Neuropeptide Y co-opts neuronal ensembles for memory lability and stability

Neuropeptide Y co-opts neuronal ensembles for memory lability and stability | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Neuropeptide Y co-opts neuronal ensembles for memory lability and stability Tian-Le Xu, Yan-Jiao Wu, Xue Gu, Yalei Kong, Shuo Yang, Huan Wang, and 16 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4347593/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 31 Mar, 2026 Read the published version in Nature Neuroscience → Version 1 posted You are reading this latest preprint version Abstract Memory engrams are formed by activity-dependent recruitment of distinct subsets of excitatory principal neurons (or neuronal ensembles) whereas inhibitory neurons pivot memory lability and stability. However, the molecular logic for memory engrams to preferentially recruit specific type of interneurons over others remains enigmatic. Using activity-dependent single-cell transcriptomic profiling in mice with training of cued fear memory and extinction, we discovered that neuropeptide Y (NPY)-expressing (NPY+) GABAergic interneurons in the ventral hippocampal CA1 (vCA1) region exert fast GABAergic inhibition to facilitate the acquisition of memory, but bifurcate NPY-mediated slow peptidergic inhibition onto distinct sub-ensembles underlying the extinction of single memory trace. Genetically encoded calcium and NPY sensors revealed that both calcium dynamics of NPY+ neurons and their NPY release in vCA1 ramp up as extinction learning progresses while behavioral state switches from “fear-on” to “fear-off”. Bidirectional manipulations of NPY+ neurons or NPY itself demonstrated that NPY is both necessary and sufficient to control the rate and degree of memory extinction by acting on two physically non-overlapping sub-ensembles composed of NPY1R- and NPY2R-expressing neurons. CRISPR/Cas9-mediated knockout of NPY2R or NPY1R further unravels that NPY co-opts its actions on these two sub-ensembles to gate early fast and late slow stages of extinction. These findings exemplify the intricate spatiotemporal orchestration of slow peptidergic inhibitions from single subtype of GABAergic interneurons to fine-tune engram lability verse stability of memory. Biological sciences/Neuroscience/Learning and memory/Extinction Health sciences/Diseases/Psychiatric disorders/Post-traumatic stress disorder Ventral hippocampal CA1 (vCA1) Neuropeptide Y NPY receptor Fear engrams Extinction engrams Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Main Memory traces arise when a sparse subset of neurons is recruited during learning to form an engram 1 . Activity-dependent tagging has shown that such ensembles are distributed in space and time and re-engaged during recall 2-5 . Within these ensembles, excitatory principal neurons—biased by heightened intrinsic excitability and synaptic plasticity—form the backbone of inter-ensemble connectivity, supporting flexible expression of learned behaviors 3, 5, 6 . In parallel, GABAergic interneurons shape which principles cells are admitted into the trace by suppressing non-engram activity, stabilizing network dynamics, and regulating transitions in memory fates 7-10 . Although specific interneuron subtypes have been implicated in engram-level control 5, 11, 12 , the molecular logic by which defined inhibitory neurons bias ensemble composition and tune memory lability versus stability remains unclear. GABAergic interneurons are diverse in position and output, and many co-release peptides—including somatostatin (SST), cholecystokinin (CCK), neuropeptide Y (NPY)—alongside GABA 13-15 . This raises two unresolved questions with direct relevance to engram control: (i) which inhibitory interneuron subtypes are recruited during memory expression versus extinction, a canonical state switch 16, 17 ; and (ii) under what conditions does peptide co‑release—as opposed fast GABA alone—shape ensemble selection and behavioral output. Addressing these questions requires linking cell type, receptor identity, and peptide dynamics to behavior on appropriate time scales. Cued fear learning and extinction provide a tractable framework for this goal. Fear conditioning binds a neutral cue to an aversive outcome to produce a robust, retrieval trace 18 ; extinction training subsequently suppresses expression of fear by forming a new memory that does not erase the original 17, 19, 20 . The coexistence and reversibility of fear and extinction traces—together with their dysregulation in anxiety and trauma-related disorders 21, 22 and their capacity for spontaneous recovery 23-25 —make this paradigm ideal for probing how inhibitory circuits govern the reweighting of competing ensembles. The ventral hippocampal CA1 (vCA1) is a critical node in this circuitry 26-30 , yet it is unknown how peptidergic inhibition in vCA1 organizes receptor-defined neuronal sub-ensembles across extinction. In this study, we discovered that fear extinction, compared to fear memory retrieval, preferentially activates NPY-expressing (NPY + ) interneurons in vCA1 leading to increased NPY release. Strikingly, NPY release bifurcates onto two spatially distinct sub-ensembles, each expressing unique NPY receptors. These two sub-ensembles are interlocked in time and space, enabling the switching of memory traces between a labile and stable state. This framework links peptide kinetics, receptor identity, and laminar organization to the dynamic reallocation of engram membership, providing a principled mechanism for how inhibitory circuits switch a memory between changeable and durable states. Results Single-cell transcriptomic mapping in the vCA1 uncovers upregulation of Npy gene in inhibitory neurons during fear extinction To investigate how distinct cell types flexibly dictate memory-related behaviors, we employed an auditory cued fear conditioning and extinction paradigms in wild-type mice. On Day 1, mice received five foot shocks paired with a neutral auditory tone (conditioned stimulus, CS) in context A. On Day 2, these mice underwent extinction training in a novel context B (Ext group), receiving 20 CS presentations without foot shocks (context B). Although this training significantly reduced fear responses to the tone, it did not completely eliminate context A-associated fear, as evidenced by the recovery of conditioned fear responses when mice were re-exposed to context A, regardless of the presence of the tone (Supplementary Fig. 1). To clearly distinguish the ensemble underlying fear memory from that for extinction, we focused on auditory cued fear in a setting designed to minimize confounding contextual fear. As a control for fear extinction, we subjected mice to only two CS presentations in context B (No Ext group), which proved insufficient to induce extinction. To precisely calibrate potential paradigm differences between Ext and No Ext groups, we designed three variants of No Ext conditions: (1) a short exposure group receiving two CS presentations immediately before removal from context B; (2) a group receiving two CS presentations followed by an extended waiting period (equivalent to the other 18 CS presentations in the Ext group) but without additional tone exposure; and (3) a group that remained in context B for the duration equivalent to the first 18 CS presentations in the Ext group but without actual tone presentations, then received two CS presentations prior to removal. We collected vCA1 samples either 20 min or 2 h after training for subsequent analysis. Consistent with Fos as a transient marker of neuronal activation to salient stimuli 25, 31, 32 , Fos mRNA expression peaked at 20 min post-training and substantially declined by 2 h across all experimental conditions. Crucially, mice undergoing Ext showed significantly higher Fos expression than all variants of the No Ext control at both 20 min and 2 h time points, highlighting sustained ensemble activation specifically associated with extinction (Supplementary Fig. 2). Guided by these findings, we selected tissue samples collected 2 h after training from both the Ext group and the short-exposure No Ext group, to ensure optimal temporal matching for detailed analyses. To identify the cell types recruited by fear memory retrieval versus extinction (Fig. 1a, b), we conducted unbiased, high-throughput, droplet-based single-cell RNA sequencing (scRNA-seq) using an activity-sensitive single-cell preparation method 33 . This approach enabled us to profile transcriptomic changes at the single-cell level, revealing molecular signatures associated with two distinct memory engrams 34-36 while minimizing ex vivo transcriptional artifacts and preserving behaviorally induced gene expression. In total, we profiled 46,894 individual cells, comprising 20,263 cells from No Ext group and 26,631 cells from the Ext group, each derived from 4 male mice. To analyze this dataset, we applied uniform manifold approximation and projection (UMAP) for dimensionality reduction, to visualization of the high-dimensional single-cell data in a biologically meaningful manner. This analysis revealed 35 distinct clusters and 9 cell types, identified based on established cell-type-specific markers (Fig. 1c, Supplementary Fig. 3). Among these, 8 clusters were identified as excitatory neurons, while Cluster 6 was initially categorized as the sole inhibitory neuron type. The remaining clusters corresponded to glial cells and other non-neuronal cell types (Supplementary Fig. 3a). The number and proportion of most cell types were broadly comparable, though notable differences in certain cell populations between the two conditions were observed (Supplementary Fig. 4). To rigorously control for these discrepancies, we implemented a random down-sampling procedure in which the larger dataset was repeatedly subsampled (1,000 permutations; see Methods) to match the smaller one, followed by re-analysis of the transcriptional responses. Despite this normalization, several clusters retained significant differential responsiveness to Ext compared with No Ext (Supplementary Fig. 3b), along with distinct patterns of immediate early gene (IEG; see Methods) 33 expression across various cell types and clusters (Supplementary Fig. 5). Among the neuronal populations, Ext paradigm specifically upregulated 6 genes and downregulated 2 genes across all excitatory neuron clusters combined; and upregulated 5 genes and downregulated 2 genes within cluster 6, which represents the inhibitory neuron type (Fig. 1d). Within excitatory neurons, Ext elicited cluster-specific transcriptional responses. Volcano analyses revealed the largest sets of differentially expressed genes (DEGs) in clusters 4 and 10, whereas other clusters showed few or no DEGs (Supplementary Fig. 6a–h). Moreover, Ext produced a marked increase in Fos expression in clusters 10 and 32 compared with No Ext (Supplementary Fig. 6i). Given our aim in elucidating how memory fates transition from fear persistence to extinction, we focused on GABAergic interneurons that likely plays a pivotal role in governing memory extinction process. Within inhibitory neurons of cluster 6, Ext paradigm prominently upregulated both Pcp4 and Npy by approximately 2-fold (Fig. 1d). The proportion of Npy ⁺ interneurons within the inhibitory neuron type 15 increased from 12.2% in the No Ext group to 24.6 ± 2.2% in the Ext group (Fig. 1e). This increase presumably reflects the recruitment of neurons that previously did not express Npy or expressed it only at minimal, undetectable levels. Additionally, the Ext group exhibited elevated overall Npy transcript levels per cell compared with the No Ext group (Fig. 1f). To further dissect the relationship between Npy expression and IEG activity, we specifically examined the subset of inhibitory neurons displaying particularly high levels of IEG expression (defined as cells with at least one IEG expressed in the 90 th -percentile among all inhibitory neurons). Both the proportion of Npy + /IEG + interneurons and the Npy expression within these IEG + interneurons were significantly higher in the Ext group than in the No Ext group (Supplementary Fig. 7). These findings provide strong justification for a detailed functional interrogation of Npy + GABAergic interneurons in fear extinction, in which potentially orchestrate the shift between fear and extinction memories. Across the vCA1 transcriptome, Npy expression was largely confined to inhibitory neurons and was essentially absent from other cell types, including excitatory neurons (Supplementary Fig. 8a). While initial unsupervised clustering identified a single inhibitory neuron cluster (cluster 6), further refined clustering analysis delineated 5 distinct subclusters (Supplementary Fig. 8b, c). Npy ⁺ interneurons were distributed across several of these subclusters, with enrichment in Npy ⁺ and Npy ⁺/ SST ⁺ (somatostatin-expressing) subclusters. Ext paradigm increased Npy transcript abundance across most inhibitory subclusters (with the exception of one subcluster, Supplementary Fig. 8d), implicating broad recruitment of Npy + interneurons during extinction. Npy + inhibitory neurons are selectively activated in vCA1 during extinction learning The scRNA-seq dataset did register trace-level Npy reads in a minority of excitatory clusters, with a small upward shift in Ext relative to No Ext condition (Supplementary Fig. 8e). Because these values were near the detection floor, we sought after independent validation with other approaches. Dual-color fluorescence in situ hybridization confirmed that Npy transcripts were almost exclusively detected in Vgat + (vesicular GABA transporter-expressing) neurons and absent in Vglut1 + (vesicular glutamate transporter 1-expressing) neurons, confirming the selective expression of Npy within vCA1 GABAergic interneurons. Furthermore, Npy signals were predominantly localized to the stratum oriens (SO) of the CA1 region 37 (Supplementary Fig. 9). dFISH analyses revealed a significant increase in the proportion of Npy + and Npy + / Fos + co-labeled neurons in the vCA1 of Ext compared to No Ext mice. Moreover, the expression levels of Npy and Fos in the vCA1 were significantly correlated in the Ext group but not in the No Ext group (Supplementary Fig. 10). Collectively, these results validate that Npy + GABAergic interneurons become transcriptionally activated in the vCA1 during fear extinction, highlighting their potential role as a key cell type for gating extinction learning. Dynamics of vCA1 extracellular NPY during extinction learning To directly monitor extracellular NPY dynamics in the vCA1 region during fear extinction, we employed a genetically encoded fluorescent GRAB (G protein–coupled receptor activation‒based) NPY sensor, NPY1.0 (Fig. 1g) 38 , specially targeted to vCA1 neurons (Fig. 1g, h and Supplementary Fig. 11). Mice underwent fear conditioning followed by extinction training, resulting in a gradual reduction in freezing behavior. Real-time in vivo quantification of NPY release using fiber photometry revealed no significant changes during exposure to the auditory tone (CS), however, the signal gradually decreased following the delivery of five foot shocks (US) during fear conditioning (Extended Data Fig. 1). During extinction learning, a clear contrast in CS-evoked NPY release was observed between the No Ext and Ext groups. In the No Ext group, the first two CS presentations in the Block 1—characterized by heightened fear responses—elicited a significant decrease in NPY release. Conversely, in the Ext group, the last two CS presentations in the Block 10—associated with diminished fear responses—showed a modest but statistically significant increase in CS-evoked NPY release (Fig. 1i, j). Furthermore, NPY release progressively ramped up in response to repeated CS presentations during extinction learning, correlating with the attenuation of fear responses (Fig. 1k). During each 30-s period CS trial in extinction training, mice alternated between freezing (ON) and non-freezing (OFF) epochs (Fig. 1l), with corresponding decreases and increases in the NPY signal (Fig. 1m–o). Interestingly, prolonged NPY release was detected 2 h after Ext compared to the No Ext condition (Extended Data Fig. 2), mirroring the observed increase in Npy gene expression at the same time point. Moreover, a significant increase in cued NPY release was observed during Ext Retr on Day 3 (Extended Data Fig. 1g–i). In contrast, a control group exposed to the CS without foot shocks (CS only) did not exhibit significant NPY release during either conditioning or extinction (Supplementary Fig. 12), suggesting that the extinction-induced NPY release is specifically associated with the extinction of the CS rather than mere repeated exposures or habituation to the cue. Moreover, no detectable NPY signal changes were observed in mice expressing a non-ligand-binding mutant sensor (NPYmut) in wild-type mice or in mice expressing the GRAB-NPY1.0 sensor in Npy knockout backgrounds during extinction learning (Supplementary Fig. 13). These findings demonstrate that NPY release in the vCA1 is dynamically regulated during extinction learning and is specifically associated with the extinction of fear memory. Dynamics of vCA1 NPY + neuronal activity during extinction learning To test whether vCA1 NPY + interneurons drive NPY release, we co-expressed a Cre-dependent red-shifted opsin (ChrimsonR) and the GRAB-NPY1.0 sensor in NPY-Cre mice and recorded fiber photometry in vCA1 (Fig. 2a and Supplementary Fig. 14a). Brief 635-nm stimulation produced a stimulus-locked downward deflection in both the 470-nm sensor channel and the 405-nm isosbestic control—consistent with transient extracellular pH-dependent quenching commonly seen with GRAB sensors 38, 39 —followed by a selective post‑stimulus increase in the 470-nm signal (Fig. 2b, c and Supplementary Fig. 14b, c). Quantitatively, the 5-s “during-light” AUC was negative for both channels, whereas the 5-s “after-light” AUC was positive only in the 470-nm channel (Fig. 2c and Extended Data Fig. 16c). Together with the spectral separation between photometry excitation (470/405 nm) and optogenetic light (635 nm), these data indicate that the illumination-period dip reflects pH quenching rather than suppressed release, and that activating NPY + interneurons is sufficient to evoke NPY release in vCA1. To further investigate the neuronal activity of vCA1 NPY + interneurons, we expressed the Cre-dependent genetically encoded calcium indicator GCaMP6m in these cells and continuously recorded calcium transients using fiber photometry (Fig. 2d and Supplementary Fig. 14d). During fear conditioning, NPY + interneurons exhibited sustained and robust calcium responses to foot shocks (US). Auditory cues (CS) elicited progressively stronger calcium responses across successive learning trials, indicating heightened neuronal activity associated with fear learning (Extended Data Fig. 3). Notably, calcium responses during the Ext state were significantly elevated compared to the No Ext state (Fig. 2e, f). As extinction progressed, calcium activity in NPY + neurons continued to increase, exhibiting a significant negative correlation with the diminishing fear response during intermittent CS presentations (Fig. 2g). Consistent with the GRAB-NPY1.0 sensor data, we observed a decrease in calcium signals during freezing (ON epochs) and an increase during non-freezing (OFF epochs) periods (Fig. 2h–k). Moreover, prolonged neuronal activity was detected 2 h post-Ext compared to the No Ext condition (Extended Data Fig. 4), paralleling the sustained increase in NPY release observed with the GRAB-NPY1.0 sensor at the same time point. We also observed a significant elevation in cued NPY + neuronal calcium activity during Ext Retr on Day 3 (Extended Data Fig. 3g–i). In contrast, a control group exposed to the CS without foot shocks (CS only) did not exhibit significant NPY + neuronal calcium activity during either conditioning or extinction phases (Supplementary Fig. 15). These findings, in conjunction with the dynamic changes in NPY release specifically associated with extinction, suggest that vCA1 NPY + interneurons actively drive NPY dynamics during fear extinction, highlighting their potential role in facilitating progressive extinction. However, the absence of a corresponding association between NPY dynamics and neuronal activity during fear learning indicates that vCA1 NPY + interneurons may engage through alternative mechanisms in this scenario. This distinction underscores the specialized role of NPY + interneurons in modulating memory extinction, separate from their functions during the initial fear learning phase. Subsequently, we performed a comparative analysis of the excitability of NPY + interneurons between the No Ext and Ext groups using patch-clamp electrophysiology. Extinction training resulted in a substantial increase in the frequency of action potentials evoked by identical current step injections in the Ext group, while the amplitude of these action potentials remained unchanged (Fig. 2l–n). Furthermore, the frequency of spontaneous excitatory postsynaptic currents (sEPSCs), indicative of global excitatory synaptic input to NPY + interneurons, was significantly elevated in the Ext group, although changes in amplitude were less pronounced (Fig. 2o–q). These electrophysiological findings demonstrate that extinction learning enhances both the excitatory drive and intrinsic excitability of NPY + interneurons in the vCA1. Collectively, these data indicate that during extinction, activated NPY + GABAergic interneurons in the vCA1, through endogenous NPY release, play a pivotal role in modulating the dynamics of fear and extinction memory. Manipulation of vCA1 NPY + interneuron activity controls memory extinction To evaluate the necessity and efficacy of NPY + neuronal activity in extinction learning, we introduced halorhodopsin (NpHR) or channelrhodopsin-2 (ChR2) into NPY + interneurons in the vCA1 and manipulated their activity using spatially targeted light (Fig. 3a–c). Optogenetic inhibition of NPY + interneurons during CS presentation in extinction learning intensified the tone-shock association, sustaining a higher fear response (Fig. 3d). Even during extinction retrieval without optogenetic inhibition, the NpHR group maintained a higher level of conditioned fear. Conversely, optogenetic activation of NPY + interneurons during CS presentation in extinction learning disassociated the tone-shock association, resulting in reduced fear responses both in extinction learning and retrieval (Fig. 3e). Further analyses of extinction time course showed two stages of decline in cued freezing levels over 10-block epochs (20 CS): Stage 1, i.e., the first 5 blocks, being rapid, likely reflects the lability of fear memory to extinction; and Stage 2, the last 5 blocks, being much slower, likely measures the stability of fear memory (Fig. 3f and Supplementary Fig. 16a–d). Optogenetic inhibition of NPY + interneurons largely ablated Stage 1 extinction and precluded its progression to Stage 2; and conversely activation of NPY + neurons robustly augmented the onset of Stage 1 extinction but had little effect on the stable setpoint of freezing level for Stage 2 (Fig. 3f and Supplementary Fig. 16a–d). These observations indicated that Stage 1 and 2 of extinction are interlocked processes, for which the activity level of NPY + interneurons may confer sequentially labile and stable states of fear memory. Notably, when NPY + neuronal activity was selectively elevated optogenetically during the CS-induced freezing ON epoch without affecting other periods (Fig. 3g), significant facilitatory effects on both extinction learning and retrieval were observed. However, this targeted elevation did not eliminate the difference in extinction slope between the first and second halves of training, suggesting that a physiologically scaling-up activation of NPY + neurons on demand preserves the balance between memory liability and stability during extinction learning (Fig. 3h–j and Supplementary Fig. 16e, f). To further validate the role of NPY + neuronal activity in modulating the lability and stability of fear extinction, we conducted a complementary optogenetic activation experiment. In this case, NPY + neurons were activated following fear retrieval (two CS exposure) without subsequent 18 CS exposure for extinction training (Fig. 3k). Remarkably, this activation accelerated Stage 1 extinction and achieved fear reduction effects comparable to those in the full extinction protocol (control extinction group, Fig. 3l). Importantly, the effects of endogenous NPY levels were specific to fear extinction, as locomotor behavior in these mice remained unaffected by the same manipulations (Supplementary Fig. 17). These results demonstrate that vCA1 NPY + interneurons play a pivotal role in dictating the progression of extinction, even in the absence of the full extinction protocol. Furthermore, their activity appears most influential during the behavioral state switch from “fear-on” to “fear-off” epochs during extinction learning, underscoring their critical contribution to the dynamic regulation of memory liability and stability. NPY itself is both necessary and sufficient for fear extinction To determine whether the NPY peptide itself fulfills a similar function to that of vCA1 NPY + neuron activation in facilitating fear extinction, we microinjected exogenous NPY into the vCA1 through an implanted cannula 15 min prior to extinction leaning. This intervention significantly accelerated and enhanced the rate and extent of fear extinction (Fig. 3m, n and Supplementary Fig. 18a, b), effectively mirroring the effect of optogenetic activation of NPY + interneurons (Fig. 3e). Conversely, in mice carrying a global or partial loss of Npy using Npy gene knockout ( Npy –/– ) or knockdown ( Npy +/– ) manipulations, where NPY expression was undetectable or markedly reduced, fear extinction was impaired while fear conditioning remained intact. Moreover, the normal two-phase (Stage 1 vs. Stage 2) decline in freezing during extinction was disrupted in these mutants (Supplementary Fig. 18c–i). To further examine the cell-type-specific contribution of NPY, we employed a Cre-dependent CRISPR-Cas9 strategy to selective disrupt Npy in vCA1 NPY + interneurons in NPY-Cre::Cas9 [lox-stop-lox-Streptococcus pyogenes Cas9 (SpCas9)] double transgenic mice. By targeting exons 2 and 3 of the Npy gene, Cas9 abolished the Npy expression in these neurons. Adeno associated virus (AAV) delivery of Npy -targeting guide RNAs significantly reduced Npy expression in NPY + neurons compared to a control AAV containing scramble guide RNAs (Fig. 3o). Location-specific Npy knockout in vCA1 NPY + neurons markedly impeded extinction but had no effect on fear conditioning (Fig. 3p). Notably, the two-phase decline in freezing was eliminated during extinction (Fig. 3q and Supplementary Fig. 18j, k). Thess results indicate that, consistent with optogenetic manipulations of vCA1 NPY + interneuron activity, altering NPY peptide levels in this region bidirectionally modulates the efficacy of fear extinction. Given that NPY + neurons in the vCA1 are GABAergic (Supplementary Figs. 8 and 9), we investigated whether co-released GABA by these cells is also involved in the regulation of fear and extinction. To this end, we expressed Cre-dependent ChrimsonR along with the GRAB GABA0.8 sensor in NPY-Cre mice (Extended Data Fig. 5a). Optogenetic stimulation of NPY + neurons triggered GABA release (Extended Data Fig. 5a, b), in parallel with NPY release (Fig. 2a–c). Selective ablation of GABA release from vCA1 NPY + neurons—via Cre-dependent CRISPR-Cas9 targeting the Slc32a1 gene (which encodes the vesicular GABA transporter)—prevented fear conditioning (Extended Data Fig. 5c–i). Similarly, optogenetic inhibition of NPY + neuron activity during conditioning suppressed both NPY and GABA co-release and led to normal fear conditioning and reduced fear responses at retrieval (Extended Data Fig. 5j, k). In contrast, genetic ablation of the entire NPY + interneuron population in vCA1 attenuated fear conditioning and subsequent memory retrieval (Extended Data Fig. 5l, m). Collectively, these findings illustrate distinct yet complementary roles of GABA and NPY in memory processes. Moreover, manipulating NPY + neurons in the hypothalamic arcuate nucleus (ARC) did not affect fear memory (Supplementary Fig. 19), reinforcing the region-specific functions of these neurons. Altogether, our data reveal a dual mechanism wherein the fast GABA release from vCA1 NPY + interneurons is essential for encoding fear memory, whereas the slower NPY-mediated signaling is predominantly required for extinction. Non-overlapping neurons with distinct NPY receptor subtypes To further investigate the downstream targets of NPY in fear extinction, we analyzed the single-cell transcriptomic data presented in Fig. 1a and identified three primary types of NPY receptor genes— Npy1r , Npy2r , and Npy5r —across various cell types in the vCA1. Both excitatory and inhibitory neurons predominantly expressed Npy1r and Npy2r , whereas Npy5r was scarce (Supplementary Fig. 20a–c). A cross-cell-type dot-plot further showed that Npy1r transcripts were detected in neurons and, most prominently, in pericytes, while Npy2r was largely confined to neurons; Npy5r appeared only at trace levels across all cell types, with occasional reads in excitatory and inhibitory neurons (Supplementary Fig. 20d). Given our focus on neuronal engrams, we opted to subsequent analyses to Npy1r and Npy2r in neurons and did not pursue Npy5r further. Notably, Npy1r and Npy2r were predominantly found in non-overlapping neuronal populations (Fig. 4a, b and Supplementary Fig. 20e) and were negligibly distributed within Npy + inhibitory neurons (Supplementary Fig. 20f), suggesting that NPY does not functionally overlap with its receptor subtypes in a cell-autonomous manner. To resolve receptor specificity at single-cell resolution, we combined genetic tagging with multiplex RNAscope. NPY2R-Cre mice received AAV-DIO-EGFP in vCA1 to label NPY2R-expressing neurons. Two weeks later, high-magnification confocal z-stacks with 3D surface rendering allowed unambiguous assignment of Npy1r and Npy2r puncta to individual EGFP + somata (Fig. 4c–f). Within the anatomically defined layers of vCA1 region 40 , layer-resolved quantification revealed a pronounced laminar bias: Npy1r puncta were enriched in the superficial (sps) and middle (spm) pyramidal layers and declined toward the deep pyramidal layer (spd), whereas Npy2r puncta peaked in spd and were scarce in stratum oriens (SO) (Fig. 4e and Supplementary Fig. 20g). At the cellular level, analysis of 446 EGFP-labelled neurons showed that