Abstract
Fear discrimination—the ability to distinguish threatening stimuli from non-threatening ones—is an essential adaptive behavior, yet the neural mechanisms underlying individual differences in discrimination precision remain unclear. Here, we show that parvalbumin-positive interneuron (PV-IN)-mediated feedforward inhibition to fear engram neurons (ENs) in the lateral amygdala is selectively enhanced following auditory fear conditioning, driven by increased postsynaptic expression of α1 subunit (Gabra1)-containing GABAA receptors. Disruption of PV-IN activity during fear retrieval—via optogenetic or chemogenetic inhibition—or selective induction of inhibitory long-term depression at PV → EN synapses impaired fear discrimination, increasing freezing to novel tones while sparing responses to the conditioned stimulus. Mechanistically, we identify inducible cAMP early repressor (ICER) as a transcriptional regulator that limits inhibitory plasticity by downregulating Gabra1 expression, thereby promoting fear generalization. These findings establish a previously unrecognized form of inhibitory synaptic plasticity that governs the precision of fear memory and suggest potential therapeutic targets for maladaptive fear overgeneralization in disorders such as post-traumatic stress disorder.
Similar content being viewed by others
Introduction
Fear enables animals to associate environmental cues with aversive experiences and respond to potential threats. In auditory fear conditioning (AFC), a neutral auditory conditioned stimulus (CS) is paired with an aversive stimulus such as a foot shock, leading to defensive responses, including freezing, upon later CS presentation [1,2]. While most individuals can discriminate the CS from dissimilar cues, some exhibit fear generalization to non-threatening novel stimuli [3,4,5]. This variability highlights the need to understand neural mechanisms that preserve fear memory precision.
Neural engrams are learning-activated neuronal ensembles that undergo enduring neuroplastic changes to encode memories [6,7,8]. In the lateral amygdala (LA), a critical site for fear memory formation, sensory inputs from thalamic and cortical pathways converge to associate environmental cues with aversive stimuli [9]. Although excitatory synaptic plasticity in LA engram neurons (ENs) is well established as a mechanism for fear memory formation and expression [10,11], the role of inhibitory synaptic plasticity in regulating EN activity remains unclear.
Emerging evidence suggests that inhibitory circuits and their plasticity are critical for sharpening memory specificity and preventing fear generalization [12,13,14]. In particular, feedforward inhibition (FFI), mediated by parvalbumin-positive interneurons (PV-INs), contributes to memory precision in the hippocampus and the amygdala [15,16], whereas disruption of inhibitory transmission or plasticity promotes generalized fear [17,18]. At the transcriptional level, cAMP response element-binding protein (CREB) regulates the expression of genes essential for synaptic plasticity and engram formation [19,20,21,22], whereas inducible cAMP early repressor (ICER) antagonizes CREB activity and can suppress genes involved in inhibitory plasticity, including Gabra1 that encodes the α1 subunit of GABAA receptors (GABAARs) [23,24,25].
In this study, we investigate how synaptic plasticity in AFC-activated ENs of the LA contributes to fear discrimination. Our study reveals that PV-IN-mediated FFI (PV-FFI) to ENs is postsynaptically strengthened after AFC to support fear discrimination, a process that is impaired in mice exhibiting generalized fear responses. Furthermore, we identify ICER as a key factor that disrupts inhibitory synaptic plasticity in ENs through suppression of Gabra1, thereby promoting fear generalization. Altogether, these findings provide mechanistic insights into fear memory precision and suggest potential therapeutic targets for maladaptive fear overgeneralization associated with post-traumatic stress disorder (PTSD).
Materials and methods
Detailed methods are provided in the Supplementary Material.
Animals
Male and female adult mice aged 3–5 months were used in this study. Intracranial viral injections were performed when mice were older than 10 weeks. Both sexes were included in all experiments unless otherwise specified. All procedures were approved by the POSTECH Institutional Animal Care and Use Committee (POSTECH-2025-0105).
AFC and fear discrimination tests
Mice underwent AFC consisting of four pairings of a 20 kHz tone with a foot shock. A fear discrimination test was subsequently performed in a modified context using the original CS and novel tones. Discrimination performance was quantified using a discrimination index (DI).
EN labeling
To label AFC-activated ENs, we used the cFos-tTA/TRE system. Doxycycline (DOX) was withdrawn for 3 days before AFC and reinstated at a high dose immediately after AFC.
Ex vivo electrophysiology
Acute coronal amygdala slices were prepared for whole-cell recordings from LA pyramidal neurons to measure thalamic-evoked excitatory postsynaptic currents (EPSCs) and inhibitory postsynaptic currents (IPSCs). PV-, somatostatin (SST)-, vasoactive intestinal peptide (VIP)-, or vesicular GABA transporter-positive (VGAT-) INs were optogenetically inhibited during afferent stimulation. Miniature EPSCs (mEPSCs) and IPSCs (mIPSCs) were recorded for 15 min.
Statistical analysis
Statistical analyses were performed using GraphPad Prism 10. Data are presented as mean ± SEM, with sample sizes and statistical tests reported in each figure legend.
Results
Postsynaptic strengthening of PV-FFI to ENs
To examine synaptic changes in fear training-induced ENs, we labeled AFC-activated neurons using cFos-tTA mice injected with AAV-TRE-mCherry during a DOX-off period (Fig. 1A, B). The labeled neuronal population showed preferential reactivation upon re-exposure to the CS compared with exposure to a novel tone, suggesting that these cells were enriched for ENs participating in fear memory encoding (Fig. S1A–H). Whole-cell recordings were made from mCherry-positive ENs and mCherry-negative pyramidal neurons (non-ENs) (Fig. 1C). We first measured miniature synaptic events to obtain an initial assessment of overall excitatory and inhibitory synaptic properties in ENs compared with neighboring non-ENs. Consistent with previous reports [11,26], mEPSC event rates were higher in ENs than in non-ENs (Fig. 1D, F, and G). Notably, mIPSC amplitudes, but not event rates, were also larger in ENs (Fig. 1E, H, and I), indicating that fear learning enhances both excitatory and inhibitory synaptic transmission in ENs [27]. Electrical stimulation of thalamic axon fibers in the internal capsule evoked short-latency monosynaptic EPSCs and longer-latency disynaptic IPSCs (Fig. 1J and Fig. S1I, M), as previously demonstrated [28]. Across multiple stimulation intensities, both EPSCs and IPSCs were larger in ENs than in non-ENs (Fig. 1K, L). Accordingly, inhibitory/excitatory transmission ratios (I/E) remained comparable between ENs and non-ENs (Fig. 1M, N). Thus, increased FFI, reflected by larger disynaptic IPSCs, appears to offset the heightened excitatory transmission to ENs after AFC, thereby maintaining synaptic balance.
