Report

Reactivation of recall-induced neurons contributes to remote fear memory attenuation

See allHide authors and affiliations

Science  15 Jun 2018:
Vol. 360, Issue 6394, pp. 1239-1242
DOI: 10.1126/science.aas9875

The mechanisms of fear attenuation

Surprisingly little is known about how remote fearful memories are stored and attenuated. Khalaf et al. used independent fear memory attenuation paradigms, engram-based tagging techniques, and chemogenetic tools to alter neuronal activity (see the Perspective by Frankland and Josselyn). They found that a discrete subset of neurons within an ensemble is engaged during recall after memory attenuation, which correlated with fear reduction. Memory updating and extinction mechanisms thus likely coexist to make this happen. These findings support the notion that effective memory attenuation is mediated by a rewriting of the original memory trace of fear toward one of safety.

Science, this issue p. 1239; see also p. 1186

Abstract

Whether fear attenuation is mediated by inhibition of the original memory trace of fear with a new memory trace of safety or by updating of the original fear trace toward safety has been a long-standing question in neuroscience and psychology alike. In particular, which of the two scenarios underlies the attenuation of remote (month-old) fear memories is completely unknown, despite the impetus to better understand this process against the backdrop of enduring traumata. We found—chemogenetically and in an engram-specific manner—that effective remote fear attenuation is accompanied by the reactivation of memory recall–induced neurons in the dentate gyrus and that the continued activity of these neurons is critical for fear reduction. This suggests that the original memory trace of fear actively contributes to remote fear attenuation.

Traumatic memories develop after the experience of grave physical or psychological harm. Such memories are extraordinarily robust and difficult to treat, with an estimated lifetime prevalence of fear- and stress-related disorders of close to 29% (1). Despite these circumstances, little is known about how long-lasting (remote) fearful memories are stored and can be attenuated. In particular, it is unclear whether the attenuation of remote fearful memories represents an inhibition of the original trace of fear with a new memory trace of safety, a process termed extinction (24), or an unlearning of the original fear memory trace toward safety following memory recall, termed reconsolidation-updating (5, 6).

To investigate this question, we visualized memory traces implicated in remote fear memory storage and attenuation by using double transgenic TetTag mice, which provide a c-fos–based labeling system of neuronal activity (7) (Fig. 1A). TetTag mice show no difference from wild-type mice in general anxiety and overall locomotion, making them ideally suited for studying acquired fear-related behaviors (fig. S1). We focused on the dentate gyrus (DG) because of its importance in the encoding (8), recall (9, 10), and attenuation (11) of contextual fear and because of the recently demonstrated implication of DG granule cells in fear memory retrieval by c-Fos–based engram studies (10, 1214).

Fig. 1 Remote fear attenuation is characterized by reactivation of memory recall–induced neurons in the DG but not CA3.

(A) Schematic representing the inducible double transgenic TetTag mouse (top) and the experimental design used in the spaced extinction paradigm (bottom). TRE, tetracycline response element; CFC, contextual fear conditioning; d, days; ′, minutes; IHC, immunohistochemistry. (B) Confocal images of the DG and CA3 regions. Activated neurons at remote memory recall express β-gal (green), whereas those activated by the last extinction session express endogenous c-Fos (red). Scale bars, 20 μm. Smaller panels below show enlarged areas within the dotted squares. Arrows indicate recall-induced neurons reactivated by extinction training. Scale bars, 10 μm. (C) The reactivation rate shows a significant increase in the DG (n = 17 animals) but not in CA3 (n = 9 animals) over chance levels (dashed lines). Data are means ± SEM compared by a two-tailed, unpaired t test. ***P = 0.001; n.s., not significant. (D) The reactivation rate in the DG (circles) but not that in CA3 (triangles) is positively correlated with the degree of fear memory attenuation (Δ freezing). DG, n = 17 animals, correlation coefficient r = 0.7585, P = 0.0004; CA3, n = 9 animals, r = 0.1805, P = 0.6421. (E) Schematic of the experimental design used in the massed extinction paradigm for wild-type mice. (F) Confocal images of the DG and CA3 regions. Activated neurons at remote memory recall express cytoplasmic Homer1a (cyt H1a) mRNA (green), whereas those activated during extinction training express nuclear (nuc) c-fos mRNA (red). Scale bars, 20 μm. Smaller panels below show enlarged areas within the dotted squares. The arrow indicates a representative recall-induced neuron reactivated by extinction training. Scale bars, 10 μm. (G) The reactivation rate shows a significant increase in the DG (n = 8 animals) but not in CA3 (n = 8 animals) over chance levels (dashed lines). Data are means ± SEM compared by a two-tailed, unpaired t test (****P < 0.0001). (H) The reactivation rate in the DG (circles) but not that in CA3 (triangles) is positively correlated with the degree of fear memory attenuation (expressed as Δ freezing). DG, n = 8 animals, r = 0.7125, P = 0.0474; CA3, n = 8 animals; r = 0.0055, P = 0.9896.

