Engram Cells, Memory Formation & Alzheimer’s Disease: The Competition of Neurons in Becoming Part of the Past

(1)

L

ITERATURE

T

HESIS

Engram Cells, Memory Formation & Alzheimer’s Disease

The Competition of Neurons in Becoming Part of the Past

M

ELISSA DE

R

EUS

,

B.S

C

.

Student number : 10000423 Brain and Cognitive Sciences, track Behavioural Neuroscience

IIS, ABC, University of Amsterdam Supervisors: Dr. H.J. KRUGERS (SILS-CNS) Co-Assessor & UvA-representative: Dr. C. LANSINK (SILS-CNS) Date of submission report: 2 July 2019

(2)

2

Abstract

Memories contain the entire past of an organism. Memory retrieval is necessary for an

or-ganism to be able to recognize safe and dangerous situations, to search for food, to navigate

through new and familiar environments. The medial temporal lobe contains the structure

necessary for the storage and retrieval of memory. The neural substrate is to be found in the

MTL while representing the memory trace, or memory engram. This concept, the memory

engram theory, was established by Richard Semon in 1902. Back then, Semon already

con-sidered the memory storage (engraphy) and retrieval (ecphory) as underlying concepts for

memory processing. A memory engram consists of a subset of selected neurons, so-called

engram cells. Engram cells should represent the lasting changes of its physical and/or

chemi-cal properties established during learning experience and will continue during rechemi-call and

re-trieval of that memory. Nowadays, this concept has been thoroughly studied and with the

current innovations and technologies, it is now possible to combine various approaches in

order to identify, manipulate and characterize the engram cells. For example, optogenetics

label the neurons that are activated during new learning experiences. This allows researchers

to manipulate the neurons at later time points and to find the underlying role in the memory

storage and retrieval. Different types of studies can be considered using these techniques,

namely observational, loss-of-function and gain-of-function studies, in which the role of the

engram in memory is established. The aim of this thesis is to find the underlying cellular

mechanisms of the memory engram (memory trace). The three types of studies answer how

and why specific neurons get allocated for the memory engram. Together, it reveals the

spe-cific properties the engram cells. This contributes to a better understanding of the process of

learning and memory and the underlying cellular mechanisms. Besides, during memory

for-mation and updating, the engram is very susceptible to disruption causing memory

impair-ments. In the case of Alzheimer’s Disease, memories are considered to be lost. However, with

the contemporary approaches it is possible to rescue to those forgotten memories. This could

provide new insights for potential treatments of memory recovery in Alzheimer’s Disease.

(4)

4

1. Introduction

The memory is defined by the dictionary [1]_{as “The mental capacity or faculty of retaining and reviving} facts, events, impressions, or of recalling or recognizing previous experiences”. Mental capacity ena-bles an organism to recognize safe and dangerous situations, to search for food, to navigate through new and familiar environments. In other words, the ability to recall experiences from the memory contributes to the organism’s survival in the environment.

Since the 1950’s the complex mechanisms that enable the organism of storing and recalling memories have been thoroughly studied1,2_{. Especially in the medial temporal lobe (MTL, Figure 1) structures,} consisting of the hippocampal areas and the parahippocampal region including the entorhinal, perirhinal and postrhinal cortex (Figure 2).These structures are considered major and necessary com-ponents of declarative memory, the organism’s ability to recall everyday facts and events2,3_{. In that} decade, Scoville and Milner corroborated the underlying function of the MTL for memory through the famous case-study of Henry Molaison (H.M.)1_{. After a bicycle accident H.M. suffered from epileptic} seizures. These were to be remedied by the removal of the MTL unilaterally, including the hippocam-pus, entorhinal and perirhinal cortices. Removal of the MTL resulted in an impaired declarative memory1,4_{. Hereafter, H.M. suffered from anterograde and graded retrograde amnesia, as he was} unable to form learning-induced memories and to recall of memories from recent events before the onset of the amnesia2_{. Other cognitive and perceptual capacities remained unaffected, leading to the} idea that hippocampal structures are necessary at the onset and during the learning process. It con-firmed specialized role of the MTL structures in the consolidation of declarative memories.

Figure 1. The anatomical position of the medial temporal lobe. A; shows the MTL in the rat brain laterally, termed as the postrhinal cortex. B; displays the MTL in the monkey brain ventrally. C, dis-plays the MTL in the human brain ventrally, with the borders repre-sented of perirhinal cortex in grey, the entorhinal cortex with diag-onal stripes, and the parahippocampal cortex with mottled shading (adapted from Burwell et al. 1996)3_.

The onset of memory consolidation is presumed to happen directly after learning, triggering the stor-age of new memories. Additionally, consolidation is the updating and stabilization of previously formed memories into long-term memories5_{. Since the fifties, plenty of researchers have been} cor-roborating the role of the MTL in memory encoding, formation and retrieval. For instance, the hippo-campal area CA1 (cornu ammonis 1) was demonstrated to be essential for memory formation, as lesioning this region bilaterally impaired the memory similarly to patients with amnesia6_{. Additional} findings to the role of the MTL and the hippocampus were added by Moser and Moser in 19987_{. They} demonstrated that the hippocampus functions as an integrative hub for various memory traces, which are evidently differentiated along the hippocampus’ dorsoventral axis7_{. Eichenbaum and} col-leagues validated this important role of the MTL, and emphasized the functional organization of many different memory processes (like spatial information, object recognition and organization of time) in all the regions of the MTL8_.

How are the memories represented in the brain? German scientist Richard Semon proposed a con-cept for the underlying mechanisms of the memory and its formation already in the beginning of the twentieth century: the memory engram theory4,9_{. Based on his theory, the memory engram is defined} as a physical representation of the memory in the brain, also known as the memory trace9_{. In addition} to the memory engram concept, he established two hypotheses: the Laws of Engraphy and Ecphory, representing memory storage and memory retrieval, respectively4_{. The idea of this framework is that}

(5)

5

the memory results from the synergy between the already stored memory engram and retrieval cues. Semon’s work stimulated several researchers to study the underlying mechanisms of the memory very closely, allowing them to describe the memory engram as the lasting changes of the physical and/or chemical properties of a neuron established during learning experience and continued during recall and retrieval of the memory4,10_.

