Opto-extinction of a threat memory in mice

Abstract

Memories of past experiences guide future behaviour. Sparse ensembles of neurons, known as engrams, are thought to store memories in the brain. Neurons involved in a particular engram (“engram neurons”) are necessary for subsequent memory expression as memory retrieval is thought to be initiated by an external sensory cue reactivating engram neurons. However, conditions or environments are dynamic, such that future behaviour should be flexible. The role of engrams in mediating flexible behaviour is not understood. Here we examined this question using one type of flexible behaviour, extinction of a threat response. An initially neutral tone is first paired with an aversive footshock such that the tone alone induces defensive freezing. After subsequent repeated tone presentations without the footshock, rodents no longer freeze to the tone. Because the tone cue is thought to reactivate the engram to induce memory retrieval, we examined whether it is possible to induce an extinction-like behavioural effect by optogenetically reactivating the lateral amygdala component of the engram alone (without tone re-exposure). Similar to tone-induced extinction, mice showed decreased freezing to optogenetic stimulation of the lateral amygdala engram in the “extinction training” session. Moreover, “opto-extinguished” mice showed decreased freezing to the tone when subsequently tested for retrieval of the extinction training in the same context, suggesting that the opto-extinction transferred to the actual sensory stimulus. However, unlike tone extinction, in which mice showed renewal of tone-induced freezing when tested in a novel context, opto-extinguished mice continued to show a deficit in tone-induced freezing. Extinction has been characterized as new learning that inhibits the original memory or a phenomenon in which the original memory is “unlearned”. Our findings suggest that opto-extinction may silence the original engram to “unlearn” the original memory.

1. Introduction

“I have never seen Francis Crick in a modest mood”. So wrote James Watson in the opening sentence of his book “The Double Helix” describing his personal take on the discovery of DNA structure (Watson, 1981). I (SAJ) have deep concerns about Watson’s political/-scientific/sociological views as well as his minimization of Rosalind Franklin’s contribution to this discovery. But this was the first book I encountered that showed science to be a very human endeavour. And I loved it! The very idea that fundamental insights that change our view of the world could be made by people who, in addition to being amazing scientists, can also be funny, kind, petty, driven, confident, shy, and all-too-human, inspired me.

I am delighted to be included in this Special Issue dedicated to my long-time friend, Dr. Karim Nader. I believe Dr. Nader made a paradigm-shifting discovery about memory. Here, I will provide my personal take on Dr. Nader and his findings, and then my team and I will describe some data from our memory lab conducted in the spirit of Dr. Nader’s research style.

I have never seen Dr. Karim Nader in a modest mood. We first met in grad school. And Karim had swagger even then. This swagger allowed him to take on dogma in the established memory field and ask what some might characterize as heretical questions on how the brain stores and uses information. The textbook thinking in the late 1990 s was that experiences were initially encoded as short-term memories that can be modified and disrupted before being consolidated into long-term memories that were more immune to modification and disruption. This process, referred to as cellular consolidation, required protein synthesis to somehow stabilize the memory. In support of cellular consolidation were the results of many experiments showing that administering a protein synthesis inhibitor before or shortly after a training experience blocked the consolidation of long-term memory, while leaving short-term memory intact. One such paper used Pavlovian threat conditioning, a well-characterized paradigm popularized by Joe LeDoux among others. LeDoux was Karim’s post-doctoral mentor. In a typical conditioned threat memory experiment, an initially innocuous tone (the conditioned stimulus, CS) is paired with an aversive footshock (the unconditioned stimulus, US) (LeDoux, 2000; Maren, 2003; Davis, 1992). Memory is examined by the time rodents spent freezing (the conditioned response, CR) when the tone is replayed. Freezing is a robust, sophisticated species-specific response to threat and not simply a passive response (Fanselow and Lester, 1988; Blanchard and Blanchard, 1969; Roelofs and Dayan, 2022).

