Aversive experiences induce valence plasticity of instructive signals to change future learning rules in mice

Tohyama, Suguru; Nagashima, Takashi; Higashino, Iori; Arima-Yoshida, Fumiko; Hiyoshi, Kanae; Nagase, Masashi; Yada, Yuichiro; Honda, Naoki; Watabe, Ayako M.

doi:10.1038/s42003-025-08367-3

Download PDF

Article
Open access
Published: 04 July 2025

Aversive experiences induce valence plasticity of instructive signals to change future learning rules in mice

Communications Biology volume 8, Article number: 1002 (2025) Cite this article

1699 Accesses
1 Altmetric
Metrics details

Subjects

Abstract

Aversive signals such as pain serve an instructive role in aversive learning to promote animal survival. While negative valence of aversive signals is considered to be innately assigned, the valence can be scaled by internal state and previous experiences. However, the neuronal mechanisms underlying state and experience-dependent valence modulation remain unexplored. Previous studies demonstrated synaptic potentiation in instructive signal pathways following robust aversive learning. Here, we hypothesized that long-term potentiation (LTP) in the parabrachial-to-central amygdala (PB-CeC/L) pathway, an important nociceptive circuit for producing pain and emotional learning, enhances the negative valence and thereby alter future learning rules. To test this hypothesis, we developed pathway-specific in vivo LTP induction methods and mathematical models. Our results suggest that LTP in the PB-CeC/L pathway alters aversive valence and future learning rules by enhancing subsequent learning and memory generalization. These results may help to identify a therapeutic target for post-traumatic stress disorder.

Plasticity of neuronal dynamics in the lateral habenula for cue-punishment associative learning

Article Open access 06 July 2023

Genetically identified amygdala–striatal circuits for valence-specific behaviors

Article 18 October 2021

Reward and aversion processing by input-defined parallel nucleus accumbens circuits in mice

Article Open access 21 October 2022

Introduction

Adaptive behaviors based on past experiences are important for animal survival. Noxious stimuli such as pain play an essential role in producing aversive learning, leading to the development of adaptive behaviors. Pavlovian threat conditioning is one of the most widely used behavioral paradigms to explore the neural mechanisms underlying associative aversive learning. In Pavlovian threat conditioning, sensory signals with negative valence function as an unconditioned stimulus (US) to produce learning when paired with an emotionally neutral conditioned stimulus (CS)^{1,2,3,4,5,6,7}. The central amygdala (CeA) serves as a hub for the CS-US association by integrating various sensory signals and orchestrating appropriate defensive behaviors^8,9,10,11. However, despite extensive research on the neural mechanisms underlying this association, US signal coding and its valence modulation by past experiences remain incompletely understood. Previous studies have demonstrated that the lateral parabrachial nucleus (PB) in the pons receives nociceptive signals from the dorsal horn and then projects directly to the CeA, especially to the capsular and lateral subdivision (CeC/L). This PB-CeC/L pathway is vital for transmitting aversive signals that act as the US in aversive learning^{12,13,14,15,16,17,18,19,20}.

Long-term potentiation (LTP) of synaptic transmission is an essential mechanism underlying learning and memory. Research has demonstrated LTP occurrence in CS pathways, such as those from the auditory cortex and thalamus to the lateral amygdala (LA), following associative aversive learning^3,21,22. Moreover, LTP in the LA causally influences aversive learning outcomes^23,24,25. In addition to findings of plasticity in CS pathways which underlies associative learning memory, LTP has also been observed in the US pathways across various animal models. For example, synaptic transmission in the PB-CeC/L pathway is significantly enhanced in acute and chronic pain models following robust and repeated aversive learning experiences in rats and mice^{26,27,28,29,30,31,32,33,34,35,36,37}. However, the specific causal involvement of LTP in the US pathway in associative aversive learning remains unexplored, posing a fundamental question regarding the physiological role of LTP in the regulation of adaptive behavior.

We hypothesized that LTP in the US pathway changes the valence of US signals, thereby updating learning rules, such as altering learning threshold and generalization, in subsequent aversive encounters. While learning rules can be altered based on past experiences and internal states^38,39,40, the involvement of US pathway-specific synaptic plasticity remains largely unexplored. To address this issue, we employed a novel behavioral paradigm integrating optogenetic, chemogenetic, electrophysiological and mathematical approaches.

Results

Electrophysiological characterization and optogenetic LTP induction in the PB-CeC/L pathway

We first confirmed the electrophysiological characterization of the US pathway, PB-CeC/L. Following adeno-associated virus (AAV) injection (AAV-Syn-Chronos:GFP) into the bilateral PB, the whole-cell patch-clamp recordings in acute brain slices were performed (Fig. 1a). Light-evoked excitatory post-synaptic currents (EPSCs) were detected from CeC/L neurons (Fig. 1b). We recorded EPSCs evoked by a 5 ms light pulse in voltage-clamp modes at −60 mV (Fig. 1c–e). These EPSCs were blocked by the voltage-dependent sodium channel blocker, tetrodotoxin (TTX; 1 μM). The application of 4-aminopyridine (4-AP; 100 μM) recovered EPSCs, while subsequent addition of 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX; 10 μM) and D-2-amino-5-phosphonovalerate (APV; 50 μM) abolished EPSCs. These results indicate that PB-CeC/L projections are monosynaptic and glutamatergic transmission.

**Fig. 1: Characterization of the PB-CeC/L pathway and optogenetic LTP induction protocols.**

To directly address our hypothesis that LTP in the US pathway changes future learning rules, LTP induction protocol in vivo needs to be established. For this purpose, we first performed whole-cell patch-clamp recordings in acute brain slices to estimate the optimal stimulus frequency to induce LTP in the PB-CeC/L pathway by optogenetic approach. After obtaining a stable baseline for more than 10 min, we applied optogenetic high-frequency stimulation (oHFS) to the PB-CeC/L pathway using Chronos, a channelrhodopsin suitable for HFS⁴¹(Fig. 1f–h). An oHFS consisting of five trains (0.1 Hz, each train at 20 Hz [n = 5 cells], 40 Hz [n = 14 cells], or 50 Hz [n = 8 cells], 0.5 s) was applied. While 40 Hz and 50 Hz HFS significantly potentiated light-evoked excitatory postsynaptic currents (leEPSCs), 20 Hz HFS did not induce such an effect (Fig. 1h). Thus, we used a 40 Hz stimulation frequency to induce LTP in the PB-CeC/L pathway in the rest of the work.

Next, we examined whether in vivo 40 Hz stimulation could induce LTP in the PB-CeC/L pathway. Following AAV injection (AAV-Syn-Chronos:GFP or AAV-Syn-eGFP) bilaterally into the PB, dual light-emitting diode (LED) cannulae were implanted over the CeC/L for optogenetic activation of axonal terminals (Fig. 1i). Optogenetic HFS (40 Hz, 2 s on/3 s off, for 7 min) was applied in the context A. One day after in vivo LTP induction, we made acute brain slices from oHFS and no-stimulation mice and conducted electrophysiological analyses including synaptic input-output relationship and NMDAR/AMPAR ratio. The input-output curve was left-shifted in the in vivo LTP induction mice, and leEPSC amplitude was significantly increased especially at low intensity stimulation (Fig. 1j, k). The NMDAR/AMPAR ratio did not exhibit significant differences between the groups (Fig. 1l). In contrast, we found that there is a significant reduction in paired-pulse ratio after successive threat conditioning (Fig. 1m, n). Taken together, these results indicate that the LTP of the PB-CeC/L was successfully induced by our in vivo LTP paradigm and may involve a presynaptic mechanism.