How do these lasting alterations form the memory engram leading to the given memory? Various lines of evidence have convincingly demonstrated that synaptic communication is essential for memory formation10_{. During memory formation and (re)consolidation the brain undergoes selective} modifications of learning and memory substrates at the level of synaptic plasticity. The underlying synaptic strengthening is a form of plasticity that tweaks the strength of connections between neuronal popula-tions10_{. The alterations of synaptic strength are} ex-pressed in long-term potentiation (LTP) and long-term depression (LTD)11–13_{. The process of LTD weakens the} synaptic basis selectively by activity-dependent reduc-tion, from which the smallest fraction can already cause memory inactivation disabling its formation11,12_{. LTP} promotes synaptic strength by a long-lasting increase of signal transduction generated through high-frequency stimulation (HFS)12,14_{. The activation patterns transform} the temporary, susceptible and labile state (early in the memory formation process) into more long-lasting sta-ble states, known as the memory engram10,15_. Altera-tions in chemical synapses are an essential contribution to LTP as the trafficking and modifications of the glu-tamatergic receptors AMPA and NMDA[2]_{are key in the} excitatory transmission14,16_{. It provides the increase of} presynaptic glutamatergic release and postsynaptic al-terations of the glutamate receptor activity leading to short-term memory formation17_{. The transformation to stable long-term memories is depending on} de novo synthesis of the proteins. Successful synthesis will lead to the growth of new dendritic spines and synaptic connections. In turn, it will improve the connections between the pre- and postsynaptic neurons16_{. After initial consolidation the memory engram can be reactivated again, by restabilizing} the neuronal bases of the memory, referred to as reconsolidation. It represents the state between active retrieval and encoding, besides it is essential for refining and recovering the formed memory. Reconsolidation was demonstrated to be of great importance, as blocking this process using protein synthesis inhibitors lead to memory impairments10_{. Hence, inhibition of the protein synthesis makes} the memory very vulnerable and easily disrupted during (re)consolidation of short-term memories causing retrograde amnesia5,16_.

With the current innovations of technologies and approaches, including the recent view on Semon’s engram theory, researchers are nowadays able to characterize and follow neurons of the engram. Therefore, this thesis focusses on when a specific neuron is allocated for the memory engram? Why and how are particular neurons identified as engram cells? What are the cells underlying properties? Furthermore, with amnesia as a defining feature of Alzheimer’s Disease (AD), it remains unclear what disruptions underly to the inability of memory retrieval or even memory loss. Thus, here it is ques-tioned, whether engram cells and its specific properties have altered underlying mechanisms in AD? If so, what are characteristic disruptions of the engram in AD? Firstly, it is discussed what the current methodologies and technologies are and how the engram is identified, characterized and

2_{AMPA is α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid. NMDA is N-methyl-D-aspartat.}

Figure 2. The Medial Temporal Lobe struc-tures: A schematic view. The structures im-portant for declarative memory and their connections. S = subicular complex; DG = dentate gyrus; CA1 and CA3 = the CA fields of the hippocampus. (adapted from Burwell

(6)

6

ed using those. Next, description of the molecular and cellular properties of an engram cell will be discussed thoroughly. Finally, to find out what the underlying mechanism is of learning and memory and to contribute to the understanding of amnesia in AD.

2. Identification, Characterization and Manipulation of the Engram Cells

To study the engram and the engram cells, the neurons involved must be identified, isolated and ma-nipulated. What techniques and approaches are used to identify, characterize and manipulate the memory engram cells? The studies demonstrating these methods were able to find what is needed for neurons to be selected for an engram and thus represent the corresponding memory.

2.1 Behavioral Paradigms

The memory-related expressed behavior is usually studied with distinct behavioral paradigms. To identify and characterize memory engram cells, the approach requires a simple learning paradigm including fear conditioning (FC). In most memory studies, the Pavlovian FC paradigm is applied (Figure 3). Pavlovian FC paradigm involves the exposure of a neutral conditioned stimulus (CS) paired with an aversive unconditioned stimulus (US)4,10_{. The CS can be for instance a new context (in contextual FC)} or an auditory cue (in cued FC). The aversive US can be a foot shock, its frightful characteristic will result in a conditioned response (CR) observed as fearful behavior: freezing. The freezing behavior of the rodent comprises the absence of all activity of the rodent except for respirational movement. The US will probe the rodent to link the earlier neutral CS to the fearful element of the US, resulting in eliciting the same fearful CR after subsequent exposure to solely the CS10_.

Figure 3. Pavlovian fear condi-tioning. In fear conditioning the rodent is placed in a particular context and a neutral condi-tioned stimulus (CS) in the form of a tone of a new context is coupled to an aversive uncondi-tioned stimulus (US), resulting in the tone-driven conditioned fear response (CR). This re-sponse is measured in form of freezing behavior. As soon as the rodent is subsequently exposed to the CS it will result in the CR expressed as fear behavior. From Goshen (2014)18_.

Hereafter, a time of consolidation is provoked lasting for long periods of time (from minutes to months) even after single conditioning trials18_{. This allows the accurate determination of the brain} locations during the conditioning process. The relation between the neutral CS and aversive US is mostly mediated by the amygdala. The hippocampus is more involved with the more contextual in-formation memory storage, while the auditory cues (such as the tone) is hippocampal-independent memory18_{. The hippocampal-dependent memory (contextual FC) is confirmed by repeating the trial} such that the rodent is located back in the initial conditioning context and the freezing behavior is measured. The freezing behavior represents a successful recall depending on the successful memory storage of the hippocampus and the amygdala. Cued FC is confirmed to represent the same auditory

(7)

7

cues as in the initial FC in a different context. The exposure of the auditory cue alone should elicit the same freezing response for a successful recall and an unaffected amygdala4,10,18_.

Pavlovian FC is a well-established, but simple approach which integrates the thoroughly explored anatomy of the brain areas involved in fear memory (amygdala, hippocampal area), the cellular changes as synaptic strength and the measurable changes of phenotypic behavior (freezing)18_{. It} func-tions after single-trials and is therefore easily reproducible4,10_{. A combination of the FC paradigm with} optogenetics and transgenics allow the determination of the learning moment, and the changes on cellular and behavioral level. This can accurately locate the causal and temporal neuronal ensembles in the fear memory-related structures and their functional connections. The FC paradigms will provide the association between underlying mechanisms as synaptic plasticity, engram cell formation and the memory trace4,10_.

2.2. Techniques and Methodological approaches

How to approach the identification, characterization and manipulation of the engram cells? In the search for when neurons are considered the engram cells, studies by Tonegawa (2015) and Josselyn (2015) assessed that the allocated neurons should be activated by learning experience, have lasting alterations chemically and molecularly, and its reactivation should cause the retrieval or recall of the initially formed memory4,10_{. Those criteria are studied in three different types of studies:} observation-al, loss-of-function, gain-of-function studies4_{. The approaches examine the functionality of the} memory engram (cells), but also methods disrupting the formation or reactivation of the engram, the consequences of these manipulations and possible solutions to fix the disruptions4,10_{. All to show the} relevance for the specific memories expressed behaviorally. Observational studies represent the cor-relation of the activity of the studied neurons with the associated behavioral expression. Gain-of-function studies are shown to obtain stronger evidence for the memory engram definitions. Since it reveals the sufficiency of a studied population of neurons in relation to the memory expressed in be-havior. It displays how a specific cell population is sufficiently activated to induce the related memory expression. Lastly, the loss-of-function studies contain the strongest evidence for the memory engram definition, as it indicates the necessity of the particular population of neurons studied for the associ-ated behavioral memory expression4_.

