Introduction

Previously acquired information is thought to be stored in sparse ensembles of neurons referred to as engrams [1,2,3,4,5]. Most rodent studies examining engrams have focused on conditioned threat or fear memory. The results from these studies indicate that neurons involved in engrams (engram neurons) supporting conditioned fear memories in several brain regions, including the CA1 of the dorsal hippocampus [6,7,8], dentate gyrus [9,10,11], various cortical regions [12,13,14,15,16] and the amygdala [17,18,19,20,21,22], are both necessary and sufficient for behavior consistent with expression of a conditioned fear memory.

The neural processes involved in the formation of an engram supporting a conditioned fear memory have also been examined. Engrams are sparsely encoded. Therefore, several groups investigated the mechanism (or mechanisms) by which only a sparse population of neurons is recruited or allocated to a specific engram. Results showed that eligible neurons within a given brain region are allocated to an engram supporting a conditioned fear memory based on relative neuronal excitability at the time of the aversive learning event [19, 23]. In addition to conditioned fear, engrams are important in supporting other types of memory, including appetitive or rewarding memory [24,25,26,27,28]. By contrast, much less is understood about how engrams supporting a rewarding memory are formed and used to guide motivated behavior.

Memory allows past experiences to guide present and future decision making and behavior [29, 30]. However, circumstances and environments are dynamic such that previously useful stored information may no longer be relevant. This previously stored information, though, may again become relevant as environments continue to change. For example, repeated exposure to a cue previously predictive of a rewarding outcome, such as delivery of a rewarding drug, in the absence of the rewarding drug, becomes less likely to motivate behavior (extinction). However, the cue may again become relevant and guide behavior if the animal is re-exposed to the rewarding drug in a different context, in a phenomenon referred to as drug-induced reinstatement [31,32,33]. Although reinstatement has been well-studied at the behavioral and pharmacological levels [31,32,33,34], little is understood about whether or how the engram supporting the original cue-drug memory may mediate this phenomenon.

The goals of the present set of studies were twofold. First, to examine how a sparsely encoded engram supporting a rewarding memory is formed, and second, to examine whether and how this original sparse engram is required when circumstances change such as with extinction and drug-induced reinstatement. We focused on the lateral nucleus of the amygdala (LA), a region important for fearful, as well as appetitive or rewarding, memories [35,36,37,38]. For example, in a classic paper, Hiroi and White (1991) showed that the LA is critical for the expression of amphetamine-induced conditioned place preference memory in which a contextual cue was paired with amphetamine administration. Consistent with this, retrieval of a cue-amphetamine memory was associated with expression of the activity-dependent gene Arc in the LA of rats [39]. Therefore, we used a similar paradigm to assess how LA neurons are allocated to an engram supporting a rewarding cue-cocaine memory. We then examined whether neurons allocated to this original engram are necessary and sufficient for behavior consistent with reinstatement of a cue-cocaine memory following extinction.

Materials and methods

Mice

Experiments were conducted in 8–12 week old male and female hybrid (C57BL/6Ntac × 129S6/SvEvTac) mice. Mice were bred at the Hospital for Sick Children and group housed (3–5 mice per cage) on a 12 h light/dark cycle with food and water available ad libitum. Experiments were conducted during the light-phase. All procedures were conducted in accordance with the policies of the Hospital for Sick Children Animal Care and Use Committee and conformed to both the Canadian Council on Animal Care (CCAC) and National Institutes of Health (NIH) Guidelines on the Care and Use of Laboratory Animals.

HSV vectors

cDNAs for wild-type full-length CREB fused to N-terminal GFP (kindly provided by Dr. Satoshi Kida, Tokyo University, Tokyo, Japan), GFP (subcloned from pEGFP-N1, Clontech Laboratories), hM4Di (kindly provided by Dr. Bryan Roth, UNC), iC++ (an engineered chloride-conducting channelrhodopsin, kindly provided by Dr. Karl Deisseroth, Stanford University) [40], were subcloned into bi-cistronic replication-defective herpes simplex virus (HSV) vectors containing the constitutive HSV promoter IE4/5 and a CMV promoter. The first named transgene (GFP-CREB or GFP) was driven by the IE 4/5 promoter, while the second (iC++ or hM4Di) was driven by the CMV promoter. For vectors containing CREB alone, GFP-CREB fusion expression was driven by the IE4/5 promoter. HSV-NpACY contains both enhanced channelrhodopsin (ChR2-H134R) fused to enhanced yellow fluorescent protein (eYFP) and halorhodopsin 3.0 (eNpHR3.0) to enable bidirectional control of neuronal activity. Opsin genes were connected in the viral vector using a p2A self-cleavage linker and expression was driven by the endogenous HSV promoter IE 4/5. ChR2 and eNpHR3.0 are spectrally compatible; neurons can be excited (ChR2) by blue light (BL, 473 nm) and inhibited (eNpHR3.0) by red light (RL, 660 nm). Previous whole-cell current-clamp experiments of hippocampal neurons verified the activation spectra for ChR2 and NpHR3.0 are separable by ~100 nm, allowing distinct activation of each opsin [41]. Moreover, BL (488 nm) increases, while RL (594 nm, 639 nm) decreases, the activity of cells expressing this construct, with minimal cross-talk [18, 41,42,43].