the great majority contained Npy2r puncta only; a small minority co-expressed Npy2r and Npy1r ; and only a single EGFP-positive neuron harbored Npy1r puncta without Npy2r (Fig. 4f). Consistent with the somatic location of the peptide source, Npy mRNA puncta were concentrated in SO interneurons bordering spd (Supplementary Fig. 20h, i). Together, these data integrate the scRNA-seq overview with spatially resolved histology, showing that NPY produced by SO interneurons is positioned to influence largely segregated, layer-biased pyramidal subpopulations defined by Npy1r or Npy2r in vCA1. Layer-resolved immunofluorescence for the receptor proteins revealed a related but non-identical pattern: NPY1R signal was strongest in the spd and weaker in spm and sps, while NPY2R signal was relatively enriched in sps and spd compared with spm (Extended Data Fig. 6a, b). Antibody specificity was verified by two orthogonal controls. First , in NPY2R-Cre mice expressing AAV-DIO-EGFP in vCA1, nearly all EGFP-labelled neurons were NPY2R-positive (99.2% EGFP + /NPY2R + ) and only a small minority co-labelled with NPY1R (4.8% EGFP + /NPY1R + in somata, 9.2% EGFP + /NPY1R + in neurites). High-magnification images revealed NPY2R puncta outlining somata and proximal processes (Extended Data Fig. 6c, d). Second , in Syn-Cre::Cas9 mice, Npy1r sgRNA selectively reduced NPY1R signal across vCA1 layers, whereas Npy2r sgRNA selectively diminished NPY2R staining, shifting the per-cell intensity distribution leftward with a trend toward a lower mean (Extended Data Fig. 6e–i). These controls confirm the specificity of both antibodies for quantitative analyses. High‑magnification confocal imaging with 3D compartment assignment revealed distinct subcellular biases: NPY1R immunoreactivity extended prominently along neurites, whereas NPY2R was concentrated on somatic membranes with comparatively sparse labeling in processes (Extended Data Fig. 6c–e). Compartment-assigned quantification confirmed a higher neurite‑to‑soma ratio for NPY1R than for NPY2R. Because RNAscope primarily detects soma-localized transcripts, a partial mismatch between the mRNA and protein maps is expected and most likely reflects trafficking and turnover of these G‑protein-coupled receptors (i.e., NPY1R and NPY2R) rather than protein expression in the absence of cognate mRNA. This arrangement supports a model in which NPY released from SO‑biased interneurons engages receptor fields with different subcellular reach—NPY1R distributed along neurites and NPY2R enriched at somata—across the pyramidal layers of vCA1. Next, we examined the single-cell transcriptomes of Npy1r - and Npy2r -expressing neurons to identify DEGs. Under Ext conditions, two DEGs were enriched in Npy2r -expressing neurons, whereas eight DEGs were enriched in Npy1r -expressing neurons; none of DEGs except Npy1r and Npy2r in these two neuronal populations were identified under No Ext conditions (Fig. 4g). Consistent with these findings, analysis of IEG expression revealed no significant difference between Npy2r - and Npy1r -expressing neurons under the No Ext condition. However, under Ext, Npy2r -expressing neurons showed significantly lower IEG expression than Npy1r -expressing neurons (Supplementary Fig. 21). These observations suggest that NPY released in the vCA1 during extinction engages distinct transcriptional programs in Npy1r - and Npy2r -expressing neurons, potentially giving rise to two sub-ensembles that mediate fear versus extinction engram functions. To further investigate this bivalent role of NPY in shaping memory ensembles, we optogenetically activated NPY + interneurons during the No Ext condition, when these neurons are typically less active. This manipulation increased Npy1r -expressing neurons while decreasing Npy2r -expressing neurons (Extended Data Fig. 7a–e). Conversely, optogenetically inhibiting NPY + interneurons during Ext, when they are normally highly active, reduced Npy1r -expressing neurons but elevated Npy2r -expressing neurons (Extended Data Fig. 7f–j). Thus, modulating NPY + interneuron activity at specific time point is sufficient to alter the proportions of Npy1r - and Npy2r -tagged neuronal ensembles. To clarify the roles of Npy1r - and Npy2r -expressing sub-ensembles in fear- and/or extinction-engram neurons in the vCA1, we employed the targeted recombination in active populations (TRAP) system 31, 41 in TRAP2::Ai9 (lox-stop-lox-tdTomato) double transgenic mice. Specifically, we used intraperitoneal 4-hydroxytamoxifen (4-OHT) injections to tag active neurons (i.e., label them with tdTomato) during three distinct behavioral epochs: (1) Fear retrieval (No Ext): exposure to two CS at high fear levels; (2) Extinction learning (Ext): exposure to 20 CS with progressively decreasing fear; and (3) Extinction retrieval (Ext Retr): exposure to eight CS on the day following extinction at much lower fear levels. In situ hybridization (Fig. 4h) revealed that Ext, but not Ext Retr, increased the total number of TRAPed neurons compared to fear retrieval. Moreover, fear retrieval increased both the absolute number and the relative proportion of Npy1r -expressing neurons while decreasing those of Npy2r -expressing neurons within the tagged ensemble. In contrast, extinction and extinction retrieval generated the opposite pattern, preferentially recruiting Npy2r -expressing neurons over Npy1r -expressing neurons (Fig. 4i–k). To further elucidate the mechanistic link between Npy1r - and Npy2r -expressing neuronal sub-ensembles and their roles in extinction and fear engrams, we employed a combination of TRAP tagging and Fos labeling to more definitively identify fear engram neurons as those co-expressing TRAPed tdTomato and Fos. We tagged neurons that were active during fear retrieval (No Ext) in TRAP2::Ai9 double transgenic mice, then performed in situ hybridization to reveal the expression of Fos , Npy , NPY receptors Npy1r and Npy2r (Extended Data Fig. 8a). As expected, a significantly higher proportion of tdTomato + neurons in the No Ext condition were reactivated ( tdTomato + and Fos + ) compared to the Ext condition, suggesting that these overlapping populations encode fear memory (Extended Data Fig. 8b, c). Notably, among these reactivated neurons in No Ext, we observed a much greater overlap with Npy2r than with Npy1r or Npy (Extended Data Fig. 8d–f), reinforcing the idea that Npy2r -expressing sub-ensembles form the core of the fear memory engram. Because the transition from fear to extinction is largely irreversible within this time window, we were unable to assess the overlap between reactivated extinction-engram neurons and NPY or its receptors in the comparable paradigm. Nevertheless, together with previous findings, these observations raised an intriguing possibility that NPY differentially engages two receptor-defined neuronal sub-ensembles to support fear versus extinction memory. While extinction learning and retrieval appeared to modestly impact the overall size of the engram relative to fear retrieval, they drove a more pronounced switch from Npr2r - to Npr1r -expressing ensembles. This shift underscores the mechanistic bifurcation of NPY signaling in regulating the balance between memory lability and stability. Genetic deletion or pharmacological perturbations of Npy1r or Npy2r sub-ensembles selectively impact memory liability and stability To dissect the functional roles of Npy1r and Npy2r sub-ensembles in the vCA1, we took advantage of TRAP2::Cas9 (lox-stop-lox-Cas9) double transgenic mice in which Cas9 is specifically expressed in fear- or extinction-tagged neurons. Within these TRAPed populations, guide RNAs targeting the Npy2r or Npy1r genes abolished the expression of these genes, as confirmed by post hoc analyses (Supplementary Fig. 22). We thus evaluated how fear- and extinction-tagged ensembles are affected by selectively knocking out Npy1r or Npy2r (Fig. 5a). In fear-tagged neurons, Npy1r deletion did not alter either Stage 1 or Stage 2 of extinction. In contrast, Npy2r deletion significantly slowed Stage1 and diminished Stage 2 (Fig. 5b, c and Supplementary Fig. 23a–c), mirroring the effect of optogenetically inhibiting NPY + interneurons (Fig. 3c, d). Leveraging the phenomenon of spontaneous recovery of extinguished fear over time 23, 25 , we next evaluated extinction relearning and retrieval after deleting Npy1r or Npy2r in extinction-tagged neurons that were active during the initial round of extinction learning. Notably, Npy1r knockout enhanced both the rate and magnitude of extinction across both Stage 1 and Stage 2, whereas Npy2r knockout had minimal effects (Fig. 5d, e and Supplementary Fig. 23d–f). No-shock controls at matching time points did not modify the ratio of Npr2r - and Npr1r - expressing ensembles observed in Fig. 4h–k; moreover, TRAPed neuron-specific Npy1r or Npy2r deletion in the absence of shock failed to induce significant freezing, ruling out non-specific behavioral influences (Supplementary Fig. 24). Together, these results suggest that NPY2R in fear-tagged ensembles and NPY1R in extinction-tagged ensembles exert opposing influences on extinction: one prolonging fear and the other facilitating extinction. To directly probe the roles of NPY1R or NPY2R in vCA1, we locally delivered receptor-specific agonists or antagonists via an implanted cannula immediately prior to extinction learning (Fig. 5f). Strikingly, pharmacological activation of NPY2R led to an overly rapid extinction in Stage 1 with attenuated Stage 2, whereas NPY1R agonism slightly delayed Stage 1 extinction and stabilized freezing at a higher level (Fig. 5g, h and Supplementary Fig. 25a–c). Conversely, antagonizing NPY1R accelerated extinction, whereas NPY2R antagonism hindered it (Fig. 5i, j and Supplementary Fig. 25d–f), consistent with the results from the genetic deletion experiments and electrophysiology analyses of neuronal excitability and synaptic transmission in these mice (Extended Data Fig. 9 and 10). Pharmacological manipulation of NPY5R in the vCA1 had no discernible effect on fear extinction (Supplementary Fig. 26). These findings underscore distinct yet complementary actions of NPY1R and NPY2R in the fear extinction process and support the idea that NPY interneuron co-opts its roles in fear memory and extinction by activity-dependent release of NPY itself to sequentially act upon Npy2r- and Npy1r -expressing neurons, thereby permitting the transition from fear to extinction sub-ensembles. Specifically, NPY2R sub-ensemble engages in Stage 1 extinction, gating memory lability, while NPY1R functions in Stage 2 to determine the setpoint of memory stability during extinction Supplementary Fig. 27). Discussion Memories persist or update according to how their constituent ensembles are recruited and reweighted over time. Activity-dependent tagging has established that sparsely distributed ensembles are re-engaged at recall 1, 42 , yet this IEG-based framework (e.g., Fos, Arc) offers limited insight into how inhibitory motifs and neuromodulators bias an engram toward lability or stability. Here we identify a peptidergic control axis in vCA1: NPY + interneurons co-opt receptor-defined neuronal subpopulations to tilt a single memory trace from fear expression toward extinction. In this scheme, fast GABAergic output from NPY + cells facilitates acquisition, whereas slower NPY release bifurcates onto two largely non-overlapping targets—NPY2R + and NPY1R + neurons—thereby rebalancing ensemble dominance across time. This linkage between real-time peptide dynamics and cell-type specificity provides a molecular logic for how inhibitory interneurons determine which neuronal ensembles govern behavior. A central feature of this logic is temporal staging, but we do not view the late NPY signal to mediate a second phase of freezing. During extinction, NPY + interneuron activity and peptide release rise in step with transitions from “fear-on” to “fear-off”, consistent with a rapid, cue-locked component that modulates behavior on the timescale of seconds. By contrast, the elevation of extracellular NPY that persists for ~2 h after training is better viewed as a slower, tonic molecular state—coincident with increased Npy transcription—than as continued moment-to-moment control of freezing. We therefore regard the prolonged signal as a consolidation-related aftereffect that likely supports synaptic and network plasticity (for example, by sustaining peptide availability, influencing receptor occupancy/trafficking, or altering interneuron-pyramidal neuron coupling) and prepares the circuit for future encounters with the cue, rather than as a direct driver of rapid behavioral transitions. In this light, our “two-component” description is intended to distinguish a fast, phasic NPY pulse that updates behavior online from a slower, metaplastic NPY tone that biases subsequent retrieval and limits relapse. Because our single-cell profiling was sampled 2 h after training, it most plausibly captures this consolidation-associated state; higher-temporal-resolution tagging will be required to resolve the precise sequence of neuronal recruitment. Taken together, we propose a model in which a brief NPY surge loosens fear-biased representations during learning, while a lingering NPY milieu stabilizes the extinction bias for future expression. TRAP-based analyses anchor our model to receptor identity within engrams. Fear‑tagged neuronal ensembles were enriched for Npy2r , whereas extinction‑tagged ensembles were enriched for Npy1r . Causal manipulations aligned with this segregation: genetic or pharmacological disruption of NPY2R selectively blunted the early, labile phase of extinction, while engaging NPY1R set the late, stable asymptote reached after training. Thus, the two behavioral phases map onto receptor‑defined subpopulations embedded within the tagged ensembles. More broadly, these findings argue that engrams are not specified solely by which neurons fire together, but also by a modulator-receptor scaffold that biases the eligibility of specific neuronal sub-ensembles to carry the trace at different stages. In this view, IEG-based tagging identifies the engram membership, while peptidergic inhibition specifies the timing and strength of memory trace—steering the same network toward lability (NPY→NPY2R) or stability (NPY→NPY1R). The observed shift from Npy2r - to Npy1r -dominated ensembles across extinction is therefore mechanistic rather than merely descriptive: it shows how a single interneuron class can orchestrate a reversible transformation of memory state. The circuit architecture and biophysics of signaling help explain this progression. NPY is produced locally by SO interneurons in vCA1, whereas NPY2R- and NPY1R-expressing targets are stratified across deep versus superficial/middle pyramidal layers, with only rare co-expression. Protein distributions are broader than soma-localized mRNA—consistent with receptor trafficking—and NPY acts by volume transmission, diffusing from its release sites to receptor fields at different radii. Superimposed on this geometry is receptor pharmacology 43, 44 : the higher affinity of NPY2R predicts preferential engagement when peptide levels are low, with NPY1R increasingly recruited as NPY accumulates. Together, these factors naturally yield a stage-dependent receptor switch during extinction. Perturbations of the interneuron source further support this logic. Optogenetic activation of NPY + interneurons in the low-activity No Ext state increased the fraction of Npy1r -expressing cells and reduced Npy2r -expressing cells within the tagged population, facilitating extinction. Conversely, inhibiting NPY + interneurons during Ext had the reciprocal effect—reducing Npy1r and increasing Npy2r —and attenuated extinction. Thus, extinction involves not only changes in engram size but, more critically, a reweighting of receptor-defined sub-ensembles from Npy2r to Npy1r . Functionally, we propose that NPY2R + neurons act as a lability gate: early, modest NPY release preferentially engages this high-affinity population to loosen fear-biased representations and lower the threshold for updating. As extinction proceeds and peptide levels build, recruitment of NPY1R + neurons stabilizes the extinction-biased set point, limiting over-generalization and conferring persistence. This receptor-stage choreography reframes extinction as more than global inhibition; it is a peptidergic switch that rebalances competing ensembles across time—first loosening the old trace, then fastening the new one—consistent with the dissociation we observe between early rate changes and late asymptotic performance. These principles suggest several general interpretations. First , interneuron‑derived neuropeptides can endow inhibitory circuits with time‑scale diversity, coupling millisecond GABA to minutes‑to‑hours peptide action to bridge online control and consolidation. Second , receptor heterogeneity partitions neurons into functionally addressable sub‑ensembles, allowing a single interneuron class to steer the same trace along distinct trajectories. Third , ensemble “identity” is not fixed; it is re‑weighted by the momentary neurochemical milieu, providing a mechanistic substrate for the coexistence of fear and extinction memories within shared circuitry. The model yields testable predictions and future directions. Temporally gated activity-dependent tagging should reveal a leading-edge recruitment of NPY2R + cells that gives way to NPY1R + cells across extinction trials. Closed-loop manipulations that boost NPY specifically during “fear-on” epochs should accelerate the NPY2R-dominated lability phase, whereas late-epoch NPY1R augmentation should preferentially improve retention. Quantifying NPY diffusion and receptor occupancy in vivo —together with longitudinal single-cell transcriptomics and spatial proteomics—will clarify how peptide kinetics, receptor trafficking, and laminar architecture shape ensemble reweighting. Beyond vCA1, analogous peptidergic switches may generalize to other forms of adaptive learning and to other interneuron peptides (SST, CCK) 13-15, 38 , raising the prospect of a broader neuropeptidergic grammar for engram control. NPY is broadly acting cotransmitter 45-47 that signals through Gi/o‑coupled receptors 48 . Within vCA1, our data define a region‑ and cell‑type‑specific mechanism by which NPY dynamically tunes memory: peptide released from NPY + interneurons sequentially engages largely non‑overlapping NPY2R + and NPY1R + neuronal sub‑ensembles, biasing the same trace toward lability or stability. This spatiotemporal control layer complements activity‑dependent engram tagging by linking peptide dynamics to ensemble selection, providing a mechanistic bridge from interneuron signaling to behavioral adaptation and aligning with the therapeutic promise of NPY pathways in stress‑ and trauma‑related states 49 . The stage specificity of this control suggests clear translational avenues. Brief, well-timed enhancement of NPY→NPY2R signaling should transiently increase engram lability and thereby potentiate exposure-based therapy, whereas the late NPY→NPY1R arm constrains extinction by stabilizing a conservative set point—implying that timed NPY1R antagonism, rather than engagement, would be expected to deepen extinction and reduce relapse. Any such strategy will demand precise control of timing, circuit targeting, and receptor selectivity, together with rigorous cross-species validation. Even so, aligning intervention windows with the two-component NPY dynamics as we described—an early, NPY2R-dominated phase that loosens fear-biased representations and a later, NPY1R-dominated phase that caps further plasticity—offers a concrete framework for future preclinical intervention. In conclusion, NPY + interneurons in vCA1 do not merely suppress activity; they co‑opt receptor‑defined neuronal sub-ensembles to govern memory liability and stability. By tiling a fast modulatory pulse to a slower, gene‑dependent wave, the NPY system provides a molecular continuum for gating memory fate. Methods Mice All animal procedures were ethically approved by the Animal Ethics Committee of Shanghai Jiao Tong University School of Medicine and the Institutional Animal Care and Use Committee (Department of Laboratory Animal Science, Shanghai Jiao Tong University School of Medicine; Policy Number DLAS-MP-ANIM. 01–05). Mice were housed in groups under a 12-h light/dark cycle with unrestricted access to food and water. Adult male mice (7–12 weeks old, C57BL/6J background) were selected for all experiments. The study utilized the following mouse strains: NPY-Cre ( Npy -IRES-Cre, JAX, catalog number 027851), NPY2R-Cre ( Npy2r -ires-Cre, JAX, catalog number 029285), Fos2A-iCreER (TRAP2, JAX, catalog number 030323), Rosa26-Ai9 tdTomato reporter (JAX, catalog number 007909), and Npy -hrGFP (JAX, catalog number 006417), and CRISPR/Cas9 knock-in (JAX, catalog number 024857) mice. Npy knockout ( Npy –/– ), heterozygous ( Npy +/– ), and wild-type ( Npy +/+ ) littermates were obtained by intercrossing Npy +/– mice (GemPharmatech). C57BL/6J mice were purchased from the Shanghai Laboratory Animal Center (SLAC), Chinese Academy of Sciences. Fear conditioning, extinction, and extinction retrieval All auditory fear conditioning and extinction procedures were conducted using the Ugo Basile Fear Conditioning System (UGO Basile S.R.L., Italy) according to established protocols 24 . Initially, mice were acclimated and habituated to the conditioning chamber for three consecutive days. The conditioning chambers (17 × 17 × 25 cm), equipped with stainless-steel shocking grids, were linked to a precision feedback current-regulated shocker (UGO Basile S.R.L., Italy). During habituation and fear conditioning, the chamber walls were adorned with black-and-white checkered wallpapers (context A) and cleaned with 75% ethanol. On Day 1, individual mice underwent conditioning in context A, with five pure tones (conditioned stimulus, CS; 4 kHz, 76 dB, 20 s each) presented at fixed intervals (140 s), each paired with a foot shock (unconditioned stimulus, US; 0.5 mA, 2 s). Auditory tones and foot shocks were autonomously controlled by the ANY-maze software (version 7.20, Stoelting Co.). Following conditioning, mice were returned to their home cages 60 s after the final tone, with individual cage floors and walls sanitized with 75% ethanol. On Day 2 (24 h post-conditioning), the no-extinction training group (No Ext) received two CS presentations (4 kHz, 76 dB, 30 s each) within a 5-min period, while the extinction training group (Ext) received 20 CS presentations within a 25-min period, without foot shocks. Both No Ext and Ext tests were conducted in a chamber with a gray, non-shocking plexiglass floor and dark gray wallpaper (context B), and the testing environment was cleaned with a 4% acetic acid solution between individual mouse tests. To minimize CS anticipation during extinction, a 30-s tone duration (different from the 20-s conditioning duration) was used. To control for time differences between groups No Ext and Ext groups, we designed three variants of the No Ext conditions (1) a short exposure group receiving two CS presentations immediately before removal from context B, (2) a group receiving two CS presentations followed by an extended waiting period equivalent to the remaining 18 CS presentations in the Ext group, but without additional tone exposure, and (3) a group that remained in context B for a period to 18 additional CS presentations (without tone), followed by two CS presentations before removal. On Day 3, extinction retrieval (Ext Retr) was performed, where mice underwent eight CS presentations in context B. Both No Ext and Ext conditions, as well as extinction retrieval tests, combined two CS presentations into a single block. Throughout testing, the chamber was placed in a sound-attenuating enclosure with a ventilation fan and a house light (UGO Basile S.R.L., Italy). Mouse locomotion within the chamber was recorded using a near-infrared camera and analyzed in real-time by ANY-maze software. The freezing score, a dimensionless metric generated by the software, identified freezing periods based on predefined threshold settings. A fear response was characterized by measurable behavioral freezing, defined as a cessation of movement lasting more than 2 s. The duration of freezing was quantified for analysis. For conditioned fear response analysis, the total duration of freezing during each tone (CS) presentation was divided by the sound exposure time (30 s) and multiplied by 100%. For contextual fear response analysis, the total fear duration in the specific context was divided by the contextual exposure time and multiplied by 100%. Optogenetic manipulations A wired optogenetic system was used to modulate neuronal activity during behavioral assays, utilizing a light-emitting diode (LED) (Hangzhou Newdoon Technology Co. Ltd) connected to an optic patch cord via connectors at both ends. Blue light (470 nm, 4–5mW) was delivered in 10-ms pulses at 20 Hz during extinction training. For optogenetic inhibition, red light (638 nm, 8–10 mW) was consistently administered. The light pulse duration was set to exceed 5 s before and after the CS presentations to ensure full CS exposure during extinction training (30-s duration for each tone). To activate NPY + neurons during freezing response to CS presentations during extinction, the ANY-maze software detected a fear response and automatically trigger blue light (470 nm, 4–5mW). The blue light was turned off once the fear response ceased. The final output power varied based on the light transmission efficacy of the optical fiber used. Generation of single-cell suspensions Two hours after undergoing fear memory retrieval (No Ext) or extinction learning (Ext), the mice were deeply anesthetized with 1% sodium pentobarbital and then decapitated. Brains were quickly dissected and chilled in ice-cold N-methyl-D-glucamine-artificial cerebrospinal fluid (NMDG-ACSF) containing the following components (in mM): 93 N-methyl-D-glucamine, 2.5 KCl, 1.2 NaH 2 PO 4 , 30 NaHCO 3 , 20 HEPES, 25 D-glucose, 5 Sodium ascorbate, 2 Thiourea, 3 Sodium pyruvate, 10 MgSO 4 , 11.9 N-Acetyl-L-cysteine, 1 CaCl 2 , and 6 ml HCl (pH 7.35–7.45). Coronal brain slices (400 μm thick) containing regions of the vHPC were cut using a vibratome (Leica VT1000S, Germany), then promptly transferred to ice-cold NMDG-ACSF and incubated for 15 min. Actinomycin D (8 μM; A1410, Sigma-Aldrich) and Triptolide (10 μM; T3652, Sigma-Aldrich) were added in this step to minimize artificially induced activation of early immediate genes. The region of vHPC were carefully microdissected using fine tweezers on ice. After tissue collection, it was transferred to an Eppendorf tube containing a digestion solution. This solution consisted of ACSF supplemented with 1 mg/ml pronase (P6911, Sigma-Aldrich), Actinomycin D (8 μM), Triptolide (10 μM), TTX (1 μM) and D-APV (100 μM; A8054, Sigma-Aldrich), and the tissue was digested for 20 min at 34 °C. The tissue was pipetted periodically every 10 min during this digestion. Then, the cell suspension was transferred to Sterile Earle’s Balanced Salt (EBSS) Solution (LK003150, JinpanWorthington) supplemented with Actinomycin D (8 μM), Triptolide (10 μM), TTX (1 μM) and D-APV (100 μM). Finally, it was washed with washing solution consisting of ACSF supplemented with Actinomycin D (8 μM), Triptolide (10 μM), TTX (1 μM, C239, Sigma-Aldrich) and D-APV (100 μM). All solutions were continually bubbled with O 2 and CO 2 (95% O 2 /5% CO 2 , v/v). Single-cell library preparation and sequencing Single cell RNA-seq library were prepared from the single cell suspensions via the chromium single cell gene expression system (Chromium Single Cell 3’ Reagent Kits v2, 10× genomics), using the default protocols provided by 10× genomics. The NovaSeq platforms were used for sequencing libraries. Genome alignment and Seurat analysis Fastq files were aligned to the mouse reference genome mm10 and converted into gene expression matrices using the 10× Genomics Cell Ranger software (v3.0.0). Low-quality cells were filtered out, including those with fewer than 700 transcripts or 200 genes, more than 15,000 transcripts or 2,500 genes, or greater than 5% mitochondrial expression. Genes expressed in fewer than three cells were also excluded. To visualize the data, dimensionality reduction and clustering analysis were performed on the raw count matrices using the Seurat (v4.3.0) package in R (v4.1.3) 50 . The FindIntegrationAnchors and IntegrateData functions were used to integrate the No Ext and Ext datasets. Data were reduced using principal component analysis (PCA). Clusters were identified using the FindNeighbors and FindClusters functions with the resolution of 0.5, resulting in 35 cell clusters. Marker genes for each cluster were identified using the FindAllMarkers function. The following cell types were annotated using selected marker genes listed below: excitatory neurons ( Thy1 , Slc17a7 ), inhibitory neurons ( Slc32a1 , Gad1 ), astrocyte ( Gfap ), oligodendrocyte ( Olig1 ), oligodendrocyte precursor cells ( Pdgfra ), microglia ( Cx3cr1 ), pericytes ( Vtn ), endothelium ( Cldn5 ), and macrophages ( Mrc1 ). Uniform Manifold Approximation and Projection (UMAP) dimensional reduction was applied to visualize the cell classifications, resulting in 9 cell types. To explore the heterogeneity of inhibitory neurons, we extracted all inhibitory neurons and performed clustering with FindNeighbors and FindClusters functions at a resolution of 0.1, yielding 5 subclusters characterized by the expression of specific markers ( Slc32a1 , Gad1 , Gad2 , Cck , Vip , Npy , Sst , and Calb2 ). To normalize cell numbers between two conditions across all cell types and clusters, we downsampled the larger cell type or cluster to match the size of the smaller one. To reduce stochastic sampling error, we performed 1000 resampling iterations to obtain robust results. Differential expression analysis between the No Ext and Ext was carried out using the FindMarkers function with the method of MAST 51 . When identifying differentially expressed genes (DEGs) across different cell types and clusters, we used a threshold of |log 2 FC| ≥ 0.585, FDR < 0.05 (Benjamini–Hochberg correction) with ≥ 90% resampling support. For specific subpopulations, we selected cells that expressed a particular gene above a threshold value of zero. Npy1r - and Npy2r -expressing neurons were classified based on this criterion. Specifically, we observed that only a very small fraction of neurons coexpressed Npy1r and Npy2r in both excitatory and inhibitory neurons. For differential expression analysis, we therefore compared the Npy1r -specific and Npy2r -specific