Available online at www.sciencedirect.com ScienceDirect Memory engram storage and retrieval 1,2 1 1 Susumu Tonegawa , Michele Pignatelli , Dheeraj S Roy and 1,2 Toma´ s J Ryan A great deal of experimental investment is directed towards likely distributed nature of a memory engram throughout questions regarding the mechanisms of memory storage. Such the brain [6 ]. Here we review recent experimental stud- studies have traditionally been restricted to investigation of the ies on the identification of memory engram cells, with a anatomical structures, physiological processes, and molecular focus on the mechanisms of memory storage. A more pathways necessary for the capacity of memory storage, and comprehensive review of recent memory engram studies have avoided the question of how individual memories are is available elsewhere [7 ]. stored in the brain. Memory engram technology allows the labeling and subsequent manipulation of components of Memory function and the hippocampus specific memory engrams in particular brain regions, and it has The medial temporal lobe (MTL), in particular the hip- been established that cell ensembles labeled by this method are both sufficient and necessary for memory recall. Recent pocampus, was implicated in memory of events or episodes research has employed this technology to probe fundamental by neurological studies of human clinical patients, where questions of memory consolidation, differentiating between its direct electrophysiological stimulation evoked the recall mechanisms of memory retrieval from the true neurobiology of of untargeted episodic memories [8]. Subsequent study of memory storage. humans lacking large regions of the MTL showed dramatic amnesia for episodic memories [9]. Rodent behavioral Addresses 1 studies have since established that the hippocampus is a RIKEN-MIT Center for Neural Circuit Genetics at the Picower Institute for Learning and Memory, Department of Biology and Department of central brain region for contextual memory storage and Brain and Cognitive Sciences, Massachusetts Institute of Technology, retrieval [10,11]. Much is now known about brain struc- Cambridge, MA 02139, USA 2 tures, neural circuits, and molecules involved in memory Howard Hughes Medical Institute, Massachusetts Institute of encoding and consolidation [12 ,13,14], but comparatively Technology, Cambridge, MA 02139, USA few studies have attempted to investigate how individual Corresponding author: Tonegawa, Susumu ([email protected]) memory engrams are stored in the brain [15]. Current Opinion in Neurobiology 2015, 35:101–109 Synaptic plasticity as a mechanism of memory This review comes from a themed issue on Circuit plasticity and Lasting memories have long been hypothesized to be memory encoded as structural changes at synaptic junctions of Edited by Thomas Mrsic-Flogel and Alessandro Treves sparse neuronal assemblies [16]. Ramo´n y Cajal originally proposed that the strengthening of synaptic connections of existing neurons might be a mechanism of memory storage [17], but it was Donald Hebb’s theoretical inte- http://dx.doi.org/10.1016/j.conb.2015.07.009 gration of neurophysiology and psychology that created # 0959-4388/ 2015 The Authors. Published by Elsevier Ltd. This is an the modern paradigm for memory research [16]. Hebb open access article under the CC BY-NC-ND license (http:// proposed that neuronal assemblies linked by adaptable creativecommons.org/licenses/by-nc-nd/4.0/). synaptic connections could encode informational content in the brain. Empirical research into the physiological nature of memory storage has been dominated by various Introduction versions of Hebbian synaptic plasticity [18]. The typical Memory refers to the storage of learned information in the experimental model of synaptic plasticity is long-term brain, and is crucial for adaptive behavior in animals [1]. potentiation (LTP) [19], most studies of which rely on in Understanding the material basis of memory remains a vitro experimental paradigms where synaptic stimulation central goal of modern neuroscience [2]. The hypothetical patterns are substituted for behavioral training. It is clear material basis of learned information, the memory engram, that memory and synaptic plasticity have many proper- was first conceived by Richard Semon who theorized that ties in common [20 ]. NMDA receptor function is nec- learning induces persistent changes in specific brain cells essary for the encoding of many types of memory, as well that retain information and are subsequently reactivated as for the induction of synaptic plasticity [13,21]. More- upon appropriate retrieval conditions [3 ,4,5]. However, over, both memory consolidation and LTP have a late, experimental searches for specific memory engrams and protein