard, synthetic modifications of the so-called synaptic activity-responsive element (SARE), first identified as an activity-responsive enhancer for the Arc locus, have been used to generate versatile viral reporters of synaptic activity,^108,109 and longitudinal two-photon imaging with these reagents has been successfully used to monitor how CA1 plasticity patterns evolve and stabilize across repeated encounters with different environments.¹¹⁰

Scalable strategies to screen and test the utility of candidate enhancers have already yielded enhancer-based viral drivers with impressive cell-type-specificity.^111–115 Such approaches should be extended to assess excitation-enhancer activation dynamics; for example, through the use of emerging spatially resolved single-cell epigenomic profiling methods,¹¹⁶ employed in conjunction with calcium indicator imaging of neural activity. Access to an arsenal of such finely tuned enhancer-driven gene delivery tools will significantly advance efforts to dissect activity-dependent circuit responses and may have therapeutic potential for treating intractable forms of epilepsy or related disorders of aberrant circuit activity.¹¹⁷

Linking enhancers with plasticity

Unlike cell-type-specific and developmentally regulated genes, which are categorically either on or off in a given cell type or developmental stage, activity-regulated genes are typically expressed in a temporally graded fashion, such that they are transcribed to a greater or lesser degree in a given cell at different points in time. This distinction has important consequences for examining the cellular functions of activity-regulated gene expression for neuronal plasticity. Disrupting a gene is the most definitive way to establish its function, but constitutive knockouts are problematic when genes have pleiotropic functions in different contexts. Cre drivers and intersectional viral strategies provide a straightforward way to conditionally restrict gene disruption to a particular cell type, which can be further restricted to specific developmental times via the use of drug-inducible recombinases. By contrast, since activity-regulated genes are typically expressed to some degree at all times in a given cell, dissociating their functions at baseline from their functions in activity-dependent plasticity is substantially more complex. Thus, with a few notable exceptions,^118,119 it has proven challenging to isolate the specific biological effects of the activity-regulated component of a given gene’s expression.

Importantly, given the evidence that individual enhancers can exhibit restricted activity profiles that collectively give rise to that gene’s composite expression pattern, disruption of specific enhancers has the potential to dissociate activity-dependent from activity-independent functions of activity-regulated genes in circuit dynamics. In this regard, one early study showed that binding of the activity-responsive transcriptional regulator SRF to the SARE element in the distal Arc enhancer is required for late-phase long-term depression (LTD) in mouse cultured cerebellar Purkinje cells.¹²⁰ The authors demonstrated that reintroduction of a wild-type bacterial artificial chromosome (BAC) containing the Arc locus together with its surrounding regulatory regions rescued the LTD defect observed in ARC-deficient cells. Mutation of the SRF-binding site in the distal SARE element blocked this rescue, whereas co-transfection of a form of SRF engineered to recognize the mutated site was sufficient to restore late-phase LTD in this system. While additional studies are needed to confirm these findings in an in vivo context, to our knowledge this represents the first direct implication of activity-responsive enhancer function in any form of activity-dependent plasticity.

More recently, Spiegel and colleagues have examined the consequences of conditionally excising two activity-responsive enhancers that cooperatively mediate sensory-driven transcription of Insulin-like growth factor 1 (Igf1) in vasoactive intestinal peptide (VIP)-expressing interneurons in the murine visual cortex.¹²¹ Prior work had shown that visual experience induces Igf1 expression selectively in VIP interneurons, which increases inhibitory synaptic input onto these cells and regulates visual acuity in an experience-dependent manner,¹²² and identified two activity-responsive enhancer clusters ~40 and 100 kb upstream of the Igf1 gene.⁸³ Using targeted enhancer deletions, Roethler et al. demonstrated that these elements cooperatively drive sensory-induced expression of Igf1 in the adult visual cortex. Igf1 enhancer loss eliminated acute sensory-induced increases in inhibitory inputs onto VIP interneurons, leading to increased excitation of these cells, elevated activity in layer 2/3 pyramidal cells via disinhibitory mechanisms, and altered spatial frequency tuning in awake, behaving animals. In the case of Igf1, excision of the relevant enhancers effectively ablated activity-induced Igf1 expression without affecting basal expression of the gene.¹²¹ Similar approaches have also been used to investigate enhancers driving activity-dependent Bdnf gene induction.^123,124 In vivo manipulation of individual stimulus-responsive enhancer elements thus represents a powerful, yet largely untapped approach to dissect the contribution of discrete aspects of a gene’s expression profile to plasticity phenomena.