A Schematic of the AAV injection strategy and experimental timeline. B Representative image showing mCherry-expressing ENs (red) and DAPI-stained cell nuclei (blue) in the LA, outlined by dashed lines. Scale bar, 200 μm. C Schematic of mEPSC and mIPSC recordings from ENs and non-ENs. D, E Representative traces of mEPSCs D and mIPSCs E in ENs (top, red) and non-ENs (bottom, gray). F, G Mean mEPSC amplitudes F and event rates G in ENs (n = 15 cells from N = 3 mice) and non-ENs (n = 13 cells from N = 3 mice). H, I Mean mIPSC amplitudes H and event rates I in ENs (n = 19 cells from N = 5 mice) and non-ENs (n = 22 cells from N = 5 mice). Each dot represents the mean value obtained from an individual recorded cell. Data are analyzed using the Mann-Whitney U test, **p < 0.01; *p < 0.05. J Schematic of EPSC and IPSC recordings from ENs and non-ENs evoked by thalamic stimulation (left), with representative traces (right). K, L Mean EPSC K and IPSC L amplitudes in ENs (n = 38 cells from N = 10 mice) and non-ENs (n = 35 cells from N = 10 mice) across increasing stimulation intensities of 25–100 µA. Data are analyzed using a mixed-effects model followed by the two-stage Benjamini-Krieger-Yekutieli (BKY) procedure for multiple comparisons, *p < 0.05; **p < 0.01. M Scatter plot of EPSC and IPSC amplitudes in ENs (n = 28 cells from N = 7 mice) and non-ENs (n = 24 cells from N = 7 mice), with linear regression lines and 95% confidence intervals (shaded). Each dot represents an individual recorded cell. N I/E ratios in ENs (n = 28 cells from N = 7 mice) and non-ENs (n = 24 cells from N = 7 mice). Each dot represents an individual recorded cell. Data are analyzed using Mann-Whitney U test, p > 0.05. Data are presented as mean ± SEM. ns not significant.
To identify which type of inhibitory neurons could primarily mediate FFI in the LA, we examined three major GABAergic IN populations: PV-, SST-, and VIP-INs (Fig. S2A). We optogenetically suppressed each halorhodopsin (NpHR3.0)-expressing IN subtype during thalamic stimulation and quantified its contribution to FFI as the reduction in IPSC amplitude (ΔIPSC; Fig. S2B, C). PV-INs contributed the largest fraction of FFI to LA pyramidal neurons (49.3 ± 6.2%), followed by SST-INs (19.7 ± 6.7%) and VIP-INs (1.7 ± 3.4%). As a reference, inhibition of VGAT+ inhibitory neurons reduced IPSCs by approximately 78%, with PV-INs again constituting the largest proportion (Fig. S2D, E). However, this analysis was made in naïve mice without fear training. We next generated two double-transgenic mice, PV-Cre/cFos-tTA and SST-Cre/cFos-tTA, to examine whether PV- or SST-INs accounted for the enhanced FFI to ENs after AFC. We injected AAV-DIO-NpHR3.0 into the LA of these mice and then compared ΔIPSCs between ENs and non-ENs (Fig. 2A, B). PV-FFI was significantly larger in ENs than in non-ENs (Fig. 2C, D), whereas SST-FFI did not differ between the two populations (Fig. 2F, G). Importantly, optogenetic inhibition of PV-INs equalized IPSC amplitudes between ENs and non-ENs (Fig. 2E), indicating a major role for PV-INs in enhancing FFI to ENs after AFC.
A Schematic of the AAV injection strategy and experimental timeline. B Schematic of thalamic stimulation-evoked IPSC recordings from ENs and non-ENs during photoinhibition of PV-INs or SST-INs. C, D Representative IPSC traces before and during PV-IN photoinhibition C and changes in IPSC amplitude following PV-IN inhibition (ΔIPSC; D) in ENs (n = 16 cells from N = 4 mice), non-ENs (n = 13 cells from N = 4 mice), and LA pyramidal neurons from naïve controls (n = 15 cells from N = 4 mice). Each dot in D represents an individual recorded cell. Data are analyzed using one-way ANOVA followed by Tukey’s post hoc test, **p < 0.01. E Residual IPSC amplitudes during PV-IN photoinhibition in ENs (n = 16 cells from N = 4 mice) and non-ENs (n = 13 cells from N = 4 mice). Each dot represents an individual recorded cell. Data are analyzed using Mann-Whitney U test. F, G Representative IPSC traces before and during SST-IN photoinhibition F and changes in IPSC amplitude following SST-IN inhibition (ΔIPSC; G) in ENs (n = 11 cells from N = 3 mice), non-ENs (n = 8 cells from N = 3 mice), and LA pyramidal neurons from naïve controls (n = 13 cells from N = 3 mice). Each dot in G represents an individual recorded cell. Data are analyzed using one-way ANOVA followed by Tukey’s post hoc test. H Schematic of the AAV injection strategy and experimental timeline. I Schematic of optogenetically evoked PV-IPSC recordings. J, K Representative PV-IPSC traces J and paired-pulse ratios (PPRs; K) in ENs (n = 14 cells from N = 5 mice) and non-ENs (n = 16 cells from N = 5 mice) at inter-stimulus intervals of 50, 100, and 150 ms. Data are analyzed using two-way ANOVA followed by Šídák’s post hoc test. L–N Representative traces of asynchronous quantal PV-IPSCs (qIPSCs), with the shaded region indicating the 50–450-ms analysis window L, and mean qIPSC amplitudes M and frequencies N in ENs (n = 20 cells from N = 5 mice) and non-ENs (n = 21 cells from N = 5 mice). Each dot in M, N represents an individual recorded cell. Data are analyzed using Mann-Whitney U test, *p < 0.05. O Representative immunofluorescence images showing NeuN (blue), GABAAR α1 (green), an mCherry-labeled EN (red), and merged channels in the LA of cFos-tTA mice injected with AAV-TRE-mCherry. Mice underwent AFC during the DOX-off period and were sacrificed one day after AFC. Dashed outlines indicate the perisomatic regions of an EN and a non-EN, respectively. NeuN+/GABAAR α1+ double-positive puncta are highlighted in pseudocolor (orange). Scale bar, 5 μm. P Quantification of mean perisomatic GABAAR α1+ punctum size (left) and density (right) in ENs (n = 16 cells from N = 3 mice) and non-ENs (n = 18 cells from N = 3 mice). Each dot represents an individual analyzed cell. Data are analyzed using Mann-Whitney U test, ***p < 0.001, **p < 0.01. Data are presented as mean ± SEM.
We sought to examine the nature of synaptic enhancement from PV-INs to ENs by expressing channelrhodopsin-2 (ChR2) in PV-INs (Fig. 2H, I). Paired-pulse ratios (PPRs) of PV-IN-mediated IPSCs (PV-IPSCs) were similar between ENs and non-ENs, arguing against presynaptic changes (Fig. 2J, K). We also performed quantal analysis by replacing calcium with strontium in artificial cerebrospinal fluid [29], which induces asynchronous GABA release after brief photostimulation (Fig. 2L). Quantal PV-IPSC (qIPSC) amplitudes were larger in ENs than in non-ENs, while the frequencies were comparable (Fig. 2M, N). Thus, enhanced PV-FFI to ENs arises most likely through postsynaptic mechanisms at EN synapses innervated by PV-INs.
GABAAR α1-mediated enhancement of inhibitory transmission from PV-INs to ENs
The postsynaptic strengthening of inhibitory transmission observed in our physiological analyses might reflect increased recruitment of GABA receptors to the postsynaptic sites of ENs. To determine whether and which GABAAR subunits are primarily recruited to inhibitory synapses of ENs after AFC, we re-analyzed previously published single-cell RNA sequencing (scRNA-seq) data from amygdala tissue collected at 0, 2, 8, 24, and 28 h after AFC in the same mouse strain used in the present study [30]. As previously reported [31,32], we identified a putative LA pyramidal neuronal population as a gastrin-releasing peptide-positive (Grp+) neuronal cluster that also expressed copine 8 (Cpne8), consistent with the anatomical localization of Grp and Cpne8 mRNAs within the LA (Fig. S3A). This Grp+ cluster localized within the vesicular glutamate transporters-expressing (VGLUT+) neuronal population in the scRNA-seq dataset (Fig. S4A–C). Optogenetic activation of AFC-tagged Grp+ LA neurons was sufficient to induce freezing in a novel context, supporting their involvement in fear memory encoding (Fig. S3B, C). We then analyzed the expression of GABAAR subunits in Grp+ neurons and found that Gabra1 expression was significantly upregulated at 2, 8, and 24 h after AFC (Fig. S4D). We further classified Grp+ neurons into immediate early gene (IEG)-high ( ≥ 0.5) and IEG-low ( < 0.5) populations based on IEG module scores [30,33] (Fig. S4E, F). Importantly, IEG-high Grp+ neurons, representing putative ENs, exhibited significantly higher Gabra1 expression than IEG-low Grp+ neurons, while other GABAAR subunits did not differ between the two groups (Fig. S4G). This suggests that Gabra1 upregulation in ENs would underlie the observed postsynaptic enhancement of PV-FFI, consistent with previous evidence that Gabra1 regulates LA inhibitory transmission and fine-tunes fear memory [34].
