We all know that we are able to store sensory experiences for later recall; in a very real sense, collection of memories acquired during life represents our personal identity. However, the neuroscience of memory formation is still in its infancy. We have, at best, a rudimentary understanding of a few of the many complex processes involved. Most memories last for a few brief moments, and are then lost forever. It was thought that long-term memory formation involved structural changes in neural connections via protein synthesis; especially, memories associated with fear seemed to involve structural changes in the lateral nucleus of the amygdala. Glutamate binds to NMDA receptors which in turn activates gene expression and protein synthesis via protein kinases such as MAPK and PKA. The proteins generated during the consolidation into long-term memory may be involved in restructuring the shape of the axon, others may increase the number of receptors on the receiving dendrite to lower the threshold needed to fire across the synapse. It was once thought that such long-term memories, more than 24hrs-old, were permanent, since they resulted from structural changes in the brain.
The standard model for memory formation has been long-term potentiation (LTP). (The opposing process is referred to as long-term depression or LTD). LTP refers to the reinforcement of synaptic connections increasing the responsiveness of post-synaptic neurons to pre-synaptic stimulation. Such alterations in synaptic strength are referred to as “plasticity”, and are thought to form the basis for memory and learning. LTP, for example, explains why the repetition of information improves our ability to recall it. One central problem with this model of memory is the fact that synaptic configurations are transitory. The proteins that form the synaptic receptors and ion channels that comprise synaptic connections degrade over time and are constantly recycled. Despite this, memories are durable, many lasting a lifetime. Although synaptic plasticity may play a fundamental role in the formation and recall of memories, long-term memories can not reside within the synaptic configurations. They must be stored elsewhere, and must be capable of regulating the synaptic connections. Where are memories stored and how is this information able to regulate the formation of synaptic connections? I would like to discuss some recent advances that shed light on these questions, and discuss some implications for therapeutic approaches to prevent memory loss, or perhaps even intentionally “erase” them.
Epigenetic modifications are various types of changes which alter the expression of the genes, which regulate the formation of specific proteins which then affect cellular function. Recent research has shown that sensory stimuli can produce epigenetic changes. Three types of epigenetic modifications have been shown to play a role in memory formation, DNA methylation, histone modification, and microRNA regulation. MicroRNAs can regulate protein formation by binding to mRNA after it is transcribed, and preventing it from being translated into amino acids. DNA methylation refers to the addition of methyl groups, or “marks” to DNA; such marks do not alter the genes themselves, but are usually associated with repression or reduced expression of those genes. DNA strands are wrapped around spools or nucleosomes which are composed of eight protein cores called “histones”. The tail ends of these core proteins can be modified by the addition or removal of compounds such as acetyl groups, which can increase or decrease the expression of those genes.
Histone Modifications: HATs and HDACs
In 2004, the first demonstration that histone modifications were involved in long-term memory formation was provided, see; Regulation of Histone Acetylation during Memory Formation in the Hippocampus
Researchers discovered that two different types of memory both resulted from the acetylation of two distinct histones in the hippocampus, a brain area long-known to be crucial for the consolidation of long-term memories. Acetylation of histone H3 was associated with fear memory, and acetylation of H4 with latent inhibition. As the name suggests, “fear memory” is the ability to remember that certain stimuli are associated with an adverse event. “Latent inhibition” refers to the brains ability to learn that certain sensory stimuli are of no importance, and do not need to be remembered. It was soon discovered that histone phosphorylation also plays a role in memory formation (ref, ref). Histone acetylation results from specialized enzymes, histone acetyltransferases (HATs). Trangenic mice with reduced HAT activity display various deficits in long-term memory formation.(ref, ref) The enzymes responsible for removing acetyl groups from histones are histone deacetylases (HDACs). If inhibiting HAT activity interferes with memory formation, one is led to wonder if inhibiting HDAC activity might improve memory. The answer is yes. HDAC inhibition has been shown to improve various types of memory, as well as treat a wide range of neurological disorders. See:
Targeting HDACs: A Promising Therapy for Alzheimer’s Disease
Multiple roles of HDAC inhibition in neurodegenerative conditions
However, a major challenge is the problem of specificity. Altering the acetylation of histones is likely to have very different effects in different cell types. Moreover, other proteins such as transcription factors, signalling proteins, etc. are likely to be affected with unpredictable consequences.(ref)
Erasing Memories: an approach to treating PTSD
Post-traumatic Stress Disorder (PTSD) is a devastating condition with profound implications for the future quality of life of the patient and family members. Military cases, in the U.S. alone, are known to exceed 200,000, more than at any other time in history, including during two world wars (according to VA statistics). The total number of cases is likely very under-reported; and this figure does not include non-military patients, such as rape victims.