HSV virus was packaged using a replication-defective helper virus, purified on a sucrose gradient, pelleted, and resuspended in 10% sucrose, as previously described [44, 45]. The average titer of the virus stocks was 4.0 × 107 infectious units/ml. Transgene expression using this viral system typically peaks 3 d, and dissipates within 10–14 days after injection [23, 44, 46,47,48,49,50].

AAV vectors

AAV(DJ) expressing ChR2 driven by an αCaMKII promoter was generated in- house (1013 infectious units/ml).

Surgery

Mice were pre-treated with atropine sulfate (0.1 mg/kg, i.p.), anesthetized with isofluorane-oxygen mix (3% isofluorane for initial induction and 1–2.5% through nose cone thereafter), administered meloxicam (2 mg/kg, s.c.) for analgesia, and placed in a stereotaxic apparatus. HSV vectors were infused bilaterally (2 μL/side, flow rate 0.12 μL/min) into the LA (AP: −1.3 mm, ML: ±3.4 mm, DV: −4.8 mm relative to bregma) and optical fibers implanted above the LA (AP: −1.3 mm; ML: ±3.4 mm, DV: −4.3 mm). Optical fibers were constructed in-house by attaching a 10 mm piece of 200-μm, optical fiber (with a 0.37 numerical-aperture, NA) to a 1.25-mm zirconia ferrule (fiber extended 5 mm beyond ferrule). Optical fibers were stabilized to the skull with screws and black dental cement.

Verifying location of vector injection and extent of viral infection

Mice were transcardially perfused with 0.1 M PBS followed by 4% PFA. Brains were post-fixed for 2 h and transferred to a 30% sucrose solution. Coronal brain slices (50 µm) across the anterior-posterior extent of LA were collected. Every 2nd section was mounted on a gel-coated glass slide and coverslipped with Vectashield fluorescence mounting medium containing DAPI (Vector Laboratories Inc., Burlingame, CA).

Consistent with previous reports from several labs [51,52,53], microinjection of HSV vectors produced robust localized transgene expression with minimal tissue damage around the site of injection. Native GFP-immunofluorescence (which did not differ across vectors) was used to determine placement and extent of the viral infection for each mouse. Each mouse was classified as a “hit” or “miss” by an examiner unaware of the treatment condition and behavioral results. Mice were defined as “hits” if bilateral GFP expression was observed in LA in at least 5 consecutive brain sections (across the anterior-posterior plane). Only mice determined to be a bilateral “hit” were included in subsequent data analysis. The same criteria were used to verify viral infection in AAV experiments.

Drugs

Cocaine HCl (Health Canada) was dissolved in sterile PBS and delivered i.p at the appropriate dose (please see individual experiments for precise dosing information). CNO (Toronto Research Chemicals, TRC) was made in a stock solution of 10 mg/ml in DMSO and then diluted in saline to desired concentration. CNO was injected at a dose of 1 mg/kg i.p., 1 h before behavioral experiment.

Cue-cocaine memory training and testing

To examine a rewarding cue-cocaine memory we used a cocaine-induced conditioned place preference using an unbiased, counterbalanced protocol [54, 55]. The conditioned place preference apparatus consisted of two 15 cm × 20 cm Plexiglas chambers connected by a guillotine door (open during pre-training habituation and test sessions, but closed during conditioning sessions). Each chamber had a unique combination of visual, tactile, and olfactory properties (one side had white walls and a transparent rough floor while the other side had black and white striped walls, a smooth white floor that was wiped with 3% acetic acid (0.2 ml) before each conditioning and test trial). We balanced the two chambers in terms of initial, baseline preference such that during the habituation phase, all groups of mice spent a similar amount of time in each chamber. The activity and location of mice was monitored by an overhead CCD camera, connected to a computer running Limelight software (Colbourn Instruments).

The conditioned place preference procedure consists of 3 sequential phases: habituation, conditioning and test. In the habituation phase, drug-free mice were allowed access to the two chambers for 10 min. Time spent in each chamber (defined as the time when mice fully entered the chamber) was calculated for each mouse. During the conditioning day, one chamber was paired with SAL in the morning (15 min) and 5 h later, COC (either a low or high dose) was paired with the other chamber (15 min). The dose of COC [either 15 mg/kg (low) or 30 mg/kg (high)] was chosen based on the question asked during each experiment. We used the lower dose if we wanted to allow for an increase in memory and the higher dose if we wanted to ensure that all groups showed similar high levels of memory. 18–24 h after training, memory was assessed in a test session, in which drug-free mice were given free access to the two chambers (as in habituation) for 10 min.

To assess whether optogenetic activation of allocated engram neurons was sufficient for behavior consistent with memory recall in the absence of external recall cues, we used a real-time place preference task. Mice were given cue-COC pairing as described above. Immediately before COC administration, experimental mice expressing ChR2 were given blue light stimulation for 30 sec. Control mice received blue light stimulation 24 h before COC administration. To assess whether optogenetic activation of allocated neurons was sufficient to induce behavior consistent with recall of a rewarding memory, 24 h later, a real-time place preference test was performed in a novel 2-chamber apparatus, differing from the original apparatus in visual and tactile properties. Specifically, one chamber had white walls with colored circles and a floor of thin metal rails, while the other chamber had black walls with a single white square centered on each wall and a smooth metal floor with small circular holes.