In vivo LTP induction in the PB-CeC/L pathway enhances subsequent aversive learning and avoidance behavior

Next, we performed in vivo LTP induction in the PB-CeC/L pathway and examined the artificial manipulation of learning rules. Following AAV injection (AAV-Syn-Chronos:GFP or AAV-Syn-eYFP) bilaterally into the PB, LED cannulae were implanted over the CeC/L (Fig. 2a, b). oHFS was applied in the context A. The next day, the mice received parings of an auditory CS and a weak shock US, a training regimen that was insufficient to produce a strong fear memory⁴², and retrieval tests were performed on the following day (Fig. 2c). Interestingly, the Chronos group exhibited higher CS-evoked freezing than the eYFP group, while baseline freezing levels were comparable between these two groups in the retrieval test (Fig. 2d, e). This suggests that oHFS in the PB-CeC/L pathway is sufficient to promote subsequent aversive learning. Using the von Frey, marble burying, and open-field tests, we also examined whether optogenetic LTP induction in the PB-CeC/L pathway altered nociceptive responses or anxiety levels, as these factors can influence aversive learning. In the von Frey test, the paw withdrawal thresholds for mechanical stimuli did not significantly differ between the two groups (Supplementary Fig. 1a), suggesting that the thresholds for mechanical nociceptive stimuli were not significantly affected by LTP induction in the PB-CeC/L pathway. Additionally, no significant differences were observed between the eYFP and Chronos groups in the marble burying test (Supplementary Fig. 1b), suggesting that anxiety levels were also not markedly affected. In the open-field test, the total distance travelled and the time spent in the central region were not significantly different between the eYFP and Chronos groups (Supplementary Fig. 1c–e), supporting the notion that anxiety levels were not changed. Taken together, these results indicate that LTP induction in the PB-CeC/L pathway promotes subsequent aversive learning with no detectable anxiogenic effects.

**Fig. 2: In vivo LTP induction in the PB-CeC/L pathway enhances avoidance behavior and subsequent aversive learning.**

Additionally, we addressed whether LTP in the PB-CeC/L pathway changes the valence of US signals using real-time place aversion (RTPA) test. We newly prepared mice expressing Chronos or eGFP in PB neurons and applied oHFS to the PB-CeC/L under the same conditions as above (Fig. 2c), and RTPA test was performed in a two-compartment chamber on the following day (Fig. 2f). As we expected, the Chronos group exhibited strong avoidance behavior against shock area and high aversion index (Fig. 2g–i). In addition, we measured the footshock stimulus threshold for responses of flinching, vocalization, and jumping (Supplementary Fig. 2a–d). The threshold for responses of flinching was significantly decreased by in vivo LTP induction. These results suggest that in vivo LTP induction in the PB-CeC/L pathway enhances the US signals and provide an idea that LTP in the US pathway enhances subsequent aversive learning via an increase of US signals.

Past fearful experience promotes subsequent aversive learning

Previously, some groups demonstrated that PB-CeC/L pathway is enhanced in acute and chronic pain models in rodents^26,33,35. Furthermore, we have demonstrated that LTP induction occurs in the PB-CeC/L following robust aversive learning, suggesting a potential mechanism of learning rule alteration in an experience-dependent manner³⁷. Here, we hypothesized that past fearful experience could define appropriate future learning rules via LTP in the instructive signal pathway, PB-CeC/L, and designed following experiments to investigate learning rule alteration in subsequent aversive learning. One group of mice received nine footshocks as a fearful experience (Shock group), while another group experienced context exposure only (Control group). First, we performed successive threat conditioning in mice. In this test, one group of mice received nine footshocks as a fearful experience (Shock group), while another group experienced context exposure only (Control group). Subsequently, both groups underwent weak threat conditioning (Fig. 3a). The retrieval test in a novel context revealed no significant differences in baseline freezing between the two groups. However, the Shock group exhibited higher levels of auditory CS-evoked freezing compared with the Control group (Fig. 3b, c), demonstrating that prior fearful experiences promote subsequent aversive learning. These results are consistent with previous study reporting that stronger intensity of footshock enhances freezing responses⁴³. Subsequently, we examined whether nociceptive responses or anxiety levels changed after the aversive experience. No significant differences were observed between the two groups in the von Frey test (Supplementary Fig. 3a), suggesting that the mechanical nociceptive thresholds were not significantly affected by past fearful experiences. In contrast, the Shock group buried a higher number of marbles under beddings in the marble burying test (Supplementary Fig. 3b), indicating a potential increase in anxiety levels. Additionally, in the open-field test, while the total distance travelled remained equivalent in the two groups, the Shock group exhibited a significant decrease in the time spent in the central region, suggesting that past fearful experiences can enhance anxiety-like behavior without affecting locomotor activity (Supplementary Fig. 3c, d). Overall, our findings suggest that past fearful experiences modulate subsequent learning rules with some alteration in affective states.

In addition, we investigated a change of the US signals in an experience-dependent manner. Nine footshocks as a fearful experience were applied in the context A, and RTPA test was performed in a two-compartment chamber on the following day (Fig. 3d). The Shock group exhibited strong avoidance behavior compared to Control group in the RTPA test (Fig. 3e–g), suggesting that fearful experience enhances the US signals. Moreover, we measured the footshock stimulus threshold for responses of flinching, vocalization, and jumping (Supplementary Fig. 4a–d). The threshold for responses of flinching and jumping was significantly decreased by in vivo LTP induction, suggesting that fearful experience enhances the US signals. Collectively, our results provide an idea that fearful experience enhances subsequent aversive learning via an increase of US valence.

Past fearful experience promotes fear memory generalization in subsequent learning

Memory generalization is facilitated when strong footshocks are used during learning in threat conditioning studies^43,44,45,46. To examine whether the generalization of fear memory can also be enhanced by a prior shock experience, we employed a differential threat conditioning paradigm⁴⁷. In this successive paradigm, both the Shock and Control groups received interleaved presentations of a tone paired with a US (CS1) and an unpaired tone (CS2) after the shock experience, and then CS1 and CS2 were reintroduced in a discrimination test on the following day (Fig. 4a). While the Shock group exhibited similar freezing ratio as the Control group during CS1, they showed a significantly higher freezing ratio than the Control group during the unpaired CS2 (Fig. 4b–d). Interestingly, individual mice in the Shock group showed similar levels of freezing to both CS1 and CS2, whereas those in the Control group exhibited significantly higher freezing to CS1 than to CS2 (Fig. 4e). Moreover, the discrimination index for the CS tone (CS1-CS2/CS1 + CS2) was significantly lower in the Shock group than in the Control group (Fig. 4f). These findings show that past fearful experiences induce discrimination deficits and promote generalization in subsequent aversive learning scenarios.

**Fig. 4: Past fearful experience promotes fear memory generalization in future learning.**

PB is required for generalization following past fearful experience

To investigate whether the PB activity during a past fearful experience is required for enhancing generalization of subsequent aversive learning, we chemogenetically silenced the PB activity during the initial shock experience using a Designer Receptors Exclusively Activated by Designer Drugs (DREADD) system⁴⁸. Specifically, we delivered the inhibitory DREADD hM4Di or mCherry (as a control) into the PB using AAV (Fig. 5a, b) and systemically administered the DREADD agonist clozapine N-oxide (CNO) before the delivery of nine footshocks (Fig. 5c). The following day, all mice underwent differential threat conditioning without drug administration (Fig. 5c). Notably, chemogenetic inhibition of the PB did not significantly influence the freezing ratio during the CS1 period in the discrimination test; however, it reduced the freezing ratio during the CS2 period (Fig. 5d, e). While mCherry control mice showed similar levels of freezing during CS1 and CS2 (Fig. 5f), consistent with Fig. 4e, chemogenetic inhibition of the PB significantly decreased the freezing ratio during the CS2 period, compared with that during the CS1 period (Fig. 5e). Additionally, the discrimination index of the hM4Di mice was significantly higher than that of the mCherry mice (Fig. 5f). These results suggest that the PB activity during a past fearful experience is necessary to alter subsequent learning rules, as shown in the generalization of fear memory.

**Fig. 5: The PB is required for the generalization of fear memory following past fearful experience.**

PB-CeC/L pathway mediates generalization of fear memory following past fearful experience

To investigate the necessity of the PB-CeC/L pathway during fearful experiences in generalization, we used the inhibitory opsin ArchT and a dual LED cannula unit in the CeC/L for pathway-specific optogenetic inhibition (Fig. 6a). Optogenetic inhibition of the PB-CeC/L pathway was achieved through continuous light exposure throughout the initial shock experience (Fig. 6b). The freezing ratio did not remarkably change during CS1 (Fig. 6c), whereas that tended to decrease during CS2 by optogenetic inhibition (Fig. 6d). The freezing ratio during CS2 was significantly lower than that during CS1 in the ArchT group (Fig. 6e). Furthermore, the discrimination index in the ArchT group was significantly higher than that in the eYFP group (Fig. 6f), consistent with the findings from the chemogenetic inhibition of the PB (Fig. 5). Taken together, these results suggest that the alteration of learning rules, as shown by the enhanced generalization of fear memory, is mediated by the activation of the PB-CeC/L pathway during the past fearful experience.