Multiple lines of evidence implicated the amygdala, particularly the lateral nucleus of the amygdala (LA), in Pavlovian threat conditioning (LeDoux, 2000; Davis, 1992; Clugnet and LeDoux, 1990; Campeau and Davis, 1995; Fanselow and Gale, 2003; Maren and Fanselow, 1996; Josselyn et al., 2001). In the literature, the LA is sometimes included in a larger basolateral complex (including the LA, basal nucleus of the amygdala and sometimes the accessory basal nucleus), but Nader and colleagues in the LeDoux lab showed that the LA, but not the basal nucleus of the amygdala or the accessory basal, is required for expression of a threat memory (Nader et al., 2001). Importantly, other findings from the LeDoux lab found that blocking protein synthesis in the LA immediately after training disrupted the consolidation of long-term conditioned threat memory (Schafe and LeDoux, 2000).

Enter Dr. Nader into the mix. His idea was that consolidated memories could be modified, and even disrupted, very much like newly acquired information, if the consolidated memory was reactivated, rather than just lying dormant. Using deceptively simple experimental procedures, Karim, along with fellow LeDoux lab post-doc, Glenn Schafe, took on the dogma of the time head-on and showed that consolidated memories could be disrupted (Nader et al., 2000). Specifically, Nader and colleagues administered the protein synthesis inhibitor directly into the LA after replay of the tone in the memory test (when the memory would be recalled and active) rather than after training. Subsequent memory was disrupted. The finding of reconsolidation, as it became known, also disrupted the memory field. Since the time of Nader’s original publication, a myriad of studies replicated the basic reconsolidation effect using several pharmacological interventions to disrupt many types of consolidated memories in several species. Boundary conditions for reconsolidation were identified and studies examined the efficacy of the reconsolidation phenomenon as a jumping off point for treatments of memory disorders in humans (Kida et al., 2002; Suzuki et al., 2004; Tronson et al., 2006; Sekeres et al., 2012; Morris et al., 2006; Eisenberg and Dudai, 2004; Eisenberg et al., 2003; Walker et al., 2003; Kindt et al., 2009; Monfils et al., 2009). The field’s thinking about how memory works changed. Textbooks were updated (reconsolidated?). None of this would have been possible without Karim’s swagger.

Inspired by this swagger, my lab and I embarked on our own journey to examine the neural basis of memory, focusing, not on reconsolidation but on the neuronal ensembles that are important in storing a memory (the engram). An engram is thought to be sparsely encoded across several brain regions, with some regions serving as critical hubs (Tonegawa et al., 2015; Josselyn et al., 2015, 2017; Eichenbaum, 016; Denny et al., 2017). We, and others, previously showed it is possible to tag and manipulate critical components of an engram that support a specific memory to modify the specific underlying memory (Han et al., 2009, 2007; Rashid et al., 2016; Zhou et al., 2009; Kim et al., 2014; Kim and Cho, 2020; Choi et al., 2021; Guskjolen et al., 2018; Hsiang et al., 2014; Lau et al., 2020; Yiu et al., 2014).

Our attention focused primarily on the role of the sparse population of LA neurons allocated to an engram supporting a specific conditioned threat memory. Ablating LA neurons allocated to an engram supporting a Pavlovian threat memory, and not a similar number of random neurons at any time after training disrupted subsequent memory recall (Han et al., 2009). That memory could be impaired by killing allocated neurons even in the absence of the tone CS (and memory retrieval or an “active engram”), indicates that a process distinct from reconsolidation is at work. Since our initial findings, similar results were shown in other brain regions and other memories from many labs (Rashid et al., 2016; Kim et al., 2014; Tanaka et al., 2014; Park et al., 2016; Ghandour et al., 2019; Liu et al., 2012; Denny et al., 2014; Lacagnina et al., 2019; Cowansage et al., 2014; Matos et al., 2019; Kitamura et al., 2017; Sano et al., 2014; Abe et al., 2020). Together, these results argue that the engram neurons (neurons critical to a given engram) are in some way necessary for subsequent memory retrieval. We hypothesized that silencing engram neurons disrupted memory retrieval because external sensory retrieval cues initiate memory retrieval by reacting engram neurons. Consistent with this, behaviour consistent with memory retrieval is produced by optogenetic stimulation of critical components of an engram alone, in the absence of the sensory retrieval cue (Guskjolen et al., 2018; Liu et al., 2012; Frankland et al., 2019; Ramirez et al., 2015). These results argue that engram neurons are in some way sufficient for subsequent memory retrieval.