In accordance with the transcriptomic findings, we detected a higher abundance of GABAAR α1 in ENs compared with non-ENs (Fig. 2O, P). Perisomatic α1-positive (α1+) puncta apposed to VGAT were significantly larger and denser in ENs than in non-ENs (Fig. S5A–D). Immunogold electron microscopy further confirmed a higher density of α1+ gold particles in the perisomatic synaptic regions of ENs than in non-ENs (Fig. S5E, F). In contrast, staining for synaptotagmin-2 (Syt2), a presynaptic marker of PV-IN terminals [35,36], revealed no difference between ENs and non-ENs (Fig. S5G, H). Altogether, these results indicate that AFC enhances PV-FFI to ENs primarily through postsynaptic mechanisms involving upregulation of Gabra1 in ENs.
Increased PV-FFI required for normal fear discrimination
Amygdala INs are known to regulate fear memory and discrimination [12,17,37,38]. However, how they can control LA principal neurons, including ENs, to support precise memory expression remains unclear. Given the AFC-induced strengthening of FFI to ENs, we examined the roles of PV-, SST-, or cholecystokinin-expressing (CCK-) INs in fear discrimination. To this end, we chemogenetically inhibited each IN type during fear retrieval (Fig. S6). PV-Cre and SST-Cre mice were injected with Cre-dependent hM4Di or control viruses, while CCK-Cre mice received Dlx promoter-driven, Cre-dependent AAVs to selectively target CCK-INs (Fig. S6A). Clozapine-N-oxide (CNO, 5 mg/kg, i.p.) was administered 30 min before the discrimination test using the 20 kHz CS and a novel 3 kHz tone (Fig. S6A). Inhibition of PV-INs impaired fear discrimination by increasing freezing to the novel tone without affecting responses to the CS, thereby reducing the DI (Fig. S6B), whereas inhibition of SST- or CCK-INs had no significant effect (Fig. S6C, D).
PV-INs appear to play a major role in suppressing nonspecific freezing to novel stimuli. To achieve temporal control of PV-IN activity, we next optogenetically inhibited PV-INs by expressing archaerhodopsin (Arch) and assessed freezing responses across a range of tone frequencies (Fig. 3A–E). Auditory tones were presented serially at 2, 8, 14, and 20 kHz, with progressively increasing frequency proximity to the 20 kHz CS, while PV-INs were optogenetically inhibited during tone presentation. Laser-ON and laser-OFF conditions were counterbalanced across trials to minimize order bias in the timing of optical inhibition (Fig. 3A). This experimental design allowed us to characterize the frequency-dependent freezing profile relative to the CS [39,40] and to assess the frequency range over which PV-IN inhibition affects fear responses. PV-IN suppression selectively increased freezing to the 2 kHz tone, which elicited minimal freezing under control conditions. This effect was time-locked to light illumination (Fig. 3B, C) and was absent in EGFP-expressing controls (Fig. 3D). Accordingly, PV-IN suppression significantly reduced the DI between the 2 and 20 kHz tones (Fig. 3E).
A Schematic of the AAV injection strategy and experimental timeline for freezing-spectrum tests with optogenetic inhibition of PV-INs. B Representative image showing Arch-EGFP expression (green) and DAPI staining (blue) in the LA. Scale bar, 200 μm. C, D Freezing responses to 2, 8, 14, and 20 kHz tones under laser-OFF (only tone) and laser-ON (tone + 532 nm laser) conditions in PV-Arch C and PV-EGFP D mice (N = 13 mice per group). Data are analyzed using two-way ANOVA followed by Šídák’s post hoc test, **p < 0.01. E DI between the 2 and 20 kHz tones under laser-OFF (only tone) and laser-ON (tone + 532 nm laser) conditions in PV-Arch and PV-EGFP mice. Each dot represents an individual mouse. Data are analyzed using two-way ANOVA followed by Bonferroni’s post hoc test, ***p < 0.001. F Schematic of the AAV injection strategy and experimental timeline for PV-iLTD induction and fear discrimination testing. G, H Freezing responses to the 2 and 20 kHz tones before and after PV-iLTD induction in hChR2 (G; N = 14 mice) and EYFP (H; N = 10 mice) groups. Data are analyzed using two-way ANOVA followed by Šídák’s post hoc test, ***p < 0.001. I DI between the 2 and 20 kHz tones before and after PV-iLTD induction in hChR2 and EYFP groups. Each dot represents an individual mouse. Data are analyzed using two-way ANOVA followed by Šídák’s post hoc test, ***p < 0.001. Data are presented as mean ± SEM.
While we have thus far relied on global manipulation of PV-IN activity, we next tested whether the AFC-induced enhancement of PV-FFI specifically to ENs is required for fear discrimination. To selectively reduce PV → EN inhibition, we developed an optogenetic stimulation protocol that induced PV inhibitory long-term depression (PV-iLTD), significantly reducing PV-IPSCs in ENs while having negligible effects in non-ENs (Fig. S7). To test the behavioral consequences of this EN-selective PV-iLTD, we expressed ChR2 in PV-INs of PV-Cre mice and delivered the PV-iLTD protocol via an optic cannula implanted above the LA. Following AFC, mice underwent two fear recall tests—before and after PV-iLTD induction (Fig. 3F). Before PV-iLTD, freezing to the novel 2 kHz tone was minimal. After PV-iLTD induction, freezing to the 2 kHz tone increased, while responses to the CS remained unaffected, resulting in a reduced DI (Fig. 3G, I). No such changes were observed in EYFP controls (Fig. 3H, I), and PV-iLTD did not alter anxiety-related behaviors or locomotion (Fig. S8). Collectively, these results indicate that inhibitory strengthening occurring at PV-IN → EN synapses plays a permissive role in precluding inappropriate freezing to irrelevant stimuli and thereby supports the precision of fear memory.
Aberrant reactivation of ENs prevented by PV-IN-mediated inhibition
Since reactivation of ENs drives memory recall [41,42], enhanced PV-FFI to ENs may prevent their inappropriate reactivation by irrelevant stimuli. To test this notion, we labeled ENs with EYFP, expressed hM4Di in PV-INs, and then analyzed c-Fos expression in the LA of PV-Cre/cFos-tTA mice (Fig. 4A). Following administration of JHU37160 (J60), an hM4Di agonist, 20 kHz-conditioned mice were re-exposed to either the 20 kHz CS or a novel 2 kHz tone in a distinct context. Chemogenetic suppression of PV-INs during 2 kHz tone presentation increased both the proportion of c-Fos/EYFP double-labeled neurons and freezing responses compared with vehicle controls (Fig. 4B, D, and E), suggesting that ENs could be aberrantly reactivated by novel stimuli when PV-IN-mediated inhibition is released. However, PV-IN suppression during CS presentation did not alter c-Fos expression in ENs or freezing behavior (Fig. 4C, F, and G). The total number of c-Fos+ LA neurons (Fig. 4H, J) and of EYFP+ ENs (Fig. 4I, K) remained unchanged despite PV-IN inhibition, arguing against expansion of engram size as a mechanism for impaired fear discrimination [43,44], at least in the adult mice we tested. Chemogenetic inhibition of PV-INs did not affect anxiety-like behavior, locomotion, or innate fear in either naïve or fear-conditioned mice (Fig. S9), indicating that the observed increase in freezing to novel tones was not due to a general elevation of LA pyramidal neuron activity. Therefore, PV-INs can prevent nonspecific reactivation of ENs by irrelevant stimuli, thereby maintaining fear memory precision.