subpopulations after excluding coexpressing neurons. To account for differences in population size, subpopulations were balanced by resampling. Differentially expressed genes were defined using thresholds of |log2FC| ≥ 0.25 and FDR < 0.05 (Benjamini–Hochberg correction), and only genes supported in ≥ 90% of resampling iterations were retained. A curated list of 139 IEGs was selected based on previous literature 33 , which includes genes known to be activated in response to various stimuli, including stress, fear, and learning. These genes were identified in different experimental conditions where they were shown to be reliably upregulated upon neuronal activation. To identify activated neurons (IEG + ), we calculated the 90 th percentile expression for each of the selected IEGs based on the integrated gene expression data from both No Ext and Ext conditions. A cell was considered activated if its expression of any of these IEGs exceeded the 90 th percentile value, which indicates a higher level of gene activation compared to the majority of cells. This threshold was chosen to identify cells that exhibited significant transcriptional changes associated with neuronal activation, as supported by the literature 35 . This method of identifying activated neurons provides a robust approach to capture a broad range of neuronal responses during key processes such as fear conditioning and extinction, ensuring that our findings accurately reflect the activation of relevant neuronal populations. Virus constructs The following viruses were utilized: rAAV-EF1α-DIO-NpHR3.0-EGFP (AG26966), rAAV-EF1α-DIO-ChR2-mCherry (AG20297), rAAV-EF1α-DIO-ChR2-EGFP (M3026), rAAV-EF1α-DIO-EGFP (H3303), rAAV-EF1α-DIO-mCherry (AG20299) were procured from Obio Technology Co., Ltd. (Shanghai). rAAV-EF1α-DIO-ChR2-E123T/T159C-mCherry (PT-0015), rAAV-EF1α-DIO-GCaMp6m (PT-0283), rAAV-hSyn-DIO-ChrimsonR-mCherry (PT-1374), and rAAV-hSyn-DIO-hM4D(Gi)-mCherry (PT-0020) were obtained from Brain VTA (Wuhan). rAAV-hSyn-NPY1.0 (YL011001), rAAV-hSyn-NPYmut (YL011003), and rAAV-hSyn-GABA0.8 (YL022001) were purchased from WZ Biosciences Inc (Jinan). rAAV2/9-3x(hU6-sgRNA.sp)(m Npy1r )-hEF1a-DIO-EGFP (WY4060), rAAV2/9-3x(hU6-sgRNA.sp)(m Npy2r )-hEF1a-DIO-EGFP (WY3769), rAAV2/9-hU6-sgRNA-hEF1a-DIO-EGFP (S0883), rAAV2/9-H1-sgRNA.sp(m Slc32a1 )x3-CAG-DIO-mCherry (WY3889), rAAV2/9-H1-sgRNA.sp(m Npy )x3-CAG-DIO-mCherry (WY3909), rAAV2/9-H1-sgRNA.sp(NC)-CAG-DIO-mCherry (S0804) and rAAV2/9-CAG-DIO-taCaspase3 (S0236) were generated by Shanghai Taitool Bioscience Co., Ltd. (Shanghai). All viruses are serotype 2/9. All viral vectors were stored in aliquots at –80°C until further use, with viral titers for injection exceeding 10 12 viral particles per ml. Surgery Mice, aged 7–8 weeks, were anesthetized with 1.5–2.0% isoflurane (R510-22, RWD Life Science) in 100% oxygen and subsequently stabilized in a stereotaxic apparatus using atraumatic ear bars (RWD Life Science), ensuring precise alignment on a digital stereotactic frame. A midline scalp incision was made, and small bilateral craniotomies were created using a microdrill with 0.5-mm burrs. Glass pipettes (tip diameter: 10–20 μm) were crafted using a P-1000 Micropipette Puller (Sutter glass pipettes, Sutter Instrument Company) for AAV microinjections. Initially filled with silicone oil, the microinjection pipettes were connected to a microinjector pump (RWD Life Science) to ensure complete air exclusion. AAV-containing solutions were loaded into the pipette tips and injected at specified coordinates (in mm): vCA1 (anteroposterior to bregma, AP, −3.14; lateral to the midline, ML, ±3.20; below the bregma, DV, −4.00) and ARC (AP, −1.70; ML, ±0.30; DV, −5.80). Virus-containing solutions were injected bilaterally into the vCA1 (0.4 μl/side) and unilaterally into the ARC (0.3 μl/side) at a rate of 0.1 μl/min. After injection, the pipette remained in place for an additional 10 min to allow for adequate diffusion of the injectant. Mice were given a minimum of two weeks to recover before undergoing behavioral and other tests, and injection sites were examined at the experiment’s conclusion by assessing the expression of fluorescent proteins, such as EGFP or mCherry. In optogenetic experiments, ceramic fiber optic cannulas [200 μm in diameter, 0.37 numerical aperture (NA), Hangzhou Newdoon Technology] were surgically implanted above the vCA1 (coordinates: AP, –3.14; ML, ±3.20; DV, –3.90) and ARC (coordinates: AP, –1.70; ML, ±0.30; DV, –5.70). These cannulas were secured in place using acrylic dental cement and skull screws. Open field test Mice were acclimated to the experimental room from their home cages and given a minimum of 1 h for habituation before the start of the experiment. Individual mice were then placed in the outer area of a square Plexiglas open-field apparatus (40 × 40 × 35 cm), divided into a central field (20 × 20 cm) and an outer field. The total distance traveled was quantified using Noldus EthoVision XT (version 16.0, Noldus Information Technology, Netherlands). NPY-Cre mice underwent injections of DIO-ChR2, -NpHR, or -Control virus to modulate the activity of NPY + neurons, with customized light stimulation protocols for either activation or inhibition of these neurons. The open-field test consisted of an 18-min session divided into six 3-min epochs, alternating between light off and light om periods, starting with a light-off epoch. For analysis and charts representing only “off” and “on” conditions, the three “off” epochs and three “on” epochs were combined, respectively. To ensure consistency, the open-field arena was thoroughly cleaned with 70% ethanol between each set of tests. Fiber photometry Fiber photometry experiments were performed using a system from Thinker Tech. Fluorescent signals, generated by a 488-nm laser (OBIS 488LS; Coherent), were reflected by a dichroic mirror (MD498; Thorlabs), focused by a 10× objective lens (NA = 0.3; Olympus), and coupled to a rotary joint (FRJ_1 × 1_FC-FC, Doric Lenses). An optical fiber (200 mm O.D., NA = 0.37) was implanted into the vCA1 or ARC during virus injection and remained in place for two weeks for virus expression. Laser power at the fiber tip was adjusted to 40–50 μW to minimize bleaching of GCaMP6m, NPY1.0/NPYmut or GABA0.8 sensor. Excitation fluorescence was collected by the same multi-mode optical fiber and converted into electrical signals by low-light detectors at the detection end to capture neural activity information. Signals were digitized at 100 Hz using a Power 1401 digitizer and Spike 2 software (CED, Cambridge). Throughout behavioral tests, including fear learning, extinction training, or extinction retrieval, fluorescent intensities of GCaMP6m or sensors were recorded, with the pre-sound signal serving as the baseline. Average Ca 2+ or sensor responses were computed using MATLAB, and data were exported to MATLAB mat files from Spike2 for further analysis. Permutational tests were employed for statistical significance, and ΔF/F values were visualized as heatmaps or per-event plots, with shaded areas indicating the standard error of the mean (s.e.m.). The averaged ΔF/F during CS presentation and freezing/no-freezing states was defined, and the area under the curve (AUC, ΔF/F × s) was calculated for the same dataset. Cannula implantation and local drug injection Following anesthesia, mice were securely fixed on a stereotaxic apparatus (RWD Life Science). Stainless steel guide cannulas (RWD Life Science) were bilaterally implanted into the vCA1, with the cannula tips positioned at the following coordinates (in mm): AP, –3.14; ML, ±3.20; DV, –3.90. The cannulas were firmly attached to the skull using acrylic cement and two skull screws. Stainless steel obturators (33 gauges) were inserted into the guide cannulas to prevent obstruction until drug infusion. Animals were allowed a 2-week recovery period post-surgery before undergoing behavioral tests. Mice were familiarized with the infusion procedure three days prior to drug injection. During drug infusion, mice were briefly head-restrained, the stainless-steel obturators were removed, and injection cannulas (33 gauges, RWD Life Science) were inserted into the guide cannulas. The injection cannulas protruded 0.50 mm from the guide cannula tips. The infusion cannula was connected via PE20 tubing to a microsyringe driven by a microinfusion pump (KDS 310, KD Scientific). A total of 0.5 μl of drugs were bilaterally infused into the vCA1 at a controlled flow rate of 0.1 μl per minute. After completing the drug injection, the injection cannulas were left in place for 2 min to facilitate solution diffusion from the cannula tips. Subsequently, the stainless-steel obturators were reinserted into the guide cannulas, and the mice were returned to their home cages for a 15-min recovery period before behavioral tests. Drugs and concentrations Neuropeptide Y (1153, Tocris Bioscience). NPY1R agonist, [Leu31, Pro34]-Neuropeptide Y (porcine) (HY-P0208, MedChemExpress). NPY2R agonist, N-Acetyl- [Leu 28 , Leu 31 ]-Neuropeptide Y Fragment 24-36 (N9398, Sigma). NPY5R agonist, [cPP1-7, NPY19-23, Ala31, Aib32, Gln34]-hPancreatic Polypeptide (HY-P1324, MedChemExpress). The peptide and peptide agonists for NPY receptors were applied at a concentration of 0.25 mM for 0.5 μl. NPY1R antagonist, BIBO 3304 trifluoroacetate (2412, Tocris Bioscience) was applied at a concentration of 0.4 mM for 0.5 μl. NPY2R antagonist, BIIE 0246 hydrochloride (7377, Tocris Bioscience) was applied at a concentration of 2 mM for 0.5 μl. NPY5R antagonist, CGP 71683 hydrochloride (2199, Tocris Bioscience) was applied at a concentration of 5 mM for 0.5 μl. Engram labeling Activity-dependent recombination was triggered using 4-hydroxytamoxifen (4-OHT, H6278, Sigma-Aldrich). A 20 mg/ml solution of 4-OHT was prepared in ethanol by shaking at 37 °C for 30 min. Subsequently, twice the volume of corn oil (C8267, Sigma-Aldrich) was added to achieve a final concentration of 10 mg/ml of 4-OHT, and the ethanol was removed by vacuum under centrifugation. All injections were administered intraperitoneally. Mice were transferred from the vivarium to an adjacent holding room at least 3 h before the test. For fear and extinction memory, activity-dependent neuronal tagging was initiated by a single intraperitoneal injection of 4-OHT (20 mg/kg for mice) after exposure to two CS or 20 CS, respectively. Subsequently, mice were returned to the vivarium, maintaining a regular 12-h light/dark cycle for the remainder of the experiment. Slice electrophysiology Whole-cell recordings were conducted in acute brain slices obtained from behaviorally trained mice. Mice were deeply anesthetized with 1% sodium pentobarbital and subsequently decapitated. Brains were quickly dissected and chilled in ice-cold N-methyl-D-glucamine-artificial cerebrospinal fluid (NMDG-ACSF) containing the following components (in mM): 93 N-methyl-D-glucamine (NMDG), 2.5 KCl, 1.2 NaH 2 PO 4 , 30 NaHCO 3 , 20 HEPES, 25 D-glucose, 5 Sodium ascorbate, 2 Thiourea, 3 Sodium pyruvate, 10 MgSO 4 , 11.9 N-Acetyl-L-cysteine, 1 CaCl 2 , and 6 ml HCl (pH 7.35–7.45). Coronal brain slices (300 μm thick) containing regions of the vCA1 were cut with a vibratome (Leica VT1000S, Germany). Slices were initially recovered for 5–10 min in NMDG-ACSF at 31 °C. They were then transferred to another ACSF solution (in mM): 125 NaCl, 2.5 KCl, 12.5 D-glucose, 1 MgCl 2 , 2 CaCl 2 , 1.25 NaH 2 PO 4 , and 25 NaHCO 3 (pH 7.35–7.45) for an additional 2 hours of recovery at 31 °C before recordings. The slice was subsequently transferred to a recording chamber and continuously superfused with ACSF at a rate of 1–2 ml per minute. All solutions were continually bubbled with O 2 and CO 2 (95% O 2 /5% CO 2 , v/v). Neurons in the vCA1 were patched under visual guidance using infrared differential-interference contrast microscopy (BX51WI, Olympus), and an optiMOS camera (QImaging, Teledyne Imaging Group). Whole-cell patch-clamp recordings were performed using an Axon 200B amplifier (Molecular Devices). Membrane currents and potentials were sampled and analyzed using a Digidata 1440 interface and a personal computer running Clampex and Clampfit software (pCLAMP 10.5, Molecular Devices). Access resistance was maintained between 10–20 MΩ, and only cells with a change in access resistance < 20% were included in the analysis. In specific recording situations, NPY (1 μM) was added to the ACSF to activate NPY receptors in the vCA1. Spike firing. The spiking activity and membranous properties of cell populations in the vCA1 were quantified using an internal solution comprising the following concentrations (in mM): 145 potassium gluconate, 5 NaCl, 10 HEPES, 2 MgATP, 0.1 Na 2 GTP, 0.2 EGTA, and 1 MgCl 2 (280–300 mOsm, pH 7.2 with KOH). Subsequent data analysis was performed using the MiniAnalysis Program (Version 6.0.1, Synaptosoft) with an amplitude threshold set at 40 mV. Spontaneous excitatory postsynaptic currents (sEPSCs) . For recording sEPSCs in cell populations within the vCA1, a holding potential of –70 mV was maintained. Patch pipettes were filled with a Cs + -based solution containing the following concentrations (in mM): 132.5 Cs-gluconate, 17.5 CsCl, 2 MgCl 2 , 0.5 EGTA, 10 HEPES, 2 Na 2 ATP, with pH adjusted to 7.3 using CsOH, and osmolarity set at 280–290 mOsm. sEPSCs were recorded for 5-10 min and analyzed from 300 to 400 s after the establishment and stabilization of the recording. Data analysis was conducted using the Mini-analysis Program (Synaptosoft) with an amplitude threshold set at 10 pA, while other parameters remained at their default values. Optical stimulation response. Optical stimulation of ChR2- or NpHR-expressing neurons was conducted using a collimated LED (Lumen Dynamics Group Inc) with peak wavelengths of 473 or 638 nm, respectively. The LED was connected to an Axon 200B amplifier to initiate photostimulation. The brain slice in the recording chamber was illuminated through a 40× water-immersion objective lens (LUMPLFLN 40XW, Olympus). The intensity of photostimulation was directly regulated by the stimulator (2–18 mW/mm 2 ), while the duration was set through Digidata 1440 and pClamp 10.5 software. The functional potency of the ChR2-expressing virus was validated by measuring the numbers of action potentials (APs) elicited at different frequencies of blue-light stimulation (1 ms; 5, 10, and 20 Hz) and the inward photocurrents (500-ms pulse) mediated by ChR2 in brain slices. To validate the functional efficacy of NpHR-mediated optogenetic inhibition, red light (638 nm, 500-ms pulse) was administered to attenuate spikes under current clamp mode and induce the outward photocurrents (500-ms pulse) mediated by NpHR. CRISPR/Cas9-mediated gene knockout The CRISPR-associated endonuclease Cas9 from Streptococcus pyogenes (SpCas9) was employed to induce indels in the Npy (NC_109648), Slc32a1 (NC_22348), Npy1r (NC_18166), and Npy2r (NC_18167) genes. Corresponding single guide RNAs (sgRNAs) for each gene were designed and produced by Taitool Bioscience Co. Ltd. using the GRCM38 (Mus musculus) reference genome. Ten sgRNAs were selected for each gene based on computed specificity and efficiency scores (CHOPCHOP). Subsequently, these sgRNAs were cloned into a 2-in-1 reporter vector (CRPT001, Taitool), individually expressing a single sgRNA along with a cut-activated EGFP gene. The sgRNA target sequence was positioned ahead of the EGFP ORF within the vector. The reporter vectors were co-transfected with a Cas9-expressing plasmid into 293T cells. Functional Cas9/sgRNA complexes would cut the target on the reporter vector and rescue the expression of EGFP. Subsequently, the top three sgRNAs were selected and cloned into an adeno-associated virus (AAV) vector tandemly. The chosen sgRNAs were as follows: 5’-CGTATGCACCCACCTCTGTC-3’, 5’-AACAAGCGAATGGGGCTGTG-3’, and 5’-CAGAAAACGCCCCCAGAACA-3’ for the Npy gene targeting exon 2 of the reverse strand, exon 2 of the forward strand, and exon 3 of the forward strand of the Npy coding region, respectively. 5’-GCCACCGATGAGGAAGCGGT-3’, GCGCGTGCGGGACTCGTATG and 5’-GCACGCGATGAGGATCTTGC-3’ for Slc32a1 gene to target exon 2 of the forward strand, exon 3 of the forward strand and exon 3 of the reverse strand of the Slc32a1 coding region. 5’-GATGGAGACACACTGTACAA, GATGGACCACTGGGTCTTCG-3’, and 5’-GGCCGAAATACTGCAGCACC-3’ for the Npy1r gene targeting exon 2 of the reverse strand, exon 2 of the forward strand, and exon 2 of the reverse strand of the Npy1r coding region, respectively. 5’-GTTGGATGCCATTCACTCGG-3’, 5’-GAATCCAGGAATTTCAAGCA-3’, and 5’-GAGAAAGATATGATGCCCAG-3’ for the Npy2r gene targeting exon 2 of the forward strand, exon 1 of the reverse strand, and exon 2 of the reverse strand, respectively (according to transcript variant 1). A non-targeting negative control guide was designed with a scrambled sequence with no known targets in the genome (5’-GCTGAGTACTTCGAAATGTC-3’). sgRNAs for Npy and Slc32a1 was cloned into the pAAV2-H1-sgRNA.sp(MCS)x3-CAG-DIO-mCherry-WPRE-pA vector (AC610-C, Taitool), while sgRNAs for Npy1r or Npy2r were cloned into the pAAV-(U6-sgRNA.sp(MCS)) x 3-hEF1a-DIO-EGFP-WPRE-pA vector (AC705-C, Taitool). These plasmids were co-transfected into HEK293 cells with AAV9 Cap. Rep and adenovirus helper to produce recombinant AAV vectors. Finally, AAV vectors were administered via stereotactic injection. In situ hybridization Fluorescent in situ hybridization was conducted using the ACDBio V2 RNAScope kit (Advanced Cell Diagnostics). The following products were employed: RNAscope Multiplex Fluorescent Detection Reagent V2 (323110), Pretreatment Reagents (322381 and 322000), RNAscope Wash Buffer Reagents (310091), probes for Npy (313321-C1), Npy1r (427021-C3), Npy2r (315951-C2), Slc32a1 (319191-C3), Slc17a7 (416631-C2), Fos (316921-C1, 316921-C2, 316921-C3) and tdTomato (317041-C4), along with the PerkinElmer TSA Plus Fluorescence Palette Kit (NEL760001KT). Mice were deeply anesthetized with 1% sodium pentobarbital and transcardially perfused with ice-cold phosphate-buffered saline (PBS) before being fixed in 4% paraformaldehyde for 24 hours at 4 °C. After dehydration, brain tissue was embedded in optimal cutting temperature (OCT) and frozen at –80 °C. Coronal brain slices encompassing the entire vCA1 were sectioned at a thickness of 16 μm using a cryostat (Leica CM 1950). Slides were baked for 30 min at 60 °C. Subsequently, 5–8 drops of Hydrogen Peroxide were added to cover the entire section, followed by an incubation for 10 min at room temperature (RT). The slides were then washed three times in distilled water. The tissues were brought to 1X Target Retrieval Reagent and maintained for 5 min at 99–100 °C. Following this, the slides were washed three times in distilled water. They were then transferred to 100% alcohol for 3 min. Approximately 5 drops of Protease III were added to each section, and the samples were incubated for 30 min at 40 °C. Excess liquid was removed from the slides, and 4–6 drops of the appropriate probe mix were added to entirely cover each slide at 40 °C for two hours. To amplify the signals, tissues were sequentially immersed in AMP1, AMP2, and AMP3 at 40 °C. AMP1 and AMP2 were incubated for 30 min each, while AMP3 was incubated for 15 min. After each incubation, the appropriate HRP channel was chosen, and 4–6 drops of HRP-C1 or HRP-C2 or HRP-C3 or HRP-C4 were added to entirely cover each slide at 40 °C for 15 min, followed by incubation with HRP blocker for 15 min at 40 °C. For a specific HRP channel, corresponding diluted Opal™ Dye was added to each slide and incubated for 30 min at 40 °C. Npy , Npy1r, Npy2r , Slc32a1 , Slc17a7, Fos and tdTomato was labeled with Opal™ 520, Opal™ 570, Opal™ 620 and Opal™ 690. Between all steps, slides were washed three times in 1× RNAscope wash buffer for 2 min. Subsequently, the slides were incubated in ACDBio DAPI for 10 min, washed, dried for 20 min, cover-slipped with mounting media (0100-01, SouthernBiotech), and left to dry overnight before imaging. Confocal images of RNAscope-labeled slices were acquired using Nikon A1 (×40), FLUOVIEW FV3000 (×20, Olympus), or SLIDEVIEW VS200 (×20, Olympus) microscopes. Z-stacks were processed in Imaris (Bitplane, Oxford Instruments) to generate 3D reconstructions, and mRNA puncta were rendered using the Surface module. Neurons were segmented on the DAPI channel in ImageJ, and RNA signals were binarized based on fluorescence intensity for quantification. Histology and fluorescent immunostaining Mice were deeply anesthetized with sodium pentobarbital (1%, i.p.) and transcardially perfused with ice-cold PBS followed by 4% paraformaldehyde (PFA/PBS). Brains were post-fixed in 4% PFA at 4 °C overnight, and 45 µm coronal sections were cut using a vibratome (VT1000S, Leica). For NPY immunostaining, free-floating sections were blocked in 5% normal goat serum/0.3% Triton X-100/PBS for 1 h at room temperature (RT), then incubated with rabbit anti-NPY (1:300, 11976, Cell Signaling Technology) in blocking solution at 4 °C for 48 h. After PBS washes, sections were incubated with Alexa Fluor 568- or 488-conjugated donkey anti-rabbit IgG (1:500, A-21206 or A10042, Thermo Fisher Scientific) for 2 h at RT, counterstained with DAPI (1:2000), and mounted with glass coverslips using mounting media (0100-01, SouthernBiotech). For dual NPY1R/NPY2R detection in paraffin-embedded tissue, 5-µm sections underwent sequential deparaffinization in xylene (3 × 10 min), rehydration through graded ethanol, and antigen retrieval via microwave heating in citrate buffer (pH 6.0). Endogenous peroxidases were quenched with 3% H 2 O 2 (25 min, RT). After blocking with 10% donkey serum, sections were incubated with rabbit anti-NPY2R (1:200, ER1914-09, HuaBio) at 4 °C for 24 h, followed by HRP-conjugated goat anti-rabbit IgG (1:500, RT, 50 min, GB23303, Servicebio) and iF488-Tyramid (G1231, Servicebio) signal amplification (1:500, 10 min, RT). The same sections were then subjected to microwave-based epitope recovery and stained with rabbit anti-NPY1R (1:200, ER64221, HuaBio) using identical secondary/TSA procedures with CY3-Tyramide (G1223, Servicebio). All sections were coverslipped with mounting media (0100-01, SouthernBiotech) and imaged on the SLIDEVIEW VS200 (Olympus) with 20×/0.75 NA objective. Image analysis was performed using ImageJ software, where neurons were segmented based on the DAPI channel, and the NPY1R and NPY2R signals were binarized according to intensity. During paraffin embedding and sectioning, endogenous EGFP fluorescence was quenched; therefore, immunostaining was performed to recover the EGFP signal. Sections were incubated with chicken anti-GFP (1:500, A10262, Invitrogen) at 4 °C for 24 h, followed by HRP-conjugated donkey anti-chicken IgG (1:500, RT, 50 min; ECS010161, Jackson) and visualized with Alexa Fluor 488. All sections were coverslipped with mounting media (0100-01, SouthernBiotech) and were imaged with Nikon A1 (×40) or SLIDEVIEW VS200 (×20, Olympus) microscopes. Confocal z-stacks showing single optical planes and maximum projections. Z-stacks were processed in Imaris (Bitplane, Oxford Instruments) to generate 3D reconstructions, and protein puncta were visualized via Isosurface renderings using the Surface module. Neurons were segmented on the DAPI channel in ImageJ, and protein signals were binarized based on fluorescence intensity for quantification. Statistical analyses Statistical analyses were conducted using IBM SPSS Statistics 25, Origin 2022 Software, and Office 2019 (Microsoft). The graphs were generated using Origin Software. Data are presented as the mean ± s.e.m. unless indicated otherwise. Most histograms display individual data points representing the values and numbers of individual samples for each condition. Data distributions were tested for normality, and homogeneity of variance among groups was assessed using Levene’s test. Statistical comparisons were carried out using unpaired Student’s t -test, two-tailed paired Student’s t -test, as well as one-way analyses of variance (ANOVAs) or two-way repeated measures ANOVAs. For post hoc analysis, Bonferroni’s corrections for multiple comparisons were applied. A significance level of P < 0.05 was considered statistically significant. Significance is primarily denoted as * P < 0.05, ** P < 0.01, and *** P < 0.001. In some cases, it is indicated as # P < 0.05, ## P < 0.01, and ### P < 0.001 for multiple comparisons. NS, denotes non-significant values. Detailed information on statistical tests is provided in Source Data files. Declarations Data availability All data supporting the findings of this study are available within the paper and its Extended Data/Supplementary Information. The single-cell RNA-seq dataset generated here has been deposited in the Genome Sequence Archive (GSA) at the National Genomics Data Center (NGDC) under accession CRA023358 and is publicly accessible. Source data underlying all main and Extended Data/Supplementary Information figures—together with detailed statistics for each panel—are provided with the paper (Source Data files). Code availability Custom code used to perform the analysis is available at https://github.com/YaleiKong/NPY-fear-extinction. Acknowledgments We thank Ms. Yongling Wang (Fudan University) for her technique assistance. We are grateful to Drs. Xiajing Tong (ShanghaiTech University) and Jinfei Ni (Fudan University) for providing transgenic mice used in this study in a generous manner. This study was supported by grants from the National Natural Science Foundation of China (32430040, 32200821, 32371078 and 32300843), the STI2030-Major Projects (2021ZD0202800), the Key Discipline Project of Shanghai Municipal Health Commission (No. 2024ZDXK0050), the Science and Technology Commission of Shanghai Municipality (22XD1420700 and 23YF1433900), the Shanghai Municipal Health Commission (2022XD046), and innovative research team of high-level local universities in Shanghai. AUTHOR CONTRIBUTIONS Y.-J.W., X.G., Q.L., W.-G.L., and T.-L.X. conceived the project, designed the experiments, and interpreted the results. Y.-J.W. and X.G. performed the majority of behavioral experiments, animal surgery, immunohistochemistry, and data analysis. M.X., Q.W., X.Y., Z.-J.L., Z.-H.J., H.C., X.-Y.Z., and X.B. assisted with some of the behavioral experiments and conducted viral injections. Q.J. and Ying Li performed genotyping. Y.-J.W., H.W., and Yulong Li performed NPY dynamics measurement. Y.-J.W., X.G., Y.K., S.Y., J.H., and Q.L. did single-cell preparation and analyses. M-X.Z. commented on data interpretation. Y.-J.W., L.-Y.W., W.-G.L. and T.