Essays in Biochemistry

Epigenetic regulation of genes in learning and memory

Tania L. Roth, Eric D. Roth, J. David Sweatt

Abstract

Rapid advances in the field of epigenetics are revealing a new way to understand how we can form and store strong memories of significant events in our lives. Epigenetic modifications of chromatin, namely the post-translational modifications of nuclear proteins and covalent modification of DNA that regulate gene activity in the CNS (central nervous system), continue to be recognized for their pivotal role in synaptic plasticity and memory formation. At the same time, studies are correlating aberrant epigenetic regulation of gene activity with cognitive dysfunction prevalent in CNS disorders and disease. Epigenetic research, then, offers not only a novel approach to understanding the molecular transcriptional mechanisms underlying experience-induced changes in neural function and behaviour, but potential therapeutic treatments aimed at alleviating cognitive dysfunction. In this chapter, we discuss data regarding epigenetic marking of genes in adult learning and memory formation and impairment thereof, as well as data showcasing the promise for manipulating the epigenome in restoring memory capacity.

Introduction

The brain possesses an amazing capacity for signal integration and information storage. It is appreciated that neurons receive an enormous number of synaptic inputs, and in response to stimuli, whether internal or extrinsic, the physiology of neuronal synapses can be modified for days, months or potentially a lifetime. However, the molecular mechanisms in place in neurons to drive and sustain these changes are poorly defined. One molecular substrate that is responsive to environmental influences and can remain stable for the lifetime of a neuron is chromatin, or the DNA–histone protein complex in the nucleus that compacts and contains the entire genome.

We and others have hypothesized that chromatin serves as an ideal substrate for both the dynamic signal integration necessary for learning and the stable storage of information underlying long-term memory (for other recent reviews, see [13]). One fact making chromatin an ideal substrate for learning and memory is that the human genome contains well over 20000 protein-coding genes, with each gene possessing numerous sites for transcriptional regulation that might subserve signal integration. Secondly, chromatin is not continually degraded and resynthesized, which renders it a stable molecular substrate that might subserve long-term information storage.

Recent studies highlight that the DNA and histone proteins that comprise the core chromatin particle are an integral component of learning and memory processes, and that DNA methylation, DNA demethylation and histone modifications appear to help encode the salience of intrinsic and extrinsic signals. Overall, this is a new way to understand how intrinsic and extrinsic signals evoke not only transcriptional changes, but also structural and functional changes in the physiology of the CNS (central nervous system). For our purposes, in this chapter, we define epigenetics as the covalent modification of chromatin that influences activity-dependent changes in gene transcription that drive cognitive processes. In the following sections, we briefly review these mechanisms, and then discuss evidence for their role in synaptic plasticity, learning and memory, and cognitive dysfunction. We also discuss the promise of drug targets of epigenetic machinery to enhance and even restore memory.

DNA methylation and demethylation

One way the genome can be epigenetically regulated is through DNA methylation (Figure 1). Catalysed by a class of enzymes known as DNMTs (DNA methyltransferases), DNA methylation is the direct chemical modification of a cytosine side chain by the covalent addition of a methyl (-CH3) group. DNMTs transfer methyl groups to cytosine residues at the 5-position of the pyrimidine ring [4], and only cytosines that are immediately followed by a guanine are targets for such methylation. These CpG sequences (the ‘p’ simply indicates that a phosphodiester bond connects the cytosine and guanine nucleotides) are highly under-represented in the genome, and often occur in small clusters commonly referred to in the literature as CpG islands [4].

Figure 1 DNMTs transfer methyl groups to cytosine-guanine dinucleotides

MeCP2 binds methylated DNA and recruits chromatin-remodelling co-repressor complexes, including HDACs. This condenses chromatin, which then limits accessibility of the transcriptional machinery to gene promoters, and thus suppresses gene transcription.