2.2.1 Observational Studies

Observational studies identify the potential neurons as engram cells, demonstrating learning-induced alterations in the brain representing the encoding phase as well as the prediction events during re-trieval4,10_{. Capture studies belong to observational studies and use methods to study the expression of} immediate early genes (IEGs), which are induced by neural activity during memory encoding10_{. This} method, along with other IEG-based methods, has regularly been used for the identification of poten-tial neuronal ensembles. The advantage of this method is taking into consideration the inipoten-tial repre-sentation of the memory and the visualization of the reactivation at later times, while having a high temporal specificity10_{. In capture studies the learning-induced activated neurons are labeled with a} protein marker at times of the initial learning trials, whereas at subsequent learning trials a different marker is applied. The markers function as indicators for neurons involved in the encoding of the memory and thus a potential engram cell4,10,17_{. This strategy uses innovative technologies such as} transgenics and optogenetics, allowing researchers to permanently label putative engram cells, even long time after encoding.

The TetTag method (Box 1) is such an approach, as it manipulates the expression of the IEGs, e.g. c-Fos and ZIF268, but also the activity-regulated cytoskeleton protein, Arc4_{. Previous studies found} in-creased Arc and c-Fos expression directly after a learning event. The increase of these IEGs indicated

(8)

8

synaptic plasticity and LTP4,10_{. These two proteins are therefore proper targets to accurately label} neuron populations involved in an engram17_{. Instead using the IEG itself, its promoter can be used to} mark the neural activity of a neuron with LacZ or green fluorescent proteins (GFPs)10,19_{. The TetTag} approach can be used to examine the neuronal activity of the hippocampal dentate gyrus (DG) when exposed to familiar and novel contexts20_{. Deng et al. (2013) applied this method and found increased} reactivated levels of DG granule cells, which were initially activated during learning20_{. In other words,} the neurons active during retrieval were the same neurons activated during the original learning ex-perience. These correlations were significant with the freezing behavior measured during contextual fear conditioning. Furthermore, exposure to a novel environment stimulated a new subset of DG granule representing the novel input20_.

Similarly, reactivation in the prefrontal cortex (PFC) can be found with the catFISH method. This method uses fluorescence in situ hybridization (FISH) of the cellular compartment analysis of tem-poral activity(cat)21_{. The catFISH method includes the arrangement of neuronal activity for the} expres-sion of Arc mRNA to assess the role of, for instance, the dorsal hippocampus (dHPC), basolateral amygdala (BLA) or the medial prefrontal cortex (mPFC) in contextual fear conditioning21_{. It localizes} Arc mRNA within the cytoplasm, that appears for two minutes and maintains until 16 minutes after initial neuronal activation. The Arc mRNA within the soma can also be identified though its levels rises much later, about 20-45 minutes after neuronal activation21_{. Arc mRNA identification thereby} differ-entiates between temporal events as both the Arc mRNA transcription and translation will be

detect-Box 1. TetTag

TetTag is an approach to identify potential engram cells by labelling learning-induced activated neurons with a transgen-ic marker. It captures activated neurons controlled by tetracycline with the use of self-activated tTA-TetO system com-bined with the c-Fos promoter. With transgenic c-Fos- tetracycline (tet)-off transcriptional activator (tTA) applied in rodents, the putative neuronal ensembles are characterized using the Fos promoter. TetTag controls the Fos-tTA transgene to promote the tTA protein expression in neurons activated during initial learning experiences. Hereafter, the marker gene expression is stimulated through binding of the tTA protein of the first transgene to a tet operator in the promoter of a second transgene (see figure below). The tTA activation can occur before and after training in the activat-ed neurons. The tTA activation is manipulatactivat-ed with doxycycline (Dox) regulation via the diet of the rodent. Presence of Dox in the diet will inhibit the marker gene expression (figure below left panel), as soon as the Dox is removed from the rodent, the marker gene is expressed again (figure below center panel). This is frequently applied at the first behavioral learning trial. The marker gene, for example LacZ or GFP, will be expressed in the neurons active during the trial. Admin-istration of Dox to the rodent via the diet will block marker gene expression again (figure below right panel). The labeling of the gene expression in the preferred time window is approached by immunohistochemical tagging of the protein products using c-Fos or Zif268. The latter is an IEG endogenously expressed by activated cannabinoid receptor type 2’s, involved in neuroplasticity and long-term memory formation19_{. This TetTag method is a well-established approach for}

identification and characterization of neuron population activity during fear conditioning25,30_.

Labelling of activated neurons using the TetTag method. The tTA-TetO system is activated when c-Fos promoter is evoked at the withdrawal of doxycycline (Dox). The presence of Dox will block the expression of c-Fos promoter. TauLacZ, a marker gene, is expressed in the excited neuron. Dox is administered again to prevent from more labeling of other neurons, though the previous labeled neurons remained marked. The red flash represents neural activity. A blue-colored neuron represents tagging during activation.

(9)

9

ed21_{. Because of this property, Zelikowsky et al. (2014) demonstrated the same population of neurons} is associated in the expression of fear memory as the neurons associated with encoding the original contextual fear memory21_{. Additionally, this approach enabled to obtain information about the} dis-tinctive functions in contextual encoding and emotional processing in the HPC, BLA and medial PFC21_. An important limitation of the observational studies is its inability to monitor the cellular and molecu-lar expression representing the memory of the learned behavior over multiple hours or days. Yet, these observational studies were able to demonstrate the correlation between neural activity of the selected neurons and the associated memories, also described in Tonegawa et al. (2015) and Josselyn et al. (2015)4,10_{. Hence, suggesting observational studies as proper studies for identifying whether the} engram cells are found in the studied neuron populations4,10_.

2.2.2 Loss-of-function Studies

The loss-of-function type of studies are considered a proper approach for characterization of engram cells, as it uses the manipulation of the neuronal circuit by elimination or inhibition of necessary memory-associated neuronal substrates. Blocking reactivation of engram cells verifies if these manip-ulations induce memory impairments4,10_{. The manipulation includes the ablation, activation or} silenc-ing of engram cells dursilenc-ing learnsilenc-ing-experiences and at later time points dursilenc-ing consolidation and re-trieval experiences.

Multiple learning and memory studies have manipulated neurons expressing the transcription factor CREB, cyclic AMP-responsive element-binding protein (Figure 4). Increased levels of CREB induced the expression of downstream proteins, regulating gene expression by binding to the DNA sequence CRE (cAMP response element)4,10,22_{. The changes in protein levels, for example of CREB or c-Fos, can} con-tribute to fluctuations in excitability of the neuron. Overexpression or downregulation of CREB is found to induce changes in neuronal plasticity and thus memory formation and retrieval10,17_.

The use of viral vectors is commonly used as a neuronal tracer by expressing exogenous genes in vivo examining the role of CREB in memory processes. An example is the replication-defective herpes sim-plex viral vector (HSV-mediated) administered via microinjection in the region of interest23_. HSV-mediated studies found increased levels of CREB after exposure to FC23,24_{. More specifically, studies} found auditory FC-induced activated CREB in about 20% of neurons in the lateral amygdala (LA)23,24_. The manipulation of CREB was performed by injection of HSV-expressing endogenous CREB (CREBWT₎ or dominant-negative CREB (CREBS133A_{) in a portion of the LA neurons. The expected freezing behavior} was observed in the animals with CREBWT_{, while the animals with CREB deficiency showed little} freez-ing behavior23_{. Moreover, overexpressing of CREB in mice with CREB deficiency, revealed the recovery} of the fear memory of only 3.2% of CREB activated neurons, while the freezing behavior was fully recovered similarly to the WT animals. Interestingly, the injection of CREBWT_{into the LA of the} CREB-deficiency mice did not recover the memory impairment23_.