Enhancers also provide a useful lens through which to interpret the effects of naturally occurring genetic variation on adaptive circuit plasticity. Large-scale genetic efforts are increasingly uncovering genomic variants associated with neurodevelopmental and neuropsychiatric disorders that impact the function of specific brain circuits, at least some of which likely interfere directly with circuit plasticity mechanisms. Yet, while providing invaluable insight into the underlying genetic architecture of these neurological conditions, such mapping studies cannot on their own distinguish gene regulatory networks that serve as acute mediators of neuroplasticity from static elements involved in generating the permissive substrates upon which these plasticity mechanisms act. Here, acute responsiveness to neural stimuli may serve to distinguish cis-regulatory elements that directly mediate synaptic plasticity, just as it has at the level of the neuronal transcriptome. However, the annotation of activity-responsive enhancers will require experimental systems that are amenable to acute stimulation. In this regard, recent studies have begun to characterize the landscape of human activity-responsive gene regulatory elements using pluripotent stem cell (PSC)-derived neurons.^125,126 Although one study reports significant heritability enrichment for schizophrenia in activity-responsive enhancers active in human PSC-derived glutamatergic and GABAergic neuronal cultures,¹²⁶ work with additional human neuronal subtypes is needed to fully appreciate the extent to which human variation within activity-dependent gene regulatory elements contributes to neurological and psychiatric disease.

At the present time, we have a similarly limited knowledge of the extent to which evolutionary changes within enhancers contributed to adaptations in mammalian brain circuitry and plasticity. In this regard, investigators have reported multiple human-specific changes in neural enhancer activity,^127–130 including a number associated with human accelerated regions (HARs), conserved genomic loci with elevated divergence in humans.^131–133 Likewise, studies using PSC-derived neurons have identified differences in activity-induced gene programs between human and rodent neurons.^134–136 In one study, Osteocrin (OSTN), a well-conserved gene encoding a secreted factor expressed in bone and muscle, was found to gain binding sites for the activity-responsive MEF2 TF family at a distal enhancer element, resulting in the acquisition of activity-dependent expression in neurons of the primate lineage.¹³⁶ Indeed, OSTN expression was found to be induced in layer 4 of the macaque visual cortex in an activity-dependent manner. Rapidly expanding genomic information from diverse mammalian lineages promises to further advance our understanding of the evolutionary trajectory of neural enhancer sequences and their contribution to species-level phenotypic changes.^137,138

Enhancers in metaplasticity

The studies we have reviewed thus far show how enhancers tune the expression of their target genes to drive functionally relevant activity-dependent changes in neural circuits. Yet, thresholds for plasticity are themselves subject to activity-dependent regulation. Indeed, it is well appreciated that neurons and their synapses can shift their response to plasticity-inducing stimuli in a manner that is dependent on the recent history of the cell, with various activity-dependent processes—collectively referred to as metaplasticity—proposed to regulate the extent and timing of plasticity.¹³⁹ Might then enhancers themselves also serve as loci for cellular plasticity?

As described above, the activity state of enhancers can be defined by their chromatin features, which include their accessibility, their levels of DNA methylation, their looping and long-range chromatin interactions, and the local accumulation of specific combinations of histone modifications.^140,141 Many of the chromatin regulatory proteins that establish this epigenomic context are themselves targets of regulation by neuronal activity-dependent signaling cascades, and neuronal activity-dependent changes in chromatin features have been well documented. Thus, sustained stimulus-induced changes in enhancer chromatin could serve to maintain some stimulus-responsive enhancers in a persistent primed (or refractory) state, thereby altering subsequent enhancer-dependent gene expression upon restimulation. Such an epigenomic imprint would alter subsequent excitation-transcription coupling, leading to a shift in the expression dynamics of key plasticity-inducing (or -suppressing) genes. These gene expression changes might in turn regulate the extent and timing of subsequent plasticity, thereby constituting a form of cell-wide metaplasticity (Figure 2).