**Fig. 6: The PB-CeC/L pathway is required for the generalization of fear memory following past fearful experience.**

Optogenetic LTP induction promotes generalization of fear memory

Having shown that LTP induction in the PB-CeC/L pathway promotes subsequent aversive learning (Fig. 2), we next examined whether fear generalization is induced by optogenetic LTP protocol (Fig. 7a, b). No significant differences in CS1-evoked freezing were observed between the groups (Fig. 7c). However, the Chronos group exhibited significantly higher CS2-evoked freezing than the eYFP group (Fig. 7d). The eYFP group had higher freezing in response to CS1 than to CS2, indicating discrimination between CS1 and CS2. In contrast, the Chronos group exhibited no significant difference in the freezing responses between CS1 and CS2 presentations (Fig. 7e). Furthermore, the discrimination index was reduced in the Chronos group compared to that in the eYFP control group (Fig. 7f), consistent with data from real fearful experience experiments (Fig. 3). These results suggest that LTP induction in the PB-CeC/L pathway is sufficient to promote the generalization of fear memory.

**Fig. 7: Optogenetic LTP induction promotes the generalization of fear memory.**

LTP induction in the PB-CeC/L pathway alters emotional value of US

We further evaluated the emotional value of the US using behavioral data based on a mathematical model. Specifically, we employed the Rescorla–Wagner (RW) model, a mathematical framework for understanding associative learning in classical conditioning⁴⁹. In the RW model, the CeC/L is activated by CS through synaptic input, leading to a monotonic increase in freezing behavior correlated with CeC/L activity. The synaptic weight is updated to minimize the discrepancy between the CeC/L activity and US value (Fig. 8a, b). Using the RW model, we estimated how the value of the US varied based on past experiences by fitting the freezing ratios observed during successive threat conditioning tests (Supplementary Fig. 5a). Parameters were estimated using data from both the Control and Shock groups (Fig. 3e–g). The central tendency of estimated US values in the Shock group was higher compared to the Control group (Fig. 8c, mean: 7.22 vs. 2.00, respectively). Importantly, the RW model with estimated parameters successfully explained changes in freezing ratios during learning (Supplementary Fig. 5a). Similar results were obtained in the comparison of PB-CeC/L pathway-specific LTP-induced (Chronos) and control (eYFP) mice (Fig. 2e–g); the central tendency of estimated US values in the Chronos group was higher compared to the eYFP group (Supplementary Fig. 5b and Fig. 8d, mean: 3.73 vs. 1.98, respectively).

**Fig. 8: LTP induction in the PB-CeC/L pathway alters the emotional value of aversive unconditioned stimuli.**

Notably, the distributions of US values for the Control and eYFP groups were nearly identical, suggesting that the mice without synaptic plasticity interventions did not show any change in the US value. In contrast, the Chronos group showed an increase in US values, although not as pronounced as observed in the Shock group. These findings indicate that alterations in learning rules due to past fear experiences (shock) and the induction of PB-CeC/L pathway-specific LTP can be explained by an increase in the value of the US.

Furthermore, we extended the RW model that accounts effects of CS1-CS2 generalization (Fig. 8e, f). We estimated how the value of the US varied by fitting the freezing ratios in CS1-CS2 generalization (Supplementary Fig. 5c, d). Parameters were estimated using data from both the Control and Shock groups (Fig. 4). The central tendency of estimated US values in the Shock group was higher compared to the Control group (Supplementary Fig. 5c and Fig. 8g, mean: 5.14 vs. 4.87, respectively). Similar results were obtained in the comparison of PB-CeC/L pathway-specific LTP-induced (Chronos) and control (eYFP) mice (Supplementary Fig. 5d and Fig. 8h); the central tendency of estimated US values in the Chronos group was higher compared to the eYFP group (mean: 5.45 vs. 4.30, respectively).

Discussion

Adaptive behaviors based on past experiences significantly increase the probability of survival. Aversive experiences are critical in regulating memory formation and subsequent learning rules^50,51. Previous studies demonstrated that US-driven aversive teaching signals, which are necessary and sufficient to produce a conditioned response, contribute to threat conditioning^52,53. In the present study, our findings demonstrated that aversive learning was enhanced following both optogenetic LTP induction in the US pathway and fearful experiences (Figs. 2e and 3c). These results are consistent with the idea that the effectiveness of US teaching signals is modulated by previous aversive experiences^51,53. To investigate the mechanisms underlying this learning enhancement, we employed the RW model (Fig. 8a–d), a mathematical framework commonly used to explain Pavlovian conditioning and suitable for estimating the US value in threat conditioning paradigms. Our mathematical analyses showed that fearful experiences increased the US value during subsequent learning (Fig. 8c, d). Indeed, learning rules are influenced by US characteristics, such as the intensity or frequency of footshocks. Threat conditioning is facilitated by higher intensity and frequency of footshocks, whereas lower intensity shocks or a single footshock may not suffice to form a robust fear memory^54,55, suggesting that increased US valence of aversive signals can lower the learning threshold and subsequently enhance aversive learning in natural conditions. Previous studies demonstrated that PB-CeC/L pathway is enhanced in acute and chronic pain models in rodents^26,33,35. Furthermore, we have previously shown that aversive learning induces synaptic potentiation in the US pathway³⁷. Collectively, we propose that past fearful experiences can change the valence of the US via synaptic potentiation in the US pathway, thereby influencing learning rules by lowering the threshold for aversive learning and enhanced generalization.

While animals typically discriminate between shock-paired CS1 and shock-unpaired CS2 under normal conditions⁴⁷, high-intensity US can disrupt fear discrimination, leading to generalization⁴⁵. Previous studies have demonstrated that repeated exposure to aversive stimuli promotes generalization in subsequent aversive learning in rodents and humans^56,57,58 and strong intensity of footshock induces generalized fear in mice⁴³. Moreover, evidence suggests that US intensity affects memory generalization in humans⁵⁹. In the present study, we observed that past fearful stimuli promoted generalization in subsequent aversive learning (Fig. 4), consistent with findings from previous studies. Using the behavioral protocols, we can detect potential discrimination deficiency between CS1 and CS2 for discriminative learning, while subtle changes in learning threshold for single CS associative learning. In our experimental paradigm, generalization is also observed in the group without prior fearful experience or in vivo LTP induction, but our findings indicate that at least the US pathway contributes to promote generalization. Mechanisms that effectively regulate generalization are crucial in facilitating efficient learning processes. Our mathematical analyses revealed that past fearful experiences enhance the perceived value of the US, thereby promoting memory generalization (Fig. 8e–h). Taken together, these findings indicate that past fearful experiences may increase the US value in subsequent learning, decrease discrimination ability, and ultimately promote generalization.