Fear memories, like many other types of memory, go through a multi-step process, from the original association of sensory stimuli with an adverse event to the formation of long-term memory. Traumatic experiences are associated with the production of the stress hormones adrenaline and cortisol. Adrenergic activation is known to enhance memory formation, consolidation, and recall, making stressful memories particularly vivid and persistent. PTSD was historically treated by antipsychotic and antianxiety drugs. Then antidepressants often combined with various forms of psychotherapy became the the preferred treatment protocol. None were very successful. Currently, research is focusing on ways to reduce the association between the conditioned stimulus and the fear response, often called “exposure therapy”. This can be achieved by two different mechanisms, enhanced fear extinction, or disrupted reconsolidation (sometimes called “consolidation blockade.”)
Blocking consolidation relies on the fact that there is a delay between the adverse event and the long-term consolidation of that memory. If a drug that disrupts consolidation can be given before the adverse event, or soon afterwards, then the long-term fear association will not be established. Since adrenergic activation is involved in this process, adrenergic blockers have been used effectively to reduce, or even eliminate the fear memory in animal models. Memory of the event itself is not erased. Beta-blockers like propanolol and alpha-blockers like prazosin have both been used. The technique can even be effective years later. This is because each time a memory is recalled, the synaptic connections associated with that response are thought to be reconsolidated, making the memories susceptible to alteration or extinction.(ref) In fact, it has been shown that memory recall induces protein degradation, actively destabilizing the memory, thus allowing it to be either extinguished or reconsolidated.(ref, ref)
Therapeutically blocking this reconsolidation involves asking the patient to recall or retell the taumatic event while an adrenergic antagonist is given. This results in depression rather than potentiation of the synaptic connections. The fear memory is (partially) forgotten. Although early studies originally met with great enthusiasm; so far, human clinical trials have produced disappointing results. Adrenergic blockers can prevent traumatic stress, if taken in advance; but they appear ineffective at treating PTSD.(ref)
The mechanism of memory extinction is nearly the opposite of consolidation, though the final objective is the same, to lessen the conditioned fear memory. Extinction results from learning a new memory to replace the fear memory. This approach makes use of compounds that enhance memory formation and synaptic plasticity, rather than interfere as do adrenergic antagonists. Different synaptic connections are formed to replace the previous ones associated with the fear memory. In this case, PTSD is viewed as a failure of normal memory extinction. D-cycloserine (DCS) has been discussed as a possible compound for enhancing memory extinction. DCS is an NMDA agonist, so it increases glutamatergic activity, which is necessary for new memory formation, and is thought to be deficient in PTSD patients.(ref) Another potential target of pharmaceutical intervention to enhance fear extinction is FGF2 (Fibroblast Growth Factor 2), which has been shown, in at least one rodent study, to improve fear extinction.(ref) Future studies will need to show whether or not either of these two targets, glutamate or FGF2, can improve fear extinction in humans.
A 2010 study suggests another potential target – BDNF.
Induction of Fear Extinction with Hippocampal-Infralimbic BDNF
Brain-drived neurotrophic factor (BDNF) is known to play a crucial role in the synaptic plasticity necessary for the consolidation of new memories, and possibly the extinction of old ones. One remarkable result of this study is that unlike FGF2, BDNF when infused into rodent brains resulted in extinction of conditioned fear memory without any extinction training. (FGF2 merely enhanced the effect of extinction training.) Rats who failed to learn extinction were found to have reduced BDNF in the hippocampal pathway leading to the IL mPFC (Intralimbic medial Prefrontal Cortex). The IL mPFC is the region where the BDNF infusion was given. It was previously known that electrical stimulation of this region reduced conditioned fear, and enhanced extinction learning.(ref) When BDNF was increased in these pathways, these rats also spontaneously learned conditioned fear memory extinction, indicating an important role for BDNF in both the mPFC and in the connective pathways to the hippocampus. It is interesting to note that PTSD patients are known to have reduced brain volume in both the hippocampus and the mPFC.(ref) The researchers also show that extinction of the fear memory did not erase the memory of the traumatic event. It is further demonstrated that BDNF-induced fear extinction depends upon NMDA receptor activity.