To date, DNA methylation has been in most cases associated with the negative regulation of gene transcription. As it is currently best understood, methylation of CpG dinucleotides triggers the recruitment of transcriptional suppressors, such as methyl-DNA-binding proteins [such as MeCP2 (methyl CpG-binding protein 2)] and HDACs (histone deacetylases). This then results in a higher-affinity interaction between DNA and the histone core through localized regulation of the three-dimensional structure of DNA and its associated histone proteins [4]. As depicted in Figure 1, this cascade ultimately suppresses gene transcription. However, whereas DNA methylation is predominately associated with suppression of gene activity, recent studies provide evidence that DNA methylation status alone is not an accurate measure of whether a gene is suppressed, as it appears that, in some cases, MeCP2 can also be associated with transcriptional activators and active gene transcription [5,6].

DNA demethylation has been historically viewed as a passive and largely irreversible process in differentiated cells, whereby several rounds of cell division without DNMT-mediated remethylation are necessary to erase epigenetic marks. Whether there is active DNA demethylation in post-mitotic cells, including neurons, is still under debate. Several active DNA demethylation pathways have been proposed, including the direct removal of methyl groups via MBD2 (methyl-CpG-binding domain protein 2) [7] or a Gadd45 (growth-arrest and DNA-damage-inducible protein 45)-coupled DNA repair-like process [8,9].

Histone modifications

Another way the genome can be epigenetically regulated is through post-translational histone modifications (Figure 2). In the nucleus, DNA is wrapped around a core of eight histone proteins (histones 2A, 2B, 3 and 4, with two copies of each molecule). The N-terminal tails of histones protrude beyond the DNA and can be post-translationally modified [10]. At present, we understand these modifications to include acetylation, phosphorylation, methylation, ubiquitination and SUMOylation. Whereas acetylation is typically linked with gene activation, SUMOylation is associated with gene repression. However, some histone modifications, particularly methylation, are more complex and can be associated with either gene activation or repression depending on the nature of the modifications [10].

Figure 2 Acetylation of lysine residues by HATs decreases the affinity between histones and DNA, thereby relaxing chromatin and making DNA accessible for transcription

The opposite reaction is carried out by HDACs.

Of the histone modifications, histone acetylation has been the most studied with regard to learning and memory. Acetylation of histones occurs at lysine residues, via enzymes called HATs (histone acetyltransferases), effectively decreasing the affinity between the protein tail and DNA, and relaxing the chromatin structure, which allows the recruitment of transcriptional machinery [10]. As Figure 2 illustrates, histone acetylation is a reversible process, and the enzymes that catalyse the reversal of histone acetylation are the HDACs.

Epigenetic mechanisms in synaptic plasticity

Biologists have known for decades that epigenetic mechanisms, particularly DNA methylation, are involved in cellular differentiation, helping to encode, for example, a pattern of gene expression that enables one cell to be a skin cell, and then a distinct pattern that enables another cell to be a liver cell. Biologists have also recognized that heritable patterns of DNA methylation enable a cell to retain its identity through subsequent cell divisions. Thus DNA methylation has historically been viewed as static process in modifying the genome.

Perhaps some of the best evidence that DNA methylation is actively regulated in post-mitotic cells and that the genome remains responsive to stimuli from the environment came by way of work from Michael Meaney’s group in 2004 [11]. Using rodents that displayed various levels of maternal care, they demonstrated that the quality of maternal care received during the first postnatal week (high levels of care compared with low levels of care) directly influenced DNA methylation patterns of the glucocorticoid receptor gene in offspring, and this in turn programmed the expression of this gene throughout the animal’s lifespan. These data, along with those regarding the substantial expression of DNMTs (specifically, DNMT3a and DNMT1) in non-diving neuronal cells [1214] were among the first to suggest that the DNA methylation machinery remains active and responsive in the CNS. Such observations provided the impetus for several laboratories, including our own, to begin to investigate the capacity of changes in DNA methylation to regulate synaptic plasticity and memory in adult animals.