Previous studies have reported that aversive learning or generalization is regulated by various neuromodulators, such as noradrenaline⁶⁰, dopamine⁶¹, dynorphin⁴², and corticotropin-releasing factor⁵⁵. However, the specific synaptic targets of these molecules in the context of generalization remain unclear. In the present study, our findings suggest that the PB-CeC/L pathway may be a target for these neuromodulators in aversive learning and generalization. Moreover, the PB-CeC/L pathway transmits nociceptive information, and synaptic potentiation is observed in various chronic pain models in rats and mice^{26,27,28,29,30,31,32,33,34,35,36}. How pathway-specific synaptic plasticity modulates learning rules through these molecules could be a topic for future investigation. Intriguingly, no obvious changes in the mechanical nociceptive threshold were observed after either the shock experience or optogenetic LTP induction of the PB-CeC/L pathway in the present study (Supplementary Figs. 1a and 3a). These superficially contradictory results may suggest that the previous shock experiences and optogenetic LTP induction were insufficient to alter the threshold to nociceptive stimuli. An alternative and more plausible possibility is that the behavioral phenotypes observed in the present study does not come from changes in peripheral mechanical threshold, but rather come from central synaptic plasticity in US pathway. Furthermore, it is plausible that synaptic potentiation in the PB-CeC/L pathway is not causally involved in chronic pain per se but rather modulates susceptibility to chronic pain following future damage, such as peripheral nerve injury and/or inflammation. In addition, leEPSC it is not easy to detect LTP-induced synapses under some experimental conditions because optogenetic verification can also be affected by opsin expression. An exploration of this possibility represents a crucial area for future investigation, which could contribute to our understanding of psychiatric disorders associated with pain. Notably, chronic pain and inflammation are intricately linked to psychiatric disorders, such as post-traumatic stress disorder (PTSD)^62,63,64. Indeed, patients with PTSD exhibit heightened generalization of fear to stimuli unrelated to traumatic events⁶⁵. These observations imply that aversive experiences induce plastic changes in the PB-CeC/L pathway, thereby lowering the learning threshold for traumatic events and generalization of similar events. This aligns with our findings of enhanced aversive learning and generalization based on LTP in the PB-CeC/L pathway. Such insights could bridge the gap between chronic pain and PTSD symptoms regarding their neural bases, offering opportunities for translational research on PTSD.

In this study, we optogenetically stimulated the PB-CeC/L pathway (Figs. 2 and 7). However, the possibility of antidromic activation cannot be completely ruled out due to collateral projections of the PB neurons^17,48. Previous studies have suggested that pathway-specific terminal stimulation of the PB neurons induces different behavioral phenotypes^17,48. In addition, while terminal inhibition of the PB-CeC/L pathway attenuated the enhancement of generalization (Fig. 6f), terminal activation promoted it. Thus, the behavioral phenotype induced by the terminal stimulation of the PB-CeC/L pathway may not be attributable to antidromic activation of the PB.

The neuronal bases for aversive learning and generalization have been extensively studied. Recently, it was demonstrated that generalization by auditory threat conditioning is mediated by an overlapping population of LA neurons, especially when both events occur in close temporal proximity⁶⁶. However, the circuit basis for the generalization via the US pathway at the synaptic level remains largely unknown. Our pathway-specific manipulation demonstrated that LTP in the US pathway, especially in the PB-CeC/L, promoted generalization (Figs. 7 and 8). While memory generalization in fear learning is mediated by extrasynaptic inhibition in protein kinase C-δ-expressing (PKCδ + ) inhibitory neurons in the CeC/L, somatostatin-expressing (SOM + ) neurons in the CeC/L enhance aversive learning rather than generalisation^47,67. Moreover, SOM+ and SOM− neurons show distinct modifications in a US experience-dependent manner⁶⁸, and repeated US experience enhances CeC/L SOM+ responses⁶⁹. These findings suggest that plasticity in CeC/L SOM+ and PKCδ+ neurons may contribute to different aspects of behavioral regulation based on past fearful experiences. Given that multiple neuron types, including PKCδ+ and SOM+ neurons, in the CeC/L receive input from PB neurons^15,70, LTP induction in the PB-CeC/L pathway may change the local circuitry within the CeC/L. The mechanisms underlying aversive learning and generalization via the CeC/L local circuitry represent an interesting topic for further study.

Previous studies have demonstrated the modulation of US values and learning rules based on past US experiences in both invertebrates and vertebrates⁷¹. In the invertebrate Lymnaea, past US exposure modulates future learning, and both perception of weak US and memory acquisition are promoted after a strong US exposure^72,73, which is consistent with our findings (Figs. 2 and 3). Additionally, in the nematode Caenorhabditis elegans, past stressful experiences, such as starvation, heat shock, and exposure to pathogens, modulate the neuronal circuits required for associative learning^74,75,76. In Drosophila, some neurons respond differently to electric shock intensities and encode aversive values⁷⁷. These results are consistent with our findings in this study (Figs. 4 and 7). Taken together, experience-dependent alteration of learning rules via US values appears to be widely conserved in various organisms.

Our findings indicate that LTP in the PB-CeC/L pathway enhances subsequent learning and generalization by increasing affective US values. This provides the mechanism of experience-dependent learning rule alteration by the US pathway, which may be a framework for understanding learning rules in various organisms and for potentially treating stress-related disorders.

Methods

Animals

Male C57BL/6 J mice (Japan SLC, Inc., Shizuoka, Japan) were subjected to the behavioral tests and electrophysiological recordings. The mice were group-housed and provided with food and water ad libitum on a 12 h light/ 12 h dark cycle. All the experimental protocols in this study, including the use of animals, were approved by the Institutional Animal Care and Use Committee of the Jikei University (Tokyo, Japan; Approval No. 2018-072 and 2019-045). All experiments complied with the Guidelines for Proper Conduct of Animal Experiments by the Science Council of Japan (2006) and those recommended by the International Association for the Study of Pain. All experiments were carried out with continuous confirmation that no severe motor deficits or behavioral abnormalities emerged during the experimental period. All efforts were made to reduce the number of animals used and the suffering of the animals.

Adeno-associated virus (AAV)

AAVs used for microinjection were as follows: AAV1-hSyn-Chronos:GFP and AAV5-CaMKIIa-eArchT3.0-eYFP (UNC Vector Core, North Carolina, USA), AAVDJ-Syn-eYFP, AAVDJ-Syn-eGFP, AAVDJ-Syn-mCherry (a generous gift from Prof. Toshihisa Ohtsuka, University of Yamanashi), and AAV5-hSyn-hM4D(Gi):mCherry (Addgene Plasmid #50475).

Stereotaxic surgery

Five-week-old male C57BL/6 J mice were intraperitoneally anaesthetized with a mixture of medetomidine hydrochloride (0.75 mg/kg; Zenoaq, Fukushima, Japan), midazolam (4.0 mg/kg; Astellas, Tokyo, Japan), and butorphanol tartrate (5.0 mg/kg; Meiji Seika Pharma, Tokyo, Japan) and fixed on a stereotaxic device. The AAV vector solution (0.5 μl) was microinjected bilaterally into the PB (6.4 mm posterior to the bregma, 1.25 mm lateral to the midline, and 3.2 mm ventral to the skull surface, with a 20° anterior-to-posterior angle to avoid damaging the superficial arteries during surgery) using a Hamilton microsyringe (1701RN Neuros Syringe, 33 G, 10 μl; Hamilton Company, Reno, NV, USA), as previously described in ref. ⁷⁸. The injection speed (50 nl/min) was controlled by a microsyringe pump (UMP3; UltraMicroPumpII with SYS-Micro4 Controller, UMP2, UMC4; World Precision Instruments, Sarasota, FL, USA). Injection syringes were left in place for 10 min before withdrawing.

After 4–6 weeks, a second surgical procedure was performed for the placement of a bilateral LED cannula unit consisting of dual optical fibres (0.25 mm in diameter, 4.5 mm in length, and 6.0 mm in spacing) attached to an LED body (470 nm, TeleLCD-B-4.5-250-6.0; 590 nm, TeleLCD-Y-4.5-250-6.0, Bio Research Center, Tokyo, Japan). The LED cannula unit was stereotactically inserted to target the CeC/L (1.3 mm posterior to the bregma, 3.0 mm lateral to the midline, and 4.4 mm ventral to the skull surface) and fixed to the skull with dental cement (GC Fuji I; GC Corporation, Tokyo, Japan). The mice were allowed to recover for several days.

Drugs

The DREADD agonist clozapine N-oxide (CNO) (Sigma Aldrich) was diluted in saline to 0.5 mg/mL and intraperitoneally administered to the mice at 5 mg/kg 30 min before each experimental session.

Behavioral experiments

Shock experience

The mice were placed in Context A (triangular shapes, 170 mm width × 100 mm depth × 100 mm height, acrylic black wall, 10 lux, 50 dB background white noise) surrounded by a sound-attenuating box. Following a 120-s observation period and conditioning with nine pairings of a 20 s white noise tone (65 dB) that terminated concurrently with a footshock (0.6 mA, 2 s). Footshocks were delivered at 40- or 50-s inter-trial intervals using the floor grid of the chamber using a shock generator (O’Hara & Co., Ltd, Tokyo, Japan). The Control groups were subjected to the same context without footshock. In chemogenetic inhibition, the mice were intraperitoneally injected with CNO before 30 min in shock experience. In pathway-specific optogenetic inhibition, the mice received the light stimulation (590 nm, continuous illumination) during a shock experience. One sample from the Chronos group, three samples from the ArchT group, and three samples from the hM4Di group were excluded due to unclear AAV expression or placement of the LED cannula tip.