This ingenious study successfully identified BDNF as the “key molecular mediator” of fear extinction. The fact that the fear extinction occurred spontaneously without any extinction training shows that it does not result from long-term potentiation, or from latent inhibition. Researchers further demonstrate the necessity of NMDA receptor activity for this effect. When the BDNF was co-administered with an NMDA antagonist results were the same as in the control group.
Previously, research on fear conditioning has focused on the amygdala, the center for emotions. Although the amygdala plays a central role in emotional learning and the expression of fear. PTSD patients have hyperactive amygdala and hypoactive mPFC and hippocampus. Clearly, the PFC must process information from many brain regions; however, the hippocampal pathway was shown to be of primary importance in fear extinction learning, not the amygdala.
The researchers even go on to suggest a potential pharmcological approach to increase BDNF in PTSD patients. Acetylation of histone H4 in the hippocampus results in increased expression of BDNF, and is correlated with extinction. (ref) Valproic acid (VPA) has been effectively used to stabilize mood and to reduce fear. VPA also happens to be an HDAC inhibitor. It appears likely that the fear extinction effect of VPA results from increased BDNF expression caused by acetylation of H4.(ref) It is likely that other HDAC inhibitors with greater specificity could be much more effective at increasing BDNF, and thereby treating or even curing PTSD. Regardless of the approach used, these findings establish BDNF as a primary target for successfully treating PTSD.
Environmental Enrichment and Acetylation: Recovering Lost Memories
It has long been known that increased environmental stimulation improves memory.(ref) In a 2007 study, Recovery of learning and memory is associated with chromatin remodelling researchers confirmed the memory benefits of environmental enrichment (EE), and demonstrated that EE could not only enhance memory, but also recover lost memories, even after severe neuronal and synaptic loss had occurred due to extensive brain atrophy. (These results further demonstrate that long-term memories are not stored in synaptic connections.) EE resulted in increased histone acetylation in the hippocampus, strongly suggesting that EE improves memory by means of this epigenetic mechanism. To add further support to this hypothesis, researchers demonstrated that the memory effects of EE could be replicated by the use of an HDAC inhibitor. This strongly implies that HDAC inhibition could also allow humans to recover lost memories, even after substantial neurodegeneration. If epigenetic treatments can indeed recover lost memories after substantial brain atrophy, an intriguing question is what kinds of memories might be recovered in a person with a healthy brain?
DNA methylation is catalyzed by enzymes known as DNA methyltransferases (DNMTs). Memory researcher, David Sweatt, summarizes his research on DNA methylation: “In a recent series of studies my laboratory has investigated the capacity of DNA methylation, the other major epigenetic molecular mechanism besides histone modification, to regulate synaptic plasticity and memory in adult animals (55,56). In our first series of studies in this area we found that inhibitors of DNMTs that likely block the net effects of both maintenance and de novo DNMTs could alter DNA methylation in adult CNS tissue and block hippocampal Long-term Potentiation (LTP) in physiologic studies vitro (55). In additional more recent studies we found that de novo DNMT gene expression (DNMT3a and DNMT3b) is upregulated in the adult rat hippocampus following contextual fear conditioning, and that generalized DNMT inhibition blocks memory formation in this same paradigm (56). In addition, fear conditioning was associated with rapid methylation and transcriptional silencing of the memory suppressor gene Protein Phosphatase 1 (PP1) and demethylation and transcriptional activation of the synaptic plasticity gene reelin. These findings have the surprising implication that both DNA methylation and demethylation might be involved in long-term memory consolidation. Overall these results suggest that DNA methylation is dynamically regulated in the adult nervous system and that this cellular mechanism is a crucial step in memory formation.” The result that both DNA methylation and demethylation are involved in memory consolidation is surprising, because no enzymes have yet been identified to demethylate DNA.