-L.X. wrote the manuscript with contributions from all authors. All authors read and approved the final manuscript. DECLARATION OF INTERESTS The authors declare no competing interests. Additional information Extended data includes 10 Extended Data and 27 Supplementary Information figures and their legends are available for this paper online. References Josselyn, S.A., Kohler, S. & Frankland, P.W. Finding the engram. Nat Rev Neurosci 16, 521–534 (2015). Josselyn, S.A. & Tonegawa, S. Memory engrams: Recalling the past and imagining the future. Science 367 (2020). Sun, X., et al. Functionally Distinct Neuronal Ensembles within the Memory Engram. Cell 181, 410–423 e417 (2020). Guskjolen, A. & Cembrowski, M.S. Engram neurons: Encoding, consolidation, retrieval, and forgetting of memory. Mol Psychiatry 28, 3207–3219 (2023). Tome, D.F., et al. Dynamic and selective engrams emerge with memory consolidation. Nat Neurosci 27, 561–572 (2024). Choi, J.H., et al. Interregional synaptic maps among engram cells underlie memory formation. Science 360, 430–435 (2018). Gallinaro, J.V., Gasparovic, N. & Rotter, S. Homeostatic control of synaptic rewiring in recurrent networks induces the formation of stable memory engrams. PLoS Comput Biol 18, e1009836 (2022). Vogels, T.P., Sprekeler, H., Zenke, F., Clopath, C. & Gerstner, W. Inhibitory plasticity balances excitation and inhibition in sensory pathways and memory networks. Science 334, 1569–1573 (2011). Davis, P., Zaki, Y., Maguire, J. & Reijmers, L.G. Cellular and oscillatory substrates of fear extinction learning. Nat Neurosci 20, 1624–1633 (2017). Wang, Q., et al. Insular cortical circuits as an executive gateway to decipher threat or extinction memory via distinct subcortical pathways. Nat Commun 13, 5540 (2022). Kim, T., et al. Activated somatostatin interneurons orchestrate memory microcircuits. Neuron 112, 201–208.e204 (2024). Lin, Z.J., et al. Stimulation of an entorhinal-hippocampal extinction circuit facilitates fear extinction in a post-traumatic stress disorder model. J Clin Invest 134, e181095 (2024). d'Aquin, S., et al. Compartmentalized dendritic plasticity during associative learning. Science 376, eabf7052 (2022). Topolnik, L. & Tamboli, S. The role of inhibitory circuits in hippocampal memory processing. Nat Rev Neurosci 23, 476–492 (2022). Tzilivaki, A., et al. Hippocampal GABAergic interneurons and memory. Neuron 111, 3154–3175 (2023). Dunsmoor, J.E., Niv, Y., Daw, N. & Phelps, E.A. Rethinking Extinction. Neuron 88, 47–63 (2015). Bouton, M.E., Maren, S. & McNally, G.P. Behavioral and Neurobiological Mechanisms of Pavlovian and Instrumental Extinction Learning. Physiol Rev 101, 611–681 (2021). Izquierdo, I., Furini, C.R. & Myskiw, J.C. Fear Memory. Physiol Rev 96, 695–750 (2016). Trouche, S., Sasaki, J.M., Tu, T. & Reijmers, L.G. Fear extinction causes target-specific remodeling of perisomatic inhibitory synapses. Neuron 80, 1054–1065 (2013). Monfils, M.H., Cowansage, K.K., Klann, E. & LeDoux, J.E. Extinction-reconsolidation boundaries: key to persistent attenuation of fear memories. Science 324, 951–955 (2009). Ressler, K.J., et al. Post-traumatic stress disorder: clinical and translational neuroscience from cells to circuits. Nat Rev Neurol 18, 273–288 (2022). Lebois, L.A.M., Seligowski, A.V., Wolff, J.D., Hill, S.B. & Ressler, K.J. Augmentation of Extinction and Inhibitory Learning in Anxiety and Trauma-Related Disorders. Annu Rev Clin Psychol 15, 257–284 (2019). Vervliet, B., Craske, M.G. & Hermans, D. Fear extinction and relapse: state of the art. Annu Rev Clin Psychol 9, 215–248 (2013). Li, W.G., et al. Input associativity underlies fear memory renewal. Natl Sci Rev 8, nwab004 (2021). Gu, X., et al. Dynamic tripartite construct of interregional engram circuits underlies forgetting of extinction memory. Mol Psychiatry 27, 4077–4091 (2022). Peters, J., Dieppa-Perea, L.M., Melendez, L.M. & Quirk, G.J. Induction of fear extinction with hippocampal-infralimbic BDNF. Science 328, 1288–1290 (2010). Wang, Q., et al. Fear extinction requires ASIC1a-dependent regulation of hippocampal-prefrontal correlates. Sci Adv 4, eaau3075 (2018). Meyer, H.C., et al. Ventral hippocampus interacts with prelimbic cortex during inhibition of threat response via learned safety in both mice and humans. Proc Natl Acad Sci U S A 116, 26970–26979 (2019). Szadzinska, W., et al. Hippocampal Inputs in the Prelimbic Cortex Curb Fear after Extinction. J Neurosci 41, 9129–9140 (2021). Nguyen, R., Koukoutselos, K., Forro, T. & Ciocchi, S. Fear extinction relies on ventral hippocampal safety codes shaped by the amygdala. Sci Adv 9, eadg4881 (2023). Allen, W.E., et al. Thirst-associated preoptic neurons encode an aversive motivational drive. Science 357, 1149–1155 (2017). Pettit, N.L., Yap, E.L., Greenberg, M.E. & Harvey, C.D. Fos ensembles encode and shape stable spatial maps in the hippocampus. Nature 609, 327–334 (2022). Wu, Y.E., Pan, L., Zuo, Y., Li, X. & Hong, W. Detecting Activated Cell Populations Using Single-Cell RNA-Seq. Neuron 96, 313–329 e316 (2017). Chen, M.B., Jiang, X., Quake, S.R. & Sudhof, T.C. Persistent transcriptional programmes are associated with remote memory. Nature 587, 437–442 (2020). Hochgerner, H., et al. Neuronal types in the mouse amygdala and their transcriptional response to fear conditioning. Nat Neurosci 26, 2237–2249 (2023). Sun, W., et al. Spatial transcriptomics reveal neuron-astrocyte synergy in long-term memory. Nature 627, 374–381 (2024). Fanselow, M.S. & Dong, H.W. Are the dorsal and ventral hippocampus functionally distinct structures? Neuron 65, 7–19 (2010). Wang, H., et al. A tool kit of highly selective and sensitive genetically encoded neuropeptide sensors. Science 382, eabq8173 (2023). Wu, Z., et al. Neuronal activity-induced, equilibrative nucleoside transporter-dependent, somatodendritic adenosine release revealed by a GRAB sensor. Proc Natl Acad Sci U S A 120, e2212387120 (2023). Dong, H.W., Swanson, L.W., Chen, L., Fanselow, M.S. & Toga, A.W. Genomic-anatomic evidence for distinct functional domains in hippocampal field CA1. Proc Natl Acad Sci U S A 106, 11794–11799 (2009). DeNardo, L.A., et al. Temporal evolution of cortical ensembles promoting remote memory retrieval. Nat Neurosci 22, 460–469 (2019). Garner, A.R., et al. Generation of a synthetic memory trace. Science 335, 1513–1516 (2012). Wieland, H.A., Willim, K. & Doods, H.N. Receptor binding profiles of NPY analogues and fragments in different tissues and cell lines. Peptides 16, 1389–1394 (1995). Loh, K., Herzog, H. & Shi, Y.C. Regulation of energy homeostasis by the NPY system. Trends Endocrinol Metab 26, 125–135 (2015). Marcos, P. & Covenas, R. Regulation of Homeostasis by Neuropeptide Y: Involvement in Food Intake. Curr Med Chem 29, 4026–4049 (2022). Gotzsche, C.R. & Woldbye, D.P. The role of NPY in learning and memory. Neuropeptides 55, 79–89 (2016). Tasan, R.O., et al. The role of Neuropeptide Y in fear conditioning and extinction. Neuropeptides 55, 111–126 (2016). Michel, M.C., et al. XVI. International Union of Pharmacology recommendations for the nomenclature of neuropeptide Y, peptide YY, and pancreatic polypeptide receptors. Pharmacol Rev 50, 143–150 (1998). Raut, S.B., et al. Diverse therapeutic developments for post-traumatic stress disorder (PTSD) indicate common mechanisms of memory modulation. Pharmacol Ther 239, 108195 (2022). Hao, Y., et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587 e3529 (2021). Finak, G., et al. MAST: a flexible statistical framework for assessing transcriptional changes and characterizing heterogeneity in single-cell RNA sequencing data. Genome Biol 16, 278 (2015). Additional Declarations There is NO Competing Interest. Supplementary Files ExtendedDataFiguresNNA86504A20250827.docx Extended Data Figures SupplementaryInformationNNA86504A20250827.docx Supplementary Information Cite Share Download PDF Status: Published Journal Publication published 31 Mar, 2026 Read the published version in Nature Neuroscience → Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4347593","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":523329662,"identity":"1dffb480-27ec-499a-8788-840fcd7dcd72","order_by":0,"name":"Tian-Le Xu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABA0lEQVRIiWNgGAWjYHACNhDBA2Z+YDgAEeMhVgvjDFK0gAEzDzFaDG4ksD34ueOOjDl7++PPNn/u2OvOSGB88LaNQd4chxbJGQnshr1nnvFY9pwxk85te5a47UYCs+HcNgbDnQ3YtfBL5H+T4G07zGNwI4eNObfhcIIZ0F5p3jaGBIMDODwikcAm+Rek5f7zx58t/hy2B2ph/41PC78E2EyQLQwG0gxshxmBDmNjxqdFsucBm7Qs0C8GZ3LMJHvbDiduO/OwWXLOOQnDDTi0GBwHOuztjjv2BsePP/7wA+Sw48kHP7wps5HHZQsYMDYcQOUCCQk86jG1jIJRMApGwShABQCxFV6rxV6UogAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0002-1438-0038","institution":"Shanghai Jiao Tong University School of Medicine","correspondingAuthor":true,"prefix":"","firstName":"Tian-Le","middleName":"","lastName":"Xu","suffix":""},{"id":523329663,"identity":"a0c702dc-9cf2-4579-8bf9-bee88a795275","order_by":1,"name":"Yan-Jiao Wu","email":"","orcid":"","institution":"Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Yan-Jiao","middleName":"","lastName":"Wu","suffix":""},{"id":523329664,"identity":"c217c42a-ecd6-4a00-8ad0-cbce9a284a6e","order_by":2,"name":"Xue Gu","email":"","orcid":"","institution":"Collaborative Innovation Center for Brain Science, Department of Anatomy and Physiology, Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Xue","middleName":"","lastName":"Gu","suffix":""},{"id":523329665,"identity":"d5164050-11e2-4791-b516-74786034c4a6","order_by":3,"name":"Yalei Kong","email":"","orcid":"","institution":"Center for Brain Science, Shanghai Children's Medical Center, School of Medicine, Shanghai Jiao Tong University","correspondingAuthor":false,"prefix":"","firstName":"Yalei","middleName":"","lastName":"Kong","suffix":""},{"id":523329666,"identity":"b51f4161-a048-44eb-9445-391b1cf69768","order_by":4,"name":"Shuo Yang","email":"","orcid":"","institution":"Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Shuo","middleName":"","lastName":"Yang","suffix":""},{"id":523329667,"identity":"d527920e-5368-4151-b1a0-370b52a1ad49","order_by":5,"name":"Huan Wang","email":"","orcid":"https://orcid.org/0000-0002-7472-0057","institution":"Peking University","correspondingAuthor":false,"prefix":"","firstName":"Huan","middleName":"","lastName":"Wang","suffix":""},{"id":523329668,"identity":"b08559bd-32f7-48aa-97ef-004dc525de71","order_by":6,"name":"Miao Xu","email":"","orcid":"","institution":"Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Miao","middleName":"","lastName":"Xu","suffix":""},{"id":523329669,"identity":"fad93a5d-f827-4f3b-82f2-e991232ab7bb","order_by":7,"name":"Qi Wang","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Qi","middleName":"","lastName":"Wang","suffix":""},{"id":523329670,"identity":"cf97dc44-3191-4d25-b6fa-a13118a84169","order_by":8,"name":"Xin Yi","email":"","orcid":"","institution":"Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Xin","middleName":"","lastName":"Yi","suffix":""},{"id":523329671,"identity":"e8945b90-a439-46d9-9b8a-5a1713f4ce6a","order_by":9,"name":"Ze-Jie Lin","email":"","orcid":"","institution":"Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Ze-Jie","middleName":"","lastName":"Lin","suffix":""},{"id":523329672,"identity":"476a35f5-2b1e-4fc9-915e-66095624f266","order_by":10,"name":"Zhi-Han Jiao","email":"","orcid":"","institution":"Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Zhi-Han","middleName":"","lastName":"Jiao","suffix":""},{"id":523329673,"identity":"307a4e75-7405-4776-ba01-1fadfeb041e6","order_by":11,"name":"Hoiyin Cheung","email":"","orcid":"","institution":"Ruijin Hospital, Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Hoiyin","middleName":"","lastName":"Cheung","suffix":""},{"id":523329674,"identity":"ee24234e-1902-4964-bb78-06749c749192","order_by":12,"name":"Xin-Yu Zhao","email":"","orcid":"","institution":"Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Xin-Yu","middleName":"","lastName":"Zhao","suffix":""},{"id":523329675,"identity":"15d32db2-3ab9-4790-8fd0-692162365107","order_by":13,"name":"Xin Bian","email":"","orcid":"","institution":"Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Xin","middleName":"","lastName":"Bian","suffix":""},{"id":523329676,"identity":"038abb71-5cbb-46fd-a45c-b0e0d480ac4d","order_by":14,"name":"Qin Jiang","email":"","orcid":"","institution":"Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Qin","middleName":"","lastName":"Jiang","suffix":""},{"id":523329677,"identity":"7421850e-de07-438a-94ce-69a375b29bf4","order_by":15,"name":"Ying Li","email":"","orcid":"","institution":"Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Ying","middleName":"","lastName":"Li","suffix":""},{"id":523329678,"identity":"7f57247b-8160-48b2-a781-8632757514c3","order_by":16,"name":"Michael Zhu","email":"","orcid":"https://orcid.org/0000-0002-5676-841X","institution":"The University of Texas Health Science Center at Houston","correspondingAuthor":false,"prefix":"","firstName":"Michael","middleName":"","lastName":"Zhu","suffix":""},{"id":523329679,"identity":"841cbd31-41e3-4929-9155-167cc43169eb","order_by":17,"name":"Lu-Yang Wang","email":"","orcid":"https://orcid.org/0000-0002-5399-944X","institution":"SickKids Research Inst \u0026 Dept Physiol, Univ Toronto","correspondingAuthor":false,"prefix":"","firstName":"Lu-Yang","middleName":"","lastName":"Wang","suffix":""},{"id":523329680,"identity":"2cde62a1-c332-491e-a0e3-bd8b7038d70e","order_by":18,"name":"Yulong Li","email":"","orcid":"https://orcid.org/0000-0002-9166-9919","institution":"Peking University","correspondingAuthor":false,"prefix":"","firstName":"Yulong","middleName":"","lastName":"Li","suffix":""},{"id":523329681,"identity":"ce46d258-3c35-49df-880a-e9832ca2553c","order_by":19,"name":"Ju Huang","email":"","orcid":"","institution":"Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Ju","middleName":"","lastName":"Huang","suffix":""},{"id":523329682,"identity":"8b394953-6c6e-47dd-8e52-e5a53edba632","order_by":20,"name":"Qian Li","email":"","orcid":"https://orcid.org/0000-0003-1300-3377","institution":"Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Qian","middleName":"","lastName":"Li","suffix":""},{"id":523329683,"identity":"5a67defd-9264-4502-85bd-beb0e291c702","order_by":21,"name":"Wei-Guang Li","email":"","orcid":"https://orcid.org/0000-0002-3442-7518","institution":"Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Wei-Guang","middleName":"","lastName":"Li","suffix":""}],"badges":[],"createdAt":"2024-04-30 08:45:23","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4347593/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4347593/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41593-026-02235-x","type":"published","date":"2026-03-31T04:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":94588971,"identity":"25b9cd8c-a8cf-408f-897a-2f8b97e2398a","added_by":"auto","created_at":"2025-10-28 18:19:54","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":228434,"visible":true,"origin":"","legend":"","description":"","filename":"CleanTextNNA86504A20250827.docx","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/a4b6276148af1b203a25a818.docx"},{"id":94589001,"identity":"da12b710-09d0-48d4-b1dd-2af6e1f2a26e","added_by":"auto","created_at":"2025-10-28 18:19:55","extension":"json","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":20582,"visible":true,"origin":"","legend":"","description":"","filename":"NNA86504B.json","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/852f23aa02b49c6ac5053113.json"},{"id":94589570,"identity":"ed94458f-43e5-4a9b-b460-1cde71f72b6f","added_by":"auto","created_at":"2025-10-28 18:20:28","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6662967,"visible":true,"origin":"","legend":"","description":"","filename":"ExtendedDataFiguresNNA86504A20250827.docx","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/7e9fa41b5dfb361b3daddd21.docx"},{"id":94589286,"identity":"b527096f-bdd1-4083-91f5-31837b7a3433","added_by":"auto","created_at":"2025-10-28 18:20:06","extension":"docx","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9877878,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryInformationNNA86504A20250827.docx","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/4d8eaee2a4b67b4a5a30ed2c.docx"},{"id":94588410,"identity":"273c52cc-c28b-41e9-b676-f95179a2a3e7","added_by":"auto","created_at":"2025-10-28 18:19:17","extension":"xml","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":208481,"visible":true,"origin":"","legend":"","description":"","filename":"NNA86504B2enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/7668b34f53d213c60c68b381.xml"},{"id":94588723,"identity":"6c2908a6-24c4-4cae-a534-3a47e97a56f6","added_by":"auto","created_at":"2025-10-28 18:19:42","extension":"pdf","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8906837,"visible":true,"origin":"","legend":"","description":"","filename":"Figure.1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/ff1df80b9293b8ddd249626f.pdf"},{"id":94589038,"identity":"a0a72497-643e-4db4-bb44-44c46a45ecfe","added_by":"auto","created_at":"2025-10-28 18:19:58","extension":"pdf","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9114627,"visible":true,"origin":"","legend":"","description":"","filename":"Figure.2.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/0ec13fbc044593d837d7522d.pdf"},{"id":94589279,"identity":"1295b1d1-a4e0-41cb-bbfd-70851ab082c7","added_by":"auto","created_at":"2025-10-28 18:20:05","extension":"pdf","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5893502,"visible":true,"origin":"","legend":"","description":"","filename":"Figure.3.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/5784d1d95b410f72844c89c3.pdf"},{"id":94589288,"identity":"d604c2c7-2ef1-4b73-baef-160ccfb2572a","added_by":"auto","created_at":"2025-10-28 18:20:07","extension":"pdf","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":69241827,"visible":true,"origin":"","legend":"","description":"","filename":"Figure.4.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/0c82cb75ea67f5911a5316c1.pdf"},{"id":94588492,"identity":"1d9d107c-f977-4d00-8cb9-6ad960b82c0e","added_by":"auto","created_at":"2025-10-28 18:19:26","extension":"pdf","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":714289,"visible":true,"origin":"","legend":"","description":"","filename":"Figure.5.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/f73a3e7393c4f456c81fccbb.pdf"},{"id":94589004,"identity":"49235b0e-604b-4608-a582-b35cb4ce7ace","added_by":"auto","created_at":"2025-10-28 18:19:55","extension":"xml","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":204384,"visible":true,"origin":"","legend":"","description":"","filename":"NNA86504B2structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/d90888fc695ab9d9e20e8029.xml"},{"id":94589569,"identity":"87689802-4291-4b52-970a-ddd1daec1fc8","added_by":"auto","created_at":"2025-10-28 18:20:27","extension":"html","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":229225,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/032a6aa768939b79546a6781.html"},{"id":94589037,"identity":"ada7b856-3447-4bea-9e69-f827661bb9f1","added_by":"auto","created_at":"2025-10-28 18:19:58","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":642796,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003evCA1 NPY dynamics and single‑cell transcriptomics during extinction.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e, Behavioral design. Mice underwent cued fear conditioning (Day 1, Cond) and, on Day 2, either received No Ext (2 CS; 1 block) or Ext (20 CS; 10 blocks) before vHPC dissociation and activity-sensitive single-cell RNA sequencing (Act-Seq) 2 h later. \u003cstrong\u003eb\u003c/strong\u003e, Time course of CS-evoked freezing (mean ± s.e.m.). n = 4 mice per group. \u003cstrong\u003ec\u003c/strong\u003e, Uniform Manifold Approximation and Projection (UMAP) of 33,092 vHPC cells resolved into 9 major cell types. \u003cstrong\u003ed\u003c/strong\u003e, Volcano plots for excitatory (left) and inhibitory (right) neurons showing extinction-regulated genes (|log\u003csub\u003e2\u003c/sub\u003e fold change| ≥ 0.585, FDR \u0026lt; 0.05). To control for unequal group sizes, the larger dataset was randomly down-sampled to the smaller one (1,000 permutations). Only genes meeting this criterion in ≥ 90% of resamples were called “extinction-regulated” and are shown in red and annotated. Point size reflects the resampling support fraction. \u003cstrong\u003ee\u003c/strong\u003e, Pie charts showing the fraction of \u003cem\u003eNpy\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e cells within the inhibitory neuron type. To correct for unequal sample sizes, the Ext inhibitory pool was randomly down-sampled 1,000 times to match the No Ext inhibitory-cell count; the pie for Ext depicts the mean proportion across resamples (segment labels report mean ± s.e.m., %). \u003cstrong\u003ef\u003c/strong\u003e, \u003cem\u003eNpy\u003c/em\u003e transcript abundance per \u003cem\u003eNpy\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e inhibitory neuron. For display, the violin plots correspond to the resample whose \u003cem\u003eP\u003c/em\u003e-value lies at the 90\u003csup\u003eth\u003c/sup\u003e percentile of the 1,000-permutation distribution. Dots are single cells. \u003cstrong\u003eg\u003c/strong\u003e, Schematic of the GRAB\u003csub\u003eNPY1.0\u003c/sub\u003e sensor and representative expression in vCA1. Scale bar, 100 μm. \u003cstrong\u003eh\u003c/strong\u003e, Fiber photometry configuration for recording NPY1.0 signals during extinction. \u003cstrong\u003ei\u003c/strong\u003e, Left: freezing in No Ext (block 1) and Ext (block 10). Right: averaged NPY1.0 ΔF/F traces aligned to CS (mean ± s.e.m.). n = 8 mice. \u003cstrong\u003ej\u003c/strong\u003e, Area Under the Curve (AUC) of NPY1.0 ΔF/F during No Ext \u003cem\u003evs.\u003c/em\u003e Ext conditions. n = 8 mice. \u003cstrong\u003ek\u003c/strong\u003e, Negative correlation between increased NPY1.0 response and diminished freezing levels during extinction. \u003cstrong\u003el\u003c/strong\u003e, Schematic of alternating behavioral states during extinction (freezing ON/OFF). \u003cstrong\u003em\u003c/strong\u003e,\u003cstrong\u003e n\u003c/strong\u003e, Heatmaps (\u003cstrong\u003em\u003c/strong\u003e) and AUC (\u003cstrong\u003en\u003c/strong\u003e) of NPY1.0 ΔF/F for 5 s following freezing ON and OFF epochs. n = 8 mice and 37 events for freezing ON and OFF states, respectively. \u003cstrong\u003eo\u003c/strong\u003e, Example trace illustrating coupling of fear state and NPY1.0 signals. Exact statistic tests and \u003cem\u003eP\u003c/em\u003e values are provided in Source Data.\u003c/p\u003e","description":"","filename":"11.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/8344b99772b0d9cc7a5fdc8d.jpg"},{"id":94588484,"identity":"5df6edad-47ab-490a-a597-ca4d0e24cf72","added_by":"auto","created_at":"2025-10-28 18:19:25","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":686179,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003evCA1 NPY\u003c/strong\u003e\u003csup\u003e\u003cstrong\u003e+\u003c/strong\u003e\u003c/sup\u003e\u003cstrong\u003e interneuronal activity dynamics during extinction.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e, \u003cstrong\u003ed\u003c/strong\u003e, Viral strategies and representative expression in vCA1 of NPY-Cre mice. \u003cstrong\u003ea,\u003c/strong\u003e Co-expression of DIO-ChrimsonR-tdTomato and the GRAB\u003csub\u003eNPY1.0\u003c/sub\u003e sensor. \u003cstrong\u003ed,\u003c/strong\u003e DIO-GCaMP6m. Scale bars, 100 μm. \u003cstrong\u003eb\u003c/strong\u003e, Optogenetic protocol (635 nm, 20 Hz, 100 pulses; left) and averaged GRAB\u003csub\u003eNPY1.0\u003c/sub\u003e response aligned to light (shaded; middle). \u003cstrong\u003ec\u003c/strong\u003e, AUC of ΔF/F in 5-s windows before, during and after light delivery. n = 6 mice. The “during-light” AUC is negative (presumably pH-quenching), whereas the \u003cstrong\u003eafter\u003c/strong\u003e window shows a selective increase indicating NPY release. \u003cstrong\u003ee\u003c/strong\u003e, Freezing (left) and mean GCaMP6m ΔF/F (right) during No Ext (block 1) and Ext (block 10) conditions in the same mice. \u003cstrong\u003ef\u003c/strong\u003e, AUC of GCaMP6m ΔF/F across the 30-s CS period. n = 10 mice. \u003cstrong\u003eg\u003c/strong\u003e, Negative correlation between increased calcium response and diminished freezing levels during extinction. \u003cstrong\u003eh\u003c/strong\u003e, Schematic of alternating fear states. \u003cstrong\u003ei\u003c/strong\u003e,\u003cstrong\u003e j\u003c/strong\u003e, Calcium signals time-locked to state transitions: heatmaps (\u003cstrong\u003ei\u003c/strong\u003e) and AUC (\u003cstrong\u003ej\u003c/strong\u003e) of NPY\u003csup\u003e+\u003c/sup\u003e neuronal calcium responses for 5 s following freezing ON and OFF epochs. n = 10 mice and 36 events for freezing ON and OFF states, respectively. \u003cstrong\u003ek\u003c/strong\u003e, Example trace showing GCaMP6m signals tracking rapid switches between fear states. \u003cstrong\u003el\u003c/strong\u003e, Timeline for \u003cem\u003eex vivo\u003c/em\u003e electrophysiological recordings of NPY\u003csup\u003e+\u003c/sup\u003e interneurons 2 h after No Ext or Ext. \u003cstrong\u003em\u003c/strong\u003e, Example firing to +100 pA steps. \u003cstrong\u003en\u003c/strong\u003e, Frequency-current relationships (0–250 pA, 500 ms). No Ext, n = 26 neurons from three mice; Ext, n = 31 neurons from three mice. \u003cstrong\u003eo\u003c/strong\u003e, Representative spontaneous excitatory postsynaptic currents (sEPSCs). \u003cstrong\u003ep\u003c/strong\u003e,\u003cstrong\u003e q\u003c/strong\u003e,\u003cstrong\u003e \u003c/strong\u003eCumulative distributions of sEPSC inter-event intervals (\u003cstrong\u003ep\u003c/strong\u003e) and amplitudes (\u003cstrong\u003eq\u003c/strong\u003e). No Ext, n = 26 neurons from three mice; Ext, n = 21 neurons from three mice. Exact statistic tests and \u003cem\u003eP\u003c/em\u003e values are provided in Source Data.\u003c/p\u003e","description":"","filename":"12.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/9bfbbb14702b330ff9a3013d.jpg"},{"id":94588839,"identity":"4591df21-141a-47fa-a55b-f98a9a5f8b49","added_by":"auto","created_at":"2025-10-28 18:19:50","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":838266,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eManipulating vCA1 NPY\u003c/strong\u003e\u003csup\u003e\u003cstrong\u003e+\u003c/strong\u003e\u003c/sup\u003e\u003cstrong\u003e interneurons and NPY signaling bidirectionally controls extinction.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e, Experimental timelines. NPY-Cre mice received vCA1 injections for optical inhibition (DIO-NpHR-EGFP or DIO-EGFP) or activation (DIO-ChrimsonR-mCherry or DIO-mCherry), followed by conditioning (Cond; Day 1), extinction (Ext; 20 CSs; Day 2), and extinction retrieval (Ext Retr; 20 CSs; Day 3) as indicated. \u003cstrong\u003eb\u003c/strong\u003e, Injection sites (coronal reference at Bregma −3.16 mm, left) and representative expression (right) of EGFP (NpHR) or mCherry (ChR2). Scale bars, 100 μm. \u003cstrong\u003ec\u003c/strong\u003e, Photostimulation schedule during Ext: light was delivered for 5 s immediately before each 30-s CS, throughout the 30-s CS, and for an additional 5 s immediately after the CS (total light per trial = 40 s). \u003cstrong\u003ed–f\u003c/strong\u003e, Freezing time courses with optical inhibition (\u003cstrong\u003ed\u003c/strong\u003e; Ctrl, n = 20; NpHR, n = 16) or activation (\u003cstrong\u003ee\u003c/strong\u003e; Ctrl, n = 24; ChR2, n = 17). \u003cstrong\u003ef,\u003c/strong\u003e Extinction slope (linear fit of freezing \u003cem\u003evs.\u003c/em\u003e CS number) for Stage 1 (CS 1–10) and Stage 2 (CS 11–20). \u003cstrong\u003eg–j\u003c/strong\u003e, Closed-loop ChR2 activation restricted to freezing-ON epochs during each CS (\u003cstrong\u003eg\u003c/strong\u003e). \u003cstrong\u003eh\u003c/strong\u003e, Freezing rasters per CS for optogenetic activation of NPY\u003csup\u003e+\u003c/sup\u003e neurons. \u003cstrong\u003ei\u003c/strong\u003e, Time courses (Ctrl, n = 11; ChR2, n = 11). \u003cstrong\u003ej\u003c/strong\u003e, Extinction slopes. \u003cstrong\u003ek\u003c/strong\u003e, \u003cstrong\u003el\u003c/strong\u003e, Post-retrieval (2×CS) activation without additional CSs (18×CS): Ext + mCherry (n = 10), No Ext + mCherry (n = 14), No Ext + ChR2 (n = 11).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003em\u003c/strong\u003e, \u003cstrong\u003en\u003c/strong\u003e, Local NPY infusion into vCA1 prior to Ext onset accelerates extinction \u003cem\u003evs.\u003c/em\u003e ACSF (\u003cstrong\u003em\u003c/strong\u003e; ACSF, n = 12; NPY, n = 11); corresponding slopes in \u003cstrong\u003en\u003c/strong\u003e. \u003cstrong\u003eo–q\u003c/strong\u003e, CRISPR/Cas9 \u003cem\u003eNpy\u003c/em\u003e conditional knockout in NPY\u003csup\u003e+\u003c/sup\u003e interneurons. \u003cstrong\u003eo,\u003c/strong\u003e Strategy and representative images (scale bars, 100 µm; inset 50 µm). \u003cstrong\u003ep,\u003c/strong\u003e Freezing time courses (Ctrl, n = 13; \u003cem\u003eNpy\u003c/em\u003e cKO, n = 12). \u003cstrong\u003eq,\u003c/strong\u003e Extinction slopes. Data are mean ± s.e.m.; exact statistics and tests are reported in Source Data. Exact statistic tests and \u003cem\u003eP\u003c/em\u003e values are provided in Source Data.\u003c/p\u003e","description":"","filename":"13.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/d0156adc4d58a57befca2874.jpg"},{"id":94589319,"identity":"d6ab7c81-e708-45e3-866c-1c4357280ffa","added_by":"auto","created_at":"2025-10-28 18:20:08","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1068739,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eNPY bifurcates largely non-overlapping NPY1R\u003c/strong\u003e\u003csup\u003e\u003cstrong\u003e+\u003c/strong\u003e\u003c/sup\u003e\u003cstrong\u003e and NPY2R\u003c/strong\u003e\u003csup\u003e\u003cstrong\u003e+\u003c/strong\u003e\u003c/sup\u003e\u003cstrong\u003e sub-ensembles in vCA1.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e, Venn diagrams of scRNA-seq–identified vHPC neurons (8,338 cells) expressing \u003cem\u003eNpy1r\u003c/em\u003e (purple), \u003cem\u003eNpy2r\u003c/em\u003e (green), or both (black). Numbers and percentages indicate cell counts per class. \u003cstrong\u003eb\u003c/strong\u003e, UMAP overlays showing the distribution of \u003cem\u003eNpy1r\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e (purple), \u003cem\u003eNpy2r\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e (green), double-positive (black), and receptor-negative (grey) neurons. \u003cstrong\u003ec\u003c/strong\u003e, Strategy for cellular validation: NPY2R-Cre mice received AAV-DIO-EGFP in vHPC; two weeks later multiplex RNAscope for \u003cem\u003eNpy1r\u003c/em\u003e (cyan) and \u003cem\u003eNpy2r\u003c/em\u003e (red) was performed to read out receptor mRNAs specifically in EGFP-labelled (NPY2R\u003csup\u003e+\u003c/sup\u003e) neurons. \u003cstrong\u003ed\u003c/strong\u003e, Representative RNAscope images across vCA1 layers (sps, spm, spd, SO) with high-magnification z-stacks and 3D surface renderings (Imaris) illustrating EGFP-labelled somata and receptor puncta. Scale bars, 30 µm (overview), 10 µm (zoom), 5 µm (surface views). \u003cstrong\u003ee\u003c/strong\u003e, Layer-resolved receptor mRNA density (puncta × 10\u003csup\u003e3\u003c/sup\u003e mm\u003csup\u003e2\u003c/sup\u003e; mean ± s.e.m.). \u003cem\u003eNpy1r\u003c/em\u003e is enriched in sps/spm and declines toward spd/SO; \u003cem\u003eNpy2r\u003c/em\u003e peaks in spd and is scarce in SO. n = 9 mice. \u003cstrong\u003ef\u003c/strong\u003e, Per-cell receptor content in EGFP-labelled neurons (n = 446 cells): the majority (n = 340 cells) contained \u003cem\u003eNpy2r\u003c/em\u003e puncta only (red); a minority (n = 71 cells) co-expressed \u003cem\u003eNpy1r\u003c/em\u003e and \u003cem\u003eNpy2r\u003c/em\u003e (yellow); one neuron was \u003cem\u003eNpy1r\u003c/em\u003e-only (cyan); grey, no detectable puncta. Symbol size encodes counts at identical coordinates. \u003cstrong\u003eg\u003c/strong\u003e, Within-condition receptor-biased transcriptomics computed with permutation down-sampling. For No Ext and Ext separately, the larger receptor group was randomly subsampled to the size of the smaller group across 1,000 permutations; points plot expression in \u003cem\u003eNpy2r\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e versus \u003cem\u003eNpy1r\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e neurons. Point size reflects the resampling support (fraction of permutations meeting criteria). Genes that satisfied |log\u003csub\u003e2\u003c/sub\u003e fold change| ≥ 0.25 and FDR \u0026lt; 0.05 in ≥90% of permutations are highlighted in red and reported as DEGs. \u003cstrong\u003eh\u003c/strong\u003e, FosTRAP2 × Ai9 scheme for tagging ensembles during No Ext (Fear), Ext, or Ext Retr (4‑OHT windows indicated), followed by triple RNAscope for \u003cem\u003etdTomato\u003c/em\u003e, \u003cem\u003eNpy1r\u003c/em\u003e, and \u003cem\u003eNpy2r\u003c/em\u003e. \u003cstrong\u003ei\u003c/strong\u003e, Representative triple-label images. Light green arrows indicate \u003cem\u003eNpy2r\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e cells, light purple arrows indicate \u003cem\u003eNpy1r\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e cells, and yellow arrows indicate double‑positive cells. Dashed circles outline \u003cem\u003etdTomato\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e cells. Scale bar, 50 μm. \u003cstrong\u003ej\u003c/strong\u003e, Quantification of ensemble size and receptor-positive cell counts. Left, per‑slice values; right, per‑mouse averages. Categories: \u003cem\u003etdTomato\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e, \u003cem\u003eNpy1r\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e, \u003cem\u003eNpy2r\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e, Both, None (mean ± s.e.m.). \u003cstrong\u003ek\u003c/strong\u003e, Composition of the tdTomato‑labelled ensembles: pie charts show the proportions of \u003cem\u003eNpy1r\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e (purple), \u003cem\u003eNpy2r\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e (green), Both (yellow), and None (grey) within Fear‑, Ext‑, and Ext Retr‑tagged populations; a significant shift from \u003cem\u003eNpy2r\u003c/em\u003e‑dominant (Fear) to \u003cem\u003eNpy1r\u003c/em\u003e‑dominant (Ext/Ext Retr) is observed. Sample sizes for \u003cstrong\u003ej–k\u003c/strong\u003e: No Ext, 17 slices from 3 mice; Ext, 22 slices from 3 mice; Ext Retr, 17 slices from 3 mice. Exact statistic tests and \u003cem\u003eP\u003c/em\u003e values are provided in Source Data.\u003c/p\u003e","description":"","filename":"14.