First, we identified the cellular ensemble participating in remote (28-day-old) fear memory storage. TetTag mice were contextually fear-conditioned while on doxycycline (DOX) and then taken off DOX 3 days prior to remote memory recall, which triggered lasting β-galactosidase (β-gal) expression in recall-induced neurons (Fig. 1, A and B). Immediately after memory recall, TetTag mice were put on DOX again, and 1 day later they entered a previously described spaced extinction paradigm (15), which successfully reduced their fearful memories (fig. S2). Second, we used endogenous c-Fos expression 1 hour after the last extinction trial to identify the cellular ensemble participating in memory attenuation per se (Fig. 1, A and B).

We then evaluated whether neurons activated by remote memory recall (β-gal+ cells) became reengaged during extinction training (as indicated by c-Fos+ cells) by calculating their reactivation rate, defined as the number of double-positive β-gal+ c-Fos+ cells normalized to the total number of β-gal+ cells. In the DG, this reactivation rate was significantly higher than chance levels in the DG, unlike in hippocampal area CA3 (Fig. 1, B and C), which is of minimal importance for remote contextual fear retrieval (10). Furthermore, we found a positive correlation between the reactivation rate and the degree of fear memory attenuation (expressed as Δ freezing, the difference in freezing between the recall and the last extinction session) in the DG but not in CA3 (Fig. 1D). Compared with a home cage control group, the CS-US group (the group exposed to the conditioned stimulus in association with the unconditioned stimulus) showed an increased recall-induced activation rate (the number of β-gal+ cells normalized to the total number of cells) in both the DG and CA3, testifying to the specificity of β-gal+ as a marker of remote memory recall (fig. S3). Likewise, the DG and CA3 did not differ in the extinction learning rate (the number of extinction-induced c-Fos+ cells normalized to the total number of cells) (fig. S4), and neither did the activation or the extinction learning rate correlate with fear memory attenuation (fig. S5). These findings suggest that remote fear attenuation is facilitated as more recall-induced neurons are reactivated by the extinction paradigm.

We independently verified these findings in a different paradigm of memory attenuation by using a second mouse line and a different technique to visualize neuronal ensembles. C57Bl6/J wild-type mice were contextually fear-conditioned as before (Fig. 1E), and 28 days later they entered a previously described massed extinction paradigm (15), which successfully attenuated their remote fear memories (fig. S6) to levels comparable to those obtained in the spaced extinction paradigm (fig. S7). Then we used the intracellular spatiotemporal characteristics of immediate early gene mRNA species and cellular compartment analysis of temporal activity by fluorescence in situ hybridization (catFISH) (16, 17) to identify the neuronal populations activated during remote fear memory recall (those displaying cytoplasmic Homer1a mRNA) and during massed extinction (those displaying nuclear c-fos mRNA) (Fig. 1, E and F). Similar to the results in the spaced extinction paradigm, the reactivation rate in the DG was significantly higher than chance levels and correlated positively with fear memory attenuation, in contrast to that in hippocampal area CA3 (Fig. 1, F to H). Compared with levels in the home cage control group, the recall-induced activation rate (the total amount of Homer1a+ cells) in both areas was significantly elevated (fig. S8), indicating Homer1a+ labeling to be specific for memory recall. Similarly, there was no difference in the extinction learning rate (the total amount of c-fos+ cells) between the DG and CA3 (fig. S9), and neither the activation nor the extinction learning rate correlated with fear attenuation (fig. S10).