Changes in DNA methylation and LTP (long-term potentiation)

Synaptic plasticity refers to the process in which synapses undergo activity-dependent changes in their strength. Synaptic plasticity is a long-standing and candidate cellular mechanism for information storage in the CNS, and has been implicated in a wide range of cognitive functions. As such, several laboratories have examined whether DNA methylation (or demethylation) has a role in synaptic plasticity. LTP is a form of synaptic plasticity whereby synaptic strength is enhanced in response to high-frequency synaptic activity. Work from our laboratory provided the first demonstration that the acute application of the demethylating agents zebularine or 5-aza-2-deoxycytidine to hippocampal slices from the adult brain not only alters the methylation status of DNA [of the Bdnf (brain-derived neurotrophic factor) and reelin genes], but also disrupts the induction of LTP [15,16]. The ability of these same drugs, or the direct disruption of the DNA methylation machinery itself via a DNMT-knockout animal, to block induction of LTP have been independently confirmed and extended by others [1719]. Because both zebularine and 5-aza-2-deoxycytidine are nucleoside analogues that need to be incorporated into DNA to trap DNMT and block DNA methylation, the mechanism of how these drugs are able to alter methylation in hippocampal slices and affect LTP is not clear, but most likely involves a replication-independent event (i.e. a DNA repair-like process). As an aside, zebularine and 5-aza-2-deoxycytidine are undergoing clinical evaluation for their promise as anticancer treatments.

Histone modifications and LTP

Some of the earliest evidence for the role of histone modifications in synaptic plasticity came from studies in the laboratories of Eric Kandel, Mark Mayford and Ted Abel in which they investigated the role of a particular HAT, CBP {CREB [CRE (cAMP-response element)-binding protein]-binding protein}. CBP acetylates histones associated with CREs. In two examples of these types of studies, using genetically altered mice with impaired CBP function (but void of any effects of CBP on development), the Kandel and Abel laboratories found that CBP/HAT-deficient mice exhibited pronounced deficits in LTP [20,21]. In a complementary series of studies, we hypothesized that blocking HDACs and thus augmenting histone acetylation would enhance plasticity. Thus we found that acute application of either sodium butyrate or trichostatin A indeed enhanced hippocampal LTP [22].

Epigenetic mechanisms in learning and memory

DNA methylation and demethylation in learned fear

In our laboratory, we have recently begun examining whether associative learning and memory consolidation elicit any changes in methylation of CpG dinucleotides in the adult hippocampus. Contextual-fear conditioning is a paradigm we have commonly used to assess hippocampal-dependent learning and memory function. In this paradigm, animals learn to associate a novel context (the conditioned stimulus) with a mildly aversive unconditioned stimulus (typically, this is a brief foot-shock) that naturally elicits a freezing response (the unconditioned response). In our first study, we found that, whereas associative learning and fear memory formation elicits rapid methylation and transcriptional silencing of the memory-suppressor gene PP1 (protein phosphatase 1), this also elicits rapid demethylation and transcriptional activation of the synaptic plasticity gene reelin [23]. Additional work from our laboratory demonstrated that the associative learning and memory formation induced by contextual-fear conditioning also evokes rapid methylation and demethylation of various regulatory regions of the Bdnf gene [24]. Thus both active methylation and demethylation appear to serve as important layers, much like transcription factors, in controlling activity-dependent gene regulation necessary for memory.

We have also used various pharmacological approaches (zebularine, 5-aza-2-deoxycytidine or RG-108) to demonstrate that disrupting DNA methylation and DNMT activity before training blocks hippocampus-dependent memory formation in the contextual-fear conditioning paradigm [23,24]. In related work, Lisa Monteggia’s laboratory has shown that disruption of the function of MeCP2 impairs an animal’s ability to form long-term memory in the cued-fear conditioning paradigm [25]. Cued-fear conditioning evokes an amygdala-dependent learning process similar to that described for contextual-fear conditioning, but, in this case, an animal forms a conditioned fear response to a cue instead of a context.