In vivo optogenetic LTP induction

The mice were placed in Context A surrounded by a sound-attenuating box. Following a brief period of 180 s, in vivo LTP induction was initiated via optogenetic high-frequency stimulation (40 Hz, 2 s on/3 s off, 5 ms, 7 min).

Weak threat conditioning

The day following the initial shock experience or in vivo optogenetic LTP induction, mice underwent weak threat conditioning, as previously described, with slight modifications^14,37. The mice were placed in Context B (square shapes, 170 mm width × 100 mm depth × 100 mm height, acrylic clear wall, floor metal grids, 200 lux, 60 dB background white noise) surrounded by a sound-attenuating box. Following the 270 s observation period, the mice were conditioned with three pairings of CS tone (10 kHz, 65 dB, 30 s at 270, 440, and 570 s) that co-terminated with mild footshocks as the US (0.1 mA, 2 s). The day after weak threat conditioning, the mice were placed in Context C (square shapes, 170 mm width × 100 mm depth × 100 mm height, white acrylic plate walls scented with peppermint odor, 200 lux, 50 dB background white noise, sandpaper on the floor, 200 lux, 50 dB background white noise) and they were administered six CSs (30 s each). The CS presentation began 270 s after the test initiation.

Differential threat conditioning

The day following the initial shock experience or in vivo optogenetic LTP induction, the mice were placed in Context B. Following the 270-s observation period, the mice were conditioned with interleaved paired tone CS1 (4 kHz pure tone, 65 dB) that co-terminated with a 0.3 mA footshock and unpaired tone CS2 (12.5 kHz pulsatile tone, 65 dB) with variable inter-tone interval. CS1 and CS2 were presented three times each, as detailed in a previous report⁴⁷. Specifically, CS1 was introduced into the session at 270, 440, and 570 s, whereas CS2 was presented at 370, 520, and 660 s. The day after the differential threat conditioning experiment, the mice were placed in Context C and received three interleaved CS1 and CS2 tone presentations. CS1 was presented at 370, 520, and 660 s, whereas CS2 was presented at 270, 440, and 570 s. All tones lasted for 30 s. After the last CS1 presentation, the mice were allowed to remain in Context C for an additional 30 s.

Freezing analyses for threat conditioning

All behavioral analyses and post-mortem histological analyses were conducted by experimenters blinded to the manipulations and behavioral analyses, respectively. Mouse behavior was captured using a digital camera at 2 frames/s, and freezing behavior was analyzed using Time FZ software (O’Hara & Co., Ltd., Japan), a package based on National Institute of Health images. The movements of the mice were detected by pixel-to-pixel subtraction between two subsequent frames. The behavior in each frame was defined as “freezing” when the total number of pixels with a detectable frame-to-frame difference was <30. The freezing ratio was calculated as the proportion of time the subject spent freezing during the tone presentation divided by the total duration of the tone. This ratio ranges from 0 to 1, where 0 indicates no freezing and 1 indicates continuous freezing. The identification of freezing was pre-optimized by two independent human observers using C57BL/6 J mice.

In the discrimination test, the discrimination index was calculated as follows with reference to a previous report⁴⁷:

$${{{\rm{Discrimination\; index}}}}=\tfrac{\tfrac{\mathop{\sum }_{i=1}^{N}{CS}1i}{N}-\tfrac{\mathop{\sum }_{i=1}^{N}{CS}2i}{N}}{\tfrac{\mathop{\sum }_{i=1}^{N}{CS}1i}{N}+\tfrac{\mathop{\sum }_{i=1}^{N}{CS}2i}{N}}$$

where the parameter of $N$ is the number of animals, ${CS}1$ is the CS1-evoked freezing ratio, and ${CS}2$ is the CS2-evoked freezing ratio.

Real-time place aversion (RTPA)

RTPA test was conducted in a custom-built two-compartment apparatus (CL-2CS; O’Hara & Co., Ltd., Tokyo, Japan) placed in a sound-attenuating box (CL-M3; O’Hara & Co., Ltd.). Each compartment (170 mm width × 200 mm depth × 230 mm height) contains different wall patterns (Shock compartment, dots; Control compartment, stripes). In the baseline session (Day 1), mice were placed to the two-compartment chamber and allowed to explore freely for 10 min. On the following day (Day 2), mice underwent either shock-experience or in vivo optogenetic LTP induction protocol. On Day 3, the RTPA test was conducted for 10 min. Mice received mild footshocks (0.02 mA), whenever they entered Shock compartment. Behavioral analyses were performed using Time OFCR1 software (O’Hara & Co., Ltd., Japan).

von Frey test

The paw withdrawal threshold to mechanical stimuli was evaluated using a series of von Frey filaments (North Coast Medical, Inc., Gilroy, CA, USA) of different rigidity (0.02–2.0 g) as previously described in refs. ^14,37. Each mouse was placed on a metal mesh floor and allowed to habituate to a 500 mL glass beaker placed upside down for 30 min before the experiments. The 50% threshold was estimated using the up-and-down method⁷⁹. The mechanical threshold was determined as the average of the measurements for both hind paws per mouse.

Marble burying test

Under lighting conditions (100 lux), the mice were placed in an environment mimicking their home cage, with identical dimensions (197 mm width × 340 mm depth × 165 mm height) filled with a 5 cm thick layer of bedding chips and 15 round marbles (1 cm in diameter; 3 rows marbles × 5 marbles) 2 cm apart. The mice were placed in cages and allowed to exhibit exploratory behaviors. After 30 min, the mice were removed, and up to two-thirds of the number of marbles buried was counted independently by two observers blinded to the experimental group to prevent any subjective bias.

Open-field test

An open-field test was conducted using a square apparatus (100 lux) with grey walls (50 cm width × 50 cm length × 30 cm height) (O’Hara & Co., Ltd., Japan). The mice were placed in a corner of the apparatus and allowed to freely explore the environment for 10 min. Ambulation was recorded and analyzed using a video-computerized tracking system (TimeOFCR1; O’Hara & Co., Ltd., Japan). The entire open-field area was divided into 25 (5 × 5) squares, and the central nine (3 × 3) squares surrounded by the peripheral 16 squares were considered the central region. The total distance travelled (cm) and time spent in the central region (%) were analyzed as proxies for basic locomotor activity and general anxiety levels, respectively.

Shock sensitivity

Flinch, vocalization, and jump threshold to footshock were evaluated in the chamber used for threat conditioning (200 lux, 0 dB background white noise). Mice were placed individually in the=chamber. After a 90 s period of habituation, footshocks with increasing intensities were given in a stepwise manner (from 0.01 to 0.15 mA, in 0.01 mA steps). The time gap between shocks was 30 s, shock duration was 1 s, and each animal was tested only once. Behaviors were captured and recoded using a web camera (HD Webcam C525, Logicool) and its software (version 2.51, Logicool) inside the sound-attenuating chamber. Any detectable reaction to the shock was manually analyzed. The flinch threshold was defined as the lowest shock intensity that elicited a flinch. The vocalization threshold was defined as the lowest shock intensity that elicited an audible vocalization. The jump threshold was defined as the lowest shock intensity that elicited simultaneous removal of both hindpaws from the grid.