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/0b8226b3f206d08f5f72326d.jpg"},{"id":94589553,"identity":"10ddefd7-38ba-4c4d-aa2f-189ddc0dc91e","added_by":"auto","created_at":"2025-10-28 18:20:27","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":604140,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eReceptor-specific perturbations of NPY signaling differentially affect memory liability and stability.\u003c/strong\u003e \u003cstrong\u003ea\u003c/strong\u003e, Experimental design for receptor-restricted CRISPR editing (conditional knockout, cKO) in engram neurons. FosTRAP2 × Cas9 mice received AAVs carrying U6-sgRNAs targeting \u003cem\u003eNpy1r\u003c/em\u003e or \u003cem\u003eNpy2r\u003c/em\u003e (or scrambled control) together with DIO-EGFP. On Day 2, mice were TRAPed either after 2×CS (No Ext; Fear-tagged) or during 20×CS extinction (Ext-tagged) with 4-OHT. After allowing time for editing, animals were tested for extinction (Day 22) and extinction retrieval (Day 23). \u003cstrong\u003eb–e\u003c/strong\u003e, Freezing during conditioning (Cond), extinction (Ext), and extinction retrieval (Ext Retr) for Fear-tagged (\u003cstrong\u003eb\u003c/strong\u003e, \u003cstrong\u003ec\u003c/strong\u003e) and Ext\u003cstrong\u003e-\u003c/strong\u003etagged (\u003cstrong\u003ed\u003c/strong\u003e, \u003cstrong\u003ee\u003c/strong\u003e) cohorts. \u003cstrong\u003ec\u003c/strong\u003e, \u003cstrong\u003ee\u003c/strong\u003e, Extinction slope. \u003cstrong\u003eb\u003c/strong\u003e, \u003cstrong\u003ec\u003c/strong\u003e, Ctrl, n = 11, \u003cem\u003eNpy1r\u003c/em\u003e cKO, n = 8, \u003cem\u003eNpy2r\u003c/em\u003ecKO, n = 10; \u003cstrong\u003ed\u003c/strong\u003e, \u003cstrong\u003ee\u003c/strong\u003e, Ctrl, n = 17, \u003cem\u003eNpy1r\u003c/em\u003e cKO, n = 12; \u003cem\u003eNpy2r\u003c/em\u003ecKO, n = 8. \u003cstrong\u003ef\u003c/strong\u003e, Schematic of bilateral cannula implantation in vHPC (vCA1) and timeline for acute pharmacology; drug was microinfused immediately before extinction (red arrow). \u003cstrong\u003eg–j\u003c/strong\u003e, Effects of receptor‑selective ligands on freezing (\u003cstrong\u003eg\u003c/strong\u003e, \u003cstrong\u003ei\u003c/strong\u003e) and extinction slope (\u003cstrong\u003eh\u003c/strong\u003e, \u003cstrong\u003ej\u003c/strong\u003e). \u003cstrong\u003eg\u003c/strong\u003e, \u003cstrong\u003eh\u003c/strong\u003e, Agonists: NPY1R agonist and NPY2R agonist \u003cem\u003evs.\u003c/em\u003e ACSF. \u003cstrong\u003ei\u003c/strong\u003e, \u003cstrong\u003ej\u003c/strong\u003e, Antagonists: NPY1R antagonist and NPY2R antagonist \u003cem\u003evs.\u003c/em\u003e ACSF. \u003cstrong\u003eg\u003c/strong\u003e, \u003cstrong\u003eh\u003c/strong\u003e, ACSF, n = 13, 1R agonist, n = 10, 2R agonist, n = 11; \u003cstrong\u003ei\u003c/strong\u003e, \u003cstrong\u003ej\u003c/strong\u003e, ACSF, n = 13, 1R antagonist, n = 10, 2R antagonist, n = 11. Data are mean ± s.e.m. Exact statistic tests and \u003cem\u003eP\u003c/em\u003e values are provided in Source Data.\u003c/p\u003e","description":"","filename":"15.jpg","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/188e7f905dac99950d6b58e1.jpg"},{"id":105886555,"identity":"460be7a7-62d7-4a61-8156-dd7a32195c6c","added_by":"auto","created_at":"2026-04-01 07:30:11","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":5449969,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/7ae8ec9f-06e4-4947-baaf-a81865177c5d.pdf"},{"id":94588453,"identity":"4750ab57-f531-454d-adc6-760737aceb58","added_by":"auto","created_at":"2025-10-28 18:19:23","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":6662967,"visible":true,"origin":"","legend":"Extended Data Figures","description":"","filename":"ExtendedDataFiguresNNA86504A20250827.docx","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/5767d09f7ee5595fb193dc12.docx"},{"id":94588648,"identity":"37ca7964-80e2-4514-84e3-78b4733fcdc3","added_by":"auto","created_at":"2025-10-28 18:19:37","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":9877878,"visible":true,"origin":"","legend":"Supplementary Information","description":"","filename":"SupplementaryInformationNNA86504A20250827.docx","url":"https://assets-eu.researchsquare.com/files/rs-4347593/v1/370161c7d9f6faf30c81099d.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Neuropeptide Y co-opts neuronal ensembles for memory lability and stability","fulltext":[{"header":"Main","content":"\u003cp\u003eMemory traces arise when a sparse subset of neurons is recruited during learning to form an engram\u003csup\u003e1\u003c/sup\u003e. Activity-dependent tagging has shown that such ensembles are distributed in space and time and re-engaged during recall\u003csup\u003e2-5\u003c/sup\u003e. Within these ensembles, excitatory principal neurons—biased by heightened intrinsic excitability and synaptic plasticity—form the backbone of inter-ensemble connectivity, supporting flexible expression of learned behaviors\u003csup\u003e3, 5, 6\u003c/sup\u003e. In parallel, GABAergic interneurons shape which principles cells are admitted into the trace by suppressing non-engram activity, stabilizing network dynamics, and regulating transitions in memory fates\u003csup\u003e7-10\u003c/sup\u003e. Although specific interneuron subtypes have been implicated in engram-level control\u003csup\u003e5, 11, 12\u003c/sup\u003e, the molecular logic by which defined inhibitory neurons bias ensemble composition and tune memory lability versus stability remains unclear.\u003c/p\u003e\n\u003cp\u003eGABAergic interneurons are diverse in position and output, and many co-release peptides—including somatostatin (SST), cholecystokinin (CCK), neuropeptide Y (NPY)—alongside GABA\u003csup\u003e13-15\u003c/sup\u003e. This raises two unresolved questions with direct relevance to engram control: (i) which inhibitory interneuron subtypes are recruited during memory expression versus extinction, a canonical state switch\u003csup\u003e16, 17\u003c/sup\u003e; and (ii) under what conditions does peptide co‑release—as opposed fast GABA alone—shape ensemble selection and behavioral output. Addressing these questions requires linking cell type, receptor identity, and peptide dynamics to behavior on appropriate time scales.\u003c/p\u003e\n\u003cp\u003eCued fear learning and extinction provide a tractable framework for this goal. Fear conditioning binds a neutral cue to an aversive outcome to produce a robust, retrieval trace\u003csup\u003e18\u003c/sup\u003e; extinction training subsequently suppresses expression of fear by forming a new memory that does not erase the original\u003csup\u003e17, 19, 20\u003c/sup\u003e. The coexistence and reversibility of fear and extinction traces—together with their dysregulation in anxiety and trauma-related disorders\u003csup\u003e21, 22\u003c/sup\u003e and their capacity for spontaneous recovery\u003csup\u003e23-25\u003c/sup\u003e—make this paradigm ideal for probing how inhibitory circuits govern the reweighting of competing ensembles. The ventral hippocampal CA1 (vCA1) is a critical node in this circuitry\u003csup\u003e26-30\u003c/sup\u003e, yet it is unknown how peptidergic inhibition in vCA1 organizes receptor-defined neuronal sub-ensembles across extinction.\u003c/p\u003e\n\u003cp\u003eIn this study, we discovered that fear extinction, compared to fear memory retrieval, preferentially activates NPY-expressing (NPY\u003csup\u003e+\u003c/sup\u003e) interneurons in vCA1 leading to increased NPY release. Strikingly, NPY release bifurcates onto two spatially distinct sub-ensembles, each expressing unique NPY receptors. These two sub-ensembles are interlocked in time and space, enabling the switching of memory traces between a labile and stable state.\u0026nbsp;This framework links peptide kinetics, receptor identity, and laminar organization to the dynamic reallocation of engram membership, providing a principled mechanism for how inhibitory circuits switch a memory between changeable and durable states.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eSingle-cell transcriptomic mapping in the vCA1 uncovers upregulation of \u003cem\u003eNpy\u003c/em\u003e gene in inhibitory neurons during fear extinction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo investigate how distinct cell types flexibly dictate memory-related behaviors, we employed an auditory cued fear conditioning and extinction paradigms in wild-type mice. On Day 1, mice received five foot shocks paired with a neutral auditory tone (conditioned stimulus, CS) in context A. On Day 2, these mice underwent extinction training in a novel context B (Ext group), receiving 20 CS presentations without foot shocks (context B). Although this training significantly reduced fear responses to the tone, it did not completely eliminate context A-associated fear, as evidenced by the recovery of conditioned fear responses when mice were re-exposed to context A, regardless of the presence of the tone (Supplementary Fig. 1). To clearly distinguish the ensemble underlying fear memory from that for extinction, we focused on auditory cued fear in a setting designed to minimize confounding contextual fear.\u003c/p\u003e\n\u003cp\u003eAs a control for fear extinction, we subjected mice to only two CS presentations in context B (No Ext group), which proved insufficient to induce extinction. To precisely calibrate potential paradigm differences between Ext and No Ext groups, we designed three variants of No Ext conditions: (1) a short exposure group receiving two CS presentations immediately before removal from context B; (2) a group receiving two CS presentations followed by an extended waiting period (equivalent to the other 18 CS presentations in the Ext group) but without additional tone exposure; and (3) a group that remained in context B for the duration equivalent to the first 18 CS presentations in the Ext group but without actual tone presentations, then received two CS presentations prior to removal. We collected vCA1 samples either 20 min or 2 h after training for subsequent analysis. Consistent with \u003cem\u003eFos\u003c/em\u003e as a transient marker of neuronal activation to salient stimuli\u003csup\u003e25, 31, 32\u003c/sup\u003e, \u003cem\u003eFos\u003c/em\u003e mRNA expression peaked at 20 min post-training and substantially declined by 2 h across all experimental conditions. Crucially, mice undergoing Ext showed significantly higher \u003cem\u003eFos\u003c/em\u003e expression than all variants of the No Ext control at both 20 min and 2 h time points, highlighting sustained ensemble activation specifically associated with extinction (Supplementary Fig. 2). Guided by these findings, we selected tissue samples collected 2 h after training from both the Ext group and the short-exposure No Ext group, to ensure optimal temporal matching for detailed analyses.\u003c/p\u003e\n\u003cp\u003eTo identify the cell types recruited by fear memory retrieval versus extinction (Fig. 1a, b), we conducted unbiased, high-throughput, droplet-based single-cell RNA sequencing (scRNA-seq) using an activity-sensitive single-cell preparation method\u003csup\u003e33\u003c/sup\u003e. This approach enabled us to profile transcriptomic changes at the single-cell level, revealing molecular signatures associated with two distinct memory engrams\u003csup\u003e34-36\u003c/sup\u003e while minimizing \u003cem\u003eex vivo\u003c/em\u003e transcriptional artifacts and preserving behaviorally induced gene expression. In total, we profiled 46,894 individual cells, comprising 20,263 cells from No Ext group and 26,631 cells from the Ext group, each derived from 4 male mice. To analyze this dataset, we applied uniform manifold approximation and projection (UMAP) for dimensionality reduction, to visualization of the high-dimensional single-cell data in a biologically meaningful manner. This analysis revealed 35 distinct clusters and 9 cell types, identified based on established cell-type-specific markers (Fig. 1c, Supplementary Fig. 3).\u0026nbsp;Among these, 8 clusters were identified as excitatory neurons, while Cluster 6 was initially categorized as the sole inhibitory neuron type. The remaining clusters corresponded to glial cells and other non-neuronal cell types (Supplementary Fig. 3a).\u0026nbsp;The number and proportion of most cell types were broadly comparable, though notable differences in certain cell populations between the two conditions were observed (Supplementary Fig. 4). To rigorously control for these discrepancies, we implemented a random down-sampling procedure in which the larger dataset was repeatedly subsampled (1,000 permutations; see Methods) to match the smaller one, followed by re-analysis of the transcriptional responses. Despite this normalization, several clusters retained significant differential responsiveness to Ext compared with No Ext (Supplementary Fig. 3b), along with distinct patterns of immediate early gene (IEG; see Methods)\u003csup\u003e33\u003c/sup\u003e expression across various cell types and clusters (Supplementary Fig. 5).\u003c/p\u003e\n\u003cp\u003eAmong the neuronal populations, Ext paradigm specifically upregulated 6 genes and downregulated 2 genes\u0026nbsp;across all excitatory neuron clusters combined; and upregulated 5 genes and downregulated 2 genes within cluster 6, which represents the inhibitory neuron type (Fig. 1d). Within excitatory neurons, Ext elicited cluster-specific transcriptional responses. Volcano analyses revealed the largest sets of differentially expressed genes (DEGs) in clusters 4 and 10, whereas other clusters showed few or no DEGs (Supplementary Fig. 6a–h). Moreover, Ext produced a marked increase in \u003cem\u003eFos\u003c/em\u003e expression in clusters 10 and 32 compared with No Ext (Supplementary Fig. 6i). Given our aim in elucidating how memory fates transition from fear persistence to extinction, we focused on GABAergic interneurons that likely plays a pivotal role in governing memory extinction process. Within inhibitory neurons of cluster 6, Ext paradigm prominently upregulated both \u003cem\u003ePcp4\u003c/em\u003e and \u003cem\u003eNpy\u003c/em\u003e by approximately 2-fold (Fig. 1d). The proportion of \u003cem\u003eNpy\u003c/em\u003e⁺ interneurons within the inhibitory neuron type\u003csup\u003e15\u003c/sup\u003e increased from 12.2% in the No Ext group to 24.6 ± 2.2% in the Ext group (Fig. 1e). This increase presumably reflects the recruitment of neurons that previously did not express \u003cem\u003eNpy\u003c/em\u003e or expressed it only at minimal, undetectable levels. Additionally, the Ext group exhibited elevated overall \u003cem\u003eNpy\u003c/em\u003e transcript levels per cell compared with the No Ext group (Fig. 1f). To further dissect the relationship between \u003cem\u003eNpy\u003c/em\u003e expression and IEG activity, we specifically examined the subset of inhibitory neurons displaying particularly high levels of IEG expression (defined as cells with at least one IEG expressed in the 90\u003csup\u003eth\u003c/sup\u003e-percentile among all inhibitory neurons). Both the proportion of \u003cem\u003eNpy\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e/IEG\u003csup\u003e+\u003c/sup\u003e interneurons and the \u003cem\u003eNpy\u003c/em\u003e expression within these IEG\u003csup\u003e+\u003c/sup\u003e interneurons were significantly higher in the Ext group than in the No Ext group (Supplementary Fig. 7). These findings provide strong justification for a detailed functional interrogation of \u003cem\u003eNpy\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e GABAergic interneurons in fear extinction, in which potentially orchestrate the shift between fear and extinction memories.\u003c/p\u003e\n\u003cp\u003eAcross the vCA1 transcriptome,\u003cem\u003e\u0026nbsp;Npy\u003c/em\u003e expression was largely confined to inhibitory neurons and was essentially absent from other cell types, including excitatory neurons (Supplementary Fig. 8a). While initial unsupervised clustering identified a single inhibitory neuron cluster (cluster 6), further refined clustering analysis delineated 5 distinct subclusters (Supplementary Fig. 8b, c). \u003cem\u003eNpy\u003c/em\u003e⁺ interneurons were distributed across several of these subclusters, with enrichment in \u003cem\u003eNpy\u003c/em\u003e⁺ and \u003cem\u003eNpy\u003c/em\u003e⁺/\u003cem\u003eSST\u003c/em\u003e⁺ (somatostatin-expressing) subclusters. Ext paradigm increased \u003cem\u003eNpy\u003c/em\u003e transcript abundance across most inhibitory subclusters (with the exception of one subcluster, Supplementary Fig. 8d), implicating broad recruitment of \u003cem\u003eNpy\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e interneurons\u0026nbsp;during extinction.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eNpy\u003csup\u003e+\u003c/sup\u003e\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;inhibitory neurons are selectively activated in vCA1 during extinction learning\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe scRNA-seq dataset did register trace-level \u003cem\u003eNpy\u003c/em\u003e reads in a minority of excitatory clusters, with a small upward shift in Ext relative to No Ext condition (Supplementary Fig. 8e). Because these values were near the detection floor, we sought after independent validation with other approaches. Dual-color fluorescence in situ hybridization confirmed that \u003cem\u003eNpy\u003c/em\u003e transcripts were almost exclusively detected in \u003cem\u003eVgat\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e (vesicular GABA transporter-expressing) neurons and absent in \u003cem\u003eVglut1\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e (vesicular glutamate transporter 1-expressing) neurons, confirming the selective expression of \u003cem\u003eNpy\u003c/em\u003e within vCA1 GABAergic interneurons. Furthermore, \u003cem\u003eNpy\u003c/em\u003e signals were predominantly localized to the stratum oriens (SO) of the CA1 region\u003csup\u003e37\u003c/sup\u003e (Supplementary Fig. 9).\u003c/p\u003e\n\u003cp\u003edFISH analyses revealed a significant increase in the proportion of \u003cem\u003eNpy\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e and \u003cem\u003eNpy\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e/\u003cem\u003eFos\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e co-labeled neurons in the vCA1 of Ext compared to No Ext mice. Moreover, the expression levels of \u003cem\u003eNpy\u003c/em\u003e and \u003cem\u003eFos\u003c/em\u003e in the vCA1 were significantly correlated in the Ext group but not in the No Ext group (Supplementary Fig. 10). Collectively, these results validate that \u003cem\u003eNpy\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e GABAergic interneurons become transcriptionally activated in the vCA1 during fear extinction, highlighting their potential role as a key cell type for gating extinction learning.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDynamics of vCA1 extracellular NPY during extinction learning\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo directly monitor extracellular NPY dynamics in the vCA1 region during fear extinction, we employed a genetically encoded fluorescent GRAB (G protein–coupled receptor activation‒based) NPY sensor, NPY1.0 (Fig. 1g)\u003csup\u003e38\u003c/sup\u003e, specially targeted to vCA1 neurons (Fig. 1g, h and Supplementary Fig. 11). Mice underwent fear conditioning followed by extinction training, resulting in a gradual reduction in freezing behavior. Real-time \u003cem\u003ein vivo\u003c/em\u003e quantification of NPY release using fiber photometry revealed no significant changes during exposure to the auditory tone (CS), however, the signal gradually decreased following the delivery of five foot shocks (US) during fear conditioning (Extended Data Fig. 1). During extinction learning, a clear contrast in CS-evoked NPY release was observed between the No Ext and Ext groups. In the No Ext group, the first two CS presentations in the Block 1—characterized by heightened fear responses—elicited a significant decrease in NPY release. Conversely, in the Ext group, the last two CS presentations in the Block 10—associated with diminished fear responses—showed a modest but statistically significant increase in CS-evoked NPY release (Fig. 1i, j).\u0026nbsp;Furthermore, NPY release progressively ramped up in response to repeated CS presentations during extinction learning, correlating with the attenuation of fear responses (Fig. 1k). During each 30-s period CS trial in extinction training, mice alternated between freezing (ON) and non-freezing (OFF) epochs (Fig. 1l), with corresponding decreases and increases in the NPY signal (Fig. 1m–o). Interestingly, prolonged NPY release was detected 2 h after Ext compared to the No Ext condition (Extended Data Fig. 2), mirroring the observed increase in \u003cem\u003eNpy\u003c/em\u003e gene expression at the same time point. Moreover, a significant increase in cued NPY release was observed during Ext Retr on Day 3 (Extended Data Fig. 1g–i).\u003c/p\u003e\n\u003cp\u003eIn contrast, a control group exposed to the CS without foot shocks (CS only) did not exhibit significant NPY release during either conditioning or extinction (Supplementary Fig. 12), suggesting that the extinction-induced NPY release is specifically associated with the extinction of the CS rather than mere repeated exposures or habituation to the cue. Moreover, no detectable NPY signal changes were observed in mice expressing a non-ligand-binding mutant sensor (NPYmut) in wild-type mice or in mice expressing the GRAB-NPY1.0 sensor in \u003cem\u003eNpy\u003c/em\u003e knockout backgrounds during extinction learning (Supplementary Fig. 13). These findings demonstrate that NPY release in the vCA1 is dynamically regulated during extinction learning and is specifically associated with the extinction of fear memory.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDynamics of vCA1 NPY\u003csup\u003e+\u003c/sup\u003e neuronal activity during extinction learning\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo test whether vCA1 NPY\u003csup\u003e+\u003c/sup\u003e interneurons drive NPY release, we co-expressed a Cre-dependent red-shifted opsin (ChrimsonR) and the GRAB-NPY1.0 sensor in NPY-Cre mice and recorded fiber photometry in vCA1 (Fig. 2a and Supplementary Fig. 14a). Brief 635-nm stimulation produced a stimulus-locked downward deflection in both the 470-nm sensor channel and the 405-nm isosbestic control—consistent with transient extracellular pH-dependent quenching commonly seen with GRAB sensors\u003csup\u003e38, 39\u003c/sup\u003e—followed by a selective post‑stimulus increase in the 470-nm signal (Fig. 2b, c and Supplementary Fig. 14b, c). Quantitatively, the 5-s “during-light” AUC was negative for both channels, whereas the 5-s “after-light” AUC was positive only in the 470-nm channel (Fig. 2c and Extended Data Fig. 16c). Together with the spectral separation between photometry excitation (470/405 nm) and optogenetic light (635 nm), these data indicate that the illumination-period dip reflects pH quenching rather than suppressed release, and that activating NPY\u003csup\u003e+\u003c/sup\u003e interneurons is sufficient to evoke NPY release in vCA1.\u003c/p\u003e\n\u003cp\u003eTo further investigate the neuronal activity of vCA1 NPY\u003csup\u003e+\u003c/sup\u003e interneurons, we expressed the Cre-dependent genetically encoded calcium indicator GCaMP6m in these cells and continuously recorded calcium transients using fiber photometry (Fig. 2d and Supplementary Fig. 14d). During fear conditioning, NPY\u003csup\u003e+\u003c/sup\u003e interneurons exhibited sustained and robust calcium responses to foot shocks (US). Auditory cues (CS) elicited progressively stronger calcium responses across successive learning trials, indicating heightened neuronal activity associated with fear learning (Extended Data Fig. 3). Notably, calcium responses during the Ext state were significantly elevated compared to the No Ext state (Fig. 2e, f). As extinction progressed, calcium activity in NPY\u003csup\u003e+\u003c/sup\u003e neurons continued to increase, exhibiting a significant negative correlation with the diminishing fear response during intermittent CS presentations (Fig. 2g).\u003c/p\u003e\n\u003cp\u003eConsistent with the GRAB-NPY1.0 sensor data, we observed a decrease in calcium signals during freezing (ON epochs) and an increase during non-freezing (OFF epochs) periods (Fig. 2h–k). Moreover, prolonged neuronal activity was detected 2 h post-Ext compared to the No Ext condition (Extended Data Fig. 4), paralleling the sustained increase in NPY release observed with the GRAB-NPY1.0 sensor at the same time point. We also observed a significant elevation in cued NPY\u003csup\u003e+\u003c/sup\u003e neuronal calcium activity during Ext Retr on Day 3 (Extended Data Fig. 3g–i). In contrast, a control group exposed to the CS without foot shocks (CS only) did not exhibit significant NPY\u003csup\u003e+\u003c/sup\u003e neuronal calcium activity during either conditioning or extinction phases (Supplementary Fig. 15). These findings, in conjunction with the dynamic changes in NPY release specifically associated with extinction, suggest that vCA1 NPY\u003csup\u003e+\u003c/sup\u003e interneurons actively drive NPY dynamics during fear extinction, highlighting their potential role in facilitating progressive extinction. However, the absence of a corresponding association between NPY dynamics and neuronal activity during fear learning indicates that vCA1 NPY\u003csup\u003e+\u003c/sup\u003e interneurons may engage through alternative mechanisms in this scenario. This distinction underscores the specialized role of NPY\u003csup\u003e+\u003c/sup\u003e interneurons in modulating memory extinction, separate from their functions during the initial fear learning phase.\u003c/p\u003e\n\u003cp\u003eSubsequently, we performed a comparative analysis of the excitability of NPY\u003csup\u003e+\u003c/sup\u003e interneurons between the No Ext and Ext groups using patch-clamp electrophysiology. Extinction training resulted in a substantial increase in the frequency of action potentials evoked by identical current step injections in the Ext group, while the amplitude of these action potentials remained unchanged (Fig. 2l–n). Furthermore, the frequency of spontaneous excitatory postsynaptic currents (sEPSCs), indicative of global excitatory synaptic input to NPY\u003csup\u003e+\u003c/sup\u003e interneurons, was significantly elevated in the Ext group, although changes in amplitude were less pronounced (Fig. 2o–q). These electrophysiological findings demonstrate that extinction learning enhances both the excitatory drive and intrinsic excitability of NPY\u003csup\u003e+\u003c/sup\u003e interneurons in the vCA1.\u0026nbsp;Collectively, these data indicate that during extinction, activated NPY\u003csup\u003e+\u003c/sup\u003e GABAergic interneurons in the vCA1, through endogenous NPY release, play a pivotal role in modulating the dynamics of fear and extinction memory.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eManipulation of vCA1 NPY\u003csup\u003e+\u003c/sup\u003e interneuron activity controls memory extinction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo evaluate the necessity and efficacy of NPY\u003csup\u003e+\u003c/sup\u003e neuronal activity in extinction learning, we introduced halorhodopsin (NpHR) or channelrhodopsin-2 (ChR2) into NPY\u003csup\u003e+\u003c/sup\u003e interneurons in the vCA1 and manipulated their activity using spatially targeted light (Fig. 3a–c). Optogenetic inhibition of NPY\u003csup\u003e+\u003c/sup\u003e interneurons during CS presentation in extinction learning intensified the tone-shock association, sustaining a higher fear response (Fig. 3d). Even during extinction retrieval without optogenetic inhibition, the NpHR group maintained a higher level of conditioned fear. Conversely, optogenetic activation of NPY\u003csup\u003e+\u003c/sup\u003e interneurons during CS presentation in extinction learning disassociated the tone-shock association, resulting in reduced fear responses both in extinction learning and retrieval (Fig. 3e). Further analyses of extinction time course showed two stages of decline in cued freezing levels over 10-block epochs (20\u0026nbsp;CS): Stage 1, i.e., the first 5 blocks, being rapid, likely reflects the lability of fear memory to extinction; and Stage 2, the last 5 blocks, being much slower, likely measures the stability of fear memory (Fig. 3f and Supplementary Fig. 16a–d). Optogenetic inhibition of NPY\u003csup\u003e+\u003c/sup\u003e interneurons largely ablated Stage 1 extinction and precluded its progression to Stage 2; and conversely activation of NPY\u003csup\u003e+\u003c/sup\u003e neurons robustly augmented the onset of Stage 1 extinction but had little effect on the stable setpoint of freezing level for Stage 2 (Fig. 3f and Supplementary Fig. 16a–d). These observations indicated that Stage 1 and 2 of extinction are interlocked processes, for which the activity level of NPY\u003csup\u003e+\u003c/sup\u003e interneurons may confer sequentially labile and stable states of fear memory.\u003c/p\u003e\n\u003cp\u003eNotably, when NPY\u003csup\u003e+\u003c/sup\u003e neuronal activity was selectively elevated optogenetically during the CS-induced freezing ON epoch without affecting other periods (Fig. 3g), significant facilitatory effects on both extinction learning and retrieval were observed. However, this targeted elevation did not eliminate the difference in extinction slope between the first and second halves of training, suggesting that a physiologically scaling-up activation of NPY\u003csup\u003e+\u003c/sup\u003e neurons on demand preserves the balance between memory liability and stability during extinction learning (Fig. 3h–j and Supplementary Fig. 16e, f). To further validate the role of NPY\u003csup\u003e+\u003c/sup\u003e neuronal activity in modulating the lability and stability of fear extinction, we conducted a complementary optogenetic activation experiment. In this case, NPY\u003csup\u003e+\u003c/sup\u003e neurons were activated following fear retrieval (two CS exposure) without subsequent 18 CS exposure for extinction training (Fig. 3k). Remarkably, this activation accelerated Stage 1 extinction and achieved fear reduction effects comparable to those in the full extinction protocol (control extinction group, Fig. 3l). Importantly, the effects of endogenous NPY levels were specific to fear extinction, as locomotor behavior in these mice remained unaffected by the same manipulations (Supplementary Fig. 17). These results demonstrate that vCA1 NPY\u003csup\u003e+\u003c/sup\u003e interneurons play a pivotal role in dictating the progression of extinction, even in the absence of the full extinction protocol. Furthermore, their activity appears most influential during the behavioral state switch from “fear-on” to “fear-off” epochs during extinction learning, underscoring their critical contribution to the dynamic regulation of memory liability and stability.