During the real-time place preference test, mice were given free access to the two chambers for 10 min. Each time the mice entered a specific chamber, determined a priori in a counterbalanced and unbiased manner, blue light stimulation was applied. A preference score was calculated for each mouse [time (sec) spent in the blue light-paired chamber minus the time (sec) spent in the blue light-unpaired chamber].

To examine whether the original engram was necessary for cocaine-primed reinstatement of an extinguished cue-cocaine memory, we paired cocaine with a chamber (as above). 24 h following pairing, mice were given a “post-training test” with free access to both chambers in the absence of drug. Only mice that acquired the task (preference score of 50 sec or higher during post-training test), were included in subsequent analyses. Mice were then given 2 d of extinction training. Specifically, mice were administered SAL before being confined to the previously SAL-paired chamber for 15 min (morning), and also administered SAL before being confined to the previously COC-paired chamber for 15 min (afternoon). An extinction test was performed 3 h after each day of extinction training. During this extinction test, mice were given free access to the two chambers (as in previous tests) for 10 min.

Eighteen-24 h after the final extinction test, a reinstatement test was performed in which mice were given a priming injection of COC (15 mg/kg) immediately before being given free access to the two chambers (for 10 min). For the experiment using blue light to inhibit iC++-expressing neurons during the reinstatement test, the test was 6 min in length, with light ON or light OFF epochs of 3 min each.

Optogenetics

For all optogenetic experiments, the implanted optrodes were tethered to a laser source (Laserglow) through a split optic fiber. Blue light (473 nm, 20 Hz, 5 ms pulse width, ~1 mW output, 30 s) was used to excite neurons expressing ChR2. Red light (660 nm, ~7 mW output, square pulse) was used to inhibit neurons expressing NpHR3.0 and blue light (473 nm, ~7 mW output, square pulse) was used to inhibit neurons expressing iC++.

Preference score and statistical analysis

Preference scores for each mouse were determined [time (sec) spent in COC-paired chamber during the test minus time spent in SAL-paired chamber during the test]. Preference scores were analysed using 1 or 2-way Analyses of Variance (ANOVAs) with repeated measures using Statistica or Prism GraphPad 7 software. To determine differences between specific groups, Fisher LSD post-hoc comparison tests were used.

Results

LA principal neurons overexpressing CREB at training are selectively allocated to an engram supporting a cue-cocaine memory and are necessary for behavior indicative of subsequent memory retrieval

Previous research shows overexpressing the transcription factor CREB (cAMP/Ca2+ responsive element-binding protein) in a sparse, but random, population of LA principal neurons biases their allocation to an engram supporting an auditory conditioned fear memory acquired with a single cue-footshock pairing [19, 23]. Consistent with this, we previously found that viral expression of CREB in a similar small, random population of LA principal neurons enhanced a rewarding cue-cocaine memory acquired over a multi-day training period, suggesting that neurons overexpressing CREB are also preferentially allocated to the engram supporting this rewarding memory [55]. Here, we replicated and extended these findings using different techniques to manipulate neuronal activity in a reward conditioning protocol using a single cue-cocaine pairing (Fig. 1A, top panel) to better approximate the conditioned fear memory training conditions.

Fig. 1: A small population of random LA principal neurons overexpressing CREB at training is biased for allocation to an engram supporting a cue-cocaine memory and is necessary for behavior consistent with subsequent memory retrieval.
figure 1

A Top panel: Mice expressing viral CREB (vCREB) or GFP alone in small, random population of LA principal neurons were trained in a one-day cocaine-induced place preference paradigm. Confinement in one chamber was paired with saline (SAL) administration, confinement in the other chamber with cocaine (COC) administration. A low dose of COC was used such that an enhancement in memory could be observed. 24 h later, drug-free mice were given free access to both chambers and chamber preference assessed. Mice spending more time in the COC-paired chamber was taken as evidence of cue-cocaine memory. Middle panel: Increased cue-cocaine memory in mice expressing vCREB vs. GFP (vCREB, n = 10; GFP, n = 8). Bottom panel: Representative heatmaps of time spent in each chamber during drug-free memory test. vCREB mice spend more time (darker blue) in COC-paired chamber than GFP mice, showing greater cue-cocaine memory. B Microinjection of viral vector expressing vCREB (left column) or GFP (right column) into LA induced robust transgene expression in a small, random population of LA principal neurons. vCREB expression was confined to the nucleus, as the vCREB construct is a fusion protein of CREB-GFP. GFP-only viral expression was seen throughout the cell. Top row: 10x magnification, outline of LA in white, scale bar = 250 μm. Bottom row: 20x magnification of the same region, scale bar = 100 μm. C Top panel: Mice expressing vCREB and blue-light sensitive inhibitory opsin iC++, or iC++ alone, in small random population of LA neurons were trained similarly to A. In second half of memory test, mice received blue-light photostimulation to silence iC++ neurons. Bottom panel: Optogenetically silencing (iC++ + blue light ON) neurons expressing vCREB during memory test disrupted cue-cocaine memory expression, whereas similarly silencing random (non-vCREB expressing neurons, expressing iC++ only) did not. (vCREB-iC++, n = 7; iC++, n = 9). Data represent mean ± SEM. Circles represent individual mouse scores. *p < 0.05, ***p < 0.001, n.s. p > 0.05.