Electrophysiological experiments

Preparation of the acute brain slices

Five to 19 weeks after AAV1-hSyn-Chronos:GFP injection, the mice were deeply anaesthetized with isoflurane (5% in 100% O₂) and decapitated. For electrophysiological recordings following behavioral tests, the mice were decapitated within an hour after the behavioral test. The brains were quickly removed and placed into an ice-cold cutting solution consisting of (in mM) 92 N-methyl-D-glucamine, 2.5 KCl, 0.5 CaCl₂, 10 MgSO₄, 1.25 NaH₂PO₄, 2 thiourea, 3 sodium pyruvate, 12 N-acetyl-L-cysteine, 25 D-glucose, 5 L-ascorbic acid, 20 HEPES, and 30 NaHCO₃, equilibrated with 95% O₂ + 5% CO₂ (pH of approximately 7.4, osmolality of approximately 290 mOsm/kg). Coronal slices (300 µm) containing the amygdala were cut with a vibrating blade slicer (VT1200S, Leica). The slices were first incubated in the cutting solution at approximately 34 °C for 10–15 min, and then kept at room temperature (20–25 °C) in standard artificial cerebrospinal fluid (ACSF) consisting of (in mM) 125 NaCl, 3 KCl, 2 CaCl₂, 1.3 MgCl₂, 1.25 NaH₂PO₄, 10 D-glucose, 0.4 L-ascorbic acid, and 25 NaHCO₃ (pH of approximately 7.4 bubbled with 95% O₂ + 5% CO₂, osmolality of approximately 300 mOsm/kg) until whole-cell recordings.

Whole-cell patch-clamp recordings in acute brain slices

A slice was transferred to a recording chamber (RC-26GS, Warner Instruments, Holliston, MA, USA) and continuously superfused at a rate of 1.5–2.5 mL/min with standard ACSF at approximately 30 °C. Whole-cell patch-clamp recordings were conducted on CeC/L neurons, which were visually identified under an upright microscope with oblique illumination (BX-51WI, Olympus) as previously reported⁸⁰. A patch-clamp electrode (4–8 MΩ) made of borosilicate glass pipettes (1B150F-4, World Precision Instruments) was filled with an internal solution composed of (in mM) 122.5 potassium gluconate, 10 HEPES, 17.5 KCl, 0.2 EGTA, 8 NaCl, 2 MgATP, and 0.3 NaGTP (pH 7.2, osmolarity 290–300 mOsm). Potassium gluconate was replaced by cesium gluconate for the analysis of NMDA/AMPA ratio. To isolate leEPSCs, cells were held at −60 mV and recorded in the presence of picrotoxin (100 μM). The membrane current was recorded with a MultiClamp 700B amplifier (Molecular Devices, San Jose, CA, USA), filtered at 2 kHz, and digitized at 10 kHz with a 16-bit resolution using a PowerLab interface (AD Instruments, Colorado Springs, CO, USA) and Digidata 1440 A (Molecular Neuroscience), together with timing pulses for light stimulation. Chronos was activated using an LED illumination system mounted on the microscope (455 or 470 nm; 455L3 or M470L3 and M470L4, Thorlabs, Newton, NJ, USA; duration, 2 or 5 ms), delivered to the entire field through a 40 × water-immersion objective lens (LUMPLFLN40XW, NA 0.8; Olympus). Photostimulation timing and duration were controlled by Master-8 (A.M.P.I., Jerusalem, Israel). Paired-pulse (100 ms interval) light stimuli were applied every 20 s. The paired-pulse ratio was defined as the ratio of the second leEPSC amplitude to the first leEPSC amplitude. The recorded membrane currents were analyzed offline with Igor Pro 7 or 8 (WaveMetrics, Portland, OR, USA) and pClamp Clampfit 10.7 (Molecular Devices).

To verify that the PB–CeC/L pathway is monosynaptic and glutamatergic, tetrodotoxin (TTX; 1 μM; abcam), 4-Aminopyridine (4AP; 100 μM; Sigma), 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX; 10 μM; Sigma) and D-2-amino-5-phosphonovalerate (APV; 50 μM; Tocris) were added to the ACSF containing picrotoxin (100 μM; Sigma and FUJIFILM Wako Pure Chemical Corporation) and bath applied.

To record synaptic plasticity, leEPSCs were stabilized for 10 min, and then 5 trains of 20, 40, or 50 Hz (0.5 s; duration, 2 ms) were applied at 0.1 Hz. leEPSCs were recorded for a further 30 min. The amplitudes of three consecutive first leEPSCs, corresponding to 1 min, were averaged and normalized to the mean amplitude of the 10-minute baseline. To analyze whether the plastic changes were significant, normalized amplitudes of leEPSCs for 25–30 min after the HFS were compared to 100% using the one-sample t-test. In the electrophysiological analysis after behavioral tests, 15 consecutive leEPSCs were used to calculate the average of individual neurons.

To assess synaptic potentiation following in vivo optogenetic LTP induction, an input-output analysis was conducted on the following day. Light-evoked EPSCs were elicited by presynaptic stimulation using light intensities of 0.030, 0.125, 0.500, 2.200, and 17.8 mW/mm². The NMDA/AMPA ratio was calculated by dividing NMDA receptor-mediated EPSCs, recorded at a holding potential of +40 mV in the presence of CNQX, by AMPA receptor-mediated EPSCs, recorded at a holding potential of −60 mV.

To assess synaptic potentiation following successive threat conditioning, leEPSCs were recorded after retrieval test. Only leEPSCs with amplitudes greater than 20 pA were used for analysis. The PPR was calculated as EPSC2/EPSC1.

Mathematical model

Bayesian inference of the US value using the Rescorla–Wagner model

The RW model is a common mathematical model used to explain associative learning in classical conditioning. For successive threat conditioning, the RW model is formulated as follows:

$${w}_{n,t}={w}_{n,t-1}+\alpha \left(U{S}_{{{{\rm{Group}}}}}-{V}_{n,t-1}\right),$$

(1)

$${V}_{n,t-1}={w}_{n,t-1}{CS},$$

(2)

where the parameters$\,\alpha ,{U}{S}_{{{{\rm{Group}}}}},{and\; CS}$ are the learning rate, value of the US, and value of the CS, respectively, and the variables ${w}_{n,t}\,{{and\; V}}_{n,t}$ are the association level between the US and CS expressed as the synaptic weights in the path of CS input to CeC/L and the predicted US value expressed as CeC/L activity of sample $n$ at time $t$, respectively. Here, we assumed $\alpha$ has the same value among all mice, and $U{S}_{{{{\rm{Group}}}}}$ has the same value for the same experimental group of mice, exposed to the same experience $({{{\rm{group}}}}=\{{{{\rm{Extinction}}}},{{{\rm{Ctrl}}}},{{{\rm{Shock}}}},{{{\rm{Chronos}}}},{{{\rm{eYFP}}}}\})$. Furthermore, we considered $U{S}_{{{{\rm{group}}}}}$ as the value of the US relative to the value of the CS, and CS was set to 1.

Next, we formulated the relationship between the variables of the learning process formulated in the above RW model and an observable variable, the freezing ratio, by the following equation using the tangent hyperbolic function:

$${{{{\rm{z}}}}}_{n,t}=\tanh ({V}_{n,t})=\frac{\exp \left({V}_{n,t}\right)-\exp \left({-V}_{n,t}\right)}{\exp \left({V}_{n,t}\right)+\exp \left({-V}_{n,t}\right)},$$

(3)

$${y}_{n,t}={z}_{n,t}+{\sigma }_{y}^{{{{\rm{Exp}}}}}\epsilon ,$$

(4)

where ${{{{\rm{z}}}}}_{n,t}$ and ${y}_{n,t}$ are the mean freezing ratio and observed freezing ratio of sample $n$ at time $t$, respectively. The parameters ${\sigma }_{y}^{{{{\rm{Exp}}}}}$ and $\epsilon$ are the noise strength and Gaussian noise with zero mean and unit variance, respectively, where ${Exp}=\{{{{\rm{Extinction}}}},{{{\rm{Successive\; Conditioning}}}}\}$.

We inferred the value of the US for each of the groups Ctrl, Shock, Chronos, and eYFP using a two-step procedure. First, we performed a fear extinction experiment using a distinct group of mice (an Extinction group). In a fear extinction test, the US is not presented, which means that the value of the US for mice under a fear extinction test ${{US}}_{{{{\rm{Extinction}}}}}=0$. With this assumption, we estimated the learning rate $\alpha$ and sample-specific initial value of the CS-US association level ${w}_{n,0}$ from the observation of freezing ratios during a fear extinction test. The prior distribution of the learning rate was assumed to follow the following beta distribution:

$$p\left(\alpha \right)={{{\rm{Beta}}}}\left(\alpha {|a},b\right),$$

(5)

and the prior distribution of the sample-specific initial parameter of the CS-US association level was assumed to follow the following Gaussian distribution:

$$p\left({w}_{n,0}\right)=N\left({w}_{n,0}|0,{\sigma }_{w}^{{{{\rm{Extinction}}}}}\right).$$

(6)

Estimation of the parameters was performed using the Markov chain Monte Carlo (MCMC) with Metropolis-Hastings (M-H) method. The hyper-parameters were set as ${\sigma }_{y}^{{{{\rm{Extinction}}}}}=0.1,{a}=2,{b}=2$, and ${\sigma }_{w}^{{{{\rm{Extinction}}}}}=20.$ We adopted the mean of the samples as an estimate of the common learning rate among mice $\alpha$.