Figure 2. An epigenomic model of transcriptional metaplasticity.

a) In response to cellular stimuli, enhancers can occupy multiple discrete dispositional states characterized by persistent changes in chromatin modifications, enhancer-bound TFs, and/or DNA topology (left). Such states could influence enhancer-mediated transcriptional activation in response to subsequent stimuli, thereby altering the excitation-transcription coupling of relevant target genes (right). By thus shifting the expression dynamics of key plasticity-related genes, such a mechanism could impact the extent and timing of subsequent cellular plasticity. b) Chromatin priming may contribute to network plasticity by altering activity-regulated gene expression. Representation of findings from Gemberling et al.¹⁸⁴ showing that priming of Fos inducibility via dCas9-p300-mediated acetylation of a Fos enhancer increases the number of Fos-expressing (Fos+) cells in hippocampus following sensory experience. Primed neurons show altered current input-spike output relationships, which could generate distinct outputs from a common input. Panel created with Biorender.com.

While still somewhat speculative, this form of transcriptional regulation is not without precedent. Phenomena termed transcriptional memory, in which the rate or strength of gene expression in response to a stimulus is enhanced by previous exposure to that stimulus, have been observed in a broad range of evolutionary divergent organisms, including yeast, flies, plants, and mammals.^142,143 While a variety of mechanisms have been implicated in these processes, they largely center around epigenetic effects encoded at the level of chromatin.

In mammals, such forms of transcriptional memory have been most extensively studied in the context of trained immunity, where an initial inflammatory stimulus leads to a long-term functional modification of cells in the innate immune system that results in an elevated response to a subsequent unrelated challenge.¹⁴⁴ In this context, antimicrobial genes in pathogen-challenged innate immune cells have been shown to display a ‘priming’ behavior, resulting in heightened gene expression in response to subsequent inflammatory stimuli. More recently, these findings have been extended to stem and progenitor cells of epithelial barrier tissues outside the immune system, demonstrating that long-lived tissue cells exhibit elevated transcriptional responses stemming from prior inflammatory encounters and contribute to a broad form of inflammatory memory.¹⁴⁵ Importantly, epigenomic reprogramming of stimulus-responsive cis-regulatory elements has been repeatedly implicated as a mediator of these effects. Although most stimulus-induced genes return to their baseline epigenomic state shortly after stimulus withdrawal, a cohort of enhancers and some associated promoters retained accessibility and certain histone modifications (e.g. H3K4me1) long after the dampening of the initial transcriptional response,^146–149 and these persistent “memory domains” were associated with faster and stronger enhancer activation upon restimulation.^146,147

The extent to which changes in the ability of neuronal excitation to elicit new transcription contribute to neuronal metaplasticity, however, remains largely unexplored. While prior work has shown that different activity patterns integrated over the course of minutes can induce distinct patterns of neuronal gene expression, few studies have investigated possible activity-dependent changes in the coupling between neural activity and gene transcription over longer timescales. Limited evidence, however, does support the idea that recent behavioral history can modify the coupling between cell activity and gene induction. In this regard, Arc expression was shown to be induced in a similar proportion of rat CA1 neurons in response to a single novel exploration session or daily sessions repeated over multiple days.¹⁵⁰ In contrast, brief but repeated exposure to the same environment within a single day led to a dramatic reduction in Arc mRNA induction after several sessions, despite stable electrophysiological activity. Nevertheless, possible epigenomic alterations were not examined in this paradigm, so these changes in the sensitivity of excitation-induced gene transcription could stem instead from alterations in synapse-to-nucleus signaling cascades.

The rapid induction of epigenomic changes during neural circuit remodeling, and the importance of chromatin regulators in this process, have been amply demonstrated. While the requirement of chromatin regulators for plasticity phenomena could simply reflect their integral roles in stimulus-induced and/or basal transcription, accumulating evidence suggests that long-lasting chromatin changes do occur at neuronal activity-responsive enhancers. For example, when analyzing mouse dentate granule neurons in mice exposed to electroconvulsive shock, Song and colleagues identified an extensive network of activity-responsive enhancers in mouse dentate granule neurons that exhibit increased accessibility for at least 24 hours after the stimulus.¹⁵¹ Similarly, sustained chromatin accessibility and DNA looping changes were observed in cultured excitatory hippocampal neurons for as long as 48 hours following kainic acid stimulation, a time when the neuronal transcriptome was largely restored to its basal state.²⁶ A separate study by Marco et al.²⁷ also employed IEG-based cell-tagging methods to analyze transcriptional and chromatin-level changes in mouse hippocampal neurons following contextual fear conditioning. The authors report long-lasting activation of enhancers for up to five days following the initial stimulus. Moreover, this widespread enhancer priming during memory encoding was associated with new enhancer-promoter interactions during later phases of memory formation, particularly during memory retrieval, suggesting that enhancer-associated chromatin changes at earlier stages modulate the transcriptional changes observed at later timepoints. Finally, a variety of sustained epigenomic marks associated with maladaptive addictive behavior have also been reported.^152–155