Full size image

Mice were microinjected with a viral vector expressing either CREB [viral CREB (vCREB) fused with GFP] or GFP alone to induce robust transgene expression in a small, random population of LA principal neurons (Fig. 1B). Cue-cocaine pairing took place on a single training day in a 2-chamber apparatus. During training, mice were confined to one chamber immediately after saline (SAL) injection (in the morning) and to a second distinct chamber immediately after cocaine (COC, 15 mg/kg) injection (in the afternoon). Importantly, this dose of cocaine induces only a modest conditioned place preference such that an increase in memory may be observed. 24 h later, drug-free mice were allowed to freely explore both chambers. A preference score was computed for each mouse (subtracting time spent in SAL-paired chamber from time spent in COC-paired chamber) indicating the strength of the cue-cocaine memory. vCREB mice spent more time in the COC-paired chamber than GFP control mice (Fig. 1A, bottom panel; 1-way ANOVA, F(1, 16) = 5.87, p < 0.05). Therefore, enhancing CREB function in a small, random population of LA neurons enhances a cue-cocaine memory induced by a single pairing.

To examine if LA neurons expressing vCREB are critical components of the engram supporting this rewarding memory, we used an inhibitory optogenetic construct to silence the activity of vCREB+ neurons during the memory test, thereby allowing a within-subject comparison. Viral vectors expressing both vCREB and the blue-light sensitive inhibitory opsin, iC++ [40] or the control iC++ alone were used. Mice were trained with a high dose of COC (30 mg/kg) to ensure both groups formed a similarly strong cue-cocaine memory. In the memory test, mice were first tested without light (3 min) and then with light (3 min) to inhibit iC++-expressing neurons (Fig. 1C, top panel).

Both groups of mice showed similarly high preference for the COC-paired chamber in the light off condition. Only mice expressing the inhibitory opsin in vCREB neurons showed a decrease in preference for the COC-paired chamber when tested with the laser light on (Fig. 1C; 2-way repeated-measures ANOVA, significant Vector (vCREB-iC++, iC++) × Light (ON, OFF) interaction, F(1, 14) = 36.02, p < 0.05; post-hoc tests showed similar preference in both groups in absence of light, while vCREB-iC++, but not iC++, mice showed lower preference in light-ON epoch). That decreasing the activity of a random population of non-experimentally allocated neurons (expressing iC++ alone) did not disrupt preference for the COC-paired chamber in control mice indicates that disrupting the activity of a small population of random neurons is not sufficient to disrupt behavior consistent with retrieval of the cue-cocaine memory. Together these results indicate that neurons expressing vCREB at the time of a single cue-cocaine pairing are selectively allocated to the engram supporting this rewarding memory and are necessary for subsequent memory retrieval.

LA neurons with increased neuronal excitability at training are selectively allocated to a cue-cocaine engram; allocated neurons are both necessary and sufficient for behavior consistent with subsequent memory retrieval

Neuronal overexpression of CREB produces many physiological changes, including increased neuronal excitability [19, 23, 56]. Next, we examined whether increasing the excitability of a small population of LA principal neurons immediately before training, without directly manipulating CREB function, is sufficient to bias their allocation to an engram supporting a rewarding memory. To increase neuronal excitability, we randomly expressed the blue-light sensitive excitatory opsin ChR2 in a sparse population of LA pyramidal neurons. Brief (30 s) blue-light stimulation of ChR2-expressing neurons increases their excitability, similar to CREB overexpression [23]. In experimental mice, we photostimulated ChR2+ neurons before cue-cocaine training while control mice expressed GFP alone or expressed ChR2 but were not photostimulated before training.

A moderate dose of cocaine (15 mg/kg, as in Fig. 1A) was used in this experiment such that potential memory enhancement could be observed. Only mice expressing ChR2 and given blue light before training showed enhanced memory (Fig. 2A, bottom panel; 2-way ANOVA, significant Group × Light interaction; F(1,32) = 4.37, p < 0.05; post-hoc tests showed higher preference in ChR2+ photostimulated, compared with non-photostimulated, mice. No effect of photostimulation in GFP+ mice). These findings suggest that increasing the excitability of a small population of random LA principal neurons immediately before training is sufficient to bias their allocation to an engram supporting a cue-cocaine memory.