Memory retrieval disruption can also be initiated using the diphteria toxin(DT)-based method4,24_. Herein, an allocate-and-manipulate concept is applied in which FC-induced CREB-overexpressing neu-rons were ablated directly after FC training. As soon as the administered DT bound to the DT receptor, apoptosis was activated24_{. Mice naturally lack a functional DT receptor, therefore transgenic mice} with Cre-recombinase-inducible DT receptors (iDTR mice) were used. Fear memory before and after DT-induced cell death (only the CREB-overexpressing neurons) was monitored, and decreased freez-ing behavior was observed. However, only selectively ablatfreez-ing those CREB-overexpressfreez-ing neurons induced memory impairment, as random ablation of LA neurons did not affect the memory24_{. Besides,} the memory loss assessed before the surgery remained unaffected after ablating the CREB-overexpressing neurons. Despite the loss of those selected engram cells, re-learning of the fear

(10)

10

memory did not impair subsequent learning, demonstrating other subsets of neurons are sufficient for memory encoding10,24_{. The approaches with HSV and DTR indicated downregulating the CREB} ex-pression or deletion of CREB-overexpressing neurons affected the behavioral fear outcome and there-fore the loss of its function in fear memory23,24_{. CREB expression of the identified neurons during} ini-tial learning-induced activity influences the likelihood of a neuron becoming allocated as a fear en-gram cell. Additionally, it could assume that neural competition is at the basis of the formation of memory. The relative activity of CREB is key for the selection of neurons to the memory engram spe-cific to an associated memory experience10,23,24_.

Figure 4. CREB cascade. Cyclic AMP-responsive element-binding protein (CREB) is the cellular activity-induced transcription factor which regulates gene transcription through binding to cAMP response elements (CRE). Gene transcription, including that of c-Fos, is modulated by the CREB expression. The cascade is initiated by a signal-induced glutamatergic input on the cell surface evoking an action potential. This neural activity causes the release of cAMP or Ca2+_{, identified as second messengers, to subsequently}

stimu-late protein kinase (e.g. MAPK/ERK). Within the cell nucleus CREB is activat-ed as a result of the kinases. The CREB will link to the CRE region and CREB-binding protein (CBP) on the c-Fos promoter to enable to on-or-off switch-ing of the specific gene. The CREB cascade can hereby mediate underlyswitch-ing memory processes. It was found to be important for dendritic spine num-ber and morphology modulation, to play a role in neuronal plasticity and long-term memory formation, and to induce memory impairment via inac-tivation25_{. From Cruz et al. (2014).}

Another approach to manipulate the neurons for the loss of function is the DREADDs method (Box 2). It induces the excitation or inhibition of neurons. Taking into consideration the loss-of-function, DREADDs can be applied to inactivate selected neurons by manipulating the inhibitory neuronal sig-naling cascade (hM4Di), regulating the memory formation25,26_{. A method to approach this is the} appli-cation of the tamoxifen-inducible Cre recombinase (CreER) system. The appliappli-cation of tamoxifen in-jection into mice enabled the labeling of active neurons27_{. Transgenic mice with ArcCreER}T2_bacterial artificial chromosome were used to study the role of hippocampal subareas in memory traces27_{. The} administration of tamoxifen in combination with CFC labeled the neuronal subsets activated in the DG and CA3 at the moment of a learning experience27_{. The allocated neurons were also activated during} fear memory recall tests. After the identification of the participating neurons of that memory engram, optogenetics (Box 3) was performed to inactivate the previous allocated DG and CA3 cells of the CFC training. The impact on subsequent fear memory expression caused impairment of the previous ex-pressed fear memory4,27_.

The DREADDS method can also be applied in combination with TetTag, for instance in a double trans-genic complex of the c-Fos-tTA/tetO-EGFP-TeNT28_{. The first part of c-Fos-tTA/tetO is based on the} Cre-system, the time window of labeling the active neurons was Dox-controlled during FC learning28_. TeNT complex (tetanus toxin light chain) selectively suppressed the synaptic transmission of the allo-cated neurons. TeNT manipulates a protein needed for exocytosis, inhibiting the release of neuro-transmitters in the synaptic cleft28_{. The Dox-regulated TeNT synthesis was initiated in the allocated} neurons and formed long-term memory (LTM) during CFC. Hereafter, Dox-treatment continued to block the further TeNT synthesis. Impaired fear memory retrieval was observed by the significantly less freezing behavior 24 hours later. The application of TetTag and DREADDs was also used to assess the compensation of the brain to such neuronal inhibition and discover whether relearning was af-fected. Again Dox-regulated TeNT synthesis was stimulated for neuron allocation during CFC, with retraining in the same context to stimulate reactivation of the allocated neurons. Interestingly, there was no increase in the freezing behavior in the mice with inhibited neurons, meaning that the relearn-ing was hindered28_{. The mice with unaffected neurons showed strengthened fear memory after the}

(11)

11

same training. Lastly, the same procedure was performed with a novel, distinct context. This revealed similar freezing behaviors in healthy mice and mice with previously inactivated neuron. It proved that inhibited neuronal population impeded relearning, however it did not affect the ability of obtaining and retrieving of a novel fear memory28_.

These loss-of-function studies were able to first identify the allocated neurons of the memory en-gram, followed by inhibition or disruption the neurons to reveal their role in memory formation and retrieval. The DT method showed to be a spatial and temporal accurate approach to selectively elimi-nate the allocated neurons, causing memory impairment24_{. Others showed DREADDs as a functional} approach initiating loss-of-function27,28_{. Observations showed neuronal competition is an important} concept for cells to become allocated for the engram. Furthermore, inactivating or disrupting engram

Box 2. DREADDs

Designer Receptors Exclusively Activated by Designer Drugs (DREADDs) are genetically engineered recombinant recep-tors used for identification and manipulation of (non-)neuronal signal transduction in brain circuits associated to behav-ior, emotion, memory, perception, all underlying brain functions25_{. It genetically encodes the neuron activity modulator.}

DREADDs take over the function of the G-protein coupled receptors (GCPR) to silence or activate neuronal firing. The GPCRs are receptors activated by ligand binding and involved in e.g. the cAMP signal pathway (CREB cascade). The DREADDs are engineered GCPRs to which the natural endogenous ligands cannot bind and thus the signaling cascade is inhibited. This way the DREADDs temporally and spatially take-over the in vivo G-protein signaling. The engineered GCPR, is a modified muscarinic GPCR which is applied in the cell by viral vectors. It is used mostly the construct of a muscarinic acetylcholine receptor (M) and edited through mutagenesis prohibiting the high affinity for acetylcholine (natural ligand), while preserving it for a synthetic ligand. The engineered GPCR are produced to interact solely with synthetic ligands (designer drugs) such as clozapine N-oxide (CNO). It qualifies as inert, highly potent and has a high bioavailability in the host lacking other pharmacological specificity. The effect of CNO linkage to the DREADD replicates the effect of the acetylcholine to the M. CNO can be administered to the rodent very easily through injection, food or water intake, or via a pump25,26_.