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eNPY itself is both necessary and sufficient for fear extinction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo determine whether the NPY peptide itself fulfills a similar function to that of vCA1 NPY\u003csup\u003e+\u003c/sup\u003e neuron activation in facilitating fear extinction, we microinjected exogenous NPY into the vCA1 through an implanted cannula 15 min prior to extinction leaning. This intervention significantly accelerated and enhanced the rate and extent of fear extinction (Fig. 3m, n and Supplementary Fig. 18a, b), effectively mirroring the effect of optogenetic activation of NPY\u003csup\u003e+\u003c/sup\u003e interneurons (Fig. 3e). Conversely, in mice carrying a global or partial loss of \u003cem\u003eNpy\u003c/em\u003e using \u003cem\u003eNpy\u003c/em\u003e gene knockout (\u003cem\u003eNpy\u003csup\u003e–/–\u003c/sup\u003e\u003c/em\u003e) or knockdown (\u003cem\u003eNpy\u003csup\u003e+/–\u003c/sup\u003e\u003c/em\u003e) manipulations, where NPY expression was undetectable or markedly reduced, fear extinction was impaired while fear conditioning remained intact. Moreover, the normal two-phase (Stage 1 \u003cem\u003evs.\u003c/em\u003e Stage 2) decline in freezing during extinction was disrupted in these mutants (Supplementary Fig. 18c–i).\u003c/p\u003e\n\u003cp\u003eTo further examine the cell-type-specific contribution of NPY, we employed a Cre-dependent CRISPR-Cas9 strategy to selective disrupt \u003cem\u003eNpy\u003c/em\u003e in vCA1 NPY\u003csup\u003e+\u003c/sup\u003e interneurons in NPY-Cre::Cas9 [lox-stop-lox-Streptococcus pyogenes Cas9 (SpCas9)] double transgenic mice. By targeting exons 2 and 3 of the \u003cem\u003eNpy\u003c/em\u003e gene, Cas9 abolished the \u003cem\u003eNpy\u003c/em\u003e expression in these neurons. Adeno associated virus (AAV) delivery of \u003cem\u003eNpy\u003c/em\u003e-targeting guide RNAs significantly reduced \u003cem\u003eNpy\u003c/em\u003e expression in NPY\u003csup\u003e+\u003c/sup\u003e neurons compared to a control AAV containing scramble guide RNAs (Fig. 3o). Location-specific \u003cem\u003eNpy\u003c/em\u003e knockout in vCA1 NPY\u003csup\u003e+\u003c/sup\u003e neurons markedly impeded extinction but had no effect on fear conditioning (Fig. 3p). Notably, the two-phase decline in freezing was eliminated during extinction (Fig. 3q and Supplementary Fig. 18j, k). Thess results indicate that, consistent with optogenetic manipulations of vCA1 NPY\u003csup\u003e+\u003c/sup\u003e interneuron activity, altering NPY peptide levels in this region bidirectionally modulates the efficacy of fear extinction.\u003c/p\u003e\n\u003cp\u003eGiven that NPY\u003csup\u003e+\u003c/sup\u003e neurons in the vCA1 are GABAergic (Supplementary Figs. 8 and 9), we investigated whether co-released GABA by these cells is also involved in the regulation of fear and extinction. To this end, we expressed Cre-dependent ChrimsonR along with the GRAB\u003csub\u003eGABA0.8\u003c/sub\u003e sensor in NPY-Cre mice (Extended Data Fig. 5a). Optogenetic stimulation of NPY\u003csup\u003e+\u003c/sup\u003e neurons triggered GABA release (Extended Data Fig. 5a, b), in parallel with NPY release (Fig. 2a–c). Selective ablation of GABA release from vCA1 NPY\u003csup\u003e+\u003c/sup\u003e neurons—via Cre-dependent CRISPR-Cas9 targeting the \u003cem\u003eSlc32a1\u003c/em\u003e gene (which encodes the vesicular GABA transporter)—prevented fear conditioning (Extended Data Fig. 5c–i).\u0026nbsp;Similarly, optogenetic inhibition of NPY\u003csup\u003e+\u003c/sup\u003e neuron activity during conditioning suppressed both NPY and GABA co-release and led to normal fear conditioning and reduced fear responses at retrieval (Extended Data Fig. 5j, k). In contrast, genetic ablation of the entire NPY\u003csup\u003e+\u003c/sup\u003e interneuron population in vCA1 attenuated fear conditioning and subsequent memory retrieval (Extended Data Fig. 5l, m). Collectively, these findings illustrate distinct yet complementary roles of GABA and NPY in memory processes. Moreover, manipulating NPY\u003csup\u003e+\u003c/sup\u003e neurons in the hypothalamic arcuate nucleus (ARC) did not affect fear memory (Supplementary Fig. 19), reinforcing the region-specific functions of these neurons. Altogether, our data reveal a dual mechanism wherein the fast GABA release from vCA1 NPY\u003csup\u003e+\u003c/sup\u003e interneurons is essential for encoding fear memory, whereas the slower NPY-mediated signaling is predominantly required for extinction.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eNon-overlapping neurons with distinct NPY receptor subtypes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo further investigate the downstream targets of NPY in fear extinction, we analyzed the single-cell transcriptomic data presented in Fig. 1a and identified three primary types of NPY receptor genes—\u003cem\u003eNpy1r\u003c/em\u003e, \u003cem\u003eNpy2r\u003c/em\u003e, and \u003cem\u003eNpy5r\u003c/em\u003e—across various cell types in the vCA1. Both excitatory and inhibitory neurons predominantly expressed \u003cem\u003eNpy1r\u003c/em\u003e and \u003cem\u003eNpy2r\u003c/em\u003e, whereas \u003cem\u003eNpy5r\u003c/em\u003e was scarce (Supplementary Fig. 20a–c). A cross-cell-type dot-plot further showed that \u003cem\u003eNpy1r\u003c/em\u003e transcripts were detected in neurons and, most prominently, in pericytes, while \u003cem\u003eNpy2r\u003c/em\u003e was largely confined to neurons; \u003cem\u003eNpy5r\u003c/em\u003e appeared only at trace levels across all cell types, with occasional reads in excitatory and inhibitory neurons (Supplementary Fig. 20d). Given our focus on neuronal engrams, we opted to subsequent analyses to \u003cem\u003eNpy1r\u003c/em\u003e and \u003cem\u003eNpy2r\u003c/em\u003e in neurons and did not pursue \u003cem\u003eNpy5r\u003c/em\u003e further. Notably, \u003cem\u003eNpy1r\u003c/em\u003e and \u003cem\u003eNpy2r\u003c/em\u003e were predominantly found in non-overlapping neuronal populations (Fig. 4a, b and Supplementary Fig. 20e) and were negligibly distributed within\u0026nbsp;\u003cem\u003eNpy\u003csup\u003e+\u003c/sup\u003e\u003c/em\u003e inhibitory neurons (Supplementary Fig. 20f), suggesting that NPY does not functionally overlap with its receptor subtypes in a cell-autonomous manner.\u003c/p\u003e\n\u003cp\u003eTo resolve receptor specificity at single-cell resolution, we combined genetic tagging with multiplex RNAscope. NPY2R-Cre mice received AAV-DIO-EGFP in vCA1 to label NPY2R-expressing neurons. Two weeks later, high-magnification confocal z-stacks with 3D surface rendering allowed unambiguous assignment of \u003cem\u003eNpy1r\u003c/em\u003e and \u003cem\u003eNpy2r\u003c/em\u003e puncta to individual EGFP\u003csup\u003e+\u003c/sup\u003e somata (Fig. 4c–f). Within the anatomically defined layers of vCA1 region\u003csup\u003e40\u003c/sup\u003e, layer-resolved quantification revealed a pronounced laminar bias: \u003cem\u003eNpy1r\u003c/em\u003e puncta were enriched in the superficial (sps) and middle \u0026nbsp;(spm) pyramidal layers and declined toward the deep pyramidal layer (spd), whereas \u003cem\u003eNpy2r\u003c/em\u003e puncta peaked in spd and were scarce in stratum oriens (SO) (Fig. 4e and Supplementary Fig. 20g). At the cellular level, analysis of 446 EGFP-labelled neurons showed that the great majority contained \u003cem\u003eNpy2r\u003c/em\u003e puncta only; a small minority co-expressed \u003cem\u003eNpy2r\u003c/em\u003e and \u003cem\u003eNpy1r\u003c/em\u003e; and only a single EGFP-positive neuron harbored \u003cem\u003eNpy1r\u003c/em\u003e puncta without \u003cem\u003eNpy2r\u003c/em\u003e (Fig. 4f). Consistent with the somatic location of the peptide source, \u003cem\u003eNpy\u003c/em\u003e mRNA puncta were concentrated in SO interneurons bordering spd (Supplementary Fig. 20h, i). Together, these data integrate the scRNA-seq overview with spatially resolved histology, showing that NPY produced by SO interneurons is positioned to influence largely segregated, layer-biased pyramidal subpopulations defined by \u003cem\u003eNpy1r\u003c/em\u003e or \u003cem\u003eNpy2r\u003c/em\u003e in vCA1.\u003c/p\u003e\n\u003cp\u003eLayer-resolved immunofluorescence for the receptor proteins revealed a related but non-identical pattern: NPY1R signal was strongest in the spd and weaker in spm and sps, while NPY2R signal was relatively enriched in sps and spd compared with spm (Extended Data Fig. 6a, b). Antibody specificity was verified by two orthogonal controls. \u003cem\u003eFirst\u003c/em\u003e, in NPY2R-Cre mice expressing AAV-DIO-EGFP in vCA1, nearly all EGFP-labelled neurons were NPY2R-positive (99.2% EGFP\u003csup\u003e+\u003c/sup\u003e/NPY2R\u003csup\u003e+\u003c/sup\u003e) and only a small minority co-labelled with NPY1R (4.8% EGFP\u003csup\u003e+\u003c/sup\u003e/NPY1R\u003csup\u003e+\u003c/sup\u003e in somata, 9.2% EGFP\u003csup\u003e+\u003c/sup\u003e/NPY1R\u003csup\u003e+\u003c/sup\u003e in neurites). High-magnification images revealed NPY2R puncta outlining somata and proximal processes (Extended Data Fig. 6c, d). \u003cem\u003eSecond\u003c/em\u003e, in Syn-Cre::Cas9 mice, \u003cem\u003eNpy1r\u003c/em\u003e sgRNA selectively reduced NPY1R signal across vCA1 layers, whereas \u003cem\u003eNpy2r\u003c/em\u003e sgRNA selectively diminished NPY2R staining, shifting the per-cell intensity distribution leftward with a trend toward a lower mean (Extended Data Fig. 6e–i). These controls confirm the specificity of both antibodies for quantitative analyses.\u003c/p\u003e\n\u003cp\u003eHigh‑magnification confocal imaging with 3D compartment assignment revealed distinct subcellular biases: NPY1R immunoreactivity extended prominently along neurites, whereas NPY2R was concentrated on somatic membranes with comparatively sparse labeling in processes (Extended Data Fig. 6c–e). Compartment-assigned quantification confirmed a higher neurite‑to‑soma ratio for NPY1R than for NPY2R. Because RNAscope primarily detects soma-localized transcripts, a partial mismatch between the mRNA and protein maps is expected and most likely reflects trafficking and turnover of these G‑protein-coupled receptors (i.e., NPY1R and NPY2R) rather than protein expression in the absence of cognate mRNA. This arrangement supports a model in which NPY released from SO‑biased interneurons engages receptor fields with different subcellular reach—NPY1R distributed along neurites and NPY2R enriched at somata—across the pyramidal layers of vCA1.\u003c/p\u003e\n\u003cp\u003eNext, we examined the single-cell transcriptomes of \u003cem\u003eNpy1r\u003c/em\u003e- and \u003cem\u003eNpy2r\u003c/em\u003e-expressing neurons to identify DEGs. Under Ext conditions, two DEGs were enriched in \u003cem\u003eNpy2r\u003c/em\u003e-expressing neurons, whereas eight DEGs were enriched in \u003cem\u003eNpy1r\u003c/em\u003e-expressing neurons; none of DEGs except \u003cem\u003eNpy1r\u003c/em\u003e and \u003cem\u003eNpy2r\u003c/em\u003e in these two neuronal populations were identified under No Ext conditions (Fig. 4g). Consistent with these findings, analysis of IEG expression revealed no significant difference between \u003cem\u003eNpy2r\u003c/em\u003e- and \u003cem\u003eNpy1r\u003c/em\u003e-expressing neurons under the No Ext condition. However, under Ext, \u003cem\u003eNpy2r\u003c/em\u003e-expressing neurons showed significantly lower IEG expression than \u003cem\u003eNpy1r\u003c/em\u003e-expressing neurons (Supplementary Fig. 21). These observations suggest that NPY released in the vCA1 during extinction engages distinct transcriptional programs in \u003cem\u003eNpy1r\u003c/em\u003e- and \u003cem\u003eNpy2r\u003c/em\u003e-expressing neurons, potentially giving rise to two sub-ensembles that mediate fear versus extinction engram functions. To further investigate this bivalent role of NPY in shaping memory ensembles, we optogenetically activated NPY\u003csup\u003e+\u003c/sup\u003e interneurons during the No Ext condition, when these neurons are typically less active. This manipulation increased \u003cem\u003eNpy1r\u003c/em\u003e-expressing neurons while decreasing \u003cem\u003eNpy2r\u003c/em\u003e-expressing neurons (Extended Data Fig. 7a–e). Conversely, optogenetically inhibiting NPY\u003csup\u003e+\u003c/sup\u003e interneurons during Ext, when they are normally highly active, reduced \u003cem\u003eNpy1r\u003c/em\u003e-expressing neurons but elevated \u003cem\u003eNpy2r\u003c/em\u003e-expressing neurons (Extended Data Fig. 7f–j). Thus, modulating \u003cem\u003eNPY\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e interneuron activity at specific time point is sufficient to alter the proportions of \u003cem\u003eNpy1r\u003c/em\u003e- and \u003cem\u003eNpy2r\u003c/em\u003e-tagged neuronal ensembles.\u003c/p\u003e\n\u003cp\u003eTo clarify the roles of \u003cem\u003eNpy1r\u003c/em\u003e- and \u003cem\u003eNpy2r\u003c/em\u003e-expressing sub-ensembles in fear- and/or extinction-engram neurons in the vCA1, we employed the targeted recombination in active populations (TRAP) system\u003csup\u003e31, 41\u003c/sup\u003e in TRAP2::Ai9 (lox-stop-lox-tdTomato) double transgenic mice. Specifically, we used intraperitoneal\u0026nbsp;4-hydroxytamoxifen (4-OHT) injections to tag active neurons (i.e., label them with tdTomato) during three distinct behavioral epochs: (1) Fear retrieval (No Ext): exposure to two CS at high fear levels; (2) Extinction learning (Ext): exposure to 20 CS with progressively decreasing fear; and (3) Extinction retrieval (Ext Retr): exposure to eight CS on the day following extinction at much lower fear levels. \u003cem\u003eIn situ\u003c/em\u003e hybridization (Fig. 4h) revealed that Ext, but not Ext Retr, increased the total number of TRAPed neurons compared to fear retrieval. Moreover, fear retrieval increased both the absolute number and the relative proportion of \u003cem\u003eNpy1r\u003c/em\u003e-expressing neurons while decreasing those of \u003cem\u003eNpy2r\u003c/em\u003e-expressing neurons within the tagged ensemble. In contrast, extinction and extinction retrieval generated the opposite pattern, preferentially recruiting \u003cem\u003eNpy2r\u003c/em\u003e-expressing neurons over \u003cem\u003eNpy1r\u003c/em\u003e-expressing neurons (Fig. 4i–k).\u003c/p\u003e\n\u003cp\u003eTo further elucidate the mechanistic link between \u003cem\u003eNpy1r\u003c/em\u003e- and \u003cem\u003eNpy2r\u003c/em\u003e-expressing neuronal sub-ensembles and their roles in extinction and fear engrams, we employed a combination of TRAP tagging and Fos labeling to more definitively identify fear engram neurons as those co-expressing TRAPed tdTomato and Fos. We tagged neurons that were active during fear retrieval (No Ext) in TRAP2::Ai9 double transgenic mice, then performed \u003cem\u003ein situ\u003c/em\u003e hybridization to reveal the expression of \u003cem\u003eFos\u003c/em\u003e, \u003cem\u003eNpy\u003c/em\u003e, NPY receptors \u003cem\u003eNpy1r\u003c/em\u003e and \u003cem\u003eNpy2r\u003c/em\u003e (Extended Data Fig. 8a). As expected, a significantly higher proportion of \u003cem\u003etdTomato\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e neurons in the No Ext condition were reactivated (\u003cem\u003etdTomato\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e and \u003cem\u003eFos\u003c/em\u003e\u003csup\u003e+\u003c/sup\u003e) compared to the Ext condition, suggesting that these overlapping populations encode fear memory (Extended Data Fig. 8b, c). Notably, among these reactivated neurons in No Ext, we observed a much greater overlap with \u003cem\u003eNpy2r\u003c/em\u003e than with \u003cem\u003eNpy1r\u003c/em\u003e or \u003cem\u003eNpy\u003c/em\u003e (Extended Data Fig. 8d–f), reinforcing the idea that \u003cem\u003eNpy2r\u003c/em\u003e-expressing sub-ensembles form the core of the fear memory engram.\u003c/p\u003e\n\u003cp\u003eBecause the transition from fear to extinction is largely irreversible within this time window, we were unable to assess the overlap between reactivated extinction-engram neurons and NPY or its receptors in the comparable paradigm. Nevertheless, together with previous findings, these observations raised an intriguing possibility that NPY differentially engages two receptor-defined neuronal sub-ensembles to support fear versus extinction memory. While extinction learning and retrieval appeared to modestly impact the overall size of the engram relative to fear retrieval, they drove a more pronounced switch from \u003cem\u003eNpr2r\u003c/em\u003e- to \u003cem\u003eNpr1r\u003c/em\u003e-expressing ensembles. This shift underscores the mechanistic bifurcation of NPY signaling in regulating the balance between memory lability and stability.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGenetic deletion or pharmacological perturbations of \u003cem\u003eNpy1r\u003c/em\u003e or \u003cem\u003eNpy2r\u003c/em\u003e sub-ensembles selectively impact memory liability and stability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo dissect the functional roles of \u003cem\u003eNpy1r\u003c/em\u003e and \u003cem\u003eNpy2r\u003c/em\u003e sub-ensembles in the vCA1, we took advantage of TRAP2::Cas9 (lox-stop-lox-Cas9) double transgenic mice in which Cas9 is specifically expressed in fear- or extinction-tagged neurons. Within these TRAPed populations, guide RNAs targeting the \u003cem\u003eNpy2r\u003c/em\u003e or \u003cem\u003eNpy1r\u003c/em\u003e genes abolished the expression of these genes, as confirmed by post hoc analyses (Supplementary Fig. 22). We thus evaluated how fear- and extinction-tagged ensembles are affected by selectively knocking out \u003cem\u003eNpy1r\u003c/em\u003e or \u003cem\u003eNpy2r\u003c/em\u003e (Fig. 5a). In fear-tagged neurons, \u003cem\u003eNpy1r\u003c/em\u003e deletion did not alter either Stage 1 or Stage 2 of extinction. In contrast, \u003cem\u003eNpy2r\u003c/em\u003e deletion significantly slowed Stage1 and diminished Stage 2 (Fig. 5b, c and Supplementary Fig. 23a–c), mirroring the effect of optogenetically inhibiting NPY\u003csup\u003e+\u003c/sup\u003e interneurons (Fig. 3c, d). Leveraging the phenomenon of spontaneous recovery of extinguished fear over time\u003csup\u003e23, 25\u003c/sup\u003e, we next evaluated extinction relearning and retrieval after deleting \u003cem\u003eNpy1r\u003c/em\u003e or \u003cem\u003eNpy2r\u003c/em\u003e in extinction-tagged neurons that were active during the initial round of extinction learning. Notably, \u003cem\u003eNpy1r\u003c/em\u003e knockout enhanced both the rate and magnitude of extinction across both Stage 1 and Stage 2, whereas \u003cem\u003eNpy2r\u003c/em\u003e knockout had minimal effects (Fig. 5d, e and Supplementary Fig. 23d–f). No-shock controls at matching time points did not modify the ratio of \u003cem\u003eNpr2r\u003c/em\u003e- and \u003cem\u003eNpr1r\u003c/em\u003e- expressing ensembles observed in Fig. 4h–k; moreover, TRAPed\u0026nbsp;neuron-specific \u003cem\u003eNpy1r\u003c/em\u003e or \u003cem\u003eNpy2r\u003c/em\u003e deletion in the absence of shock failed to induce significant freezing, ruling out non-specific behavioral influences (Supplementary Fig. 24). Together, these results suggest that NPY2R in fear-tagged ensembles and NPY1R in extinction-tagged ensembles exert opposing influences on extinction: one prolonging fear and the other facilitating extinction.\u003c/p\u003e\n\u003cp\u003eTo directly probe the roles of NPY1R or NPY2R in vCA1, we locally delivered receptor-specific agonists or antagonists via an implanted cannula immediately prior to extinction learning (Fig. 5f). Strikingly, pharmacological activation of NPY2R led to an overly rapid extinction in Stage 1 with attenuated Stage 2, whereas NPY1R agonism slightly delayed Stage 1 extinction and stabilized freezing at a higher level (Fig. 5g, h and Supplementary Fig. 25a–c). Conversely, antagonizing NPY1R accelerated extinction, whereas NPY2R antagonism hindered it (Fig. 5i, j and Supplementary Fig. 25d–f), consistent with the results from the genetic deletion experiments and electrophysiology analyses of neuronal excitability and synaptic transmission in these mice (Extended Data Fig. 9 and 10). Pharmacological manipulation of NPY5R in the vCA1 had no discernible effect on fear extinction (Supplementary Fig. 26). These findings underscore distinct yet complementary actions of NPY1R and NPY2R in the fear extinction process and support the idea that NPY interneuron co-opts its roles in fear memory and extinction by activity-dependent release of NPY itself to sequentially act upon \u003cem\u003eNpy2r-\u003c/em\u003e and \u003cem\u003eNpy1r\u003c/em\u003e-expressing neurons, thereby permitting the transition from fear to extinction sub-ensembles. Specifically, NPY2R sub-ensemble engages in Stage 1 extinction, gating memory lability, while NPY1R functions in Stage 2 to determine the setpoint of memory stability during extinction Supplementary Fig. 27).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eMemories persist or update according to how their constituent ensembles are recruited and reweighted over time. Activity-dependent tagging has established that sparsely distributed ensembles are re-engaged at recall\u003csup\u003e1, 42\u003c/sup\u003e, yet this IEG-based framework (e.g., Fos, Arc) offers limited insight into how inhibitory motifs and neuromodulators bias an engram toward lability or stability. Here we identify a peptidergic control axis in vCA1: NPY\u003csup\u003e+\u003c/sup\u003e interneurons co-opt receptor-defined neuronal subpopulations to tilt a single memory trace from fear expression toward extinction. In this scheme, fast GABAergic output from NPY\u003csup\u003e+\u003c/sup\u003e cells facilitates acquisition, whereas slower NPY release bifurcates onto two largely non-overlapping targets—NPY2R\u003csup\u003e+\u003c/sup\u003e and NPY1R\u003csup\u003e+\u003c/sup\u003e neurons—thereby rebalancing ensemble dominance across time. This linkage between real-time peptide dynamics and cell-type specificity provides a molecular logic for how inhibitory interneurons determine which neuronal ensembles govern behavior.\u003c/p\u003e\n\u003cp\u003eA central feature of this logic is temporal staging, but we do not view the late NPY signal to mediate a second phase of freezing. During extinction, NPY\u003csup\u003e+\u003c/sup\u003e interneuron activity and peptide release rise in step with transitions from “fear-on” to “fear-off”, consistent with a rapid, cue-locked component that modulates behavior on the timescale of seconds. By contrast, the elevation of extracellular NPY that persists for ~2 h after training is better viewed as a slower, tonic molecular state—coincident with increased \u003cem\u003eNpy\u003c/em\u003e transcription—than as continued moment-to-moment control of freezing. We therefore regard the prolonged signal as a consolidation-related aftereffect that likely supports synaptic and network plasticity (for example, by sustaining peptide availability, influencing receptor occupancy/trafficking, or altering interneuron-pyramidal neuron coupling) and prepares the circuit for future encounters with the cue, rather than as a direct driver of rapid behavioral transitions. In this light, our “two-component” description is intended to distinguish a fast, phasic NPY pulse that updates behavior online from a slower, metaplastic NPY tone that biases subsequent retrieval and limits relapse. Because our single-cell profiling was sampled 2 h after training, it most plausibly captures this consolidation-associated state; higher-temporal-resolution tagging will be required to resolve the precise sequence of neuronal recruitment. Taken together, we propose a model in which a brief NPY surge loosens fear-biased representations during learning, while a lingering NPY milieu stabilizes the extinction bias for future expression.\u003c/p\u003e\n\u003cp\u003eTRAP-based analyses anchor our model to receptor identity within engrams. Fear‑tagged neuronal ensembles were enriched for \u003cem\u003eNpy2r\u003c/em\u003e, whereas extinction‑tagged ensembles were enriched for \u003cem\u003eNpy1r\u003c/em\u003e. Causal manipulations aligned with this segregation: genetic or pharmacological disruption of NPY2R selectively blunted the early, labile phase of extinction, while engaging NPY1R set the late, stable asymptote reached after training. Thus, the two behavioral phases map onto receptor‑defined subpopulations embedded within the tagged ensembles. More broadly, these findings argue that engrams are not specified solely by which neurons fire together, but also by a modulator-receptor scaffold that biases the eligibility of specific neuronal sub-ensembles to carry the trace at different stages. In this view, IEG-based tagging identifies the engram membership, while peptidergic inhibition specifies the timing and strength of memory trace—steering the same network toward lability (NPY→NPY2R) or stability (NPY→NPY1R). The observed shift from \u003cem\u003eNpy2r\u003c/em\u003e- to \u003cem\u003eNpy1r\u003c/em\u003e-dominated ensembles across extinction is therefore mechanistic rather than merely descriptive: it shows how a single interneuron class can orchestrate a reversible transformation of memory state.\u003c/p\u003e\n\u003cp\u003eThe circuit architecture and biophysics of signaling help explain this progression. NPY is produced locally by SO interneurons in vCA1, whereas NPY2R- and NPY1R-expressing targets are stratified across deep versus superficial/middle pyramidal layers, with only rare co-expression. Protein distributions are broader than soma-localized mRNA—consistent with receptor trafficking—and NPY acts by volume transmission, diffusing from its release sites to receptor fields at different radii. Superimposed on this geometry is receptor pharmacology\u003csup\u003e43, 44\u003c/sup\u003e: the higher affinity of NPY2R predicts preferential engagement when peptide levels are low, with NPY1R increasingly recruited as NPY accumulates. Together, these factors naturally yield a stage-dependent receptor switch during extinction. Perturbations of the interneuron source further support this logic. Optogenetic activation of NPY\u003csup\u003e+\u003c/sup\u003e interneurons in the low-activity No Ext state increased the fraction of \u003cem\u003eNpy1r\u003c/em\u003e-expressing cells and reduced \u003cem\u003eNpy2r\u003c/em\u003e-expressing cells within the tagged population, facilitating extinction. Conversely, inhibiting NPY\u003csup\u003e+\u003c/sup\u003e interneurons during Ext had the reciprocal effect—reducing \u003cem\u003eNpy1r\u003c/em\u003e and increasing \u003cem\u003eNpy2r\u003c/em\u003e—and attenuated extinction. Thus, extinction involves not only changes in engram size but, more critically, a reweighting of receptor-defined sub-ensembles from \u003cem\u003eNpy2r\u003c/em\u003e to \u003cem\u003eNpy1r\u003c/em\u003e. Functionally, we propose that NPY2R\u003csup\u003e+\u003c/sup\u003e neurons act as a lability gate: early, modest NPY release preferentially engages this high-affinity population to loosen fear-biased representations and lower the threshold for updating. As extinction proceeds and peptide levels build, recruitment of NPY1R\u003csup\u003e+\u003c/sup\u003e neurons stabilizes the extinction-biased set point, limiting over-generalization and conferring persistence. This receptor-stage choreography reframes extinction as more than global inhibition; it is a peptidergic switch that rebalances competing ensembles across time—first loosening the old trace, then fastening the new one—consistent with the dissociation we observe between early rate changes and late asymptotic performance.\u003c/p\u003e\n\u003cp\u003eThese principles suggest several general interpretations. \u003cem\u003eFirst\u003c/em\u003e, interneuron‑derived neuropeptides can endow inhibitory circuits with time‑scale diversity, coupling millisecond GABA to minutes‑to‑hours peptide action to bridge online control and consolidation. \u003cem\u003eSecond\u003c/em\u003e, receptor heterogeneity partitions neurons into functionally addressable sub‑ensembles, allowing a single interneuron class to steer the same trace along distinct trajectories. \u003cem\u003eThird\u003c/em\u003e, ensemble “identity” is not fixed; it is re‑weighted by the momentary neurochemical milieu, providing a mechanistic substrate for the coexistence of fear and extinction memories within shared circuitry.\u003c/p\u003e\n\u003cp\u003eThe model yields testable predictions and future directions. Temporally gated activity-dependent tagging should reveal a leading-edge recruitment of NPY2R\u003csup\u003e+\u003c/sup\u003e cells that gives way to NPY1R\u003csup\u003e+\u003c/sup\u003e cells across extinction trials. Closed-loop manipulations that boost NPY specifically during “fear-on” epochs should accelerate the NPY2R-dominated lability phase, whereas late-epoch NPY1R augmentation should preferentially improve retention. Quantifying NPY diffusion and receptor occupancy \u003cem\u003ein vivo\u003c/em\u003e—together with longitudinal single-cell transcriptomics and spatial proteomics—will clarify how peptide kinetics, receptor trafficking, and laminar architecture shape ensemble reweighting. Beyond vCA1, analogous peptidergic switches may generalize to other forms of adaptive learning and to other interneuron peptides (SST, CCK)\u003csup\u003e13-15, 38\u003c/sup\u003e, raising the prospect of a broader neuropeptidergic grammar for engram control.\u003c/p\u003e\n\u003cp\u003eNPY is broadly acting cotransmitter\u003csup\u003e45-47\u003c/sup\u003e that signals through Gi/o‑coupled receptors\u003csup\u003e48\u003c/sup\u003e. Within vCA1, our data define a region‑ and cell‑type‑specific mechanism by which NPY dynamically tunes memory: peptide released from NPY\u003csup\u003e+\u003c/sup\u003e interneurons sequentially engages largely non‑overlapping NPY2R\u003csup\u003e+\u003c/sup\u003e and NPY1R\u003csup\u003e+\u003c/sup\u003e neuronal sub‑ensembles, biasing the same trace toward lability or stability. This spatiotemporal control layer complements activity‑dependent engram tagging by linking peptide dynamics to ensemble selection, providing a mechanistic bridge from interneuron signaling to behavioral adaptation and aligning with the therapeutic promise of NPY pathways in stress‑ and trauma‑related states\u003csup\u003e49\u003c/sup\u003e. The stage specificity of this control suggests clear translational avenues. Brief, well-timed enhancement of NPY→NPY2R signaling should transiently increase engram lability and thereby potentiate exposure-based therapy, whereas the late NPY→NPY1R arm constrains extinction by stabilizing a conservative set point—implying that timed NPY1R antagonism, rather than engagement, would be expected to deepen extinction and reduce relapse. Any such strategy will demand precise control of timing, circuit targeting, and receptor selectivity, together with rigorous cross-species validation. Even so, aligning intervention windows with the two-component NPY dynamics as we described—an early, NPY2R-dominated phase that loosens fear-biased representations and a later, NPY1R-dominated phase that caps further plasticity—offers a concrete framework for future preclinical intervention.