Fig. 2: Increasing excitability of random LA neurons before training preferentially allocates these neurons to an engram supporting a rewarding cue-cocaine memory; allocated neurons are both necessary and sufficient for behavior consistent with subsequent cue-cocaine memory recall.
figure 2

A Top panel: Mice expressing ChR2 or GFP in small, random population of LA neurons were trained in cue-cocaine paradigm as in Fig. 1 with a low dose of cocaine (COC) such that memory enhancement could be observed. Immediately before cue-cocaine pairing, mice received 30 sec of blue light photostimulation, or no photostimulation. Bottom panel: Blue light enhanced memory in mice expressing ChR2 but not in mice expressing GFP alone. (ChR2 + blue light, n = 10; ChR2 + no blue light, n = 8; GFP + blue light, n = 8; GFP + no blue light, n = 10). B Top panel: NpACY vector expresses both blue light-responsive excitatory opsin ChR2 and red light-sensitive inhibitory opsin eNpHR3.0, allowing bidirectional (excitatory and inhibitory) manipulation of the same small population of infected LA neurons with blue or red light, respectively. Mice expressing NpACY or eNpHR3.0 alone were trained in cue-cocaine paradigm with high dose of COC such that all mice acquired strong memory. Immediately before cue-cocaine pairing, mice received 30 sec of blue-light photostimulation, or no photostimulation. Mice tested under two conditions, first with no red light and then with red light to silence eNpHR3.0+ neurons. Bottom panel: In memory test, only mice expressing NpACY that received blue light stimulation immediately before training showed impaired cue-cocaine memory during red light inhibitory stimulation. No effect of red light in other groups, showing specificity of this memory deficit. (NpACY + blue light before training, n = 8; NpACY + no blue light, n = 7; eNpHR3.0 alone + blue light, n = 9; eNpHR3.0 alone + no blue light, n = 7). C Top panel: Mice expressing ChR2 or GFP received cue-cocaine pairing (with high dose of COC) either immediately or 24 h after blue-light stimulation. Mice were tested in a novel 2-chamber apparatus. Occupancy in one chamber (randomly assigned) resulted in blue-light photostimulation. Bottom panel: Only mice expressing ChR2 that received blue light stimulation immediately (and not 24 h) before cue-cocaine pairing showed a real time place preference (RT-PP) for the photostimulated chamber. (ChR2 + blue light immediately before training, n = 13; ChR2 + blue light 24 h before training, n = 9; GFP + blue light stimulation immediately before training, n = 10). Data represent mean ± SEM. Circles represent individual mouse scores. *p < 0.05, n.s. p > 0.05.

Full size image

We next used an all-optical method to explicitly test whether neurons experimentally made more excitable before training (ChR2 + blue light) are necessary for behavior consistent with subsequent memory retrieval, and therefore critical components of the engram supporting this memory. We used a viral vector (NpACY vector) expressing both a blue-light sensitive excitatory opsin, ChR2, and a red-light sensitive inhibitory opsin, eNpHR3.0. This allowed the activity of the same sparse population of infected LA principal neurons to be increased or decreased using different wavelengths of light (Fig. 2B, top panel) [18, 43, 57]. As above, in the experimental group, NpACY-expressing neurons were photostimulated with blue-light before training to increase their excitability and bias their allocation to the engram. Control mice expressed eNpHR3.0 alone or expressed NpACY but did not receive blue light before training. A high dose of cocaine (30 mg/kg) was used to ensure all groups acquired strong cue-cocaine memory. Mice were tested under two conditions; first in the absence of light and then in the presence of red light to silence infected neurons (expressing either NpACY or eNpHR3.0 alone) (Fig. 2B, top panel). Red light disrupted cue-cocaine memory only in mice expressing NpACY that received blue light before training (NpACY + blue light) and not in control mice (expressing eNpHR3.0 alone or expressing NpACY but not allocated with blue light before training) (Fig. 2B, bottom panel; 2-way repeated-measures ANOVA, significant Group × Light interaction F(3,27) = 3.38, p < 0.05. Post-hoc tests showed decreased preference during red light only in the NpACY-BL + group, no effect of red light in all other groups). These results indicate that the activity of neurons made more excitable before formation of a rewarding memory is necessary for behavior consistent with subsequent expression of that memory and, furthermore, that silencing a similar number of non-experimentally allocated neurons (expressing eNpHR3.0 alone) does not affect memory expression.

If neurons made more excitable before cue-cocaine training are critical components of the engram supporting this memory, then artificial (optogenetic) reactivation of these neurons alone should serve as a sufficient memory retrieval cue even in the absence of external sensory retrieval cues (e.g., the chamber previously paired with cocaine) and produce behavior consistent with retrieval of this rewarding memory. To examine this, we biased the allocation of ChR2+ neurons to the engram with blue light in the experimental group (as in Fig. 2A). We included two control groups; mice similarly expressing ChR2 that received blue light stimulation 24 h (rather than immediately) before cue-cocaine pairing, and mice expressing GFP that received blue light stimulation immediately before training (Fig. 2C, top panel). To determine whether artificial reactivation of allocated neurons was sufficient to induce behavior consistent with memory retrieval in the absence of external sensory retrieval cues, we asked whether mice would show a preference for a particular chamber in a novel 2-chamber apparatus if entry into that chamber resulted in blue light optogenetic stimulation of allocated ChR2+ neurons. Importantly, mice showed no initial preference for either chamber in the novel place preference apparatus (Supplementary Fig. 1).