The application is performed by viral vectors and monitored by a recombinase-based system. Cre-dependent adeno-associated viruses (AAVs) can develop the functional DREADDs as these engineered receptors will be expressed in Cre-expressing neurons. Various transgenic mouse models are potent for restricted Cre-expression and are therefore used, or a viral construct is administered via stereotaxic microinjection in preferred location of the brain. DREADDs are linked to inhibitory or excitatory signaling cascade on which the CNO can bind. For example, the M4 muscarinic DREADD (hM4Di) is a popular form of a modified human M4 and coupled to inhibitory cascade. The cellular effects are increased K+_{efflux eliciting hyperpolarization, decrease of intracellular Ca}2+_{, cAMP and neuronal firing, silencing the neural}

activa-tion. Another popular CNO activated DREADD receptor is the hM3Dq. This is a modified form of human M3 muscarinic receptor and linked to the excitatory signaling. The linkage of CNO to the hM3Dq receptor causes the increase of intra-cellular Ca2+_{, protein kinases (ERK) and neural firing, while stimulating neuronal activity in form of burst firing}26_.

(12)

12

cells selectively erased the associated memory, while novel, distinct memories were still formed, as different subset of neurons were allocated to the associated memory engram.

2.2.3 Gain-of-function Studies

In addition to the loss-of-function studies, gain-of-function studies demonstrate the artificial activa-tion of the allocated neurons associated to learning experience4_{. Recent studies combine methods for} allocation and tagging of neurons with techniques comprising the selective activation of neurons. The identification and manipulation of brain cells activated during memory formation is enabled by com-bining optogenetics (Box 3) with activity-dependent drug-associated gene expression system (Box 1). The described Dox-regulated TetTag method is used to regulate the c-Fos expression and label the neuronal activity (Box 1)19_{. While at the same time, optogenetics is performed for light-stimulated} gene regulation. Specifically, channelrhopsin-2 (ChR2) can be expressed after administration of the viral vector AAV-TRE-ChR2-EYFP (Box 3)29_.

A viral vector like AAV-TRE-ChR2-EYFP can be expressed in the DG to identify the granule cells in-volved in the contextual fear memory30_{. As described in Box 1 and 3, the activation by contextual FC} labeled the activated neurons with ChR2 under Dox regulation. Only a sparse population of those granule cells (2-3%) were activated during CFC. Nevertheless, reactivation of the specific engram suc-ceeded when the allocated neurons received blue-light stimulation via optical fibers. This was suffi-cient to induce the freezing behavior as seen in the natural contextual testing10,30_{. Additionally, when} testing in the initial context (non-fear associated) without the light-stimulation, the freezing behavior was absent30_{. Similar to this, blue light-stimulation of allocated neurons without the labeled ChR2 also} lacked freezing behavior. This approach indicated light-stimulation is a sufficient to reactivate the prior ChR2 tagged neurons in order to evoke memory recall4,10,30_{. The findings of Liu et al. (2012)} sug-gested the initial fear-induced allocation of DG neurons is needed to retrieve a fear memory. Howev-er, it is sufficient to induce the fear memory recall by using light-stimulation of those sparse, memory-specific DG neurons30_{. This functioned as the basis for more studies using the combination of} optoge-netics and activity-dependent gene expression onto the identification and manipulation of memory engram cells.

Box 3. Optogenetics

Optogenetics is a technique using optics and genetics to manipulate and observe the gain and loss of function of accu-rately determined processes in the (brain) cells of living organisms49_{. Light sensitive proteins (opsins) are targeted by}

light-stimulation with millisecond precision. The opsins are basically light-sensitive ion channels expressed through the genetic modification of brain cells50_{. Their conformation is altered from resting into signaling state by absorbing light}

through illumination, inducing inhibition or excitation of neuronal populations. Inhibition is induced by the microbial opsins Halorhodopsin (NpHR) or Archaerhodopsin-3 (Arch). These are ion pumps activated by yellow light mediating the hyperpolarization inhibiting the neural activity50_{. The most used microbial opsin is Channelrhodopsin-2 (ChR2), as it}

effectively manipulates the kinetics, wavelength and ion selectivity imitating the neuron populations. On the contrary, this is an opsin evoking excitation upon depolarization which was induced by illuminating the light-sensitive cation chan-nels with blue light50_{. Illumination of the opsins is enabled by the implantation of an optic fiber into the rodent’s brain}

above the region of interest using stereotaxic points of reference. The light is delivered via impulses which provide milli-second precision and selected frequencies. Different wavelengths are applied as the opsins react to various types of light, for example yellow or blue light, characterized with distinct wavelengths (570–590 nm or 450–495 nm, respective-ly). To drive the expression of the preferred opsin DNA, a viral vector including a short and strong promoter of neuronal population with the opsin DNA sequence package is directly inserted into the region of interest, such as the DG, via microinjection (figure below, A). Viral vectors, among which the adeno-associated viruses (AAV), are able to induce the opsin expression in the cell membrane and even over distinct neuronal fibers in a period from days to weeks50_{. However,}

when the promoter is too weak or the virus too big, the opsin DNA is expressed in transgenic mouse under a strong promoter, for instance Cre recombinase. The transgenic mice are injected with an opsin gene in inverted orientation with two incongruous Cre sites, controlled by a ubiquitous promoter. Solely, the neurons expressing Cre transgenetically will have the DNA flipped in correct orientation, allowing transcription of the preferred genes. > Continues page 13.

(13)

13

The approach described above contributed to the possibility to couple a fear associated context to a neutral context with optogenetic stimulation, creating a false fear memory for the neutral context31_. Transgenic mice with AAV-TRE-ChR2-mCherry were created to promote the gene expression for tag-ging the fear-activated in either the DG or CA1 neurons31_{. Under the regulation of Dox the animals} were trained in a neutral context followed by a fearful context. The neutral context allocated the neu-rons, while these were optically stimulated in a new context (Figure 5). Freezing behavior in the DG-manipulated mice was significantly increased together with the c-Fos expression in those neurons after re-exposure to the neutral context. The CA1-manipulated mice did show increased c-Fos expres-sion, although it lacked increased freezing behavior. Both groups showed a little freezing behavior after being placed back in the fear conditioned context, indicating a natural response to the fear stim-ulus. The addition of a novel context (D in Figure 5) lacked the freezing response, however,

re-Box 3. Optogenetics (continued)

The most frequently used approach is accurate manipulation by the activity-dependent opsin expression (figure below A & C). A transgenic mouse line of c-Fos-tTA is used to express the tTA moderated by the c-Fos promoter expression after neuronal activity. During a learning experience tTA expression is found in activated neuron. The expression of an opsin is driven by tTA binding. This is achieved through injecting a tetracycline response element (TRE)-controlled viral virus encoding for the preferred opsin (e.g. AAV-TRE-ChR2-EYFP). The tTA expression is controlled by the Dox-rich diet, limit-ing the tTA expression in a given time window. The Dox will block the TRE from linkage to the tTA. The withdrawal of Dox will therefor induce the expression of opsins29,50_.