\u003c/p\u003e\n\u003cp\u003eIn conclusion, NPY\u003csup\u003e+\u003c/sup\u003e interneurons in vCA1 do not merely suppress activity; they co‑opt receptor‑defined neuronal sub-ensembles to govern memory liability and stability. By tiling a fast modulatory pulse to a slower, gene‑dependent wave, the NPY system provides a molecular continuum for gating memory fate.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eMice\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll animal procedures were ethically approved by the Animal Ethics Committee of Shanghai Jiao Tong University School of Medicine and the Institutional Animal Care and Use Committee (Department of Laboratory Animal Science, Shanghai Jiao Tong University School of Medicine; Policy Number DLAS-MP-ANIM. 01–05). Mice were housed in groups under a 12-h light/dark cycle with unrestricted access to food and water. Adult male mice (7–12 weeks old, C57BL/6J background) were selected for all experiments. The study utilized the following mouse strains: NPY-Cre (\u003cem\u003eNpy\u003c/em\u003e-IRES-Cre, JAX, catalog number 027851), NPY2R-Cre (\u003cem\u003eNpy2r\u003c/em\u003e-ires-Cre, JAX, catalog number\u0026nbsp;029285), Fos2A-iCreER (TRAP2, JAX, catalog number 030323), Rosa26-Ai9 tdTomato reporter (JAX, catalog number 007909), and \u003cem\u003eNpy\u003c/em\u003e-hrGFP (JAX, catalog number 006417), and CRISPR/Cas9 knock-in (JAX, catalog number 024857) mice. \u003cem\u003eNpy\u0026nbsp;\u003c/em\u003eknockout (\u003cem\u003eNpy\u003csup\u003e–/–\u003c/sup\u003e\u003c/em\u003e), heterozygous (\u003cem\u003eNpy\u003csup\u003e+/–\u003c/sup\u003e\u003c/em\u003e), and wild-type (\u003cem\u003eNpy\u003csup\u003e+/+\u003c/sup\u003e\u003c/em\u003e) littermates were obtained by intercrossing \u003cem\u003eNpy\u003csup\u003e+/–\u003c/sup\u003e\u003c/em\u003e mice (GemPharmatech). C57BL/6J mice were purchased from the Shanghai Laboratory Animal Center (SLAC), Chinese Academy of Sciences.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFear conditioning, extinction, and extinction retrieval\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll auditory fear conditioning and extinction procedures were conducted using the Ugo Basile Fear Conditioning System (UGO Basile S.R.L., Italy) according to established protocols\u003csup\u003e24\u003c/sup\u003e. Initially, mice were acclimated and habituated to the conditioning chamber for three consecutive days. The conditioning chambers (17 × 17 × 25 cm), equipped with stainless-steel shocking grids, were linked to a precision feedback current-regulated shocker (UGO Basile S.R.L., Italy). During habituation and fear conditioning, the chamber walls were adorned with black-and-white checkered wallpapers (context A) and cleaned with 75% ethanol.\u003c/p\u003e\n\u003cp\u003eOn Day 1, individual mice underwent conditioning in context A, with five pure tones (conditioned stimulus, CS; 4 kHz, 76 dB, 20 s each) presented at fixed intervals (140 s), each paired with a foot shock (unconditioned stimulus, US; 0.5 mA, 2 s). Auditory tones and foot shocks were autonomously controlled by the ANY-maze software (version 7.20, Stoelting Co.). Following conditioning, mice were returned to their home cages 60 s after the final tone, with individual cage floors and walls sanitized with 75% ethanol.\u003c/p\u003e\n\u003cp\u003eOn Day 2 (24 h post-conditioning), the no-extinction training group (No Ext) received two CS presentations (4 kHz, 76 dB, 30 s each) within a 5-min period, while the extinction training group (Ext) received 20 CS presentations within a 25-min period, without foot shocks. Both No Ext and Ext tests were conducted in a chamber with a gray, non-shocking plexiglass floor and dark gray wallpaper (context B), and the testing environment was cleaned with a 4% acetic acid solution between individual mouse tests. To minimize CS anticipation during extinction, a 30-s tone duration (different from the 20-s conditioning duration) was used. To control for time differences between groups No Ext and Ext groups, we designed three variants of the No Ext conditions (1) a short exposure group receiving two CS presentations immediately before removal from context B, (2) a group receiving two CS presentations followed by an extended waiting period equivalent to the remaining 18 CS presentations in the Ext group, but without additional tone exposure, and (3) a group that remained in context B for a period to 18 additional CS presentations (without tone), followed by two CS presentations before removal.\u003c/p\u003e\n\u003cp\u003eOn Day 3, extinction retrieval (Ext Retr) was performed, where mice underwent eight CS presentations in context B. Both No Ext and Ext conditions, as well as extinction retrieval tests, combined two CS presentations into a single block.\u003c/p\u003e\n\u003cp\u003eThroughout testing, the chamber was placed in a sound-attenuating enclosure with a ventilation fan and a house light (UGO Basile S.R.L., Italy). Mouse locomotion within the chamber was recorded using a near-infrared camera and analyzed in real-time by ANY-maze software. The freezing score, a dimensionless metric generated by the software, identified freezing periods based on predefined threshold settings. A fear response was characterized by measurable behavioral freezing, defined as a cessation of movement lasting more than 2 s. The duration of freezing was quantified for analysis. For conditioned fear response analysis, the total duration of freezing during each tone (CS) presentation was divided by the sound exposure time (30 s) and multiplied by 100%. For contextual fear response analysis, the total fear duration in the specific context was divided by the contextual exposure time and multiplied by 100%.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOptogenetic manipulations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA wired optogenetic system was used to modulate neuronal activity during behavioral assays, utilizing a light-emitting diode (LED) (Hangzhou Newdoon Technology Co. Ltd) connected to an optic patch cord via connectors at both ends. Blue light (470 nm, 4–5mW) was delivered in 10-ms pulses at 20 Hz during extinction training. For optogenetic inhibition, red light (638 nm, 8–10 mW) was consistently administered. The light pulse duration was set to exceed 5 s before and after the CS presentations to ensure full CS exposure during extinction training (30-s duration for each tone). To activate NPY\u003csup\u003e+\u003c/sup\u003e neurons during freezing response to CS presentations during extinction, the ANY-maze software detected a fear response and automatically trigger blue light (470 nm, 4–5mW). The blue light was turned off once the fear response ceased. The final output power varied based on the light transmission efficacy of the optical fiber used.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGeneration of single-cell suspensions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTwo hours after undergoing fear memory retrieval (No Ext) or extinction learning (Ext), the mice were deeply anesthetized with 1% sodium pentobarbital and then decapitated. Brains were quickly dissected and chilled in ice-cold N-methyl-D-glucamine-artificial cerebrospinal fluid (NMDG-ACSF) containing the following components (in mM): 93 N-methyl-D-glucamine, 2.5 KCl, 1.2 NaH\u003csub\u003e2\u003c/sub\u003ePO\u003csub\u003e4\u003c/sub\u003e, 30 NaHCO\u003csub\u003e3\u003c/sub\u003e, 20 HEPES, 25\u0026nbsp;D-glucose, 5 Sodium ascorbate, 2 Thiourea, 3 Sodium pyruvate, 10 MgSO\u003csub\u003e4\u003c/sub\u003e, 11.9 N-Acetyl-L-cysteine, 1 CaCl\u003csub\u003e2\u003c/sub\u003e, and 6 ml HCl (pH 7.35–7.45). Coronal brain slices (400 μm thick) containing regions of the vHPC were cut using a vibratome (Leica VT1000S, Germany), then promptly transferred to ice-cold NMDG-ACSF and incubated for 15 min. Actinomycin D (8 μM; A1410, Sigma-Aldrich) and Triptolide (10 μM; T3652, Sigma-Aldrich) were added in this step to minimize artificially induced activation of early immediate genes. The region of vHPC were carefully microdissected using fine tweezers on ice. After tissue collection, it was transferred to an Eppendorf tube containing a digestion solution. This solution consisted of ACSF supplemented with 1 mg/ml pronase (P6911, Sigma-Aldrich), Actinomycin D (8 μM), Triptolide (10 μM), TTX (1 μM) and\u0026nbsp;D-APV (100 μM; A8054, Sigma-Aldrich), and the tissue was digested for 20 min at 34 °C. The tissue was pipetted periodically every 10 min during this digestion. Then, the cell suspension was transferred to Sterile Earle’s Balanced Salt (EBSS) Solution (LK003150, JinpanWorthington) supplemented with Actinomycin D (8 μM), Triptolide (10 μM), TTX (1 μM) and D-APV (100 μM). Finally, it was washed with washing solution consisting of ACSF supplemented with Actinomycin D (8 μM), Triptolide (10 μM), TTX (1 μM, C239, Sigma-Aldrich) and D-APV (100 μM). All solutions were continually bubbled with O\u003csub\u003e2\u003c/sub\u003e and CO\u003csub\u003e2\u003c/sub\u003e (95% O\u003csub\u003e2\u003c/sub\u003e/5% CO\u003csub\u003e2\u003c/sub\u003e, v/v).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSingle-cell library preparation and sequencing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSingle cell RNA-seq library were prepared from the single cell suspensions via the chromium single cell gene expression system (Chromium Single Cell 3’ Reagent Kits v2, 10× genomics), using the default protocols provided by 10× genomics. The NovaSeq platforms were used for sequencing libraries.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGenome alignment and Seurat analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFastq files were aligned to the mouse reference genome mm10 and converted into gene expression matrices using the 10× Genomics Cell Ranger software (v3.0.0). Low-quality cells were filtered out, including those with fewer than 700 transcripts or 200 genes, more than 15,000 transcripts or 2,500 genes, or greater than 5% mitochondrial expression. Genes expressed in fewer than three cells were also excluded.\u003c/p\u003e\n\u003cp\u003eTo visualize the data, dimensionality reduction and clustering analysis were performed on the raw count matrices using the Seurat (v4.3.0) package in R (v4.1.3)\u003csup\u003e50\u003c/sup\u003e. The \u003cem\u003eFindIntegrationAnchors\u003c/em\u003e and \u003cem\u003eIntegrateData\u003c/em\u003e functions were used to integrate the No Ext and Ext datasets. Data were reduced using principal component analysis (PCA). Clusters were identified using the \u003cem\u003eFindNeighbors\u003c/em\u003e and \u003cem\u003eFindClusters\u003c/em\u003e functions with the resolution of 0.5, resulting in 35 cell clusters. Marker genes for each cluster were identified using the \u003cem\u003eFindAllMarkers\u003c/em\u003e function. The following cell types were annotated using selected marker genes listed below: excitatory neurons (\u003cem\u003eThy1\u003c/em\u003e, \u003cem\u003eSlc17a7\u003c/em\u003e), inhibitory neurons (\u003cem\u003eSlc32a1\u003c/em\u003e, \u003cem\u003eGad1\u003c/em\u003e), astrocyte (\u003cem\u003eGfap\u003c/em\u003e), oligodendrocyte (\u003cem\u003eOlig1\u003c/em\u003e), oligodendrocyte precursor cells (\u003cem\u003ePdgfra\u003c/em\u003e), microglia (\u003cem\u003eCx3cr1\u003c/em\u003e), pericytes (\u003cem\u003eVtn\u003c/em\u003e), endothelium (\u003cem\u003eCldn5\u003c/em\u003e), and macrophages (\u003cem\u003eMrc1\u003c/em\u003e). Uniform Manifold Approximation and Projection (UMAP) dimensional reduction was applied to visualize the cell classifications, resulting in 9 cell types. To explore the heterogeneity of inhibitory neurons, we extracted all inhibitory neurons and performed clustering with \u003cem\u003eFindNeighbors\u003c/em\u003e and \u003cem\u003eFindClusters\u003c/em\u003e functions at a resolution of 0.1, yielding 5 subclusters characterized by the expression of specific markers (\u003cem\u003eSlc32a1\u003c/em\u003e, \u003cem\u003eGad1\u003c/em\u003e, \u003cem\u003eGad2\u003c/em\u003e, \u003cem\u003eCck\u003c/em\u003e, \u003cem\u003eVip\u003c/em\u003e, \u003cem\u003eNpy\u003c/em\u003e, \u003cem\u003eSst\u003c/em\u003e, and \u003cem\u003eCalb2\u003c/em\u003e).\u003c/p\u003e\n\u003cp\u003eTo normalize cell numbers between two conditions across all cell types and clusters, we downsampled the larger cell type or cluster to match the size of the smaller one. To reduce stochastic sampling error, we performed 1000 resampling iterations to obtain robust results. Differential expression analysis between the No Ext and Ext was carried out using the FindMarkers function with the method of MAST\u003csup\u003e51\u003c/sup\u003e. When identifying differentially expressed genes (DEGs) across different cell types and clusters, we used a threshold of |log\u003csub\u003e2\u003c/sub\u003eFC| ≥ 0.585, FDR \u0026lt; 0.05 (Benjamini–Hochberg correction) with ≥ 90% resampling support.\u003c/p\u003e\n\u003cp\u003eFor specific subpopulations, we selected cells that expressed a particular gene above a threshold value of zero. \u003cem\u003eNpy1r\u003c/em\u003e- and \u003cem\u003eNpy2r\u003c/em\u003e-expressing neurons were classified based on this criterion. Specifically, we observed that only a very small fraction of neurons coexpressed \u003cem\u003eNpy1r\u003c/em\u003e and \u003cem\u003eNpy2r\u003c/em\u003e in both excitatory and inhibitory neurons. For differential expression analysis, we therefore compared the \u003cem\u003eNpy1r\u003c/em\u003e-specific and \u003cem\u003eNpy2r\u003c/em\u003e-specific subpopulations after excluding coexpressing neurons. To account for differences in population size, subpopulations were balanced by resampling. Differentially expressed genes were defined using thresholds of |log2FC| ≥ 0.25 and FDR \u0026lt; 0.05 (Benjamini–Hochberg correction), and only genes supported in ≥ 90% of resampling iterations were retained.\u003c/p\u003e\n\u003cp\u003eA curated list of 139 IEGs was selected based on previous literature\u003csup\u003e33\u003c/sup\u003e, which includes genes known to be activated in response to various stimuli, including stress, fear, and learning. These genes were identified in different experimental conditions where they were shown to be reliably upregulated upon neuronal activation. To identify activated neurons (IEG\u003csup\u003e+\u003c/sup\u003e), we calculated the 90\u003csup\u003eth\u003c/sup\u003e percentile expression for each of the selected IEGs based on the integrated gene expression data from both No Ext and Ext conditions. A cell was considered activated if its expression of any of these IEGs exceeded the 90\u003csup\u003eth\u003c/sup\u003e percentile value, which indicates a higher level of gene activation compared to the majority of cells. This threshold was chosen to identify cells that exhibited significant transcriptional changes associated with neuronal activation, as supported by the literature\u003csup\u003e35\u003c/sup\u003e. This method of identifying activated neurons provides a robust approach to capture a broad range of neuronal responses during key processes such as fear conditioning and extinction, ensuring that our findings accurately reflect the activation of relevant neuronal populations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eVirus constructs\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe following viruses were utilized: rAAV-EF1α-DIO-NpHR3.0-EGFP (AG26966), rAAV-EF1α-DIO-ChR2-mCherry (AG20297), rAAV-EF1α-DIO-ChR2-EGFP (M3026), rAAV-EF1α-DIO-EGFP (H3303), rAAV-EF1α-DIO-mCherry (AG20299) were procured from Obio Technology Co., Ltd. (Shanghai). rAAV-EF1α-DIO-ChR2-E123T/T159C-mCherry (PT-0015), rAAV-EF1α-DIO-GCaMp6m (PT-0283), rAAV-hSyn-DIO-ChrimsonR-mCherry (PT-1374), and rAAV-hSyn-DIO-hM4D(Gi)-mCherry (PT-0020) were obtained from Brain VTA (Wuhan). rAAV-hSyn-NPY1.0 (YL011001), rAAV-hSyn-NPYmut (YL011003), and rAAV-hSyn-GABA0.8 (YL022001) were purchased from WZ Biosciences Inc (Jinan). rAAV2/9-3x(hU6-sgRNA.sp)(m\u003cem\u003eNpy1r\u003c/em\u003e)-hEF1a-DIO-EGFP\u0026nbsp;(WY4060), rAAV2/9-3x(hU6-sgRNA.sp)(m\u003cem\u003eNpy2r\u003c/em\u003e)-hEF1a-DIO-EGFP (WY3769), rAAV2/9-hU6-sgRNA-hEF1a-DIO-EGFP (S0883), rAAV2/9-H1-sgRNA.sp(m\u003cem\u003eSlc32a1\u003c/em\u003e)x3-CAG-DIO-mCherry \u0026nbsp;(WY3889), rAAV2/9-H1-sgRNA.sp(m\u003cem\u003eNpy\u003c/em\u003e)x3-CAG-DIO-mCherry (WY3909), rAAV2/9-H1-sgRNA.sp(NC)-CAG-DIO-mCherry (S0804) and rAAV2/9-CAG-DIO-taCaspase3 (S0236) were generated by Shanghai Taitool Bioscience Co., Ltd. (Shanghai). All viruses are serotype 2/9. All viral vectors were stored in aliquots at –80°C until further use, with viral titers for injection exceeding 10\u003csup\u003e12\u003c/sup\u003e viral particles per ml.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSurgery\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMice, aged 7–8 weeks, were anesthetized with 1.5–2.0% isoflurane (R510-22, RWD Life Science) in 100% oxygen and subsequently stabilized in a stereotaxic apparatus using atraumatic ear bars (RWD Life Science), ensuring precise alignment on a digital stereotactic frame. A midline scalp incision was made, and small bilateral craniotomies were created using a microdrill with 0.5-mm burrs. Glass pipettes (tip diameter: 10–20 μm) were crafted using a P-1000 Micropipette Puller (Sutter glass pipettes, Sutter Instrument Company) for AAV microinjections. Initially filled with silicone oil, the microinjection pipettes were connected to a microinjector pump (RWD Life Science) to ensure complete air exclusion. AAV-containing solutions were loaded into the pipette tips and injected at specified coordinates (in mm): vCA1 (anteroposterior to bregma, AP, −3.14; lateral to the midline, ML, ±3.20; below the bregma, DV, −4.00) and ARC (AP, −1.70; ML, ±0.30; DV, −5.80). Virus-containing solutions were injected bilaterally into the vCA1 (0.4 μl/side) and unilaterally into the ARC (0.3 μl/side) at a rate of 0.1 μl/min. After injection, the pipette remained in place for an additional 10 min to allow for adequate diffusion of the injectant. Mice were given a minimum of two weeks to recover before undergoing behavioral and other tests, and injection sites were examined at the experiment’s conclusion by assessing the expression of fluorescent proteins, such as EGFP or mCherry.\u003c/p\u003e\n\u003cp\u003eIn optogenetic experiments, ceramic fiber optic cannulas [200 μm in diameter, 0.37 numerical aperture (NA), Hangzhou Newdoon Technology] were surgically implanted above the vCA1 (coordinates: AP, –3.14; ML, ±3.20; DV, –3.90) and ARC (coordinates: AP, –1.70; ML, ±0.30; DV, –5.70). These cannulas were secured in place using acrylic dental cement and skull screws.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOpen field test\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMice were acclimated to the experimental room from their home cages and given a minimum of 1 h for habituation before the start of the experiment. Individual mice were then placed in the outer area of a square Plexiglas open-field apparatus (40 × 40 × 35 cm), divided into a central field (20 × 20 cm) and an outer field. The total distance traveled was quantified using Noldus EthoVision XT (version 16.0, Noldus Information Technology, Netherlands). NPY-Cre mice underwent injections of DIO-ChR2, -NpHR, or -Control virus to modulate the activity of NPY\u003csup\u003e+\u003c/sup\u003e neurons, with customized light stimulation protocols for either activation or inhibition of these neurons. The open-field test consisted of an 18-min session divided into six 3-min epochs, alternating between light off and light om periods, starting with a light-off epoch. For analysis and charts representing only “off” and “on” conditions, the three “off” epochs and three “on” epochs were combined, respectively. To ensure consistency, the open-field arena was thoroughly cleaned with 70% ethanol between each set of tests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFiber photometry\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFiber photometry experiments were performed using a system from Thinker Tech. Fluorescent signals, generated by a 488-nm laser (OBIS 488LS; Coherent), were reflected by a dichroic mirror (MD498; Thorlabs), focused by a 10× objective lens (NA = 0.3; Olympus), and coupled to a rotary joint (FRJ_1 × 1_FC-FC, Doric Lenses). An optical fiber (200 mm O.D., NA = 0.37) was implanted into the vCA1 or ARC during virus injection and remained in place for two weeks for virus expression. Laser power at the fiber tip was adjusted to 40–50 μW to minimize bleaching of GCaMP6m, NPY1.0/NPYmut or GABA0.8 sensor. Excitation fluorescence was collected by the same multi-mode optical fiber and converted into electrical signals by low-light detectors at the detection end to capture neural activity information. Signals were digitized at 100 Hz using a Power 1401 digitizer and Spike 2 software (CED, Cambridge). Throughout behavioral tests, including fear learning, extinction training, or extinction retrieval, fluorescent intensities of GCaMP6m or sensors were recorded, with the pre-sound signal serving as the baseline. Average Ca\u003csup\u003e2+\u003c/sup\u003e or sensor responses were computed using MATLAB, and data were exported to MATLAB mat files from Spike2 for further analysis. Permutational tests were employed for statistical significance, and ΔF/F values were visualized as heatmaps or per-event plots, with shaded areas indicating the standard error of the mean (s.e.m.). The averaged ΔF/F during CS presentation and freezing/no-freezing states was defined, and the area under the curve (AUC, ΔF/F × s) was calculated for the same dataset.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCannula implantation and local drug injection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFollowing anesthesia, mice were securely fixed on a stereotaxic apparatus (RWD Life Science). Stainless steel guide cannulas (RWD Life Science) were bilaterally implanted into the vCA1, with the cannula tips positioned at the following coordinates (in mm): AP, –3.14; ML, ±3.20; DV, –3.90. The cannulas were firmly attached to the skull using acrylic cement and two skull screws. Stainless steel obturators (33 gauges) were inserted into the guide cannulas to prevent obstruction until drug infusion. Animals were allowed a 2-week recovery period post-surgery before undergoing behavioral tests. Mice were familiarized with the infusion procedure three days prior to drug injection. During drug infusion, mice were briefly head-restrained, the stainless-steel obturators were removed, and injection cannulas (33 gauges, RWD Life Science) were inserted into the guide cannulas. The injection cannulas protruded 0.50 mm from the guide cannula tips. The infusion cannula was connected via PE20 tubing to a microsyringe driven by a microinfusion pump (KDS 310, KD Scientific). A total of 0.5 μl of drugs were bilaterally infused into the vCA1 at a controlled flow rate of 0.1 μl per minute. After completing the drug injection, the injection cannulas were left in place for 2 min to facilitate solution diffusion from the cannula tips. Subsequently, the stainless-steel obturators were reinserted into the guide cannulas, and the mice were returned to their home cages for a 15-min recovery period before behavioral tests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDrugs and concentrations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNeuropeptide Y (1153, Tocris Bioscience). NPY1R agonist, [Leu31, Pro34]-Neuropeptide Y (porcine) (HY-P0208, MedChemExpress). NPY2R agonist, N-Acetyl- [Leu\u003csup\u003e28\u003c/sup\u003e, Leu\u003csup\u003e31\u003c/sup\u003e]-Neuropeptide Y Fragment 24-36 (N9398, Sigma). NPY5R agonist, [cPP1-7, NPY19-23, Ala31, Aib32, Gln34]-hPancreatic Polypeptide (HY-P1324, MedChemExpress). The peptide and peptide agonists for NPY receptors were applied at a concentration of 0.25 mM for 0.5 μl. NPY1R antagonist, BIBO 3304 trifluoroacetate (2412, Tocris Bioscience) was applied at a concentration of 0.4 mM for 0.5 μl. NPY2R antagonist, BIIE 0246 hydrochloride (7377, Tocris Bioscience) was applied at a concentration of 2 mM for 0.5 μl. NPY5R antagonist, CGP 71683 hydrochloride (2199, Tocris Bioscience) was applied at a concentration of 5 mM for 0.5 μl.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEngram labeling\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eActivity-dependent recombination was triggered using 4-hydroxytamoxifen (4-OHT, H6278, Sigma-Aldrich). A 20 mg/ml solution of 4-OHT was prepared in ethanol by shaking at 37 °C for 30 min. Subsequently, twice the volume of corn oil (C8267, Sigma-Aldrich) was added to achieve a final concentration of 10 mg/ml of 4-OHT, and the ethanol was removed by vacuum under centrifugation. All injections were administered intraperitoneally.\u003c/p\u003e\n\u003cp\u003eMice were transferred from the vivarium to an adjacent holding room at least 3 h before the test. For fear and extinction memory, activity-dependent neuronal tagging was initiated by a single intraperitoneal injection of 4-OHT (20 mg/kg for mice) after exposure to two CS or 20 CS, respectively. Subsequently, mice were returned to the vivarium, maintaining a regular 12-h light/dark cycle for the remainder of the experiment.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSlice electrophysiology\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWhole-cell recordings were conducted in acute brain slices obtained from behaviorally trained mice. Mice were deeply anesthetized with 1% sodium pentobarbital and subsequently decapitated. Brains were quickly dissected and chilled in ice-cold N-methyl-D-glucamine-artificial cerebrospinal fluid (NMDG-ACSF) containing the following components (in mM): 93 N-methyl-D-glucamine (NMDG), 2.5 KCl, 1.2 NaH\u003csub\u003e2\u003c/sub\u003ePO\u003csub\u003e4\u003c/sub\u003e, 30 NaHCO\u003csub\u003e3\u003c/sub\u003e, 20 HEPES, 25\u0026nbsp;D-glucose, 5 Sodium ascorbate, 2 Thiourea, 3 Sodium pyruvate, 10 MgSO\u003csub\u003e4\u003c/sub\u003e, 11.9 N-Acetyl-L-cysteine, 1 CaCl\u003csub\u003e2\u003c/sub\u003e, and 6 ml HCl (pH 7.35–7.45). Coronal brain slices (300 μm thick) containing regions of the vCA1 were cut with a vibratome (Leica VT1000S, Germany). Slices were initially recovered for 5–10 min in NMDG-ACSF at 31 °C. They were then transferred to another ACSF solution (in mM): 125 NaCl, 2.5 KCl, 12.5\u0026nbsp;D-glucose, 1 MgCl\u003csub\u003e2\u003c/sub\u003e, 2 CaCl\u003csub\u003e2\u003c/sub\u003e, 1.25 NaH\u003csub\u003e2\u003c/sub\u003ePO\u003csub\u003e4\u003c/sub\u003e, and 25 NaHCO\u003csub\u003e3\u003c/sub\u003e (pH 7.35–7.45) for an additional 2 hours of recovery at 31 °C before recordings.\u0026nbsp;The slice was subsequently transferred to a recording chamber and continuously superfused with ACSF at a rate of 1–2 ml per minute. All solutions were continually bubbled with O\u003csub\u003e2\u003c/sub\u003e and CO\u003csub\u003e2\u003c/sub\u003e (95% O\u003csub\u003e2\u003c/sub\u003e/5% CO\u003csub\u003e2\u003c/sub\u003e, v/v). Neurons in the vCA1 were patched under visual guidance using infrared differential-interference contrast microscopy (BX51WI, Olympus), and an optiMOS camera (QImaging, Teledyne Imaging Group). Whole-cell patch-clamp recordings were performed using an Axon 200B amplifier (Molecular Devices). Membrane currents and potentials were sampled and analyzed using a Digidata 1440 interface and a personal computer running Clampex and Clampfit software (pCLAMP 10.5, Molecular Devices). Access resistance was maintained between 10–20 MΩ, and only cells with a change in access resistance \u0026lt; 20% were included in the analysis. In specific recording situations, NPY (1 μM) was added to the ACSF to activate NPY receptors in the vCA1.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eSpike firing.\u003c/em\u003e The spiking activity and membranous properties of cell populations in the vCA1 were quantified using an internal solution comprising the following concentrations (in mM): 145 potassium gluconate, 5 NaCl, 10 HEPES, 2 MgATP, 0.1 Na\u003csub\u003e2\u003c/sub\u003eGTP, 0.2 EGTA, and 1 MgCl\u003csub\u003e2\u003c/sub\u003e (280–300 mOsm, pH 7.2 with KOH). Subsequent data analysis was performed using the MiniAnalysis Program (Version 6.0.1, Synaptosoft) with an amplitude threshold set at 40 mV.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eSpontaneous excitatory postsynaptic currents (sEPSCs)\u003c/em\u003e. For recording sEPSCs in cell populations within the vCA1, a holding potential of –70 mV was maintained. Patch pipettes were filled with a Cs\u003csup\u003e+\u003c/sup\u003e-based solution containing the following concentrations (in mM): 132.5 Cs-gluconate, 17.5 CsCl, 2 MgCl\u003csub\u003e2\u003c/sub\u003e, 0.5 EGTA, 10 HEPES, 2 Na\u003csub\u003e2\u003c/sub\u003eATP, with pH adjusted to 7.3 using CsOH, and osmolarity set at 280–290 mOsm. sEPSCs were recorded for 5-10 min and analyzed from 300 to 400 s after the establishment and stabilization of the recording. Data analysis was conducted using the Mini-analysis Program (Synaptosoft) with an amplitude threshold set at 10 pA, while other parameters remained at their default values.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eOptical stimulation response.\u003c/em\u003e Optical stimulation of ChR2- or NpHR-expressing neurons was conducted using a collimated LED (Lumen Dynamics Group Inc) with peak wavelengths of 473 or 638 nm, respectively. The LED was connected to an Axon 200B amplifier to initiate photostimulation. The brain slice in the recording chamber was illuminated through a 40× water-immersion objective lens (LUMPLFLN 40XW, Olympus). The intensity of photostimulation was directly regulated by the stimulator (2–18 mW/mm\u003csup\u003e2\u003c/sup\u003e), while the duration was set through Digidata 1440 and pClamp 10.5 software. The functional potency of the ChR2-expressing virus was validated by measuring the numbers of action potentials (APs) elicited at different frequencies of blue-light stimulation (1 ms; 5, 10, and 20 Hz) and the inward photocurrents (500-ms pulse) mediated by ChR2 in brain slices. To validate the functional efficacy of NpHR-mediated optogenetic inhibition, red light (638 nm, 500-ms pulse) was administered to attenuate spikes under current clamp mode and induce the outward photocurrents (500-ms pulse) mediated by NpHR.