In a conditioned threat memory experiment, the tone CS would induce the conditioned response (freezing, memory retrieval). Multiple presentations of a CS alone after the conditioned threat memory has been acquired, though, reduce freezing, in a process of extinction (Pavlov, 1927; Bouton, 1988, 2004; Myers and Davis, 2007). If the CS reactivates the engram in the LA, we asked whether optogenetic stimulation of the LA component of the engram alone, in the absence of the tone CS, would similarly induce extinction of a threat memory. Furthermore, we asked whether this “opto-extinction” would transfer to the sensory tone CS such that opto-extinguished mice would show reduced freezing to the tone.

2. Materials And Methods

Mice.

All experiments were conducted according to the guidelines provided by The Hospital for Sick Children Animal Care and Use Committee in addition to the Canadian Council on Animal Care (CCAC) and the NIH Guidelines on the Care and Use of Laboratory Animals. Adult mice (8–12 weeks of age) used in these experiments were the F1 generation of a cross between C57BL/6NTac and 129S6/SvEvTac mice. An equal number of male and female mice were used. 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. All behavioural experiments were conducted during the light cycle.

Optogenetic expression.

HSV-NpACY.

The NpACY construct encodes both enhanced channelrhodopsin (ChR2-H134R) and halorhodopsin 3.0 (eNpHR3.0) as well as eYFP. This construct enables bidirectional control of neuronal activity of the same population of NpACY⁺ neurons. These two opsins are spectrally compatible; neurons can be excited by Blue light (BL; 473 nm) activation of ChR2, and inhibited by Red light (RL; 660 nm) activation of NpHR3.0. Opsin genes were linked by a self-cleavage linker derived from porcine teschovirus (p2A). NpACY has been validated by in vitro and in vivo assays (Rashid et al., 2016; Lau et al., 2020; Vesuna et al., 2020; Stahlberg et al., 2019; Zhang et al., 2007), whole-cell current-clamp experiments showed the activation spectra for ChR2 and NpHR are separable by ~100 nm (Zhang et al., 2007), and there is minimal cross-talk between the two wavelengths of light (Rashid et al., 2016; Stahlberg et al., 2019). Expression of NpACY was driven by the HSV (herpes simplex virus) promoter IE 4/5 (Carlezon et al., 2000).

We packaged the NpACY construct in replication-defective HSVs. HSV viral vectors were packaged using a replication-defective helper virus, purified on a sucrose gradient, pelleted and resuspended in 10% sucrose, as previously described (Han et al., 2008). The average titer of the virus stocks was 4.0 × 10⁷ infectious units/ml. HSV viral vectors randomly infects approximately 10% of excitatory neurons in the LA (Yiu et al., 2014). Transgene expression from the HSV viral system typically peaks 3 d following microinjection (Carlezon et al., 2000; Carlezon and Neve, 2003; Neve et al., 2005; Barrot et al., 2002; Park et al., 2020; Cole et al., 2012).

AAV(DJ)-ChR2.

To increase the neuronal excitability of pyramidal neurons in the LA over the course of a longer experiment, we used an AAV(DJ) viral system. ChR2-H134R was fused to eYFP and this construct was driven by an αCaMKII promoter. AAV vectors were packaged in-house (approx. 10¹³ infectious units/ml). Optogenetic manipulations and behavioural experiments were performed 3–4 weeks after microinjection, at a time of high transgene expression.

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 (4 mg/kg, s.c.) for analgesia, and placed in a stereotax. Viral vectors were infused bilaterally (HSV, 1.5 μL/side, flow rate 0.12 μL/min; AAV, 0.6 μL, 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 slightly 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). Fibers were attached with epoxy resin into ferrules, cut and polished. Optical fibers were stabilized to the skull with screws and black dental cement.

Verifying location of vector injection and extent of viral infection.