Next, to test whether the sustained activity of recall-induced neurons contributes to the behavioral manifestation of fear attenuation, we used the Daun02 inactivation method (18), which reduces neuronal excitability (19) by means of daunorubicin, a derivate converted from the prodrug Daun02 exclusively by β-gal (Fig. 2A). TetTag mice were fear-conditioned and β-gal+ expression was triggered by remote recall as described above. Between training and recall, cannulas were stereotaxically implanted in the DG so that Daun02 could be readily administered 90 min post–memory recall (Fig. 2B) to inhibit the activity of recall-induced neurons. Then the spaced extinction paradigm was performed. Compared with vehicle-treated animals, Daun02-treated TetTag mice showed impaired fear memory attenuation in the DG (Fig. 2C) despite displaying similar freezing at recall (fig. S11). This was accompanied by a reduced number of c-Fos+ cells at the completion of memory attenuation (Fig. 2, D and E), consistent with the time course of 4-day-long neuronal ensemble inactivation after a single Daun02 injection in other brain areas (18, 20). In contrast, when administered in hippocampal area CA3, not characterized by reactivated recall-induced neurons after attenuation, Daun02 had no effect (fig. S12, A and B). Likewise, Daun02 had no effect in wild-type animals off DOX (fig. S12, C and D); in TetTag mice continuously on DOX (fig. S12, E and F); and in TetTag mice in which recall-induced neurons from a context B, different from the conditioning context A, were inhibited (fig. S12, G and H).

Fig. 2 Loss-of-function of memory recall–induced neurons impairs remote fear attenuation.

(A) Schematic and (B) experimental design of the Daun02 inactivation method. (C) Compared with vehicle (VEH)–treated animals (n = 9), Daun02-treated animals (n = 9) show impaired fear memory attenuation (Δ freezing) in the DG. Data are means ± SEM compared by a two-tailed, unpaired t test (*P < 0.05). (D) Confocal images of the VEH- and Daun02-treated DG showing reduced c-Fos induction in Daun02-treated animals. Scale bars, 20 μm. (E) Quantification of (D). VEH-treated animals, n = 3; Daun02-treated animals, n = 6. Data are means ± SEM compared by a two-tailed, unpaired t test (**P < 0.01).

Lastly, we tested whether the continued activation of recall-induced neurons during the extinction paradigm facilitated remote fear memory attenuation. We used single transgenic c-Fos–tTA (tetracycline-controlled transactivator) mice and virally injected the excitatory DREADD (designer receptor exclusively activated by a designer drug) hM3Dq (AAV9-TRE::hM3Dq-mCherry) (Fig. 3A) to activate remote memory recall–induced ensembles by clozapine-N-oxide (CNO) administration. c-Fos–tTA animals were contextually fear-conditioned and stereotaxically injected with the viral constructs in the DG between training and recall, and hM3Dq expression was triggered by remote recall as described above (Fig. 3B). To stimulate the activity of the recall-induced ensemble throughout the fear attenuation procedure, mice were administered CNO intraperitoneally on each of the spaced extinction days (Fig. 3B). We found that in comparison with vehicle-treated animals and as corroborated by the same control groups described above (fig. S13), DG-injected CNO-treated mice displayed facilitated memory attenuation (Fig. 3C) despite showing similar freezing at recall (fig. S14). This facilitated fear memory attenuation was accompanied by an increased amount of extinction-induced c-Fos+ cells among recall-induced mCherry+ cells (Fig. 3, D and E).