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCRISPR/Cas9-mediated gene knockout\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe CRISPR-associated endonuclease Cas9 from Streptococcus pyogenes (SpCas9) was employed to induce indels in the \u003cem\u003eNpy\u003c/em\u003e (NC_109648), \u003cem\u003eSlc32a1\u0026nbsp;\u003c/em\u003e(NC_22348), \u003cem\u003eNpy1r\u003c/em\u003e (NC_18166), and \u003cem\u003eNpy2r\u003c/em\u003e (NC_18167) genes. Corresponding single guide RNAs (sgRNAs) for each gene were designed and produced by Taitool Bioscience Co. Ltd. using the GRCM38 (Mus musculus) reference genome. Ten sgRNAs were selected for each gene based on computed specificity and efficiency scores (CHOPCHOP). Subsequently, these sgRNAs were cloned into a 2-in-1 reporter vector (CRPT001, Taitool), individually expressing a single sgRNA along with a cut-activated EGFP gene. The sgRNA target sequence was positioned ahead of the EGFP ORF within the vector. The reporter vectors were co-transfected with a Cas9-expressing plasmid into 293T cells. Functional Cas9/sgRNA complexes would cut the target on the reporter vector and rescue the expression of EGFP. Subsequently, the top three sgRNAs were selected and cloned into an adeno-associated virus (AAV) vector tandemly. The chosen sgRNAs were as follows:\u003c/p\u003e\n\u003cp\u003e5’-CGTATGCACCCACCTCTGTC-3’, 5’-AACAAGCGAATGGGGCTGTG-3’, and 5’-CAGAAAACGCCCCCAGAACA-3’ for the \u003cem\u003eNpy\u003c/em\u003e gene targeting exon 2 of the reverse strand, exon 2 of the forward strand, and exon 3 of the forward strand of the \u003cem\u003eNpy\u003c/em\u003e coding region, respectively.\u003c/p\u003e\n\u003cp\u003e5’-GCCACCGATGAGGAAGCGGT-3’, GCGCGTGCGGGACTCGTATG and 5’-GCACGCGATGAGGATCTTGC-3’ for \u003cem\u003eSlc32a1\u003c/em\u003e gene to target exon 2 of the forward strand, exon 3 of the forward strand and exon 3 of the reverse strand of the \u003cem\u003eSlc32a1\u003c/em\u003e coding region.\u003c/p\u003e\n\u003cp\u003e5’-GATGGAGACACACTGTACAA, GATGGACCACTGGGTCTTCG-3’, and 5’-GGCCGAAATACTGCAGCACC-3’ for the \u003cem\u003eNpy1r\u003c/em\u003e gene targeting exon 2 of the reverse strand, exon 2 of the forward strand, and exon 2 of the reverse strand of the \u003cem\u003eNpy1r\u003c/em\u003e coding region, respectively.\u003c/p\u003e\n\u003cp\u003e5’-GTTGGATGCCATTCACTCGG-3’, 5’-GAATCCAGGAATTTCAAGCA-3’, and 5’-GAGAAAGATATGATGCCCAG-3’ for the \u003cem\u003eNpy2r\u003c/em\u003e gene targeting exon 2 of the forward strand, exon 1 of the reverse strand, and exon 2 of the reverse strand, respectively (according to transcript variant 1).\u003c/p\u003e\n\u003cp\u003eA non-targeting negative control guide was designed with a scrambled sequence with no known targets in the genome (5’-GCTGAGTACTTCGAAATGTC-3’). sgRNAs for \u003cem\u003eNpy\u003c/em\u003e and \u003cem\u003eSlc32a1\u0026nbsp;\u003c/em\u003ewas cloned into the pAAV2-H1-sgRNA.sp(MCS)x3-CAG-DIO-mCherry-WPRE-pA vector (AC610-C, Taitool), while sgRNAs for \u003cem\u003eNpy1r\u003c/em\u003e or \u003cem\u003eNpy2r\u003c/em\u003e were cloned into the pAAV-(U6-sgRNA.sp(MCS)) x 3-hEF1a-DIO-EGFP-WPRE-pA vector (AC705-C, Taitool). These plasmids were co-transfected into HEK293 cells with AAV9 Cap. Rep and adenovirus helper to produce recombinant AAV vectors. Finally, AAV vectors were administered via stereotactic injection.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIn situ hybridization\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFluorescent in situ hybridization was conducted using the ACDBio V2 RNAScope kit (Advanced Cell Diagnostics). The following products were employed: RNAscope Multiplex Fluorescent Detection Reagent V2 (323110), Pretreatment Reagents (322381 and 322000), RNAscope Wash Buffer Reagents (310091), probes for \u003cem\u003eNpy\u003c/em\u003e (313321-C1), \u003cem\u003eNpy1r\u003c/em\u003e (427021-C3), \u003cem\u003eNpy2r\u003c/em\u003e (315951-C2), \u003cem\u003eSlc32a1\u003c/em\u003e (319191-C3), \u003cem\u003eSlc17a7\u003c/em\u003e (416631-C2), \u003cem\u003eFos\u003c/em\u003e (316921-C1, 316921-C2, 316921-C3) and \u003cem\u003etdTomato\u0026nbsp;\u003c/em\u003e(317041-C4), along with the PerkinElmer TSA Plus Fluorescence Palette Kit (NEL760001KT). Mice were deeply anesthetized with 1% sodium pentobarbital and transcardially perfused with ice-cold phosphate-buffered saline (PBS) before being fixed in 4% paraformaldehyde for 24 hours at 4 °C. After dehydration, brain tissue was embedded in optimal cutting temperature (OCT) and frozen at –80 °C. Coronal brain slices encompassing the entire vCA1 were sectioned at a thickness of 16 μm using a cryostat (Leica CM 1950). Slides were baked for 30 min at 60 °C. Subsequently, 5–8 drops of Hydrogen Peroxide were added to cover the entire section, followed by an incubation for 10 min at room temperature (RT). The slides were then washed three times in distilled water. The tissues were brought to 1X Target Retrieval Reagent and maintained for 5 min at 99–100 °C. Following this, the slides were washed three times in distilled water. They were then transferred to 100% alcohol for 3 min. Approximately 5 drops of Protease III were added to each section, and the samples were incubated for 30 min at 40 °C. Excess liquid was removed from the slides, and 4–6 drops of the appropriate probe mix were added to entirely cover each slide at 40 °C for two hours. To amplify the signals, tissues were sequentially immersed in AMP1, AMP2, and AMP3 at 40 °C. AMP1 and AMP2 were incubated for 30 min each, while AMP3 was incubated for 15 min. After each incubation, the appropriate HRP channel was chosen, and 4–6 drops of HRP-C1 or HRP-C2 or HRP-C3 or HRP-C4 were added to entirely cover each slide at 40 °C for 15 min, followed by incubation with HRP blocker for 15 min at 40 °C. For a specific HRP channel, corresponding diluted Opal™ Dye was added to each slide and incubated for 30 min at 40 °C. \u003cem\u003eNpy\u003c/em\u003e, \u003cem\u003eNpy1r, Npy2r\u003c/em\u003e, \u003cem\u003eSlc32a1\u003c/em\u003e, \u003cem\u003eSlc17a7, Fos\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;tdTomato\u003c/em\u003e was labeled with Opal™ 520, Opal™ 570, Opal™ 620 and Opal™ 690. Between all steps, slides were washed three times in 1× RNAscope wash buffer for 2 min. Subsequently, the slides were incubated in ACDBio DAPI for 10 min, washed, dried for 20 min, cover-slipped with mounting media (0100-01, SouthernBiotech), and left to dry overnight before imaging. Confocal images of RNAscope-labeled slices were acquired using Nikon A1 (×40), FLUOVIEW FV3000 (×20, Olympus), or SLIDEVIEW VS200 (×20, Olympus) microscopes. Z-stacks were processed in Imaris (Bitplane, Oxford Instruments) to generate 3D reconstructions, and mRNA puncta were rendered using the Surface module. Neurons were segmented on the DAPI channel in ImageJ, and RNA signals were binarized based on fluorescence intensity for quantification.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHistology and fluorescent immunostaining\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMice were deeply anesthetized with sodium pentobarbital (1%, i.p.) and transcardially perfused with ice-cold PBS followed by 4% paraformaldehyde (PFA/PBS). Brains were post-fixed in 4% PFA at 4 °C overnight, and 45 µm coronal sections were cut using a vibratome (VT1000S, Leica). For NPY immunostaining, free-floating sections were blocked in 5% normal goat serum/0.3% Triton X-100/PBS for 1 h at room temperature (RT), then incubated with rabbit anti-NPY (1:300, 11976, Cell Signaling Technology) in blocking solution at 4 °C for 48 h. After PBS washes, sections were incubated with Alexa Fluor 568- or 488-conjugated donkey anti-rabbit IgG (1:500, A-21206 or A10042, Thermo Fisher Scientific) for 2 h at RT, counterstained with DAPI (1:2000), and mounted with glass coverslips using mounting media (0100-01, SouthernBiotech).\u003c/p\u003e\n\u003cp\u003eFor dual NPY1R/NPY2R detection in paraffin-embedded tissue, 5-µm sections underwent sequential deparaffinization in xylene (3 × 10 min), rehydration through graded ethanol, and antigen retrieval via microwave heating in citrate buffer (pH 6.0). Endogenous peroxidases were quenched with 3% H\u003csub\u003e2\u003c/sub\u003eO\u003csub\u003e2\u003c/sub\u003e (25 min, RT). After blocking with 10% donkey serum, sections were incubated with rabbit anti-NPY2R (1:200, ER1914-09, HuaBio) at 4 °C for 24 h, followed by HRP-conjugated goat anti-rabbit IgG (1:500, RT, 50 min, GB23303, Servicebio) and iF488-Tyramid (G1231, Servicebio) signal amplification (1:500, 10 min, RT). The same sections were then subjected to microwave-based epitope recovery and stained with rabbit anti-NPY1R (1:200, ER64221, HuaBio) using identical secondary/TSA procedures with CY3-Tyramide (G1223, Servicebio). All sections were coverslipped with mounting media (0100-01, SouthernBiotech) and imaged on the SLIDEVIEW VS200 (Olympus) with 20×/0.75 NA objective. Image analysis was performed using ImageJ software, where neurons were segmented based on the DAPI channel, and the NPY1R and NPY2R signals were binarized according to intensity.\u0026nbsp;During paraffin embedding and sectioning, endogenous EGFP fluorescence was quenched; therefore, immunostaining was performed to recover the EGFP signal. Sections were incubated with chicken anti-GFP (1:500, A10262, Invitrogen) at 4 °C for 24 h, followed by HRP-conjugated donkey anti-chicken IgG (1:500, RT, 50 min; ECS010161, Jackson) and visualized with Alexa Fluor 488. All sections were coverslipped with mounting media (0100-01, SouthernBiotech) and were imaged with Nikon A1 (×40) or SLIDEVIEW VS200 (×20, Olympus) microscopes. Confocal z-stacks showing single optical planes and maximum projections. Z-stacks were processed in Imaris (Bitplane, Oxford Instruments) to generate 3D reconstructions, and protein puncta were visualized via Isosurface renderings using the Surface module. Neurons were segmented on the DAPI channel in ImageJ, and protein signals were binarized based on fluorescence intensity for quantification.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical analyses\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eStatistical analyses were conducted using IBM SPSS Statistics 25, Origin 2022 Software, and Office 2019 (Microsoft). The graphs were generated using Origin Software. Data are presented as the mean ± s.e.m. unless indicated otherwise. Most histograms display individual data points representing the values and numbers of individual samples for each condition. Data distributions were tested for normality, and homogeneity of variance among groups was assessed using Levene’s test. Statistical comparisons were carried out using unpaired Student’s \u003cem\u003et\u003c/em\u003e-test, two-tailed paired Student’s \u003cem\u003et\u003c/em\u003e-test, as well as one-way analyses of variance (ANOVAs) or two-way repeated measures ANOVAs. For post hoc analysis, Bonferroni’s corrections for multiple comparisons were applied. A significance level of \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05 was considered statistically significant. Significance is primarily denoted as \u003csup\u003e*\u003c/sup\u003e\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05, \u003csup\u003e**\u003c/sup\u003e\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.01, and \u003csup\u003e***\u003c/sup\u003e\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001. In some cases, it is indicated as \u003csup\u003e#\u003c/sup\u003e\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05, \u003csup\u003e##\u003c/sup\u003e\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.01, and \u003csup\u003e###\u003c/sup\u003e\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.001 for multiple comparisons. NS, denotes non-significant values.\u0026nbsp;Detailed information on statistical tests is provided in Source Data files.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll data supporting the findings of this study are available within the paper and its Extended Data/Supplementary Information. The single-cell RNA-seq dataset generated here has been deposited in the Genome Sequence Archive (GSA) at the National Genomics Data Center (NGDC) under accession CRA023358 and is publicly accessible. Source data underlying all main and Extended Data/Supplementary Information figures—together with detailed statistics for each panel—are provided with the paper (Source Data files).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCustom code used to perform the analysis is available at https://github.com/YaleiKong/NPY-fear-extinction.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank Ms. Yongling Wang (Fudan University) for her technique assistance. We are grateful to Drs. Xiajing Tong (ShanghaiTech University) and Jinfei Ni (Fudan University) for providing transgenic mice used in this study in a generous manner. This study was supported by grants from the National Natural Science Foundation of China (32430040, 32200821, 32371078 and 32300843), the STI2030-Major Projects (2021ZD0202800), the Key Discipline Project of Shanghai Municipal Health Commission (No. 2024ZDXK0050), the Science and Technology Commission of Shanghai Municipality (22XD1420700 and 23YF1433900), the Shanghai Municipal Health Commission (2022XD046), and innovative research team of high-level local universities in Shanghai.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAUTHOR CONTRIBUTIONS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eY.-J.W., X.G., Q.L., W.-G.L., and T.-L.X. conceived the project, designed the experiments, and interpreted the results. Y.-J.W. and X.G. performed the majority of behavioral experiments, animal surgery, immunohistochemistry, and data analysis. M.X., Q.W., X.Y., Z.-J.L., Z.-H.J., H.C., X.-Y.Z., and X.B. assisted with some of the behavioral experiments and conducted viral injections. Q.J. and Ying Li performed genotyping. Y.-J.W., H.W., and Yulong Li performed NPY dynamics measurement. Y.-J.W., X.G., Y.K., S.Y., J.H., and Q.L. did single-cell preparation and analyses. M-X.Z. commented on data interpretation. Y.-J.W., L.-Y.W., W.-G.L. and T.-L.X. wrote the manuscript with contributions from all authors. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDECLARATION OF INTERESTS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAdditional information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eExtended data includes 10 Extended Data and 27 Supplementary Information figures and their legends are available for this paper online.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eJosselyn, S.A., Kohler, S. \u0026amp; Frankland, P.W. Finding the engram. \u003cem\u003eNat Rev Neurosci\u003c/em\u003e 16, 521\u0026ndash;534 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJosselyn, S.A. \u0026amp; Tonegawa, S. Memory engrams: Recalling the past and imagining the future. \u003cem\u003eScience\u003c/em\u003e 367 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSun, X., \u003cem\u003eet al.\u003c/em\u003e Functionally Distinct Neuronal Ensembles within the Memory Engram. \u003cem\u003eCell\u003c/em\u003e 181, 410\u0026ndash;423 e417 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGuskjolen, A. \u0026amp; Cembrowski, M.S. Engram neurons: Encoding, consolidation, retrieval, and forgetting of memory. \u003cem\u003eMol Psychiatry\u003c/em\u003e 28, 3207\u0026ndash;3219 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTome, D.F., \u003cem\u003eet al.\u003c/em\u003e Dynamic and selective engrams emerge with memory consolidation. \u003cem\u003eNat Neurosci\u003c/em\u003e 27, 561\u0026ndash;572 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChoi, J.H., \u003cem\u003eet al.\u003c/em\u003e Interregional synaptic maps among engram cells underlie memory formation. \u003cem\u003eScience\u003c/em\u003e 360, 430\u0026ndash;435 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGallinaro, J.V., Gasparovic, N. \u0026amp; Rotter, S. Homeostatic control of synaptic rewiring in recurrent networks induces the formation of stable memory engrams. \u003cem\u003ePLoS Comput Biol\u003c/em\u003e 18, e1009836 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVogels, T.P., Sprekeler, H., Zenke, F., Clopath, C. \u0026amp; Gerstner, W. Inhibitory plasticity balances excitation and inhibition in sensory pathways and memory networks. \u003cem\u003eScience\u003c/em\u003e 334, 1569\u0026ndash;1573 (2011).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDavis, P., Zaki, Y., Maguire, J. \u0026amp; Reijmers, L.G. Cellular and oscillatory substrates of fear extinction learning. \u003cem\u003eNat Neurosci\u003c/em\u003e 20, 1624\u0026ndash;1633 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang, Q., \u003cem\u003eet al.\u003c/em\u003e Insular cortical circuits as an executive gateway to decipher threat or extinction memory via distinct subcortical pathways. \u003cem\u003eNat Commun\u003c/em\u003e 13, 5540 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKim, T., \u003cem\u003eet al.\u003c/em\u003e Activated somatostatin interneurons orchestrate memory microcircuits. \u003cem\u003eNeuron\u003c/em\u003e 112, 201\u0026ndash;208.e204 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLin, Z.J., \u003cem\u003eet al.\u003c/em\u003e Stimulation of an entorhinal-hippocampal extinction circuit facilitates fear extinction in a post-traumatic stress disorder model. \u003cem\u003eJ Clin Invest\u003c/em\u003e 134, e181095 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ed'Aquin, S., \u003cem\u003eet al.\u003c/em\u003e Compartmentalized dendritic plasticity during associative learning. \u003cem\u003eScience\u003c/em\u003e 376, eabf7052 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTopolnik, L. \u0026amp; Tamboli, S. The role of inhibitory circuits in hippocampal memory processing. \u003cem\u003eNat Rev Neurosci\u003c/em\u003e 23, 476\u0026ndash;492 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTzilivaki, A., \u003cem\u003eet al.\u003c/em\u003e Hippocampal GABAergic interneurons and memory. \u003cem\u003eNeuron\u003c/em\u003e 111, 3154\u0026ndash;3175 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDunsmoor, J.E., Niv, Y., Daw, N. \u0026amp; Phelps, E.A. Rethinking Extinction. \u003cem\u003eNeuron\u003c/em\u003e 88, 47\u0026ndash;63 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBouton, M.E., Maren, S. \u0026amp; McNally, G.P. Behavioral and Neurobiological Mechanisms of Pavlovian and Instrumental Extinction Learning. \u003cem\u003ePhysiol Rev\u003c/em\u003e 101, 611\u0026ndash;681 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eIzquierdo, I., Furini, C.R. \u0026amp; Myskiw, J.C. Fear Memory. \u003cem\u003ePhysiol Rev\u003c/em\u003e 96, 695\u0026ndash;750 (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTrouche, S., Sasaki, J.M., Tu, T. \u0026amp; Reijmers, L.G. Fear extinction causes target-specific remodeling of perisomatic inhibitory synapses. \u003cem\u003eNeuron\u003c/em\u003e 80, 1054\u0026ndash;1065 (2013).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMonfils, M.H., Cowansage, K.K., Klann, E. \u0026amp; LeDoux, J.E. Extinction-reconsolidation boundaries: key to persistent attenuation of fear memories. \u003cem\u003eScience\u003c/em\u003e 324, 951\u0026ndash;955 (2009).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRessler, K.J., \u003cem\u003eet al.\u003c/em\u003e Post-traumatic stress disorder: clinical and translational neuroscience from cells to circuits. \u003cem\u003eNat Rev Neurol\u003c/em\u003e 18, 273\u0026ndash;288 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLebois, L.A.M., Seligowski, A.V., Wolff, J.D., Hill, S.B. \u0026amp; Ressler, K.J. Augmentation of Extinction and Inhibitory Learning in Anxiety and Trauma-Related Disorders. \u003cem\u003eAnnu Rev Clin Psychol\u003c/em\u003e 15, 257\u0026ndash;284 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVervliet, B., Craske, M.G. \u0026amp; Hermans, D. Fear extinction and relapse: state of the art. \u003cem\u003eAnnu Rev Clin Psychol\u003c/em\u003e 9, 215\u0026ndash;248 (2013).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi, W.G., \u003cem\u003eet al.\u003c/em\u003e Input associativity underlies fear memory renewal. \u003cem\u003eNatl Sci Rev\u003c/em\u003e 8, nwab004 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGu, X., \u003cem\u003eet al.\u003c/em\u003e Dynamic tripartite construct of interregional engram circuits underlies forgetting of extinction memory. \u003cem\u003eMol Psychiatry\u003c/em\u003e 27, 4077\u0026ndash;4091 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePeters, J., Dieppa-Perea, L.M., Melendez, L.M. \u0026amp; Quirk, G.J. Induction of fear extinction with hippocampal-infralimbic BDNF. \u003cem\u003eScience\u003c/em\u003e 328, 1288\u0026ndash;1290 (2010).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang, Q., \u003cem\u003eet al.\u003c/em\u003e Fear extinction requires ASIC1a-dependent regulation of hippocampal-prefrontal correlates. \u003cem\u003eSci Adv\u003c/em\u003e 4, eaau3075 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMeyer, H.C., \u003cem\u003eet al.\u003c/em\u003e Ventral hippocampus interacts with prelimbic cortex during inhibition of threat response via learned safety in both mice and humans. \u003cem\u003eProc Natl Acad Sci U S A\u003c/em\u003e 116, 26970\u0026ndash;26979 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSzadzinska, W., \u003cem\u003eet al.\u003c/em\u003e Hippocampal Inputs in the Prelimbic Cortex Curb Fear after Extinction. \u003cem\u003eJ Neurosci\u003c/em\u003e 41, 9129\u0026ndash;9140 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNguyen, R., Koukoutselos, K., Forro, T. \u0026amp; Ciocchi, S. Fear extinction relies on ventral hippocampal safety codes shaped by the amygdala. \u003cem\u003eSci Adv\u003c/em\u003e 9, eadg4881 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAllen, W.E., \u003cem\u003eet al.\u003c/em\u003e Thirst-associated preoptic neurons encode an aversive motivational drive. \u003cem\u003eScience\u003c/em\u003e 357, 1149\u0026ndash;1155 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePettit, N.L., Yap, E.L., Greenberg, M.E. \u0026amp; Harvey, C.D. Fos ensembles encode and shape stable spatial maps in the hippocampus. \u003cem\u003eNature\u003c/em\u003e 609, 327\u0026ndash;334 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWu, Y.E., Pan, L., Zuo, Y., Li, X. \u0026amp; Hong, W. Detecting Activated Cell Populations Using Single-Cell RNA-Seq. \u003cem\u003eNeuron\u003c/em\u003e 96, 313\u0026ndash;329 e316 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen, M.B., Jiang, X., Quake, S.R. \u0026amp; Sudhof, T.C. Persistent transcriptional programmes are associated with remote memory. \u003cem\u003eNature\u003c/em\u003e 587, 437\u0026ndash;442 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHochgerner, H., \u003cem\u003eet al.\u003c/em\u003e Neuronal types in the mouse amygdala and their transcriptional response to fear conditioning. \u003cem\u003eNat Neurosci\u003c/em\u003e 26, 2237\u0026ndash;2249 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSun, W., \u003cem\u003eet al.\u003c/em\u003e Spatial transcriptomics reveal neuron-astrocyte synergy in long-term memory. \u003cem\u003eNature\u003c/em\u003e 627, 374\u0026ndash;381 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFanselow, M.S. \u0026amp; Dong, H.W. Are the dorsal and ventral hippocampus functionally distinct structures? \u003cem\u003eNeuron\u003c/em\u003e 65, 7\u0026ndash;19 (2010).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang, H., \u003cem\u003eet al.\u003c/em\u003e A tool kit of highly selective and sensitive genetically encoded neuropeptide sensors. \u003cem\u003eScience\u003c/em\u003e 382, eabq8173 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWu, Z., \u003cem\u003eet al.\u003c/em\u003e Neuronal activity-induced, equilibrative nucleoside transporter-dependent, somatodendritic adenosine release revealed by a GRAB sensor. \u003cem\u003eProc Natl Acad Sci U S A\u003c/em\u003e 120, e2212387120 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDong, H.W., Swanson, L.W., Chen, L., Fanselow, M.S. \u0026amp; Toga, A.W. Genomic-anatomic evidence for distinct functional domains in hippocampal field CA1. \u003cem\u003eProc Natl Acad Sci U S A\u003c/em\u003e 106, 11794\u0026ndash;11799 (2009).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDeNardo, L.A., \u003cem\u003eet al.\u003c/em\u003e Temporal evolution of cortical ensembles promoting remote memory retrieval. \u003cem\u003eNat Neurosci\u003c/em\u003e 22, 460\u0026ndash;469 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGarner, A.R., \u003cem\u003eet al.\u003c/em\u003e Generation of a synthetic memory trace. \u003cem\u003eScience\u003c/em\u003e 335, 1513\u0026ndash;1516 (2012).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWieland, H.A., Willim, K. \u0026amp; Doods, H.N. Receptor binding profiles of NPY analogues and fragments in different tissues and cell lines. \u003cem\u003ePeptides\u003c/em\u003e 16, 1389\u0026ndash;1394 (1995).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLoh, K., Herzog, H. \u0026amp; Shi, Y.C. Regulation of energy homeostasis by the NPY system. \u003cem\u003eTrends Endocrinol Metab\u003c/em\u003e 26, 125\u0026ndash;135 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMarcos, P. \u0026amp; Covenas, R. Regulation of Homeostasis by Neuropeptide Y: Involvement in Food Intake. \u003cem\u003eCurr Med Chem\u003c/em\u003e 29, 4026\u0026ndash;4049 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGotzsche, C.R. \u0026amp; Woldbye, D.P. The role of NPY in learning and memory. \u003cem\u003eNeuropeptides\u003c/em\u003e 55, 79\u0026ndash;89 (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTasan, R.O., \u003cem\u003eet al.\u003c/em\u003e The role of Neuropeptide Y in fear conditioning and extinction. \u003cem\u003eNeuropeptides\u003c/em\u003e 55, 111\u0026ndash;126 (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMichel, M.C., \u003cem\u003eet al.\u003c/em\u003e XVI. International Union of Pharmacology recommendations for the nomenclature of neuropeptide Y, peptide YY, and pancreatic polypeptide receptors. \u003cem\u003ePharmacol Rev\u003c/em\u003e 50, 143\u0026ndash;150 (1998).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRaut, S.B., \u003cem\u003eet al.\u003c/em\u003e Diverse therapeutic developments for post-traumatic stress disorder (PTSD) indicate common mechanisms of memory modulation. \u003cem\u003ePharmacol Ther\u003c/em\u003e 239, 108195 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHao, Y., \u003cem\u003eet al.\u003c/em\u003e Integrated analysis of multimodal single-cell data. \u003cem\u003eCell\u003c/em\u003e 184, 3573\u0026ndash;3587 e3529 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFinak, G., \u003cem\u003eet al.\u003c/em\u003e MAST: a flexible statistical framework for assessing transcriptional changes and characterizing heterogeneity in single-cell RNA sequencing data. \u003cem\u003eGenome Biol\u003c/em\u003e 16, 278 (2015).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Ventral hippocampal CA1 (vCA1), Neuropeptide Y, NPY receptor, Fear engrams, Extinction engrams","lastPublishedDoi":"10.21203/rs.3.rs-4347593/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4347593/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Memory engrams are formed by activity-dependent recruitment of distinct subsets of excitatory principal neurons (or neuronal ensembles) whereas inhibitory neurons pivot memory lability and stability. However, the molecular logic for memory engrams to preferentially recruit specific type of interneurons over others remains enigmatic. Using activity-dependent single-cell transcriptomic profiling in mice with training of cued fear memory and extinction, we discovered that neuropeptide Y (NPY)-expressing (NPY+) GABAergic interneurons in the ventral hippocampal CA1 (vCA1) region exert fast GABAergic inhibition to facilitate the acquisition of memory, but bifurcate NPY-mediated slow peptidergic inhibition onto distinct sub-ensembles underlying the extinction of single memory trace. Genetically encoded calcium and NPY sensors revealed that both calcium dynamics of NPY+ neurons and their NPY release in vCA1 ramp up as extinction learning progresses while behavioral state switches from “fear-on” to “fear-off”. Bidirectional manipulations of NPY+ neurons or NPY itself demonstrated that NPY is both necessary and sufficient to control the rate and degree of memory extinction by acting on two physically non-overlapping sub-ensembles composed of NPY1R- and NPY2R-expressing neurons. CRISPR/Cas9-mediated knockout of NPY2R or NPY1R further unravels that NPY co-opts its actions on these two sub-ensembles to gate early fast and late slow stages of extinction. These findings exemplify the intricate spatiotemporal orchestration of slow peptidergic inhibitions from single subtype of GABAergic interneurons to fine-tune engram lability verse stability of memory.","manuscriptTitle":"Neuropeptide Y co-opts neuronal ensembles for memory lability and stability","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-28 16:42:16","doi":"10.21203/rs.3.rs-4347593/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-neuroscience","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"neuro","sideBox":"Learn more about [Nature Neuroscience](http://www.nature.com/neuro/)","snPcode":"","submissionUrl":"","title":"Nature Neuroscience","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Research","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"53c6194d-5afc-4ea4-b709-f33c92740d69","owner":[],"postedDate":"October 28th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":55626476,"name":"Biological sciences/Neuroscience/Learning and memory/Extinction"},{"id":55626477,"name":"Health sciences/Diseases/Psychiatric disorders/Post-traumatic stress disorder"}],"tags":[],"updatedAt":"2026-04-01T07:30:04+00:00","versionOfRecord":{"articleIdentity":"rs-4347593","link":"https://doi.org/10.1038/s41593-026-02235-x","journal":{"identity":"nature-neuroscience","isVorOnly":false,"title":"Nature Neuroscience"},"publishedOn":"2026-03-31 04:00:00","publishedOnDateReadable":"March 31st, 2026"},"versionCreatedAt":"2025-10-28 16:42:16","video":"","vorDoi":"10.1038/s41593-026-02235-x","vorDoiUrl":"https://doi.org/10.1038/s41593-026-02235-x","workflowStages":[]},"version":"v1","identity":"rs-4347593","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4347593","identity":"rs-4347593","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}