At the completion of each behavioural experiment, mice were transcardially perfused with 0.1 M PBS followed by 4% paraformaldehyde (PFA) and brains removed. Brains were post-fixed overnight and coronally sectioned (50 μm) across the entire anterior-posterior extent of the LA. To determine placement and extent of the viral infection for each mouse, coronal brain slices were incubated with blocking solution (0.1% BSA, 2% NGS, 0.3% Triton X-100) for 2 h at room temperature, then with anti YFP (GFP) primary rabbit antibody (1:1000, ThermoFisher) for 24 h. Sections were washed, then incubated with anti-rabbit Alexa 488 secondary antibody (1:500, Invitrogen) for 2 h at room temperature (Lau et al., 2020). Following washing, sections were mounted on gel-coated glass slides, and coverslipped using DAPI-containing mounting medium (Vectashield, Vector Labs, Burlingame, CA). Images were obtained using a confocal laser scanning microscope (LSM 710; Zeiss). Each mouse was categorized as a “hit” or “miss” by an experimenter unaware of the treatment condition and behavioural results. Mice were defined as “hits” if bilateral YFP (GFP) expression was observed in LA in at least 5 consecutive brain sections (across the anterior-posterior plane). All other mice were classified as “miss” (including those with unilateral, weak or no transgene expression in the LA). Only mice determined to be a bilateral “hit” were included in subsequent data analysis.

2.1. General behavioural procedure

Conditioned tone threat training.

Mice were placed in a training chamber (Context A) and 2 min later presented with a 30-sec auditory Tone CS (2.8 kHz, 85 dB) that co-terminated with a 2-sec footshock US (0.6 mA, scrambled electrical footshock). Mice remained in the chamber for an additional 30 s and were then returned to home-cage.

For Allocation experiments, ChR2-expressing mice in the Allocation group were photostimulated with Blue light (BL+, 473 nm, 20 Hz, 5 ms pulse width, ~1 mW output, 30 s) to activate ChR2-expressing neurons, increase their excitability and bias their allocation to the engram, immediately before training.

Conditioned tone threat memory testing.

Mice were placed in a novel chamber (Context B) and, after 2 min, presented with the Tone CS for 1 min. Time spent freezing (cessation of all movement except respiration) was measured before and during the Tone via automated procedures (Han et al., 2007) or hand-scoring of videos by two experimenters unaware of mouse treatment condition.

For experiments involving optogenetics during the test, Red light (RL+, 660 nm, 7 mW output, square pulse, 1 min) was delivered during the CS to inhibit NpHR3.0 + neurons. In these experiments, freezing was assessed during two tone presentations, one tone with RL and one tone without RL, in a counterbalanced order.

Conditioned threat extinction training.

24 h after conditioned threat training, mice in the Extinction group were placed in a novel chamber (Context B) for Tone threat extinction training. Mice were presented with 16 Tones (30 s) in the absence of footshock (with a random interval between 20 and 180 s). Time spent freezing during the Tones was assessed. Mice in the No-extinction group remained in the homecage.

For “opto-extinction” training, both Allocated and Non-allocated mice were treated similarly but instead of being presented with 16 Tones, mice were given 16 BL stimulations (30 s) at the same temporal interval (random interval between 20 and 180 s). Time spent freezing during BL presentation was assessed. For the Allocated group, BL reactivated neurons allocated to an engram while in the Non-allocated group, BL activated a similar number of non-allocated neurons.

Extinction recall testing.

24 h after Tone extinction training, mice were placed in the extinction training chamber (Context B) and 2 min later, the Tone was presented (1 min). Mice in the opto-extinction experiment were tested the same (with a Tone).

Threat memory renewal testing.

24 h after extinction recall testing, mice were placed in a novel chamber (Context C) and 2 min later, the Tone was presented (1 min). The same procedure was used in the opto-extinction experiment and the time spent freezing to the Tone was assessed.

2.2. Statistical analysis

Percentage of time spent freezing was compared using an Analysis of Variance (ANOVA) Prism GraphPad 8 software or Statistica. Where appropriate, significant effects were further analyzed using post-hoc Newman-Keuls tests. Details of each analysis are presented in the Results section.