A Schematic of the AAV injection strategy and experimental timeline. B, C Representative images showing c-Fos expression (red) and EYFP-labeled ENs (green) following exposure to the 2 kHz novel tone B or the 20 kHz CS C in J60-treated (left) and vehicle-treated (right) mice one day after AFC with 20 kHz CS. White arrowheads indicate c-Fos+/EYFP+ double-labeled neurons. Scale bar, 100 μm. D Freezing responses to the 2 kHz tone in J60-treated (N = 7 mice) and vehicle-treated (N = 6 mice) groups. E Proportion of reactivated ENs, defined as c-Fos+/EYFP+ double-labeled neurons among EYFP+ ENs, following exposure to the 2 kHz tone in J60-treated and vehicle-treated groups. F Freezing responses to the 20 kHz CS in J60-treated and vehicle-treated groups (N = 6 mice per group). G Proportion of reactivated ENs following exposure to the 20 kHz CS in J60-treated and vehicle-treated groups. H Number of c-Fos+ neurons per 0.1 mm2 following exposure to the 2 kHz tone in J60-treated and vehicle-treated groups. I Number of EYFP+ ENs per 0.1 mm2 in mice exposed to the 2 kHz tone and treated with J60 or vehicle. J Number of c-Fos+ neurons per 0.1 mm2 following exposure to the 20 kHz CS in J60-treated and vehicle-treated groups. K Number of EYFP+ ENs per 0.1 mm2 in mice exposed to the 20 kHz CS and treated with J60 or vehicle. Each dot represents the mean value obtained from an individual mouse. Data are analyzed using Mann-Whitney U test, *p < 0.05. L Schematic of the AAV injection strategy and experimental timeline. M Representative image showing hChR2-EYFP expression (green) in the MGN and ACx. Scale bar, 1 mm. N Representative image showing hChR2-EYFP-expressing axonal projections (green) and mCherry-labeled ENs (red) in the LA. Scale bar, 100 μm. O Representative EPSC and IPSC traces (left) recorded from ENs and non-ENs following optogenetic stimulation of 20 kHz tone-tagged inputs, and quantification of EPSC (middle) and IPSC (right) amplitudes in ENs (n = 40 cells from N = 10 mice) and non-ENs (n = 39 cells from N = 10 mice). Each dot represents an individual recorded cell. Data are analyzed using Mann-Whitney U test, *p < 0.05; **p < 0.01. P I/E ratios (left) and scatter plot of EPSC and IPSC amplitudes (right) in ENs (red) and non-ENs (gray) following photostimulation of 20 kHz tone-tagged inputs. Solid lines indicate linear regression fits, and shaded areas indicate 95% confidence intervals. I/E ratios are analyzed using Mann-Whitney U test. Q Representative EPSC and IPSC traces (left) recorded from ENs and non-ENs following optogenetic stimulation of 2 kHz tone-tagged inputs, and quantification of EPSC (middle) and IPSC (right) amplitudes in ENs (n = 31 cells from N = 8 mice) and non-ENs (n = 32 cells from N = 8 mice). Each dot represents an individual recorded cell. Data are analyzed using Mann-Whitney U test, **p < 0.01. R I/E ratios (left) and scatter plot of EPSC and IPSC amplitudes (right) in ENs (red) and non-ENs (gray) following photostimulation of 2 kHz tone-tagged inputs. Solid lines indicate linear regression fits, and shaded areas indicate 95% confidence intervals. I/E ratios are analyzed using Mann-Whitney U test, **p < 0.01. Data are presented as mean ± SEM.
We next examined input-specific synaptic transmission in LA pyramidal neurons using cFos-CreERT2 mice [45,46]. We first validated the specificity and functional relevance of tone-responsive neuronal tagging in the medial geniculate nucleus (MGN) and auditory cortex (ACx) (Fig. S10). Photostimulation of tone-tagged MGN or ACx terminals in the LA evoked monosynaptic EPSCs and disynaptic IPSCs (Fig. S11A–F and S12B, K), and suppression of PV-INs reduced these input-evoked IPSCs by more than 50% (Fig. S12I, R), confirming that PV-INs are a major source of FFI, consistent with our earlier results (Fig. S2D, E). To determine how tone-specific inputs are differentially transmitted to ENs and non-ENs, we generated double-transgenic cFos-tTA/cFos-CreERT2 mice, enabling the labeling of both tone-specific inputs from MGN and ACx and AFC-induced ENs in the LA (Fig. 4L–N). Stimulation of 20 kHz CS-specific inputs evoked larger EPSCs and IPSCs in ENs than in non-ENs, resulting in comparable I/E ratios between the two populations (Fig. 4O, P). However, stimulation of novel 2 kHz-specific inputs evoked larger IPSCs in ENs but similar EPSCs between ENs and non-ENs, producing a higher I/E ratio in ENs (Fig. 4Q, R). Thus, CS-specific inputs proportionally enhance excitation and inhibition to ENs, whereas novel tone-specific inputs are biased toward inhibition. This inhibitory bias of novel tone-evoked synaptic responses in ENs prevents their aberrant activation by irrelevant stimuli under normal conditions.
ICER-mediated regulation of inhibitory drive in fear generalization
Given individual differences in fear discrimination [32,40], we classified mice as discriminators or generalizers using K-means clustering based on three behavioral parameters (Fig. S13A, E, I, and M). Across all timepoints after AFC, approximately 70% of mice were classified as discriminators and 30% as generalizers. The two-cluster solution consistently yielded the highest mean silhouette score, and bootstrap resampling showed that individual cluster assignments were highly stable, supporting the robustness of the discriminator–generalizer classification (Fig. S13B, F, J, and N). The proportions of discriminators and generalizers were comparable between males and females at each timepoint, with no significant association between sex and behavioral classification (Fig. S13C, G, K, and O). Interestingly, generalizers exhibited reduced size and density of perisomatic GABAAR α1+ puncta in ENs (Fig. S14A–E). A random forest classifier [47] using α1+ punctum size and density modestly distinguished discriminators from generalizers, with an area under curve (AUC) of 0.65 (Fig. S14F). Moreover, EN-specific inhibitory potentiation, characterized by enhanced PV-FFI, was observed in discriminators but was absent in generalizers (Fig. S14G–L). Together, these findings indicate that inhibitory features of ENs, particularly postsynaptic GABAAR α1, are associated with and partially predictive of behavioral classification during fear discrimination.