Artificial reactivation of experimentally-allocated ChR2-expressing neurons was sufficient to induce behavior consistent with the recall of a rewarding memory. Mice showed a preference for the novel chamber entry into which triggered blue-light stimulation, only if ChR2-expressing neurons were allocated to the previous cocaine-induced conditioned place preference memory. No real-time place preference for the blue-light chamber was observed in control mice expressing ChR2 in non-experimentally allocated neurons (blue light stimulation 24 h before cue-cocaine pairing in the original training apparatus), nor in GFP control mice (Fig. 2C; 1-way ANOVA, F(2,29) = 4.83, p < 0.05; post-hocs showed higher preference in ChR2+blue light-immediately group compared with control groups, which did not differ). These results indicate that simply photoactivating a small portion of random (non-experimentally allocated) LA principal neurons does not induce a place preference. Instead, artificial reactivation of specific LA neurons previously allocated to a rewarding engram is sufficient to induce behavior consistent with retrieval of a rewarding memory in the absence of sensory memory retrieval cues.

LA principal neurons overexpressing CREB at training are necessary for cocaine-primed reinstatement of an extinguished cue-cocaine memory

Once formed, memory expression may be modified by additional experience. For example, following acquisition of a cue-cocaine memory, repeated presentation of the cue alone in the absence of cocaine may extinguish the learned place preference. Even after extinction, though, the cue-cocaine memory is not “erased”, as the preference for the chamber cue can be re-established by a non-contingent priming dose of cocaine, perhaps serving as a memory reminder [31,32,33]. Here we used a cocaine-primed reinstatement protocol to examine the fate of neurons allocated to the original rewarding engram after extinction.

After mice acquired a cue-cocaine memory (Fig. 3A, top panel), the place preference was extinguished by repeatedly confining mice to the cocaine-paired chamber in a non-drugged state (without cocaine). After several chamber+no drug pairings, mice no longer showed a preference for the chamber previously paired with cocaine, demonstrating extinction (Supplementary Fig. 2). 24 h later, mice received a low non-contingent dose of cocaine (15 mg/kg) or saline in the homecage immediately before a memory reinstatement test in which mice had free access to the two chambers. Mice primed with cocaine showed reinstatement of cue-cocaine memory (spending greater time in the chamber originally paired with cocaine) while mice given saline did not (Fig. 3A, bottom panel; 2-way repeated-measures ANOVA, significant Test time (Extinction, Reinstatement) × Group (SAL-primed, COC-primed) interaction, F(1,24) = 6.76, p < 0.05; post-hoc tests showed no place preference in either group on final extinction test, and higher preference in reinstatement test only in COC-primed mice).

Fig. 3: LA neurons allocated to an engram at the time of training are necessary for behavior consistent with cocaine-primed reinstatement of an extinguished cue-cocaine memory.
figure 3

A Top panel: Mice received cue chamber-cocaine (COC) training, followed by extinction training in which the chamber was paired with saline (SAL) rather than COC. Once mice no longer showed a chamber preference (during final extinction test, EXT), a non-contingent priming dose of COC or SAL (SAL priming, COC priming) was delivered in the homecage immediately before a reinstatement test (REINST). Bottom panel: Place preference reinstated after extinction only in mice receiving a priming dose of COC. (n = 13 for each group). B Top panel: Mice expressing either vCREB and the inhibitory DREADD hM4Di, or hM4Di alone, received cue-cocaine training, followed by extinction training, as above. Mice received either the hM4Di ligand CNO or SAL before a COC-primed reinstatement test. Bottom panel: Mice expressing vCREB-hM4Di receiving CNO before COC-primed reinstatement test, showed no reinstatement; reinstatement observed in all control groups (vCREB-hM4Di + CNO, n = 11; vCREB-hM4Di + SAL, n = 9; hM4Di + CNO, n = 11; hM4Di + SAL, n = 11). C Top panel: Mice expressing either vCREB and the inhibitory blue-light activated optogenetic construct iC++, or iC++ alone, received cue-cocaine training, followed by extinction training as above. All mice given COC-priming before reinstatement test. Inhibitory blue-light delivered during either the first or second half of reinstatement test. Bottom panel: Only mice expressing vCREB-iC++ showed a decrease in COC-primed reinstatement when vCREB neurons were inhibited (blue light). Inhibiting neurons expressing iC++ alone (without vCREB) produced no effect in reinstatement test. (vCREB-iC++, n = 18; iC++, n = 19). Data represent mean ± SEM. Circles represent individual mouse scores *p < 0.05, **p < 0.01, n.s. p > 0.05.

Full size image

Next, we assessed whether neurons allocated to the engram supporting the original cue-cocaine memory were necessary for reinstatement behavior using two different methods to inhibit the activity of allocated engram neurons during the reinstatement test. First, we combined vCREB (to allocate neurons) and the chemogenetic construct (hM4Di) [58, 59] to inhibit the activity of allocated vCREB neurons throughout the reinstatement test. hM4Di hyperpolarizes membranes when bound by the exogenous ligand clozapine-N-oxide (CNO) [58]. Control mice expressed hM4Di alone or were administered saline rather than CNO before the reinstatement test. vCREB-hM4Di- and hM4Di-expressing mice were first given a cue-cocaine pairing (as above). The resulting place preference was then extinguished. All mice were primed with cocaine (15 mg/kg) before the reinstatement test. Mice also received either CNO (1 mg/kg), to decrease the activity of hM4Di-expressing neurons, or saline (Fig. 3B, top panel). Robust cocaine-primed reinstatement was observed in all mice except those in which neurons allocated to the original engram supporting the cue-cocaine memory were chemogenetically silenced (Fig. 3B, bottom panel; 1-way ANOVA, F(3,38) = 3.05, p < 0.05; post-hoc tests showed lower preference in vCREB-hM4di + CNO mice compared with all other groups).