From Liu et al. (2012).

In the behavioral experiment, the animals will be on Dox-diet when habituating to the context A (see figure B). During training in context B they will be taken of Dox, enabling the labeling of ChR2 to the DG neurons activated at the moment of memory encoding (see figures B & C). At testing the neurons will be illuminated with light while they are back on Dox. During the pretraining or habituation there was no freezing behavior observed, while post-training with light-stimulation freezing behavior was shown (figure D). The behavioral results proven optogenetic-induced fear memory retrieval. This is confirmed by the figures E and F representing the DG neurons of c-Fos-tTA mice labeled with AAV9-TRE-ChR2-EYFP (figure E) or AAV9-TRE-EYFP (figure F) after optogenetic stimulation. This demonstrated that the mice which received the AAV9-TRE-EYFP were not labeled with ChR229_.

(14)

14

exposure to the optical light-stimulation to the DG and CA1 neurons would evoke the freezing behav-ior again. The combination of these techniques showed retrieval of false memories including some specifics, namely it is context-specific and controlling the fear behavior response31_{. Furthermore, the} levels of c-Fos expressed in amygdalar neurons was increased during false and real fear memory re-call, indicating the optical reactivation of the DG and CA1 stimulated downstream brain areas31_.

Figure 5. Protocol and results of the creation of a false memory. Green means on Dox diet. The yellow flash represents the foot shock. The asterixis represent the neurons selected through activation by either context or light exposure. Red circles represent neuron allocation for context A, grey circles for the FC context B, and white for distinct context C. Context A, the red pyramid, is the initial neutral context in which the mice are exposed while off Dox. Hereafter, the mice withdrawal of Dox is ended. Subsequently, FC context B in which a foot shock is given and the allocated neurons from the initial context A are stimulated optogenetically with light. After 24 hours the mice are re-exposed in context A and freezing behavior is measured or placed in distinct context C and freezing behavior was also measured. From Ramirez et al. (2013).

The role of the amygdala in the memory engram can be further explored by assigning a valence to the memory. As up to now only aversive (negative) behavioral responses and therefore negative memo-ries are explored. The techniques were also used to switch the valence of those memomemo-ries32_{. Neuronal} ensembles of the dorsal DG and the BLA of transgenic mice with AAV-TRE-ChR2-mCherry were labeled with ChR2 when activated during a fear (aversive) or reward (appetitive) conditioning under Dox-regulation32_{. The fear and reward memories were evoked by two real-time optogenetic contextual} conditioning paradigms. Basically, the rodent was placed in a rectangular box in which it choose its preferred chamber by free exploration that chamber most of the time32_{. The fear memory was} formed by foot-shock (US)-induced aversive response elicited during optogenetic place avoidance test (OptoPA). The reward memory was created by female mouse interaction (US) inducing appetitive response initiated during the optogenetic place preference test (OptoPP)32_{. The same number of} neu-rons of the BLA and DG neuneu-rons were labelled with ChR2 after either aversive or appetitive contextual conditioning32_{. Optogenetic stimulation in the OptoPA and OptoPP would cause reactivation of the} fear and reward engram besides evoking aversive or appetitive behavior, respectively32_.

A memory valence switch from fear-to-reward or from reward-to-fear, related to either DG or the BLA engram, needs an induction period after the original valence reactivation. During the induction the initially allocated (ChR2-labelled) neurons were optogenetically stimulated while exposed to the op-posite cue (either foot-shock or female mice presence)32_{. Next, the optogenetic performance test of} the last presented cue (either OptoPA or OptoPP) was executed to evoke the behavioral and neuronal response. The reversal of the valence memory succeeded in mice whose DG-engram was manipulated showed a behavioral switch representing the memory valence reversal. Mice with BLA-engram ma-nipulation lacked the behavioral indication that a memory valence reversal occurred. The valence reversal was also visible on cellular level revealing a decrease in the overlap of BLA cells stimulated by

(15)

15

DG engram cell stimulation32_{. This could mean an alteration in functional connectivity between the} DG and BLA. Since the valence reversal diminished the capability of the DG engram to project success-fully to the BLA engram. The change of memory valence is able thanks to the alterations of functional connections between the amygdala and hippocampal area32_.

Current technologies and approaches as activity-dependent labeling, allows identification and manip-ulation of cells. Moreover, it allows to demonstrate specific learning-activated neurons are necessary and sufficient for the retrieval of fear memory5,19,30_{. Although, these allocated neurons also refer to} elements of the impaired memories. Moreover, recent approaches were combined with a protein synthesis inhibitor to induce amnesia by disrupting the cellular mechanisms of neuronal consolida-tion5_{. CFC induced activity-dependent cell labeling was performed with the AAV9-TRE-mCherry virus} in the DG of transgenic c-Fos-tTA mice5_{. After this neuron identification took place the CFC activated} neurons were labeled with mCherry. Amnesia was induced by administration of the protein synthesis inhibitor anisomycin (ANI) and as a control saline (SAL) was administered. Both were directly adminis-tered after foot-shock during CFC5_{. Studying the connectivity through perforant pathway was} execut-ed through AAV-CaMKIIa-ChR2-EYFP virus labeling in the presynaptic EC neurons to express ChR2 in the overlapping cells. The connectivity strength could be measured through synaptic strength with a voltage clamp on the DG cells at the same time as optogenetic stimulation of the allocated and non-allocated neurons5_{. This allowed measurement of excitatory postsynaptic currents (EPSCs) and the} AMPA/NMDA ratio, both indicating the synaptic strength. The connectivity was examined within the distributed memory engrams by measuring the spine density, potential connections (measuring opto-genetic-induced excitatory postsynaptic potentials) from DG mossy fibers to the CA3 through virus injection into the CA3 (AAV9-TRE-mCherry). The synaptic strength (the EPSCs and AMPA/NMDA ratio) and dendritic spine density were increased in the consolidated allocated neurons of the non-amnesic animals. During the consolidation of learning experiences, the protein synthesis inhibitor disrupted the synaptic strengthening and decreased the number of spines and AMPA/NMDA ratio. The ChR2 DG engram axons overlapped with the CA3 mCherry neurons showing increased potential connections after optogenetic stimulation in both amnesic and non-amnesic animals. This indicated a preferred engram cell-to-engram cell connectivity, which was unchanged by the protein synthesis inhibition. With a different CFC test, OptoPA, allocated neurons were labeled with ChR2 regulated by Dox5_{. After} the fear cue the protein synthesis inhibitor or saline was administered. Freezing behavior was ob-served in both amnesic as non-amnesic mice, though significantly decreased in the amnesic mice. When directly stimulating the allocated neurons, the freezing behavior increased in both amnesic as non-amnesic mice. This direct optogenetic-induced memory recovery was found in the DG, CA1 and the LA, despite different training and testing conditions. The disrupted freezing behavior was ob-served after disturbing the memory engram with ANI. Optogenetic stimulating the memory engram which was allocated before the administration of ANI revealed complete recovery of the fear memory behavior5_{. The DG connectivity with downstream CA3 and BLA was studied by reactivating mCherry} labeled engram cells through exposure to a natural cue or with optogenetic-stimulation in OptoPA test. The light-stimulation demonstrated an equivalent percentage of both c-Fos and mCherry positiv-ity in the amnesic and non-amnesic mice in the CA3 and BLA. Whilst the natural cue revealed signifi-cantly higher c-Fos and mCherry overlap in only those areas of the non-amnesic mice5_{. This indicated} there is a persistent and specific connectivity between the engram cells of the DG and the down-stream CA3 and BLA in both non-amnesic and amnesic brains, referring to underlying fundamental processes of memory5_{. It allowed to retrieve the artificially created false memories, by taking} ad-vantage of the unstable state of the memory engram when re-activated, followed by reconsolidation of the new association in the initial memory engram. As during the consolidation, the memories are still susceptible for disruptions, it is interesting what processes disturb the stability. Even the protein synthesis inhibitors evoking retrograde amnesia were unable to permanently erase the memory stored in the engram cells, formed before disruption5_{. These studies found the fear memory engram}