Changes in DNA methylation and long-term gene regulation

The stable nature of DNA methylation renders it an ideal substrate for the long-term cellular changes necessary for the maintenance and persistence of memory. To date, behavioural studies have examined whether there are changes in DNA methylation at what are deemed relatively short time periods following an experience. For example, the 1–2 h time frame we have examined post-learning would, at best, include the early stages of memory consolidation. The question of whether these epigenetic changes have contributed to the maintenance and persistence of a learned-fear memory has yet to be addressed. We do know that the hippocampal changes in methylation are rather short-lived, in that within 24 h of the contextual-fear conditioning, they are no longer present [23,24]. This would therefore suggest that these DNA methylation and demethylation changes brought about by contextual-fear conditioning are supporting memory formation rather than long-term storage of the fear memory in the hippocampus. This then begs for future investigations to address whether these experiences have evoked similar modifications in other regions of the brain that are more linked to the stable and long-term storage of the recently acquired information, such as the prefrontal cortex.

If we look at the developmental literature, we find some of the best evidence that experiences produce stable and long-term epigenetic changes in regulation of genes. Some of our work has demonstrated that maltreatment of an infant rat by a caregiver induces significant methylation of Bdnf DNA in the prefrontal cortex, an effect that persists throughout development and well into adulthood [26]. This hypermethylation is associated with lasting reduction in the expression of the Bdnf gene [26]. Stable changes in expression of the AVP (arginine vasopressin) gene in the paraventricular nucleus caused by maternal separation stress have also been recently linked to lasting changes in DNA demethylation of the gene [27]. If epigenetic mechanisms can contribute to the memory of early-life experiences on the cellular level, then it is certainly plausible that they can contribute to the long-term memory induced by contextual-fear conditioning.

Histone modifications in learning and memory

Some of the earliest evidence for the role of histone modifications in memory came from the same studies that investigated the role of CBP in synaptic plasticity. As alluded to above, investigators demonstrated that CBP/HAT-deficient mice had pronounced deficits in memory capacity, including long-term memory deficits for passive avoidance, novel object recognition and cued-fear conditioning [20,21,28]. Some of our initial work in this realm revealed that in mice that had learned a contextual-fear association, there were significant increases in both acetylation and phosphorylation of histone H3 that were regulated by the ERK (extracellular-signal-regulated kinase)/MAPK (mitogen-activated protein kinase) pathway [22,29]. Furthermore, we showed that the same HDAC inhibitors that enhanced hippocampal LTP also enhanced memory in these animals [22]. Together, these observations were among the first to indicate that epigenetic marking of histone tails occurs in long-term memory formation and that, surprisingly, manipulation of these processes is a viable way to alter memory capacity.

In 2007, Li-Huei Tsai’s group published a seminal study in which they demonstrated that the use of sodium butyrate is sufficient to restore learning and memory in cognitively impaired mice [30]. This beneficial effect of various HDAC inhibitors in improving learning and memory in non-diseased rodents and in other models of neurodegeneration and brain injury has continued to be replicated. For example it has now been shown by Marcelo Wood’s laboratory that HDAC inhibition (via sodium butyrate) can facilitate extinction of a cocaine-induced conditioned place preference [31]. Finally, in one of our latest series of studies, we found the systemic delivery of one of various HDAC inhibitors (valproic acid, sodium butyrate or suberoylanilide hydroxamic acid) was sufficient to rescue the cognitive deficits present in a mouse model of Alzheimer’s disease [32]. As such, there is much excitement in the field surrounding the potential clinical utility of HDAC inhibitors in treating neurodegenerative and neuropsychiatric diseases.

A handful of studies have examined the involvement of histone modifications at specific gene loci that control gene transcription in long-term memory. For example, we have shown that the memory of a contextual-fear association involves increased acetylation of histone H3 at promoter IV of Bdnf, which parallels the experience-induced changes in the expression of this specific transcript [24]. The extinction of conditioned fear in mice has been shown to evoke histone H4 acetylation around Bdnf exon IV in the prefrontal cortex [33]. Finally, Isabelle Mansuy’s group has just shown that the selective inhibition of PP1 in forebrain neurons (via a transgenic mouse) not only enhances various forms of long-term memory, but that it does so through specific histone modifications occurring at the promoter of the memory-linked CREB and NF-κB (nuclear factor κB) genes [34].