3. Results

3.1. Verifying the allocation method of engram labeling

To identify and manipulate LA neurons in an engram, we used the engram allocation approach (Josselyn et al., 2015; Josselyn and Frankland, 2018; Sehgal et al., 2018). This method takes advantage of the finding that eligible neurons compete for allocation to an engram, in part, based on their relative intrinsic excitability (Han et al., 2009, 2007; Zhou et al., 2009; Sano et al., 2014; Rogerson et al., 2016; Gouty-Colomer et al., 2015). Previously, we, and others, showed that increasing the excitability of a sparse population of pyramidal neurons (with various constructs that increase excitability, including optogenetic or chemogenetic actuators, the transcription factor CREB, or a dominant-negative K⁺-channel) immediately before training in either an aversive or rewarding task biases the allocation of these neurons into the engram supporting this memory (Han et al., 2009, 2007; Rashid et al., 2016; Zhou et al., 2009; Hsiang et al., 2014; Lau et al., 2020; Yiu et al., 2014; Park et al., 2016). Here we used sparse expression of ChR2 to bias a small population of LA pyramidal neurons to the engram supporting the tone threat conditioned memory. To confirm that neurons we excited immediately before training became critical components of the engram supporting the conditioned threat memory, we used an all-optical approach in which we silenced allocated neurons during a memory test.

We took advantage of a viral vector (NpACY) expressing both a Blue-light (BL) sensitive excitatory opsin, ChR2, and a Red-light (RL) sensitive inhibitory opsin, eNpHR3.0 (Rashid et al., 2016; Lau et al., 2020; Vesuna et al., 2020) to allocate and then silence the same population of neurons (Fig. 1 A, B). NpACY was expressed in a sparse (roughly 10%), random population of LA pyramidal neurons in two groups of mice. In one group (Allocated), we photostimulated NpACY-expressing neurons with BL immediately before training to increase the excitability of NpACY-expressing neurons and bias their allocation into the resulting engram. In the control group (Non-allocated) no BL was given before training such that NpACY-expressing neurons were not experimentally-allocated to an engram. During training, mice received a Tone CS paired with a footshock US in Context A (Fig. 1 A). 24 h later, memory was assessed in Context B as the time mice spent freezing to the Tone. The Tone was presented twice in this memory test, once in the presence of RL optogenetic inhibition to silence the activity of NpACY-expressing neurons, and once without RL (counterbalanced order).

Fig. 1.

The allocation method for tagging an engram. This method is based on previous findings that excitatory pyramidal neurons compete for allocation to a sparse engram, in part, based on neuronal excitability. We used this method to bias the allocation of infected neurons into an engram in the lateral amygdala (LA) supporting an auditory threat memory. (A) An HSV viral vector (HSV-NpACY) expressed both a blue light (BL)-sensitive optogenetic actuator (ChR2) and a red light (RL)-sensitive optogenetic inhibitor (eNpHR3.0) in a sparse random population of pyramidal LA neurons. In this way, the activity of the same population of infected neurons could be excited or inhibited. Two groups of mice were trained in auditory threat conditioning (Tone+Shock) in Context A. Immediately before training, Allocated mice received BL (BL⁺) stimulation to excite infected neurons and bias their allocation to the engram. Non-allocated mice did not receive BL (BL⁻) before training. Memory was tested 24 h later in a novel Context B. Mice received two Tone presentations, once in the presence of RL to inhibit NpACY-expressing neurons and once in the absence of RL. (B) Representative images showing robust, but sparse, expression of the NpACY construct in LA pyramidal neurons. Lateral nucleus of the amygdala (LA), basal nucleus of the amygdala (BA), central nucleus of the amygdala (CeA). (C) Both Allocated and Non-allocated mice show robust Tone freezing in absence of RL; only Allocated mice show disruption in presence of RL, verifying the importance of these allocated neurons to the engram. * ** *, p < 0.0001.

During the memory test, both groups showed low Pre-Tone freezing and high freezing when the Tone was replayed in the absence of RL. RL disrupted freezing only in Allocated mice that received BL before training (to bias their allocation to the engram) and not in Non-allocated mice that were not photostimulated before training (Fig. 1 C) [two-way repeated-measures ANOVA with between-group factor Allocation (Allocated vs. Non-allocated) and within-group factor Test phase (Pre-Tone, Tone RL⁻, Tone RL⁺), significant interaction, F_2,14 = 37.01, p < 0.0001, as well as main effects of Allocation F_1,7 = 11.66, p < 0.05, and Test phase F_2,14 = 118.30, p < 0.0001. Allocated mice (and not Non-allocated mice) froze less to the Tone in the presence, than in the absence, of RL inhibition (p < 0.0001, post-hoc Newman-Keuls)]. These results verify that the allocation method can be used to label and manipulate neurons critical to a particular engram. We used the allocation method to examine whether optogenetic activation of an LA engram alone can be used to induce memory extinction.