The interplay between epigenetic modifications is illustrated in a study, Epigenetic Alterations Are Critical for Fear Memory Consolidation and Synaptic Plasticity in the Lateral Amygdala, published last year (2011). Researchers demonstrate that both histone acetylation and DNA methylation work together to regulate emotional memory in the amygdala. In particular, they show that H3 acetylation increases expression of DNMT3A (a gene that encodes one of the DNMT enzymes) and that pharmacologic manipulation of either histone acetylation or DNA methylation “enhances or impairs, respectively, memory consolidation and associated synaptic plasticity” in the amygdala.
MicroRNA Regulation of Memory and Learning
DNA methylation and histone modifications regulate DNA transcription, the first step in the synthesis of proteins. MicroRNAs are small non-coding RNAs that usually regulate the next step, translation. One of the mechanisms by which they do this is called “RNA interference”. MicroRNAs (miRNA) bind to mRNA after transciption, and degrade them, thereby preventing translation into amino acids. For an informative video of this process, see: RNAi animation. Since miRNAs are usually associated with the suppression of new protein synthesis, one might speculate that eliminating them would enhance learning and memory function. Indeed, this seems to be the case in rodent models, at least when eliminated globally. In one study, a necessary protein, “Dicer” (depicted in the video) was deleted from adult mice forebrains. This had the effect of neutralizing all Dicer-dependent miRNAs. The mice showed improved learning and memory, as well as an increased number of a type of dendritic spine associated with memory formation.
However, many specific miRNAs have been identified that play positive roles in memory formation. MiR-132, for example, induces neurite growth and regulates their integration into the hippocampus.(ref) We have discussed fear extinction in some detail; and yes, a miRNA has been found to play an important role in the formation of fear extinction memory, miR-128b.(ref) We have also discussed BDNF, which promotes neurogenesis and synaptic plasticity. Recent research suggests that BDNF is both regulated by miRNAs, and that its effects are mediated by upregulating miRNAs.(ref, ref) For a reviews of the role of miRNAs in memory and aging, see:
MicroRNAs in Neural Stem Cells and Neurogenesis
New neurons in aging brains: molecular control by small non-coding RNAs
MicroRNA regulation of neural plasticity and memory.
Steroid Homones and miRNAs
Aging is by a characterized by a decline in sex steroid hormones, as well as an increase in circulating cortisol. These changes are accompanied by decreased cognitive capacity and memory function. Since steroid hormones have profound effects on the expression and activity of miRNAs, some researchers have concluded that many of the effects of age-related cognitive decline are mediated by miRNAs:
Importantly, estrogens and glucocorticoids are strong regulators of the miR biogenesis pathway. Both hormones have been shown to control the expression of Dicer-1 and other key enzymes in miR synthesis in different experimental systems (Yamagata et al., 2009; Smith et al., 2010b). These observations suggest that steroid hormones may be crucial in favoring the expression of miR sets or “signatures” involved in the coordination of gene networks (Castellano et al., 2009; Eendebak et al., 2011). Although the effects of steroid hormones are strongly tissue and cell type specific, these observations suggest that steroid hormone regulation of miR biogenesis could be involved in the changes in miR expression associated with aging in the brain (Somel et al., 2010; Eda et al., 2011; Khanna et al., 2011; Wang et al., 2011b).
Low levels of circulating estrogens in post-menopause females have been linked to cognitive deficits (Smith et al., 2010a). In rats, estrogen replacement after ovariectomy increases LTP and dendritic spine density in hippocampal neurons, suggesting a key role of estrogen signaling in synaptic plasticity (Smith et al., 2010a). The estrogen receptor ? (ER?) is a steroid hormone receptor that can be acetylated – and thereby activated – by p300, a target of miR-132 (Kim et al., 2006b). In addition, SIRT1 is found to promote ER? expression (Yao et al., 2010). Overall, these data indicate yet another potential pathway regulating synaptogenesis, in which miR-132 could be central.