(16)

16

representing context-specific, however plastic, connectivity pattern of engram cells. This was sup-ported by the synaptic strengthening of the neurons involved in the engram cell circuit5_.

In conclusion, the gain-of-function studies evidence for allocated neurons of the engram is obtained when a memory recall results from the manipulated neural activity29_{. It allowed to reactivate the} allo-cated population of neurons initially active during learning and later at recall. Therefore, it was possi-ble to predict the elements of a given memory which retrieval is induced at moment of stimulation of an associated engram4,29_{. The combination of optogenetics and activity-dependent gene expression} system has the spatial and temporal qualities to be able to accurately manipulate the neurons partici-pating in the memory engram4_.

3. Cellular and Molecular Properties of the Engram Cells

A specific subset of neurons distributed over multiple brain areas can be allocated as a memory en-gram associated with a specific memory. In the previous chapter, the identification and manipulation of the engram cells was discussed, which revealed the cellular and molecular properties of neurons needed to become allocated as engram cells. These properties are found to be necessary and suffi-cient to form and retrieve the memory. What are these cellular and molecular properties of the en-gram cells?

3.1. Neuronal Competition: Increased Excitability

In multiple studies it has been shown that the degree of excitability determines which neurons will be allocated to be part of the engram. Specifically, research showed the more excitable or active the neurons are at the time of training, a higher chance of being selected as an engram cell4,10_{. The} excit-ability of the neurons has been related to the expression of CREB. As manipulation of the CREB func-tion changed the neuronal reactivafunc-tion and thus altered memory retrieval5,23,24_{. Because of this, CREB} expression would also indicate whether a neuron was allocated as an engram cell. It contributed to the cells competition over neighboring cells to become selected for the engram. Nevertheless, various studies considered well-formed memory is based on neuronal competition and the strongest neurons will be part of the memory engram10_{. Competition is considered inevitable as the lack of a winner will} lead to memory impairment. The ‘winner’ neuron in turn will inhibit its neighboring cells, suppressing the other neurons to keep the memory alive33_.

The underlying role of intrinsic excitability for memory engram formation was demonstrated by artifi-cially manipulating the function of voltage-dependent potassium (K+_{) ion channels in the DG and} LA34,35_{. Since excitability depends on distribution, composition and the properties of ion channels such} as K+_{in the plasma membranes}34_{. A HSV vector}24_{was microinjected into the LA to express the Kir2.1} channel and caused reduced resistance of neuronal input, yet also diminished elicited action potential firing in small subsets of neurons34_{. This manipulation caused cell-wide consequences of the} excitabil-ity. In particular, blocking of the potassium channels ablated the over-expressing and high-excitability of neurons and disrupted the fear memory reponse4,34_{. This was to be expected since reactivation of} the memory engram induces a short-term and transient raise of the excitability of the corresponding engram cell found in both the DG and LA19,30,35_{. A high-excitability state of the DG granule cells was} found to last for about an hour in response to environmental stimuli, which was moderated by Kir2.1 channel3_{internalization (the receptor-mediated endocytosis)}35_{. The excitability raise was}

3_{Kir2 family: inward-rectifier potassium channels which can influence the membrane resistance of neurons of various brain areas. Kir2.1} blocker will suppress 68% of the current. It is presumed that downregulation of Kir2.1 channels contributes to regulation of DG cell excitabil-ity. The Kir2.1 current correlates with the membrane resistance, i.e. direct control of the specific ion channel over membrane excitability also the effect of protein-synthesis inhibitor in abolishing the return to baseline excitability level. Potassium inward rectifier Kir2.1 channels influences the stable resting membrane potential in a variety of muscle and neuronal cell-type.

(17)

17

ed to be correlated with enhanced context recognition. The engram-specific exogenous expression of DG granule Kir2.1 channels impaired context recognition35_{. Within this short period of excitability} increase, synaptic access was provided to the consecutive retrieval of the given memory when the animal was exposed to the relevant cues. This enhanced the efficacy of the memory recall, making the context recognition dependent on the plasticity of the engram cell excitability. Transient enhance-ment of context recognition depends on the plasticity of engram cell excitability. Recall of the contex-tual memory was influenced by previous yet recent activation of the same engram. From Yiu et al. (2014) and Pignatelli et al. (2019) it can be concluded that the condition of engram cell excitability (both in the DG and LA) influences various behavioral outputs in response to memory retrieval, from which the state of excitability correlated to the role in memory retrieval34,35_.

Additionally, high-excitability in the memory engram was evoked by artificially activation of the allo-cated neurons with DREADDs receptor and hM3Dq label34_{. The administration of CNO to these} neu-rons in a novel spatial context evoked fear memory retrieval, even days after training. Once activated by learning during the initial trial, it enabled the capability to become reactivated by artificial or the natural cue and induced the associated memory response. Yiu et al. (2014) showed neuronal excitabil-ity became enhanced through CREB-overexpression in the allocated LA neurons34_{. Neurons with lower} CREB levels, were excluded from the memory trace and did not influence the memory formation34_. The artificial activation of neurons with increased CREB after learning, was revealed to be sufficient for fear memory expression in behavior. The application of an artificial retrieval cue confirmed the allocated neurons (increased CREB and excitability) were essential elements of the memory engram. As described by Rao-Ruiz et al. (2019), the LA memory engram maintained the same size overall, in-dependent of the fear memory strength. This indicated a limitation of the engram size and stability once allocated36_{. This was measured by the number of Arc positive neurons after the recall test, which} did not change after varying levels of freezing or after cellular manipulations.

The CREB and voltage-dependent K+_{ion channel expression will influence the likelihood of the} alloca-tion of neurons to the memory engram. Namely, by both increasing the chance of neurons with in-creased expression to become allocated, as well as reducing the chance of neighboring neurons lack-ing or have less expression to become allocated. The competition takes place between neurons with high and low intrinsic excitability. Nonetheless, it was suggested disynaptic inhibition contributes to competition. This is feed-forward inhibition, in which the excitatory projections of principal neurons (PNs) generate interneurons inhibiting other projection cells.