While the molecular basis for these persistent epigenomic changes is not yet known, it is striking that, in each of the hippocampal studies reviewed above, unbiased motif analysis revealed a strong local enrichment for the AP-1 (FOS/JUN)-binding motif within the gene-distal cis-regulatory sites that undergo sustained chromatin reorganization.^26,27,151 Indeed, Su et al.¹⁵¹ demonstrated a significant attenuation of activity-induced chromatin opening at these elements upon Fos knockdown. These findings are particularly notable considering a recent report that AP-1 is required for the gain and retention of chromatin accessibility at inflammatory memory domains in epidermal stem cells,¹⁴⁷ consistent with the finding in other contexts that AP-1 recruits the SWI/SNF (mammalian BAF) chromatin remodeling complex to latent enhancer elements.^83,87 Subsequent maintenance of accessibility within memory domains in the epidermal stem cells appears to be driven by the binding of JUN homodimers and additional constitutive TFs that gained access upon initial chromatin opening. Provocatively, while FOS/JUN binding following the primary stimulus required the accessory stimulus-specific TF STAT3, rapid FOS recruitment to established memory domains upon a secondary challenge was STAT3-independent,¹⁴⁷ providing a potential mechanism for heightened sensitivity to secondary stimuli. Moreover, re-analysis of inflammatory memory domains from datasets across multiple hemopoietic cell types implicated AP-1 as a universal mediator of trained immunity and inflammatory memory. Collectively, these findings raise the possibility that AP-1 plays a central role in diverse forms of enhancer-mediated transcriptional memory.

Chromatin modifications can also impact gene expression via their effects on three-dimensional (3D) genome architecture. Like other features of the epigenome, the 3D folding of the genome differs by cell type¹⁵⁶ and can be remarkably stable over long periods of time.¹⁵⁷ Several studies have demonstrated that increases in neuronal activity can induce increased looping interactions between the promoters of activity-regulated genes and their validated⁹⁰ or putative¹⁵⁸ distal enhancers. Whether these architectural changes persist after their induction has not yet been assessed in neurons; however, one study that characterized activity-dependent changes in promoter-enhancer interactions in cerebellar granule neurons showed that learning behaviors normally driven by this stimulus were impaired upon conditional deletion of the cohesin subunit and chromatin architectural protein Rad21.²⁸

While the identification of long-lived chromatin changes associated with neuronal activity-responsive loci is intriguing, it will be important to exclude the possibility that these chromatin signatures merely represent causally inert byproducts of the earlier transcriptional process. Corroboration of this epigenomic metaplasticity hypothesis will require a demonstration that 1) persistent epigenomic changes at activity-responsive enhancers reflect the activity history of the cell over at least moderate timescales, as well as showing that 2) these persistent chromatin alterations alter the nature of subsequent stimulus-dependent gene induction and downstream cellular plasticity.

Addressing these issues will require the maturation of still emerging technologies. Robust single-cell multimodal epigenomic and transcriptomic profiling methods will be necessary to characterize persistent and changing chromatin states at enhancers along with their correlated effects on downstream gene transcription in specific neuronal subtypes.^159,160 In addition, locus-targeted epigenomic editing approaches are needed to specifically and dynamically manipulate persistent local enhancer chromatin states within cells.¹⁶¹ Yet, these are not distant goals; in vivo editing approaches have already been used to successfully manipulate local Fos enhancer acetylation levels within the adult mouse hippocampus.¹⁶²