We next explored the molecular mechanisms underlying these individual differences in discriminative capability. Given that CREB regulates neuronal excitability, memory allocation, and long-term memory-related transcription [20,48], we first compared Creb1 mRNA levels in LA tissue between discriminators and generalizers at 2, 6, 12, and 24 h after AFC but detected no significant differences between the two groups (Fig. 5A, B). Since CREB itself was unlikely to account for differences in fear discrimination, we next examined other members of the CREM (cAMP response element modulator) family, which includes both transcriptional activators and repressors, such as ICER, that modulate CREB-dependent transcription [49,50,51]. As ICER has been shown to suppress CREB-mediated Gabra1 expression [23,52], we quantified ICER mRNA using specific probes [53]. Generalizers showed a transient increase in ICER mRNA levels that peaked 2 h after AFC and declined thereafter, whereas discriminators maintained low ICER levels (Fig. 5B). ICER protein levels in both ENs and non-ENs peaked later, at 6 h after AFC, in generalizers but remained low in discriminators (Fig. S15C, D).
A Experimental timeline and analysis design. B Creb1 and ICER mRNA expression in the LA of discriminators and generalizers at 0, 2, 6, 12, or 24 h after AFC. C57BL/6 J mice were used (Creb1: N = 3 mice per group at each timepoint; ICER: N = 5–6 mice per group at each timepoint). Data are analyzed using mixed-effects model followed by Fisher’s post hoc test, *p < 0.05. C Schematic of the AAV injection strategy and experimental timeline for EN-specific overexpression of ICER or control mCherry. D Freezing levels during AFC in mice overexpressing ICER (ICER-OE; red, N = 16 mice) or mCherry (mCherry-OE; black, N = 14 mice) in ENs. E Freezing levels during baseline and presentations of the 2 kHz novel tone and 20 kHz CS in ICER-OE and mCherry-OE mice. Each dot represents an individual mouse. Data are analyzed using two-way ANOVA followed by Šídák’s post hoc test, *p < 0.05. F DI between the 2 and 20 kHz tones in ICER-OE and mCherry-OE mice. Data are analyzed using Mann-Whitney U test, *p < 0.05. G Schematic of thalamic stimulation-evoked EPSC and IPSC recordings from ENs and non-ENs (left), with representative EPSC and IPSC traces recorded from ENs overexpressing either mCherry (middle) or ICER (right) and from corresponding non-ENs. H EPSC amplitudes (left), IPSC amplitudes (middle), and I/E ratios (right) in ENs (n = 13 cells from N = 4 mice) and non-ENs (n = 10 cells from N = 4 mice) from the mCherry-OE group. Each dot represents an individual recorded cell. Data are analyzed using Mann-Whitney U test, *p < 0.05. I EPSC amplitudes (left), IPSC amplitudes (middle), and I/E ratios (right) in ENs and non-ENs (n = 10 cells from N = 4 mice per cell type) from the ICER-OE group. Each dot represents an individual recorded cell. Data are analyzed using Mann-Whitney U test, *p < 0.05. J Representative images showing DAPI (blue), GABAAR α1 (green), and ENs expressing either mCherry (left) or ICER-mCherry (right) (red) in the mCherry-OE and ICER-OE groups, respectively. Insets show enlarged perisomatic regions of a mCherry-OE EN (1, 1′) and an ICER-OE EN (2, 2′); mCherry+/GABAAR α1+ double-positive puncta are highlighted in pseudocolor orange (1′, 2′). Scale bars, 20 μm (main images) and 5 μm (insets). K, L Quantification of mean perisomatic GABAAR α1+ punctum size K and density L in mCherry-OE (n = 30 cells from N = 5 mice) and ICER-OE ENs (n = 31 cells from N = 5 mice). Each dot represents an individual analyzed cell. Data are analyzed using Mann-Whitney U test, ***p < 0.001. M Schematic of the AAV injection strategy and experimental timeline. N Schematic of thalamic stimulation-evoked IPSC recordings from ENs and non-ENs during optogenetic inhibition of PV-INs (left), with representative IPSC traces recorded before and during PV-IN photoinhibition from ENs overexpressing either mCherry (middle) or ICER (right) and from corresponding non-ENs. O Relative contribution of PV-FFI to total FFI, quantified as ΔIPSC/IPSC (%), in ENs from mCherry-OE mice (n = 11 cells from N = 3 mice) and ICER-OE mice (n = 9 cells from N = 3 mice), and in corresponding non-ENs from mCherry-OE mice (n = 10 cells from N = 3 mice) and ICER-OE mice (n = 8 cells from N = 3 mice). Each dot represents an individual recorded cell. Data are analyzed using mixed-effects model followed by Fisher’s post hoc test, *p < 0.05; **p < 0.01; ***p < 0.001. Data are presented as mean ± SEM.
To test whether elevated ICER in ENs affects their inhibitory synaptic organization, we overexpressed ICER in ENs using Grp-Cre/cFos-tTA double-transgenic mice (Fig. 5C and Fig. S16A, B). EN-specific ICER overexpression increased freezing to the novel 2 kHz tone without affecting freezing to the 20 kHz CS, thereby reducing the DI (Fig. 5D–F). While mCherry-expressing ENs exhibited larger EPSCs and IPSCs than non-ENs, with comparable I/E ratios (Fig. 5G, H), ICER-overexpressing ENs showed larger EPSCs but IPSCs comparable to those of non-ENs, resulting in lower I/E ratios (Fig. 5I), similar to the pattern observed in generalizers (Fig. S14J). However, nonspecific ICER overexpression in the LA did not impair discrimination (Fig. S16C–H), indicating that elevated ICER in ENs, but not in non-ENs, disrupts fear discrimination, most likely by compromising PV-FFI potentiation in ENs.
We found that ICER overexpression reduced GABAAR α1+ punctum size and density in ENs compared with mCherry controls (Fig. 5J–L). Importantly, PV-FFI, quantified as ΔIPSC/IPSC (%), was enhanced in mCherry-expressing ENs but not in ICER-overexpressing ENs, as assessed using AAV-CreOFF/FlpON-mediated expression of eNpHR3.3 [54] in PV-INs of cFos-tTA/Grp-Cre/PV-Flp triple-transgenic mice (Fig. 5M–O). To causally test whether reduced GABAAR α1 mediates ICER-induced impairment of fear memory precision, we overexpressed Gabra1 in ENs that also overexpressed ICER and assessed fear discrimination between the 2 and 20 kHz tones (Fig. S17A). Restoring GABAAR α1 levels in ICER-overexpressing ENs significantly reduced freezing to the novel 2 kHz tone and improved the DI compared with the ICER overexpression-only group (Fig. S17B–G). These results suggest that suppression of GABAAR α1 at EN synapses is a major mechanism through which ICER engenders fear overgeneralization. Conversely, ICER depletion reduced freezing to a novel 8 kHz tone and improved discrimination between the 8 and 20 kHz tones (Fig. S18A–D). Together, these findings establish ICER as a molecular regulator of inhibitory plasticity in ENs that bidirectionally controls fear discrimination and generalization.
Discussion
Our study provides circuit and molecular mechanisms by which fear memory precision is regulated through inhibitory control of EN activity in the LA. First, PV-FFI is selectively enhanced to ENs after AFC. Second, inhibitory plasticity at PV–EN synapses is induced postsynaptically via upregulation of α1-containing GABAARs, which supports fear discrimination. Third, elevated ICER expression in ENs prevents this postsynaptic inhibitory plasticity, resulting in impaired PV-FFI and fear generalization.
It is well established that enhanced excitatory inputs to ENs and their reactivation are essential for fear memory retrieval [55,56]. However, the role of inhibitory inputs—particularly to ENs—has been less explored, despite proposed roles of strengthened inhibitory synapses in maintaining I/E balance [27,57,58]. FFI can control the timing and specificity of amygdala neuron responses to sensory inputs and thereby contribute to stimulus discrimination [59,60,61]. Consistent with the importance of inhibitory regulation, reduced GABAergic inhibition in the amygdala has been associated with PTSD phenotypes in both animal models and human patients [5,62]. Transcriptomic analyses of PTSD patients have also revealed significant decreases in the expression of GABA-related neuropeptides, including PV, SST, and VIP [63]. However, it remains largely unknown which type(s) of GABAergic interneurons can primarily mediate FFI to prevent the aberrant activation of ENs in the LA and how they may contribute to fear discrimination.