Second, we used optogenetics and a within-session design to assess whether temporally silencing engram neurons at different times during the test disrupted behavior consistent with reinstatement. Neurons were biased for allocation to the engram with vCREB and their activity silenced during the reinstatement test with the inhibitory opsin iC++ (vCREB-iC++). Control mice expressed iC++ alone. Training and extinction were conducted as above (Fig. 3A, top panel). Mice were primed with cocaine before the reinstatement test and preference measured under two conditions; blue light stimulation to silence neurons expressing iC++ (light ON epoch), and with no blue light (Fig. 3C, top panel). Light presentation order was counterbalanced across groups to assess the potential contribution of extinction over the reinstatement test session. As there was no significant effect of light presentation order (Supplementary Fig. 3; 3-way ANOVA with factors Vector (CREB-iC++, iC++), Light (ON, OFF), Order (OFF-ON, ON-OFF) revealed no significant effect of Order (p = 0.44), or interaction involving Order (Order × Light × Vector, Order × Light, Order × Vector interactions, all p’s > 0.05)), we combined data from both light-order conditions (OFF-ON, ON-OFF).

Silencing CREB-allocated engram neurons, but not random (non-CREB allocated) neurons, disrupted cocaine-primed reinstatement (Fig. 3C, bottom panel; 2-way repeated-measures ANOVA, significant Vector (CREB-iC++, iC++) × Light (ON, OFF) interaction, F(1,35)=12.8, p < 0.01. Post-hoc tests showed lower preference in CREB-iC++ mice, but not mice expressing iC++ alone, during blue light epoch). Taken together, these results using two different methods to inhibit either vCREB+ neurons or a similar number of vCREB- neurons indicate that behavior consistent with cocaine-primed reinstatement requires the activation of neurons allocated to the engram during the original training event.

Artificially activating engram neurons during a drug-free test is sufficient to induce behavior consistent with reinstatement of cue-cocaine memory after extinction in the absence of cocaine-priming

Above, we showed that artificially activating LA engram neurons supporting a cue-cocaine memory was sufficient to induce behavior consistent with retrieval of a rewarding memory (Fig. 2C). Next, we examined whether similarly artificially reactivating LA engram neurons after extinction is sufficient to induce reinstatement-like behavior in the absence of cocaine priming. Because the duration of this experiment was longer than the above experiments, we used AAV (rather than HSV) to express ChR2 in a similar, small population of excitatory principal LA neurons (Fig. 4A). Control mice similarly expressed GFP alone.

Fig. 4: Post-extinction reactivation of LA neurons allocated to an engram supporting cue-cocaine memory in the absence of cocaine priming is sufficient to induce behavior consistent with reinstatement.
figure 4

A Microinjection of AAVDJ-αCaMKII-ChR2 induces robust transgene expression in random sparse population of LA pyramidal neurons (outlined by the dotted lines), similar to HSV vector. DAPI (nuclear marker) = blue, GFP = green. Outline of LA in white. Scale bar = 30 μm, 12 μm for images. B Top panel: Mice expressing ChR2 or GFP alone received blue light photostimulation, or no photostimulation, before cue chamber-cocaine (COC) pairing. Place preference was then extinguished by repeatedly presenting the COC-paired chamber with saline (SAL) until mice no longer showed a preference. Mice tested for memory reinstatement (without COC-priming) in the original apparatus. During the first half of the test, mice were tested without blue light and with blue-light photostimulation during the second half of the test session. Bottom panel: Mice expressing ChR2 showed preference for original COC-paired chamber during blue light-ON epoch of a post-extinction reinstatement test only if mice received blue light stimulation before the original cue-cocaine pairing (reactivation of original engram). In contrast, no evidence of reinstatement observed in other groups of control mice. (ChR2 + blue light, n = 14; ChR2 + no blue light, n = 12; GFP + blue light, n = 10). Data represent mean ± SEM. *p < 0.05, n.s. p > 0.05.

Full size image

As above, ChR2+ neurons were biased for allocation to the engram supporting a cue-cocaine memory with blue-light before the original training event. The resulting cocaine-induced conditioned place preference was then extinguished such that all mice no longer showed a preference for the cocaine-paired chamber. Mice were tested for reinstatement of the original cue-cocaine memory in the same apparatus (Fig. 4B, top panel). In the absence of blue light, all mice continued to not show a place preference for the chamber previously paired with cocaine. However, in the presence of blue light, only mice in which the original engram was optogenetically reactivated showed reinstatement of the original place preference while the other groups did not (Fig. 4B, bottom panel; 2-way repeated-measures ANOVA, significant Group × Light interaction F(2,33) = 5.35, p < 0.01, post-hoc tests showed increased preference in ChR2++blue-light-train group when photostimulated during test, but not in control groups). This result indicates that the artificial reactivation of a small proportion of LA neurons allocated to the engram supporting the original memory is sufficient to induce behavior consistent with reinstatement of a cue-cocaine memory after extinction in the absence of cocaine-priming.