In the second step, from the observation of freezing ratios during a successive fear condition test for each of the groups Ctrl, Shock, Chronos, and eYFP, we inferred the value of the US using the learning rate $\alpha$ estimated in the first step. The prior distribution of the US value was assumed to follow the following Gaussian distribution:

$$p\left({US}\right)=N\left({US|}{\mu }_{{US}},{\sigma }_{{US}}^{2}\right),$$

(7)

and the prior distribution of the sample-specific initial parameter of the CS-US association level was assumed to follow the following Gaussian distribution:

$$p\left({w}_{n,0}\right)=N\left({w}_{n,0}|0,{\sigma }_{w}^{{{{\rm{Succesive\; Conditioning}}}}}\right).$$

(8)

Similar to the first step, an estimation of the parameters was performed using the MCMC with M-H method. The hyper-parameters were set as ${\sigma }_{y}^{{{{\rm{Successive\; Conditioning}}}}}=0.01,{\mu }_{{US}}=1,{\sigma }_{{US}}^{2}=0.0000001$, and ${\sigma }_{w}^{{{{\rm{Successive\; Conditioning}}}}}=0.1.$

RW model of CS1-CS2 generalization is formulated as follows:

$${w}_{n,t}={w}_{n,t-1}+\alpha \left({US}-{V}_{t-1}\right),$$

(9)

$${V}_{n,t}={w}_{n,t-1}{CS}1+{w}_{n,t-1}{rCS}2,$$

(10)

$${{{{\rm{z}}}}}_{n,t}=\tanh ({V}_{n,t}),$$

(11)

$${y}_{n,t}={z}_{n,t}+{\sigma }_{y}^{{{{\rm{Exp}}}}}\epsilon ,$$

(12)

Where the parameters CS1, CS2, and $r$ are value of CS1, value of CS2, and the generalization parameter, respectively. Other parameters are same at Eqs. (1), (2). If $r$ is large, CS2 predictions are prioritized; if $r$ is small, CS1 predictions are prioritized.

We estimated parameter US using MCMC at the same distribution when RW model parameter US was estimated. The hyper-parameter were set as ${\sigma }_{y}^{{{{\rm{Successive\; Conditioning}}}}}=0.02,\,{\mu }_{{US}}=6,\,{\sigma }_{{US}}^{2}=0.05,{r}=1.5$ and ${\sigma }_{w}^{{{{\rm{Successive\; Conditioning}}}}}=0.1$.

Regarding the MCMC implementation, we performed sampling across three independent chains. To estimate the distribution of parameters, we used 100,000 samples obtained from one chain, discarding the first 1000 samples (or 30,000 samples for the CS1-CS2 generalization experiment) as burn-in. The other two chains were used to ensure reproducibility. Description of the model and implementation of the estimation were performed using a custom script in Python.

Statistics and Reproducibility

Statistical analyses were conducted using GraphPad Prism 9 software (GraphPad Software, La Jolla, CA, USA). Data in the text and figures are expressed as mean ± S.E.M. of the sample number (n). Animals were randomly assigned to the experimental conditions described in this study. All experiments were independently replicated at least once, yielding similar results. The sample sizes are indicated in the figure legends, and the statistical test used for each comparison is indicated (See also Supplementary Table 1). We used appropriate statistical tests with post hoc analyses when applicable, i.e., two-tailed unpaired t-tests and two-way repeated-measures ANOVA with Bonferroni’s multiple comparisons tests. The Mann–Whitney U test was used for normalized data such as the discrimination index. The one-sample t-test was used to compare leEPSC potentiation (%) results in the LTP induction experiments using acute brain slices. Differences were considered statistically significant when p < 0.05. Graphs were created using GraphPad Prism and Igor 8 software.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

All data generated or analyzed during this study are included in this article (See also Supplementary Data 1). The raw behavioral data used in this study are available from the corresponding author upon request.

Code availability

The code for the estimation of US values in the present study is available at https://doi.org/10.5281/zenodo.15598908⁸¹.