3.2. Extinction with a Tone CS

First, we examined the effects of “real” extinction training (repeated exposure to the Tone alone) in a conditioned Tone threat paradigm in mice (without engram manipulation). Two groups of mice (Extinction, Non-extinction) were first tone threat conditioned as above. 24 h later, Extinction mice were placed in Context B and 16 Tones were presented in the absence of the footshock to induce extinction. Non-extinction mice remained in the homecage. As expected, Extinction mice showed decreased Tone freezing over the course of extinction training; mice showed high Tone freezing early in extinction training and low Tone freezing late in extinction training (Fig. 2 B) (ANOVA, early Tone freezing vs. late Tone freezing; F_1,7 = 70.28, p < 0.0001). Both groups of mice were tested for recall of extinction training 24 h later in Context B. As expected, only mice in the Extinction group showed decreased Tone freezing, verifying the effectiveness of the extinction training protocol (Fig. 2 C) [two-way repeated-measures ANOVA with between-group factor Extinction training (Extinction vs. Non-extinction) and within-group factor Test phase (pre-Tone vs. Tone) significant interaction, F_1,7 = 64.91, p < 0.0001, and main effects of Extinction training F_1,7 = 75.08, p < 0.0001, and Test phase F_1,7 = 78.38, p < 0.0001. Extinction mice froze less to the Tone than Non-extinction mice (p < 0.0001, Newman-Keuls post-hoc)].

Fig. 2.

Optogenetic activation of a threat engram alone induces behaviour similar to extinction training. (A) Validation of Tone threat extinction in mice. Mice were trained [Tone conditioned stimulus (CS) + Shock (US) in Context A]. (B) Extinction mice received extinction training (16 Tone CSs in Context B) while Non-extinction mice did not. Extinction mice froze at greater levels to Tone early, rather than late, in extinction training, showing extinction. (C) As expected, Extinction, but not Non-extinction, mice recalled extinction training when tested subsequently with the Tone in Context B. (D) Also as expected, both groups showed robust Tone freezing in a renewal test in Context C. * ** *, p < 0.0001. (E) As the Tone CS is thought to reactivate the engram supporting the Tone threat memory, we examined whether direct optogenetic stimulation of the LA engram alone (without Tone) would produce a similar behavioural profile. The LA engram was tagged using the allocation method in mice expressing ChR2. Allocated (BL⁺ before training) and Non-allocated (no BL before training) were trained and tested as depicted. (F) Representative images for AAV-αCaMKII-ChR2 expression in LA. Lateral nucleus of the amygdala (LA), basal nucleus of the amygdala (BA), central nucleus of the amygdala (CeA). (G) Both groups received BL photostimulation (16 BL in Context B) (similar to the purely behavioural experiment in A). As with Tone extinction (B), Allocated mice froze at greater levels to BL earlier, rather than later, in the session, while Non-allocated mice did not freeze to BL. (H) Similar to (C) Allocated, but not Non-allocated, mice showed an extinction-like decrease in freezing to the Tone CS, suggesting that “opto-extinction” transferred to the actual CS. (I) However, unlike Tone extinction, opto-extinguished mice showed decreased freezing during the renewal test in Context C, suggesting some similarities and differences with Tone extinction. * ** , p < 0.001. * *, p < 0.01.

Renewal of the threat memory after extinction was examined in a novel context (Context C). Both Extinction-trained mice and mice that did not receive extinction training froze at similarly high levels to the Tone (Fig. 2 D) [Extinction training × Test phase ANOVA; no significant main effect of Extinction training F_1,7 = 1.20, p > 0.05, or Extinction training × Test phase interaction, F_1,7 = 0.07, p > 0.05, but significant effect of Test phase F_1,7 = 235.10, p < 0.0001. Both groups showed higher freezing to the Tone than pre-Tone (p < 0.0001, post-hoc Newman-Keuls)]. This finding shows renewal of the previously extinguished freezing response, and is consistent with previous findings (Lacagnina et al., 2019; Maren and Quirk, 2004) arguing that extinction learning is context-dependent (Bouton, 2004; Bouton and Bolles, 1979).