Cortisol production by the adrenals influences memory and cognition during aging. Higher cortisol levels are associated with a poorer memory performance and a higher likelihood of memory decline, especially in women. These detrimental effects of cortisol seem to be directed at the hippocampus (McEwen et al., 1999; Li et al., 2006). In healthy elderly individuals, cortisol levels seemed to be associated with cognitive impairment (Kalmijn et al., 1998). Therefore, stress and resulting increases in glucocorticoid levels may have important consequences on the degree and speed of decline in memory and other cognitive abilities in the elderly (Lamberts, 2002). Although increasing levels of glucocorticoids are not always found in aged individuals, high levels of glucocorticoids are associated with synaptic loss in the hippocampus, hippocampal atrophy, and cognitive decline during aging in some individuals. These observations have led to the suggestion that glucocorticoids may contribute to, or accelerate aspects of aging (Nichols et al., 2001). Therefore, although stress and increased glucocorticoid levels may not contribute to aging in all individuals, they could decrease structural plasticity and the brain’s vulnerability to disease (Radley and Morrison, 2005; Korosi et al., 2011) resulting in a pro-aging activity in vulnerable individuals (Wolkowitz et al., 2009).
Despite the previously discussed inhibition of AHN [adult hippocampal neurogenesis] by glucocorticoids in the DG [dentate gyrus, part of the hippocampus], the relationship between plasma glucocorticoid levels, receptor expression and AHN is complex. Interestingly, studies from our lab have demonstrated that a brief treatment with the GR antagonist mifepristone rapidly reverses the deleterious effects of chronic stress on AHN (Oomen et al., 2007), strongly suggesting that the GR is involved in chronic glucocorticoid hormone suppression of AHN. The observations that GR expression, particularly in the DG, is increased in depressed elderly women and within this group correlates positively with age, suggests that GR activity could be linked to disease mechanisms during aging (Wang et al., 2011a). Gene profiling studies in chronically stressed animals have shown that CREB is central in the signaling pathways regulated by the GR in the DG (Datson et al., 2010).
As we have discussed before, CREB is part of a central pathway in the regulation of AHN (Merz et al., 2011) and this pathway crosstalks to several miRs involved in the regulation of NSC proliferation, differentiation, and synaptogenesis (Figure ?(Figure1B),1B), in particular the neuronal activity-induced, miR-132 (Nudelman et al., 2010). Notably, GR activation suppresses miR-132 expression and results in a decrease in BDNF and glutamate receptors (Kawashima et al., 2010). These observations suggest that high glucocorticoid levels observed in many aging individuals may result in a GR-dependent inhibition of miR-132 expression and reduced glutamate receptor expression in adult-born immature neurons of the DG. This hormone and miR mediated pathway could induce significant changes (e.g., reduced synaptogenic potential) in adult-born neurons in susceptible patients that are not observed in healthy aging individuals.” (ref)
Clearly, epigenetic mechanisms play an crucial role in the formation and preservation of long-term memories. Epigenetic treatments have the potential to improve memory function and cognitive ability. This field of study is brand-new, but is advancing at a rapid rate, so I expect many exciting developments in the future. Obviously, many membrane receptors, signaling molecules and cellular structures also play important roles in memory and cognitive function, although my focus here has been, primarily, on epigenetic mechanisms.
It is quite interesting that even among individuals who are very knowledgeable of epigenetic modifications, few realize that they happen on a daily basis from mere exposure to sensory stimuli. Most “experts” in the field still seem to think that the effects result merely from maternal diet, etc. The idea that these changes take place on a daily basis throughout life, and play a role in the formation of memories of those sensory stimuli, is without a doubt, revolutionary. The old definition of epigenetic modification as semi-permanent and hereditable, may need to be expanded.
However, this does lead to an interesting question. If there really is some kind of epigenetic, or histone code for memories (see: ref, ref), and if these changes are hereditable, is it then possible that offspring maintain some type of latent memory of the sensory experiences of their ancestors? I know this sounds like Science Fiction, but we now have a known mechanism that could account for such an effect. This could further explain the observed evolutionary development of innate behavior based on environmental conditions. See: Ref. This effect, which is observed across many species in different environments, happens much to quickly, to be the result of natural selection acting upon random mutations.