3.2. Neuronal Competition: Feed-forward Inhibition

Neuronal competition contributes to the shaping of the engram architecture. Counteracting forces in the microcircuit could supply the generation and maintenance of the engrams, providing the strength, size and sparsity of the engram36_{. The sparsity of the memory engram could be explained as it refers} to the optimal number and density of neurons involved for successful memory storage. Since too many neurons per memory engram could constrain the storage capacity, while too few neurons will ameliorate the risk of fidelity of the memory storage, as it promotes disease-associated synaptic and cellular cell loss36_.

The formation and maintenance of the memory engram is thought to be relying on the excitability (see 3.1), while the high-excitability states induces the inhibition of surrounding neurons34,36_{. In} re-sponse to activated neurons in, for instance the LA (Figure 6), a selection of principal neurons (PNs) showed elevated excitability. This increased the chance of allocation relative to neighboring neurons with lower (baseline) excitability36_{. In fact, a high-excitability state of the neuron tends to triumph} during the competitive allocation. Via a disynaptic, feed-forward inhibition the neighboring PNs were excluded from being allocated for the engram. Nevertheless, the allocated PNs became indispensable

(18)

18

and the main inducing elements of the memory-supporting engram. Those neurons were assigned as necessary and sufficient for the memory engram. Additionally, within the local microcircuit GABAergic interneurons are found to be contributing to the engram architecture. As in the LA, parvalbumin[4] positive (PV+) GABAergic interneurons and in the DG, somatostatin-positive (SOM+) interneurons strengthen the excitability variations of the PNs and thereby force the engram sparsity36_.

Figure 6. Visualization of the neuronal competition of the allocation process. A, the principal neurons (PN) of the lateral amyg-dala (LA) have either baseline excitability (grey) or increased excitability (light blue) when competing to become allocated for < the engram. During learning, the excitability of some of the increased PNs have a higher chance to become allocated (blue) as competition rules out the baseline excited neurons. The allocation makes the neuron a necessary and sufficient element of the memory associated to the engram. B, the local microcircuit influences the engram morphology in a different way than the intrinsic excitability. GABAergic parvalbumin-positive (PV+) inhibitory interneurons and somatostatin-positive (SOM+) inhibitory interneurons in the LA and DG, respectively, are evoked to suppress the excitability of the other PN’s. With this the neighboring PN’s are losing the competition, contributing to the sparsity of the memory engram. From Rao-Ruiz (2019). The feed-forward inhibition is found to support the process of allocation in the hippocampal-cortical network36,37_{. CFC paradigm was able to demonstrate an increased connectivity between the engram} cells of the DG, downstream of the CA337_{. The DG neuron mossy fiber terminals form projections to} the CA3 inhibitory interneurons moderating a feed-forward inhibition to the CA3. This is thought to control the precision of memories through that inhibition37_{. The CA3 output projections and the PV}+ expressing interneurons are thought to contribute to the hippocampal-cortical network and thus memory formation. Furthermore, in aging, the CA3 is shown to be hyperexcitable, hyperactive and

4_{Parvalbumin is a calcium-binding albumin protein which are found in GABAergic interneurons in various brain areas including the} hippo-campus and cortex. They are characterized as fast spiking neurons which can generate gamma wave. Through their perisomatic connections (surrounding the soma of neurons), it controls processes like cell cycle regulation and second messenger production. The high calcium sensitivity allows to discriminate calcium ions from others. Disruptions in the function of PV+ neurons contribute to neurodegenerative diseases as AD.

(19)

19

have rigid remapping. A particular protein, ABLIM35_{, was observed to function in axon guidance and in} Guo et al. (2018) as a mossy-fiber-localized cytoskeletal factor37_{. This factor was shown to be} dimin-ished after CFC. The manipulation of the gene expression of ABLIM3 was sufficient for multiple pro-cesses related to engram maintenance and memory. Especially, the gene downregulation enhanced the connectivity between the DG engram cells and CA3 inhibitory interneurons. Other effects were, ameliorated PV-expressing CA3 inhibitor interneuron activation, alleviated feed-forward inhibition onto the CA3, and lastly contribution to the DG fear memory engram maintenance over time37_. Addi-tionally, the specific CFC-induced memory reactivation in the hippocampal-cortical and amygdalar circuits were conferred. Furthermore, it diminished the fear memory induction at remote time points. Restoration of the connectivity between DG engram cells–CA3 interneurons increased PV+ _CA3 inter-neuron activation and enhanced the precision of remote memories in the 17-month-old mice after downregulation of ABLIM3 gene expression37_{. Before downregulation of ABLIM3 the increased} learn-ing was unsuccessful. This confirmed the age-related hyperactivity of the CA3, as well as the promo-tion of feed-forward inhibipromo-tion for a successful and high precision remote memory generalizapromo-tion37_.

3.3. Plasticity

The recruitment underlying neuronal allocation is depending on increased excitability of that neuron. Neuronal activation increases the selection to be associated with a memory and is later driven by that cue. The association is enhanced by consolidation in which synaptic connections are strengthened. Reactivation during reconsolidation or recall can change the associated memory by adapting the syn-aptic connections within the engram38_{. The synaptic strength relies on synaptic plasticity and the LTP} inducing long-term memories by activating de novo synthesis and trafficking of proteins, AMPA recep-tors and genes. The plasticity of the neurons involved in the memory engram is key to the stability and strength of the associated memory.

Activity-dependent cell labelling revealed higher dendritic spine density and synaptic strength in rela-tion to the allocarela-tion of engram cells5_{. As the rodents were made amnesic by protein-synthesis} inhibi-tors, the excitability of neurons decreased. Additionally, the spine density and the connectivity be-tween engram cells were diminished. This indicates the structure of allocated neurons as engram cells are having a certain morphology and specific connection with the other engram cells5_{. In the same} study, lowered synaptic strength and AMPA/NMDA ratio between the engram cells in the amnesic animals were observed5_{. These observations were paired with impairment of the memory retrieval} process. The rapid growth in synaptic strength is proposed to allow efficient access to engram cell nucleus for reactivation and memory recall. The specific changes in connectivity, after changes in engram cell excitability, could refer to fundamental processes of the memory information storage5_. Adjusting, strengthening and eliminating the synaptic contacts between the neurons will provide im-provements and maintenance of the connections between the memory engram cells. Therefore, the persistence of the memory engram and the LTM are enhanced6,11_{. It is proposed that the catalytically} active forms of certain protein kinases are potentially contributing to the maintenance persistence of the long-term memories. Particularly, the IEG Arc is found to have its mRNA conveyed towards the activated synapses on the dendrites. Once the spines were reached, this kinase prepossesses the reg-ulation of the glutamatergic AMPA receptors’ endocytosis, and putatively also the regreg-ulation of the dendritic spines arrangement11_{. The strength of excitatory synapses is proven to rely on the synthesis} of new AMPA glutamate receptors and therefore contribute to learning16_{. AMPA receptor} modifica-tion and trafficking are found to be increased after LTP16_{. Hence, Matsuo et al. (2008) used transgenic} mice with learning-induced GluR1 subunit expression fused to GFP for visualization and controlled by c-Fos promoter (see chapter 2). Learning-induced de novo protein-synthesis of the GFP-GluR1 was