3.3. Extinction with optogenetic activation of the LA engram

Memory retrieval is thought to be initiated by a retrieval cue reactivating the engram supporting this memory. Consistent with this, optogenetic activation of a critical component of the engram induces behaviour consistent with memory retrieval even in the absence of a sensory retrieval cue (Guskjolen et al., 2018; Liu et al., 2012; Frankland et al., 2019; Ramirez et al., 2015). Therefore, we assessed whether optogenetic stimulation of the engram alone would be sufficient to induce an extinction-like effect and whether this opto-extinction would also transfer to the “real” Tone.

Allocated and Non-allocated mice (as above) were tone threat conditioned in Context A, as above. 24 h later, both groups of mice were placed in Context B and received 16 photostimulations of ChR2⁺ neurons (with a similar temporal profile to the above Tone extinction training). In the Allocated group, BL stimulation would reactivate neurons biased for allocation to the engram, whereas in the Non-allocated group, BL stimulation would activate a small population of random, non-experimentally allocated neurons. Freezing during BL stimulation was assessed. As with Tone extinction training, Allocated mice froze at high levels to BL stimulation early in the session and low levels by the end of the session. Non-allocated mice did not freeze to BL stimulation throughout the session (Fig. 2 G) [Allocation × Extinction session phase (early vs. late) ANOVA showed significant main effects of Allocation F_1,17 = 65.31, p < 0.001, and Extinction session phase F_1,17 = 33.66, p < 0.01, and a significant Allocation × Extinction session interaction, F_1,17 = 55.21, p < 0.01. Allocated mice froze less to BL stimulation late, compared to early in the session (p < 0.001, by post-hoc)]. The following day, both groups were tested in Context B for the recall of extinction training as above. Importantly, the Tone CS, rather than BL, was presented. Allocated mice showed decreased Tone freezing compared to Non-allocated mice (Fig. 2 H) [Allocation x Test phase ANOVA; significant effects of Allocation F_1,17 = 18.04, p < 0.001, and Test phase F_1,17 = 100.6, p < 0.0001, and Allocation × Test phase interaction, F_1,11 = 48.60, p < 0.0001. Allocated mice froze less to the Tone than Non-allocated mice (p < 0.0001, by post-hocs)]. This freezing profile is similar to the Tone extinction groups (Fig. 2 C), suggesting that optogenetic stimulation of the engram alone substitutes for CS presentation during extinction training. That is, opto-extinction mimics CS extinction training, at least in some regards.

In Fig. 2 D, we showed that Tone extinction training does not “erase” or degrade the conditioned threat memory, because Extinguished mice froze at high levels to the Tone in a renewal test, similar to freezing levels observed in Non-extinguished mice. In contrast, Allocated opto-extinguished mice freeze less to the Tone than Non-allocated mice in the renewal test, even though both groups received the same amount of optogenetic stimulation (Fig. 2 I) [Allocation × Test phase ANOVA, significant effects of Allocation F_1,27 = 5.96, p < 0.05, and Test phase F_1,27 = 118.30, p < 0.0001, Allocation x Test phase interaction, F_1,27 = 6.39, p < 0.05. Allocated mice froze less to the Tone than Non-allocated mice (p < 0.0001, by post-hocs)]. Therefore, unlike “real” CS-induced extinction in which the memory was recovered by presenting the CS in a novel environment (renewal), the memory was not entirely recovered in opto-induced extinction. This finding suggests that opto-extinction and CS-extinction are not always equivalent.

4. Discussion

Although much progress has been made in identifying engrams and leveraging our knowledge of engrams to modify the memories they support, much more remains unknown. In a Pavlovian conditioning paradigm, a CS is thought to induce memory recall by reactivating the underlying engram (Guskjolen et al., 2018; Liu et al., 2012; Frankland et al., 2019; Ramirez et al., 2015). As extinction of the conditioned response may be induced by multiple exposures of the CS in the absence of the US, we examined whether optogenetic stimulation of the engram alone could substitute for a CS in an extinction experiment. We used tone threat conditioning because it is a well-characterized task. Confirming previous results, we found that after tone threat conditioning, repeated presentation of the Tone CS alone during a single extinction-training session induced extinction. This extinction memory was recalled the following day when mice were tested in the extinction context. However, mice showed renewal of CS freezing when subsequently tested in a novel context. That is, mice that underwent extinction training froze at similar levels to the CS as mice that did not undergo extinction training. We compared this behavioural profile following repeated presentations of the CS to that induced by repeated optogenetic stimulation of the LA component of the conditioned threat engram.