Discussion

How engrams supporting threatening or fearful memories are formed have been well-studied. In contrast, how engrams supporting rewarding or appetitive memories are formed and their fate with additional experience is less well understood. Here, we used a combination of techniques to examine how engrams supporting a rewarding experience are formed, whether activation of engram neurons is necessary and sufficient for behavior consistent with memory retrieval, and whether neurons allocated to the original engram remain important with additional experience.

To more closely mirror the time-course of single day one-trial fear conditioning studies in which a cue is paired with footshock administration, we developed a single day one-trial reward training protocol in which a cue is paired with cocaine administration. Although it is well-known that the memory produced by single trial fear conditioning may last a lifetime [60], the persistence of memory produced by a single trial reward conditioning of this type is unclear. Overexpression of CREB or excitatory photostimulation of a random sparse population of neurons before cue-cocaine pairing was sufficient to enhance cocaine-induced conditioned place preference memory, consistent with the interpretation that these neurons are preferentially allocated to the supporting engram. To test this directly, we used two strategies to ask whether the activity of manipulated neurons was necessary and sufficient for inducing behavior consistent with retrieval of that specific rewarding memory. First, we found that chemogenetically or optogenetically inhibiting the activity of allocated neurons during the memory test disrupted memory retrieval. Second, we found that optogenetically reactivating allocated neurons is sufficient to induce behavior consistent with the retrieval of a rewarding memory in the absence of an external sensory retrieval cues. That is, a real-time place preference for a novel chamber, entry into which produced artificial activation of neurons, was observed only in mice in which neurons were previously optogenetically allocated to the engram supporting a rewarding memory. Together, these findings indicate that neurons with increased excitability are preferentially allocated to an engram supporting a rewarding memory.

The present results closely mimic the findings showing that increasing the neuronal excitability of a small random population of random pyramidal LA neurons biases their allocation to an engram supporting a conditioned fear memory [17,18,19,20,21,22]. These parallels are interesting as the manipulation of neuronal excitability in a small, seemingly random (infected), population of LA principal neurons occurs before behavioral training. Specifically, increasing the excitability of neurons (using different viral vectors and methods of increasing neuronal excitability) before either cue-footshock or cue-cocaine training biases these neurons to become allocated to an engram supporting either a fear or reward memory. The only difference is the type of training event that occurs (aversive or rewarding). Together, these results suggest that LA principal neurons are not specifically and irreversibly “tuned” based on a priori molecular identity or projection specificity to be part of an engram supporting an aversive or rewarding memory. Although “labelled lines” for aversive or rewarding memories may exist in other brain regions [61, 62], this type of coding segregation likely does not occur to the same extent in the LA. Once committed to an engram supporting a particular aversive or rewarding memory, however, LA neurons may acquire valence-specific ‘tuning’.

Previous findings examined the role of the basolateral complex of the amygdala (BLA) in reward memory retrieval. The BLA complex is often described as comprising the basal nucleus of the amygdala (BA) as well as the LA. For instance, Otis and colleagues [63] showed that β-adrenergic receptors in the more BA are not necessary for retrieval of a cocaine-induced conditioned place preference memory. By contrast, our results indicate that a subpopulation of LA neurons are necessary the retrieval of a similar memory. Differences between LA and BA in terms of anatomy and physiology [64, 65], may account for these divergent findings.

Although several studies have manipulated neurons involved initially in an engram supporting a conditioned fear memory [1,2,3,4,5], the “fate” of an engram over time, with changing circumstance or additional learning, has been less well explored. Recently, Lacagnina and colleagues [11] examined spontaneous recovery of an extinguished conditioned fear memory [66, 67]. Spontaneous recovery of an extinguished conditioned fear response may occur with the passage of time, without additional behavioral training. The activity of dentate gyrus engram neurons supporting a contextual fear memory was decreased with extinction training but increased with spontaneous recovery of the contextual fear memory. These findings suggest that the original engram was silenced by extinction, but “unsilenced” over time in the absence of any overt internal or external manipulation by spontaneous recovery.

Here we examined the criticality of LA engram neurons following extinction of a reward memory in a reinstatement paradigm. In this paradigm, a rewarding memory is first extinguished, and this behavior can be recovered by non-contingent presentation of the unconditioned stimulus (here, cocaine). We found that after extinction, cocaine-priming of reinstatement behavior required activation of neurons allocated to the engram during the original cue-cocaine pairing, indicating that the original engram was critical for reinstatement behavior. Consistent with this, directly reactivating the original engram via optogenetics was sufficient to induce reinstatement-like behavior in extinguished mice even without cocaine priming. Together these results suggest that extinction training may silence the original engram but that post-extinction administration of a priming dose of cocaine reactivates the original engram, and that without this reactivation, memory-guided reinstatement behavior does not occur. In addition to priming with cocaine, contextual cues or stress have been used to reinstate similarly extinguished behavior [68,69,70,71]. It would be interesting to examine whether these memory “reminders” also require reactivation of the original engram.

Although here we focused on allocation of LA neurons to an engram supporting a rewarding memory, we appreciate that these neurons are likely part of a larger engram across many brain regions. Because the present manipulations of LA engram neurons alone were sufficient to either impair or elicit memory-guided behavior, though, the present findings indicate that the LA is an important hub region in what is likely a much larger engram.