Fig. 3 Gain-of-function of memory recall–induced neurons facilitates remote fear attenuation.

(A) Schematic and (B) experimental design of the hM3Dq-mediated activation method. i.p., intraperitoneally. (C) Compared with vehicle (VEH)-treated animals (n = 13), CNO-treated animals (n = 16) show facilitated fear memory attenuation (Δ freezing). Data are means ± SEM compared by a two-tailed, unpaired t test (*P < 0.05). (D) Confocal images of the VEH- and CNO-treated DG. The arrow indicates an example of a recall-induced neuron (mCherry) reactivated by extinction training (c-Fos) in CNO-treated animals. Scale bars, 10 μm. (E) Quantification of (D). VEH-treated animals, n = 4; CNO-treated animals, n = 5. Data are means ± SEM compared by a two-tailed, unpaired t test (***P < 0.001).

Taken together, our data show that a subset of memory recall–induced neurons in the DG becomes reactivated after memory attenuation, that the degree of fear reduction positively correlates with this reactivation, and that consequently, the continued activity of memory recall–induced neurons is critical for remote fear memory attenuation. Although other brain areas such as the prefrontal cortex (21) and the amygdala (22, 23) are likely to be implicated in remote fear memories and remain to be investigated, these results suggest that fear attenuation at least partially occurs in memory recall–induced ensembles through updating or unlearning of the original memory trace of fear. These data thereby provide the first evidence at an engram-specific level that fear attenuation may not be driven only by extinction learning, that is, by an inhibitory memory trace different from the original fear trace, as originally proposed by Pavlov (2) and experimentally supported for recent memories by others (3, 22, 24, 25). Rather, our findings indicate that during remote fear memory attenuation both mechanisms likely coexist (26, 27), albeit with the importance of the continued activity of memory recall–induced neurons experimentally documented herein. Such activity may not only represent the capacity for a valence change in DG engram cells (28) but also be a prerequisite for memory reconsolidation, namely, an opportunity for learning inside the original memory trace (29). As such, this activity likely constitutes a physiological correlate sine qua non for effective exposure therapies against traumatic memories in humans: the engagement, rather than the suppression, of the original trauma (30). In future studies, this experimental approach may prove useful to assess boundary conditions for extinction versus reconsolidation (5, 3133) or counterconditioning (34) in an engram-specific manner and, by extension, to gauge the efficiency of other interventions to reduce traumatic memories.

Supplementary Materials

www.sciencemag.org/content/360/6394/1239/suppl/DC1

Materials and Methods

Figs. S1 to S14

Reference (35)

References and Notes

Acknowledgments: We dedicate this work to O.K.’s father, Mohamed Salah El-Dien, and J.G.’s mother, Wilma, who both sadly passed away during its completion. We thank the vector core at UNC, Chapel Hill, for providing the adeno-associated virus; S. Hausmann, O. Prat, and G. Urueña for assistance in histology and quantification; and J. V. Sanchez-Mut for comments on earlier versions of the manuscript. Funding: Research on remote memories in J.G.’s laboratory is supported by the Swiss National Science Foundation (31003A_155898), the National Competence Center for Research “Synapsy,” and the European Research Council (ERC-2015-StG 678832). J.G. is an MQ fellow (MQ15FIP100012) and a NARSAD independent investigator (through independent investigator award 24497). Author contributions: This study was planned and conceptualized by O.K. and J.G. O.K. carried out the experiments and analyzed data. S.R., V.G., and L.D. contributed to surgeries, histology, catFISH experiments, and quantification. L.G. contributed to mouse colony maintenance and the Daun02 method. The paper was written by O.K. and J.G. and commented on by all authors. Competing interests: None. Data and materials availability: pAAV-TREtight-hM3Dq-mCherry is available from J.G. under a material agreement with Imperial College London. All data necessary to understand and assess the conclusions of this research are available in the supplementary materials.
View Abstract

Subjects

Navigate This Article