We trained two groups of mice in the opto-extinction experiment. In one group, we biased the allocation of ChR2 + neurons to the engram such that these tagged neurons could be optogenetically stimulated during the extinction-like training (Allocated). In the second group, we optogenetically stimulated a similar number of neurons not experimentally allocated to the engram (Non-allocated). Similar to CS-induced extinction, Allocated mice gradually froze less to repeated optogenetic stimulation of the engram during extinction-like training. Similar to CS-induced extinction, Allocated mice also showed recall of this extinction memory when tested the next day in the extinction context. Importantly, in this test Allocated mice showed decreased freezing to the actual Tone CS even though the opto-extinction training did not involve a Tone (only optogenetic stimulation). In contrast to CS-induced extinction, however, renewal of the conditioned freezing response to the CS was impaired in a novel context. These findings point out the similarities and differences between real CS-induced extinction and opto-extinction.

The processes and mechanisms underlying extinction of a threat memory have been well-studied. The original threat memory can return with time (spontaneous recovery), re-exposure to the US (reinstatement), or re-exposure to the CS in a different context (renewal) (Lacagnina et al., 2019; Bouton, 2004; Maren and Quirk, 2004; Bouton and Bolles, 1979). These findings suggest that extinction is a form of new learning that inhibits the original threat memory (Lacagnina et al., 2019; Myers and Davis, 2007; Bouton et al., 2006). In fact, Pavlov originally characterized extinction as the “internal inhibition of conditioned reflexes” (Pavlov, 1927). However, other findings suggest that extinction involves “unlearning” of the original memory (Dunsmoor et al., 2015; Clem and Schiller, 2016).

In support of the unlearning hypothesis of extinction, for instance, are observations that extinction involves weakening of the original fear memory in the LA via synaptic depotentiation or an LTD-like process (Kim et al., 2007; Park et al., 2012). Consistent with this, Kaang and colleagues found that synapse size between engram neurons in the LA and the upstream auditory cortex increase with tone threat conditioning. The size of these synapses, however, decreased selectively with CS-induced extinction, and importantly, was restored with re-training (Choi et al., 2021). Together, these observations suggest that some extinction protocols may degrade the original threat engram.

Engrams are thought to exist in a continuum of states, from readily accessible by natural retrieval cues but not being accessed at the current time (accessible and available, but dormant) to completely inaccessible such that stored information can never be retrieved (not accessible, unavailable) (Frankland et al., 2019; Ryan and Frankland, 2022; Josselyn and Tonegawa, 2020; Tonegawa et al., 2018). Between these extremes are “silent engrams”. Silent engrams are available (the information is still present in the system) but cannot be readily accessed by natural retrieval cues. However, silent engrams can be accessed by direct optogenetic stimulation of the tagged engram showing that under some circumstances engrams can be inaccessible, but still available. For instance, a “lost” memory that cannot be revealed with natural retrieval cues may be recovered by direct optogenetic stimulation of the tagged engram (Guskjolen et al., 2018; Kitamura et al., 2017; Ryan et al., 2015; Roy et al., 2016). One process thought to silence an engram is disruption of synapses between engram neurons.

Here, we found that opto-extinction mice failed to show renewed Tone freezing in a memory test in a novel context using the natural retrieval cue. It may be that opto-extinction degraded some functional connections between engram neurons to silence this original threat engram or render the engram entirely unavailable. Future experiments are needed to determine whether optogenetic reactivation of the opto-extinguished engram would be sufficient to induce freezing (showing that the engram is silent, but can be reactivated optogenetically) or not (showing that the engram is unavailable and cannot be re-awakened even with optogenetic stimulation).

Understanding how information is encoded, stored and used in the brain is an important question in neuroscience. However, the brain in incredibly complex. We assert that to make progress in solving this great mystery requires researchers to have the swagger of Karim Nader.

Abbreviations:

CS

conditioned stimulus

US

unconditioned stimulus

LA

lateral nucleus of the amygdala

Data Availability

Data will be made available on request.