PV-INs are a major source of FFI across various brain regions [16,64]. Contextual fear conditioning has been shown to enhance excitatory inputs to fast-spiking interneurons, presumably PV-INs, thereby potentiating FFI in hippocampal CA3 and promoting memory precision [15,65]. We initially hypothesized that PV-FFI to ENs would be preferentially engaged during CS presentation, given that PV-INs are activated by CS cues [12]. However, PV-IN suppression did not affect CS-induced EN reactivation or fear retrieval, but instead increased EN reactivation and freezing in response to novel tones (Figs. 3 and 4). This may reflect a bias toward inhibition over excitation in ENs during novel-tone processing (Fig. 4L–R). How PV-INs selectively modulate EN activity remains an important question, particularly given that elevated intracellular Ca2+ levels can promote clustering and expression of GABAA receptors [66,67]. Such studies may clarify how PV-FFI constrains nonspecific fear memory expression while preserving CS-induced fear responses.
Our results further suggest that the enhanced PV-FFI observed in ENs is primarily mediated by postsynaptic, rather than presynaptic, mechanisms. Although previous studies have linked PV-IN maturation during adolescence [44] or stress-induced changes in PV-IN activity [68] to memory precision, we found no clear evidence for presynaptic changes in PV-INs underlying enhanced PV-FFI. Instead, in the adult mice ( > 13 weeks of age) used in this study, AFC increased postsynaptic α1-containing GABAA receptor expression in ENs (Fig. 2). Such postsynaptic remodeling may represent a homeostatic adaptation that maintains I/E balance [69,70,71] as excitatory inputs to ENs are strengthened during memory formation. By adjusting receptor composition and localization [72,73] specifically in activated ENs, postsynaptic inhibitory plasticity would provide a mechanism for precise, cell-type-specific control of fear memory expression.
We identified ICER as a key molecular regulator underlying the loss of PV-FFI enhancement to ENs, ultimately leading to generalized fear (Fig. 5). ICER represses CREB-dependent transcription by binding CRE sequences without a transactivation domain [25]. CREB and ICER can bidirectionally regulate Gabra1 expression: CREB homodimers promote Gabra1 transcription, whereas ICER-CREB heterodimers or ICER homodimers suppress it, reducing surface expression of α1-containing GABAARs [23,74]. Our results from ICER overexpression and depletion corroborate that ICER-mediated suppression of Gabra1 can deter the postsynaptic maturation of inhibitory synapses from PV-INs to ENs. Although ICER was also elevated in non-ENs of generalizers (Fig. S15D), its functional role in non-ENs remains to be determined. Given that efficient CREB activity depends on intracellular Ca2+ [75], ICER may have more limited effects on Gabra1 regulation in non-ENs, which presumably have lower Ca2+ levels, although this possibility requires further investigation.
Several considerations should be taken into account when interpreting our findings. First, activity-dependent cFos-based labeling systems, including cFos-tTA and cFos-CreERT2, have limited temporal precision due to the kinetics of pharmacological gating, potentially tagging neurons outside the intended behavioral windows [6,45,76,77]. However, CS-selective reactivation of labeled LA neurons and auditory populations supports their behavioral relevance (Fig. S1A–H and S10). Second, our study was confined to mice and primarily focused on amygdala mechanisms, and thus does not fully capture the heterogeneity or distributed circuit complexity of PTSD [78,79]. Rather, we identify an ICER-dependent inhibitory plasticity mechanism relevant to fear overgeneralization. Despite these caveats, the convergence of multiple independent approaches—including transgenic labeling, chemogenetic manipulation, optogenetic reactivation, and electrophysiological characterization—provides complementary evidence that ICER-controlled inhibitory plasticity contributes causally to fear memory precision.
Data availability
Data are available upon reasonable request to the corresponding author.
References
LeDoux JE. Emotion Circuits in the Brain. Annu Rev Neurosci. 2000;23:155–84.
Blundell P, Hall G, Killcross S. Preserved Sensitivity to Outcome Value after Lesions of the Basolateral Amygdala. J Neurosci. 2003;23:7702–9.
Laxmi TR, Stork O, Pape H-C. Generalisation of conditioned fear and its behavioural expression in mice. Behav Brain Res. 2003;145:89–98.
Ghosh S, Chattarji S. Neuronal encoding of the switch from specific to generalized fear. Nat Neurosci. 2015;18:112–20.
Andero R. Stress-induced changes in the molecular processes underlying fear memories: implications for PTSD and relevant animal models. Mol Psychiatry. 2025;30:2219–27.
Reijmers LG, Perkins BL, Matsuo N, Mayford M. Localization of a Stable Neural Correlate of Associative Memory. Science. 2007;317:1230–3.
Ryan TJ, Roy DS, Pignatelli M, Arons A, Tonegawa S. Engram cells retain memory under retrograde amnesia. Science. 2015;348:1007–13.
Josselyn SA, Tonegawa S. Memory engrams: Recalling the past and imagining the future. Science. 2020;367:eaaw4325.
Phelps EA, LeDoux JE. Contributions of the Amygdala to Emotion Processing: From Animal Models to Human Behavior. Neuron. 2005;48:175–87.
Sigurdsson T, Doyère V, Cain CK, LeDoux JE. Long-term potentiation in the amygdala: A cellular mechanism of fear learning and memory. Neuropharmacology. 2007;52:215–27.
Zhou Y, Won J, Karlsson MG, Zhou M, Rogerson T, Balaji J, et al. CREB regulates excitability and the allocation of memory to subsets of neurons in the amygdala. Nat Neurosci. 2009;12:1438–43.
Wolff SBE, Gründemann J, Tovote P, Krabbe S, Jacobson GA, Müller C, et al. Amygdala interneuron subtypes control fear learning through disinhibition. Nature. 2014;509:453–8.
Lange MD, Jüngling K, Paulukat L, Vieler M, Gaburro S, Sosulina L, et al. Glutamic Acid Decarboxylase 65: A Link Between GABAergic Synaptic Plasticity in the Lateral Amygdala and Conditioned Fear Generalization. Neuropsychopharmacology. 2014;39:2211–20.
Cummings KA, Lacagnina AF, Clem RL. GABAergic microcircuitry of fear memory encoding. Neurobiol Learn Mem. 2021;184:107504.
Guo N, Soden ME, Herber C, Kim MT, Besnard A, Lin P, et al. Dentate granule cell recruitment of feedforward inhibition governs engram maintenance and remote memory generalization. Nat Med. 2018;24:438–49.
Lucas EK, Jegarl AM, Morishita H, Clem RL. Multimodal and Site-Specific Plasticity of Amygdala Parvalbumin Interneurons after Fear Learning. Neuron. 2016;91:629–43.
Stujenske JM, O’Neill P-K, Fernandes-Henriques C, Nahmoud I, Goldburg SR, Singh A, et al. Prelimbic cortex drives discrimination of non-aversion via amygdala somatostatin interneurons. Neuron. 2022;110:2258–2267.e11.
Hou W-H, Jariwala M, Wang K-Y, Seewald A, Lin Y-L, Liou Y-C, et al. Inhibitory fear memory engram in the mouse central lateral amygdala. Cell Rep. 2024;43:114468.