It will also be critical to establish the time frame over which such chromatin-based metaplasticity mechanisms might operate. Over a period of hours, such mechanisms might either contribute to the co-recruitment of cells to neural ensembles encoding temporally linked experiences,^163,164 or conversely act homeostatically to restrain runaway activity-induced transcription. Alternatively, as some have proposed,¹⁶⁵ long enduring epigenomic changes might play an ongoing role in the retention of long-term memories. In this model, sustained chromatin changes at cis-regulatory elements are hypothesized to participate in generating a self-sustaining “maintenance transcriptome” that preserves the engram while setting a new altered threshold for further plasticity. Such a model seems in tension with the idea that individual neurons can be recruited over time into multiple engram networks, although such mechanisms could operate over a limited time period until the memory engram is transferred to neuronal ensembles elsewhere in the brain via systems consolidation processes. Indeed, it is notable that recent work has reported long-lasting transcriptional programs that may contribute to the consolidation of long-term memory traces.^166–168 Relatedly, analysis of the effects of negative epigenetic regulation in non-neuronal subtypes has found that repressive chromatin regulators can act stochastically via all-or-none effects at the single-cell level such that these regulators modulate the fraction of cells silenced in a cell population rather than subtly shifting the amount of gene expression within a given cell.¹⁶⁹ The operation of analogous enhancer priming mechanisms in the brain could thus serve to bias individual neurons for recruitment into a forming engram. Ultimately, progress towards understanding the contribution of persistent epigenomic mechanisms to engram formation and maintenance will require long-term monitoring of chromatin changes within engram cells, a challenging task, especially given concurrent processes of systems consolidation that involve temporal reorganization of the brain-wide engram network.¹⁷⁰ Future studies combining activity-based cell-tagging approaches with robust epigenomic and transcriptomic analyses will be needed to further characterize the scope of these enhancer-centered epigenomic changes, determine their underlying mechanisms of regulation, and directly address their functional contribution to regulated gene expression and circuit plasticity.

Expanding to non-neuronal cell types

While we have focused our discussion on neuronal gene expression, it is important to note that the establishment and plasticity of mature circuits involves coordinated signaling between neurons and a range of non-neuronal cell types. Neuronal activity, for example, has been shown to regulate a range of processes, including local oligodendrogenesis, myelination, and blood flow.^171–174 Notably, while it has been appreciated for some time that neuronal activity can upregulate Glial fibrillary acidic protein (Gfap) gene expression in astrocytes,¹⁷⁵ recent transcriptome-wide approaches have uncovered robust, acute activity-dependent changes in numerous astrocytic genes with metabolic and synaptic functions.^176–178 As of yet, the mechanisms mediating these neuronal activity-driven gene expression changes in astrocytes remain largely unexplored. However, stimulus-responsive glial enhancers likely play a central role in mediating these regulated gene expression changes. Given the important role of bidirectional neuron-astrocyte communication in regulating synapse function as part of the tripartite synapse,^179–181 further investigation of such mechanisms, including the study of glial enhancers, promises significant new insights into additional aspects of circuit maturation, refinement, and plasticity.

Future Outlook

Our knowledge with respect to the function and diversity of neural stimulus-responsive enhancers is continuing to evolve. Recent experiments have demonstrated that these regulatory elements have relevance for the study of circuit plasticity beyond a simple mechanistic understanding of neuronal gene regulation. The fact that individual enhancers typically only mediate specific aspects of a gene’s overall spatiotemporal expression profile provides the opportunity to leverage the exquisite specificity of enhancers to develop new tools for targeted imaging and genetic access to specific neuronal cell types and cell states. Similarly, targeted experimental perturbation of individual enhancer elements could also be used to test the importance of specific aspects of gene expression to plasticity phenomena. In this regard, efforts to experimentally manipulate enhancer elements within cells have begun to yield high-throughput methods for experimentally testing enhancer sequence variants in the context of their endogenous loci;^182,183 however, employing such methods in the context of the complex architecture of CNS circuitry remains challenging. Moreover, global characterization of the neuronal stimulus-responsive enhancer landscape across CNS cell types will serve as an invaluable scaffold for the interpretation of genetic data, both across evolution and within the context of human neurodevelopmental and neuropsychiatric disorders. Finally, the ability of enhancers to occupy distinct, persistent epigenomic functional states may reveal significant, as yet uncharacterized molecular mechanisms of cellular plasticity relevant to both neuronal and non-neuronal cell types.