synthesis-dependent phase [20 ,22]. Despite memory engram-bearing cells using brain lesions proved these biological commonalities, and many serious theo- inconclusive due to methodological limitations and the retical efforts to integrate memory storage and synaptic www.sciencedirect.com Current Opinion in Neurobiology 2015, 35:101–109 102 Circuit plasticity and memory plasticity [23–27], it remains a controversial subject Physiological characterization of engram cells without a clear consensus [28–30]. The storage of lasting memory in the brain must involve persistent plasticity of engram cell structure and/or Limitations of standard methodology physiology. Indeed, ex vivo characterization of DG en- Two confounds have hindered progress towards a satis- gram cells revealed two engram cell-specific properties factory synthesis of synaptic plasticity and memory. [40 ]. First, engram cells showed significantly increased First, behavioral studies of memory have relied on dendritic spine density relative to non-engram cells. the disruption of brain regions, circuits, or molecules Second, patch clamp recordings of excitatory postsynap- [12 ,13,14], and have thus addressed the importance of tic currents in paired engram and non-engram cells these structures and signaling pathways to the capacities elicited by presynaptic stimulation of perforant path of memory storage or retrieval, rather than the storage of axons showed substantially higher synaptic strength in individual memory engrams themselves. Second, typical engram cells. conceptions of memory conflate the properties of mem- ory storage and memory retrieval. But it is a fundamental The above two properties are clear cases of plasticity premise of psychology that successful memory function occurring exclusively in engram cells, and are reminiscent presupposes not only the retention of learned informa- of Hebbian plasticity. If this plasticity is representing tion, but also its successful retrieval [1]. Therefore a mnemonic information then it should be encoded by the given case of apparent memory loss (amnesia) may in specific training experience. Protein synthesis is neces- principle be due to a damaged memory engram, or an sary for late phase synaptic plasticity and memory con- inability to retrieve that particular engram [31–34]. Both solidation, and indeed when the protein synthesis of these confounds have recently been overcome inhibitor anisomycin was administered to animals imme- through the development of memory engram technology diately after fear conditioning, retrograde amnesia was [35 ]. observed one day later. Analysis of engram cells one day after anisomycin treatment showed that the anisomycin Sea change: memory engram technology abolished engram-cell specific increases in both dendritic Identification and functional activation of engram cells spine density and synaptic strength, but did not alter In order to progress in memory research it is crucial to either property in non-engram cells. Importantly, aniso- identify the engrams and engram cells for specific experi- mycin treatment one day post-training (outside the con- ences. The challenge of identifying individual memory solidation window) impaired neither the dendritic spine engrams and engram cells amidst the complexity of the density increase nor the synaptic strength augmentation brain becomes less daunting if we co-opt natural brain of engram cells [40 ]. Therefore engram cell-specific activity during learning to point us to the relevant brain structural and synaptic plasticity is protein synthesis- cells. This concept has been realized through the devel- dependent and consolidated with the target training opment of memory engram technology, which allows the experience. labeling and subsequent manipulation of engram-bearing cells [35 ]. Engram technology is based on the experi- Retrieval of lost memory from amnesia: dissociation of mental fusion of immediate early gene (IEG) labeling and engram cell plasticity and memory optogenetics. The expression of IEGs, such as c-fos or arc, Surprisingly, direct optogenetic activation of amnesic is a marker of neuronal activity [36]. Thus the promoters engram cells in mice resulted in successful retrieval of of IEGs can be co-opted to tag neurons that are active the ostensibly lost contextual fear memory. The general- during a given learning experience with an exogenous ity of the memory retrieval finding was tested in a range of target protein (Figure 1) [37 ]. Temporal specificity of experimental conditions [40 ]. First, lost memory was labeling is achieved by engineering the labeling mecha- retrieved by optogenetic stimulation of ChR2-labeled nism to be inhibited by administering doxycycline engram cells in hippocampal CA1. Second, amnesia for (DOX). When