Suzuki A, Fukushima H, Mukawa T, Toyoda H, Wu L-J, Zhao M-G, et al. Upregulation of CREB-Mediated Transcription Enhances Both Short- and Long-Term Memory. J Neurosci. 2011;31:8786–802.
Han J-H, Kushner SA, Yiu AP, Cole CJ, Matynia A, Brown RA, et al. Neuronal Competition and Selection During Memory Formation. Science. 2007;316:457–60.
Han J-H, Kushner SA, Yiu AP, Hsiang H-L(Liz), Buch T, Waisman A, et al. Selective Erasure of a Fear Memory. Science. 2009;323:1492–6.
Park A, Jacob AD, Hsiang H-L, Liz, Frankland PW, Howland JG, Josselyn SA. Formation and fate of an engram in the lateral amygdala supporting a rewarding memory in mice. Neuropsychopharmacology. 2023;48:724–33.
Hu Y, Lund IV, Gravielle MC, Farb DH, Brooks-Kayal AR, Russek SJ. Surface Expression of GABAA Receptors Is Transcriptionally Controlled by the Interplay of cAMP-response Element-binding Protein and Its Binding Partner Inducible cAMP Early Repressor. J Biol Chem. 2008;283:9328–40.
Kojima N, Borlikova G, Sakamoto T, Yamada K, Ikeda T, Itohara S, et al. Inducible cAMP Early Repressor Acts as a Negative Regulator for Kindling Epileptogenesis and Long-Term Fear Memory. J Neurosci. 2008;28:6459–72.
Borlikova G, Endo S. Inducible cAMP Early Repressor (ICER) and Brain Functions. Mol Neurobiol. 2009;40:73–86.
Choi DI, Kim J, Lee H, Kim J, Sung Y, Choi JE, et al. Synaptic correlates of associative fear memory in the lateral amygdala. Neuron. 2021;109:2717–2726.e3.
Daphne D, Ahire A, Singh SP, Wagner S, Barkai E, Reuveni I. Extended fear learning enhances excitatory and inhibitory synaptic transmission onto amygdala engram cells. Nat Commun. 2025;16:9908.
Torborg CL, Nakashiba T, Tonegawa S, McBain CJ. Control of CA3 Output by Feedforward Inhibition Despite Developmental Changes in the Excitation–Inhibition Balance. J Neurosci. 2010;30:15628–37.
Rumpel E, Behrends JC. Sr2+-dependent asynchronous evoked transmission at rat striatal inhibitory synapses in vitro. J Physiol. 1999;514:447–58.
Hochgerner H, Singh S, Tibi M, Skarbianskis N, Admati I, Ophir O, et al. Neuronal types in the mouse amygdala and their transcriptional response to fear conditioning. Nat Neurosci. 2023;26:2237–49.
Shumyatsky GP, Tsvetkov E, Malleret G, Vronskaya S, Hatton M, Hampton L, et al. Identification of a Signaling Network in Lateral Nucleus of Amygdala Important for Inhibiting Memory Specifically Related to Learned Fear. Cell. 2002;111:905–18.
Grosso A, Santoni G, Manassero E, Renna A, Sacchetti B. A neuronal basis for fear discrimination in the lateral amygdala. Nat Commun. 2018;9:1214.
Wojick JA, Paranjapye A, Chiu JK, Oswell CS, Mahmood M, Wooldridge LM, et al. A nociceptive amygdala-striatal pathway modulating affective-motivational pain. Sci Adv. 2025;11:eado2837.
Wiltgen BJ, Godsil B, Peng Z, Saab F, June H, Van Linn ML, et al. The α1 subunit of the GABA(A) receptor modulates fear learning and plasticity in the lateral amygdala. Front Behav Neurosci. 2009;3:37.
Sommeijer J-P, Levelt CN. Synaptotagmin-2 Is a Reliable Marker for Parvalbumin Positive Inhibitory Boutons in the Mouse Visual Cortex. PLoS ONE. 2012;7:e35323.
Bouhours B, Gjoni E, Kochubey O, Schneggenburger R. Synaptotagmin2 (Syt2) Drives Fast Release Redundantly with Syt1 at the Output Synapses of Parvalbumin-Expressing Inhibitory Neurons. J Neurosci. 2017;37:4604–17.
Vogel E, Krabbe S, Gründemann J, Wamsteeker Cusulin JI, Lüthi A. Projection-Specific Dynamic Regulation of Inhibition in Amygdala Micro-Circuits. Neuron. 2016;91:644–51.
Cattani A, Arnold DB, McCarthy M, Kopell N. Basolateral amygdala oscillations enable fear learning in a biophysical model. eLife. 2024;12:RP89519.
Asok A, Kandel ER, Rayman JB The Neurobiology of Fear Generalization. Front Behav Neurosci. 2019;12:329.
Bergstrom HC. Assaying Fear Memory Discrimination and Generalization: Methods and Concepts. Curr Protoc Neurosci. 2020;91:e89.
Josselyn SA, Köhler S, Frankland PW. Finding the engram. Nat Rev Neurosci. 2015;16:521–34.
Suthard RL, Senne RA, Buzharsky MD, Diep AH, Pyo AY, Ramirez S. Engram reactivation mimics cellular signatures of fear. Cell Rep. 2024;43:113850.
Morrison DJ, Rashid AJ, Yiu AP, Yan C, Frankland PW, Josselyn SA. Parvalbumin interneurons constrain the size of the lateral amygdala engram. Neurobiol Learn Mem. 2016;135:91–99.
Ramsaran AI, Wang Y, Golbabaei A, Aleshin S, de Snoo ML, Yeung BA, et al. A shift in the mechanisms controlling hippocampal engram formation during brain maturation. Science. 2023;380:543–51.
Guenthner CJ, Miyamichi K, Yang HH, Heller HC, Luo L. Permanent Genetic Access to Transiently Active Neurons via TRAP: Targeted Recombination in Active Populations. Neuron. 2013;78:773–84.
DeNardo LA, Liu CD, Allen WE, Adams EL, Friedmann D, Fu L, et al. Temporal evolution of cortical ensembles promoting remote memory retrieval. Nat Neurosci. 2019;22:460–9.
Breiman L. Random Forests. Mach Learn. 2001;45:5–32.
Park S, Kramer EE, Mercaldo V, Rashid AJ, Insel N, Frankland PW, et al. Neuronal Allocation to a Hippocampal Engram. Neuropsychopharmacology. 2016;41:2987–93.
Pittenger C, Huang YY, Paletzki RF, Bourtchouladze R, Scanlin H, Vronskaya S, et al. Reversible Inhibition of CREB/ATF Transcription Factors in Region CA1 of the Dorsal Hippocampus Disrupts Hippocampus-Dependent Spatial Memory. Neuron. 2002;34:447–62.
Aguado, Díaz-Ruiz F, Parlato R C, Martínez A, Carmona MA, Bleckmann S, et al. The CREB/CREM Transcription Factors Negatively Regulate Early Synaptogenesis and Spontaneous Network Activity. J Neurosci. 2009;29:328–33.
Lonze BE, Ginty DD. Function and Regulation of CREB Family Transcription Factors in the Nervous System. Neuron. 2002;35:605–23.
Xu K, Liu Z, Wang L, Wu G, Liu T. Influence of hippocampal low-frequency stimulation on GABAA R α1, ICER and BNDF expression level in brain tissues of amygdala-kindled drug-resistant temporal lobe epileptic rats. Brain Res. 2018;1698:195–203.