Investigators continue to show that stressors and memories of stressful events in adulthood evoke complex patterns of histone modifications. For example, chronic social defeat stress in adult mice produces a lasting down-regulation of hippocampal Bdnf transcripts III and IV that is associated with increased histone H3 lysine methylation at their promoters [35]. A psychological challenge, such as forced swimming, has been shown to evoke histone H3 Ser10 phosphorylation and Lys14 acetylation in the dentate gyrus, through the ERK/MAPK pathway [36]. Additionally, Bruce McEwen’s group has just shown that acute or chronic restraint stress evokes patterns of hippocampal histone H3 trimethylation that vary by hippocampal subregion [37]. Discovering the complexity of such histone modifications not only offers an exciting new way to understand how intrinsic and extrinsic signals are integrated by the hippocampus, but also will probably help us to understand how stress can produce such long-term damage on the hippocampus and whether this damage is reversible.

Epigenetic mechanisms in cognitive dysfunction

It is becoming increasingly clear that epigenetic mechanisms help to regulate both the dynamic and stable capacity of the CNS. Thus an epigenetic contribution to cognitive dysfunction associated with psychiatric and brain disorder in aging continues to gain traction, and evidence gathered to date supports this hypothesis. For example, a growing body of literature suggests that aberrant DNA methylation of reelin and Gad1 (glutamic acid decarboxylase 1) may underlie dysfunction of GABAergic neurons in schizophrenia, and altered GABA (γ-aminobutyric acid) activity appears to be responsible for at least some of the clinical features of schizophrenia (for a recent review, see [38]). Several studies have reported aberrant DNA methylation and histone acetylation in Alzheimer’s disease patients (for a review, see [39]). Of course, the question remains unanswered whether these epigenetic alterations are causally related to the pathogenesis of such disorders and diseases. However, data gathered from post-mortem studies will undoubtedly provide us with some very important clues in understanding the complex gene–environmental interactions that probably confer susceptibility and lead cognition awry.

Conclusions

A central tenet in the field of epigenetics has been that the epigenome is static following cellular development and differentiation. Although this may be the case for many cells, the data reviewed in this chapter indicate that this is not always the case for neurons. There is increasing evidence that neurons and specific genes within neurons retain their sensitivity to epigenetic factors beyond development, and that epigenetic regulation of gene transcription is a necessary component in CNS plasticity and memory (Figure 3). The involvement of epigenetic factors in memory is motivating the interest of an increasing number of investigators, which will undoubtedly revolutionize our understanding of how the CNS works and how we form both transient and enduring memories. Furthermore, the continued study of epigenetic mechanisms in cognition promises a future where epigenetic therapy may help to combat or alleviate memory dysfunction prominent in CNS disorders and disease.

Figure 3 Epigenetic mechanisms provide a substrate for long-term neural and behavioural modifications

Learning and remembering the salience of extrinsic stimuli, such as that occurring with contextual-fear conditioning, modulates the activity of chromatin-modifying enzymes, such as HATs, HDACs and DNMTs. This evokes activity-dependent changes in DNA methylation and histone tail modifications, and ultimately the gene transcription that is necessary for the synaptic plasticity and behavioural modifications underlying long-term fear memory.

Summary

  • Studies continue to highlight the emerging role of the epigenome in CNS function and memory in the adult.

  • Following contextual-fear conditioning, there is both rapid DNA methylation and demethylation of memory-linked genes in the hippocampus of adult rats. There is also involvement of histone modifications at specific gene loci during fear-memory formation.

  • Interfering with DNA methylation or histone acetylation disrupts both LTP and memory formation. Drugs that promote histone acetylation enhance memory capacity in cognitively impaired animals.

  • An understanding of epigenetic mechanisms in learning and memory is relevant not only to helping us better understand the processes underlying cognition, but will also be important and relevant for future therapeutics.

Acknowledgments

This work was funded by grants from the National Institutes of Health, the National Alliance for Research on Schizophrenia and Depression, Civitan International, the Rotary Clubs CART fund, and the Evelyn F. McKnight Brain Research Foundation.

References

View Abstract