Gundersen BB, Briand LA, Onksen JL, LeLay J, Kaestner KH, Blendy JA. Increased Hippocampal Neurogenesis and Accelerated Response to Antidepressants in Mice with Specific Deletion of CREB in the Hippocampus: Role of cAMP Response-Element Modulator τ. J Neurosci. 2013;33:13673–85.
Fenno LE, Mattis J, Ramakrishnan C, Hyun M, Lee SY, He M, et al. Targeting cells with single vectors using multiple-feature Boolean logic. Nat Methods. 2014;11:763–72.
Nonaka A, Toyoda T, Miura Y, Hitora-Imamura N, Naka M, Eguchi M, et al. Synaptic Plasticity Associated with a Memory Engram in the Basolateral Amygdala. J Neurosci. 2014;34:9305–9.
Gouty-Colomer LA, Hosseini B, Marcelo IM, Schreiber J, Slump DE, Yamaguchi S, et al. Arc expression identifies the lateral amygdala fear memory trace. Mol Psychiatry. 2016;21:364–75.
Xue M, Atallah BV, Scanziani M. Equalizing excitation–inhibition ratios across visual cortical neurons. Nature. 2014;511:596–600.
Barron HC, Vogels TP, Behrens TE, Ramaswami M. Inhibitory engrams in perception and memory. Proc Natl Acad Sci. 2017;114:6666–74.
Bissière S, Humeau Y, Lüthi A. Dopamine gates LTP induction in lateral amygdala by suppressing feedforward inhibition. Nat Neurosci. 2003;6:587–92.
Tully K, Li Y, Tsvetkov E, Bolshakov VY. Norepinephrine enables the induction of associative long-term potentiation at thalamo-amygdala synapses. Proc Natl Acad Sci. 2007;104:14146–50.
Ehrlich I, Humeau Y, Grenier F, Ciocchi S, Herry C, Lüthi A. Amygdala Inhibitory Circuits and the Control of Fear Memory. Neuron. 2009;62:757–71.
Mahan AL, Ressler KJ. Fear conditioning, synaptic plasticity and the amygdala: implications for posttraumatic stress disorder. Trends Neurosci. 2012;35:24–35.
Girgenti MJ, Wang J, Ji D, Cruz DA, Stein MB, Gelernter J, et al. Transcriptomic organization of the human brain in post-traumatic stress disorder. Nat Neurosci. 2021;24:24–33.
Marek R, Jin J, Goode TD, Giustino TF, Wang Q, Acca GM, et al. Hippocampus-driven feed-forward inhibition of the prefrontal cortex mediates relapse of extinguished fear. Nat Neurosci. 2018;21:384–92.
Ruediger S, Vittori C, Bednarek E, Genoud C, Strata P, Sacchetti B, et al. Learning-related feedforward inhibitory connectivity growth required for memory precision. Nature. 2011;473:514–8.
Bannai H, Niwa F, Sherwood MW, Shrivastava AN, Arizono M, Miyamoto A, et al. Bidirectional Control of Synaptic GABAAR Clustering by Glutamate and Calcium. Cell Rep. 2015;13:2768–80.
Gravielle MC. Regulation of GABAA Receptors Induced by the Activation of L-Type Voltage-Gated Calcium Channels. Membranes. 2021;11:486.
Lesuis SL, Park S, Hoorn A, Rashid AJ, Mocle AJ, Salter EW, et al. Stress disrupts engram ensembles in lateral amygdala to generalize threat memory in mice. Cell. 2025;188:121–140.e20.
Petrini EM, Ravasenga T, Hausrat TJ, Iurilli G, Olcese U, Racine V, et al. Synaptic recruitment of gephyrin regulates surface GABAA receptor dynamics for the expression of inhibitory LTP. Nat Commun. 2014;5:3921.
Bhembre N, Paolino A, Das SS, Guntupalli S, Fenlon LR, Anggono V. Learning-induced remodelling of inhibitory synapses in the motor cortex. Open Biol. 2024;14:240109.
Flores CE, Nikonenko I, Mendez P, Fritschy J-M, Tyagarajan SK, Muller D. Activity-dependent inhibitory synapse remodeling through gephyrin phosphorylation. Proc Natl Acad Sci. 2015;112:E65–E72.
Mäki-Marttunen T, Iannella N, Edwards AG, Einevoll GT, Blackwell KT. A unified computational model for cortical post-synaptic plasticity. eLife. 2020;9:e55714.
Krueger-Burg D. Understanding GABAergic synapse diversity and its implications for GABAergic pharmacotherapy. Trends Neurosci. 2025;48:47–61.
Lund IV, Hu Y, Raol YH, Benham RS, Faris R, Russek SJ, et al. BDNF Selectively Regulates GABAA Receptor Transcription by Activation of the JAK/STAT Pathway. Sci Signal. 2008;1:ra9–ra9.
Flavell SW, Greenberg ME. Signaling Mechanisms Linking Neuronal Activity to Gene Expression and Plasticity of the Nervous System. Annu Rev Neurosci. 2008;31:563–90.
Glazewski S, Bejar R, Mayford M, Fox K. The effect of autonomous alpha-CaMKII expression on sensory responses and experience-dependent plasticity in mouse barrel cortex. Neuropharmacology. 2001;41:771–8.
Álvarez-Aznar A, Martínez-Corral I, Daubel N, Betsholtz C, Mäkinen T, Gaengel K. Tamoxifen-independent recombination of reporter genes limits lineage tracing and mosaic analysis using CreERT2 lines. Transgenic Res. 2020;29:53–68.
Iqbal J, Huang G-D, Xue Y-X, Yang M, Jia X-J. The neural circuits and molecular mechanisms underlying fear dysregulation in posttraumatic stress disorder. Front Neurosci. 2023;17:1281401.
Nguyen F, Shetty AK. Post-traumatic stress disorder: pathogenesis, epidemiological characteristics, animal models, and potential therapeutic strategies. Mil Med Res. 2026;13:100005.
Acknowledgements
We thank Paul Frankland (Hospital for Sick Children) for his valuable suggestions. We are also grateful to Joshua Johansen (RIKEN/UC Irvine) and Ron Stoop (University of Lausanne) for their critical reading of and feedback on the manuscript.
Funding
This work was supported by the National Research Foundation of Korea (RS-2018-NR028477 and RS-2023-00271562 to H.L.; RS-2023-00265883, RS-2025-23324087, and RS-2018-NR030947 to J.-H.K.), and the Ministry of Food and Drug Safety (RS-2024-00332645 and RS-2026-25520698 to J.-H.K.).
Author information
Authors and Affiliations
Contributions
H.L. and S.L. performed slice electrophysiology, bioinformatic analyses, and immunohistochemistry. T.Y. conducted AAV packaging and vector subcloning. T.D.N. performed RT-qPCR and designed ICER-related AAV vectors. L.T. conducted tone-tagging experiments, AAV injections, and behavioral studies. G.Y.R. performed electron microscopy imaging. J.P. and S.K.P. provided conceptual and technical guidance. J.-H.K. conceived and designed the study, and wrote the manuscript.
Corresponding author
Ethics declarations
Competing interests
H.L., T.D.N., and J.-H.K. are listed as inventors on a Korean patent application (Korean Patent Application No. 10-2026-0155439) related to part of this work. The other authors have declared that there are no further conflicts of interest related to this study.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Lee, H., Lee, S., Yoo, T. et al. ICER-Controlled inhibitory plasticity of amygdala engram neurons regulates fear memory precision. Neuropsychopharmacol. (2026). https://doi.org/10.1038/s41386-026-02568-y
Received:
Revised:
Accepted:
Published:
Version of record:
DOI: https://doi.org/10.1038/s41386-026-02568-y