References

Maren, S. & Fanselow, M. S. The amygdala and fear conditioning: has the nut been cracked?. Neuron 16, 237–240 (1996).
Article CAS Google Scholar
Maren, S. Neurobiology of Pavlovian fear conditioning. Annu Rev. Neurosci. 24, 897–931 (2001).
Article CAS Google Scholar
Schafe, G. E., Nader, K., Blair, H. T. & LeDoux, J. E. Memory consolidation of Pavlovian fear conditioning: a cellular and molecular perspective. Trends Neurosci. 24, 540–546 (2001).
Article CAS Google Scholar
Maren, S. & Quirk, G. J. Neuronal signalling of fear memory. Nat. Rev. Neurosci. 5, 844–852 (2004).
Article CAS Google Scholar
Tovote, P., Fadok, J. P. & Lüthi, A. Neuronal circuits for fear and anxiety. Nat. Rev. Neurosci. 16, 317–331 (2015).
Article CAS Google Scholar
Ressler, R. L. & Maren, S. Synaptic encoding of fear memories in the amygdala. Curr. Opin. Neurobiol. 54, 54–59 (2019).
Article CAS Google Scholar
Choi, D. I. et al. Synaptic correlates of associative fear memory in the lateral amygdala. Neuron 109, 2717–2726.e2713 (2021).
Article CAS Google Scholar
Ciocchi, S. et al. Encoding of conditioned fear in central amygdala inhibitory circuits. Nature 468, 277–282 (2010).
Article CAS Google Scholar
Haubensak, W. et al. Genetic dissection of an amygdala microcircuit that gates conditioned fear. Nature 468, 270–276 (2010).
Article CAS PubMed Central Google Scholar
Fadok, J. P. et al. A competitive inhibitory circuit for selection of active and passive fear responses. Nature 542, 96–100 (2017).
Article CAS Google Scholar
Moscarello, J. M. & Penzo, M. A. The central nucleus of the amygdala and the construction of defensive modes across the threat-imminence continuum. Nat. Neurosci. 25, 999–1008 (2022).
Article CAS Google Scholar
Palmiter, R. D. The Parabrachial Nucleus: CGRP Neurons Function as a General Alarm. Trends Neurosci. 41, 280–293 (2018).
Article CAS PubMed Central Google Scholar
Campos, C. A., Bowen, A. J., Roman, C. W. & Palmiter, R. D. Encoding of danger by parabrachial CGRP neurons. Nature 555, 617–622 (2018).
Article CAS PubMed Central Google Scholar
Sato, M. et al. The lateral parabrachial nucleus is actively involved in the acquisition of fear memory in mice. Mol. Brain 8, 22 (2015).
Article PubMed Central Google Scholar
Han, S., Soleiman, M. T., Soden, M. E., Zweifel, L. S. & Palmiter, R. D. Elucidating an Affective Pain Circuit that Creates a Threat Memory. Cell 162, 363–374 (2015).
Article CAS PubMed Central Google Scholar
Kang, S. J. et al. A central alarm system that gates multi-sensory innate threat cues to the amygdala. Cell Rep. 40, 111222 (2022).
Article CAS PubMed Central Google Scholar
Bowen, A. J. et al. Dissociable control of unconditioned responses and associative fear learning by parabrachial CGRP neurons. eLife 9, e59799 (2020).
Article CAS PubMed Central Google Scholar
Bowen, A. J. et al. Topographic representation of current and future threats in the mouse nociceptive amygdala. Nat. Commun. 14, 196 (2023).
Article CAS PubMed Central Google Scholar
Ito, M. et al. The parabrachial-to-amygdala pathway provides aversive information to induce avoidance behavior in mice. Mol. Brain 14, 94 (2021).
Article PubMed Central Google Scholar
Nagase, M., Mikami, K. & Watabe, A. M. Parabrachial-to-amygdala control of aversive learning. Curr. Opin. Behav. Sci. 26, 18–24 (2019).
Article Google Scholar
Quirk, G. J., Armony, J. L. & LeDoux, J. E. Fear conditioning enhances different temporal components of tone-evoked spike trains in auditory cortex and lateral amygdala. Neuron 19, 613–624 (1997).
Article CAS Google Scholar
Tsvetkov, E., Carlezon, W. A., Benes, F. M., Kandel, E. R. & Bolshakov, V. Y. Fear conditioning occludes LTP-induced presynaptic enhancement of synaptic transmission in the cortical pathway to the lateral amygdala. Neuron 34, 289–300 (2002).
Article CAS Google Scholar
Kim, W. B. & Cho, J. H. Encoding of Discriminative Fear Memory by Input-Specific LTP in the Amygdala. Neuron 95, 1129–1146.e1125 (2017).
Article CAS Google Scholar
Nabavi, S. et al. Engineering a memory with LTD and LTP. Nature 511, 348–352 (2014).
Article CAS PubMed Central Google Scholar
Abdou, K. et al. Synapse-specific representation of the identity of overlapping memory engrams. Science 360, 1227–1231 (2018).
Article CAS Google Scholar
Neugebauer, V., Li, W., Bird, G. C., Bhave, G. & Gereau, R. W. IV Synaptic Plasticity in the Amygdala in a Model of Arthritic Pain: Differential Roles of Metabotropic Glutamate Receptors 1 and 5. J. Neurosci. 23, 52–63 (2003).
Article CAS PubMed Central Google Scholar
Han, J. S., Bird, G. C. & Neugebauer, V. Enhanced group III mGluR-mediated inhibition of pain-related synaptic plasticity in the amygdala. Neuropharmacology 46, 918–926 (2004).
Article CAS Google Scholar
Han, J. S. & Neugebauer, V. Synaptic plasticity in the amygdala in a visceral pain model in rats. Neurosci. Lett. 361, 254–257 (2004).
Article CAS Google Scholar
Bird, G. C. et al. Protein kinase A-dependent enhanced NMDA receptor function in pain-related synaptic plasticity in rat amygdala neurones. J. Physiol. 564, 907–921 (2005).
Article CAS PubMed Central Google Scholar
Han, J. S., Fu, Y., Bird, G. C. & Neugebauer, V. Enhanced group II mGluR-mediated inhibition of pain-related synaptic plasticity in the amygdala. Mol. Pain. 2, 18 (2006).
Article PubMed Central Google Scholar
Ren, W. & Neugebauer, V. Pain-related increase of excitatory transmission and decrease of inhibitory transmission in the central nucleus of the amygdala are mediated by mGluR1. Mol. Pain. 6, 93 (2010).
Article CAS PubMed Central Google Scholar
Neugebauer, V. in Pain Control (ed Hans-Georg Schaible) 261-284 (Springer Berlin Heidelberg, 2015).
Ikeda, R., Takahashi, Y., Inoue, K. & Kato, F. NMDA receptor-independent synaptic plasticity in the central amygdala in the rat model of neuropathic pain. Pain 127, 161–172 (2007).
Article CAS Google Scholar
Miyazawa, Y., Takahashi, Y., Watabe, A. M. & Kato, F. Predominant synaptic potentiation and activation in the right central amygdala are independent of bilateral parabrachial activation in the hemilateral trigeminal inflammatory pain model of rats. Mol. Pain. 14, 1744806918807102 (2018).
Article CAS PubMed Central Google Scholar
Torres-Rodriguez, J. M. et al. The parabrachial to central amygdala pathway is critical to injury-induced pain sensitization in mice. Neuropsychopharmacology https://doi.org/10.1038/s41386-023-01673-6 (2023).
Shinohara, K. et al. Essential role of endogenous calcitonin gene-related peptide in pain-associated plasticity in the central amygdala. Eur. J. Neurosci. 46, 2149–2160 (2017).
Article PubMed Central Google Scholar
Watabe, A. M. et al. Synaptic potentiation in the nociceptive amygdala following fear learning in mice. Mol. Brain 6, 11 (2013).
Article PubMed Central Google Scholar
Rau, V. & Fanselow, M. S. Exposure to a stressor produces a long lasting enhancement of fear learning in rats. Stress 12, 125–133 (2009).
Article Google Scholar
Nomura, H., Teshirogi, C., Nakayama, D., Minami, M. & Ikegaya, Y. Prior observation of fear learning enhances subsequent self-experienced fear learning with an overlapping neuronal ensemble in the dorsal hippocampus. Mol. Brain 12, 21 (2019).
Article PubMed Central Google Scholar
Perusini, J. N. et al. Induction and Expression of Fear Sensitization Caused by Acute Traumatic Stress. Neuropsychopharmacology 41, 45–57 (2016).
Article CAS Google Scholar
Klapoetke, N. C. et al. Independent optical excitation of distinct neural populations. Nat. Methods 11, 338–346 (2014).
Article CAS PubMed Central Google Scholar
Fellinger, L. et al. A midbrain dynorphin circuit promotes threat generalization. Curr. Biol. 31, 4388–4396.e4385 (2021).
Article CAS PubMed Central Google Scholar
Li, H -q. et al. Generalized fear after acute stress is caused by change in neuronal cotransmitter identity. Science 383, 1252–1259 (2024).
Article CAS PubMed Central Google Scholar
Dos Santos Corrêa, M. et al. Relationship between footshock intensity, post-training corticosterone release and contextual fear memory specificity over time. Psychoneuroendocrinology 110, 104447 (2019).
Article Google Scholar
Venkataraman, A. et al. Modulation of fear generalization by the zona incerta. Proc. Natl Acad. Sci. USA 116, 9072–9077 (2019).
Article CAS PubMed Central Google Scholar
Xuan, S. M. et al. mGluR5 in amygdala modulates fear memory generalization. Front Behav. Neurosci. 17, 1072642 (2023).
Article CAS PubMed Central Google Scholar
Shrestha, P. et al. Amygdala inhibitory neurons as loci for translation in emotional memories. Nature 586, 407–411 (2020).
Article CAS PubMed Central Google Scholar
Chiang, M. C. et al. Divergent Neural Pathways Emanating from the Lateral Parabrachial Nucleus Mediate Distinct Components of the Pain Response. Neuron 106, 927–939.e925 (2020).
Article CAS Google Scholar
Recorla, R. A. & Wagner, A. R. A theory of Pavlovian conditioning: variations in the effectiveness of reinforcement and nonreinforcement. In Classical Conditioning II: Current Research and Theory (eds. Prokasy, W. F. & Black, A. H.), 64–99 (1972).
Meulders, A. et al. Preexposure to (un)predictable shock modulates discriminative fear learning between cue and context: an investigation of the interaction between fear and anxiety. Int J. Psychophysiol. 84, 180–187 (2012).
Article Google Scholar
Ozawa, T. et al. A feedback neural circuit for calibrating aversive memory strength. Nat. Neurosci. 20, 90–97 (2017).
Article CAS Google Scholar
Johansen, J. P. & Fields, H. L. Glutamatergic activation of anterior cingulate cortex produces an aversive teaching signal. Nat. Neurosci. 7, 398–403 (2004).
Article CAS Google Scholar
Johansen, J. P., Tarpley, J. W., LeDoux, J. E. & Blair, H. T. Neural substrates for expectation-modulated fear learning in the amygdala and periaqueductal gray. Nat. Neurosci. 13, 979–986 (2010).
Article CAS PubMed Central Google Scholar
Cole, K. E., Lee, J., Davis, M. & Parsons, R. G. Subthreshold Fear Conditioning Produces a Rapidly Developing Neural Mechanism that Primes Subsequent Learning. eneuro 6, ENEURO.0113–0119 (2019).
Article Google Scholar
Sanford, C. A. et al. A Central Amygdala CRF Circuit Facilitates Learning about Weak Threats. Neuron 93, 164–178 (2017).
Article CAS Google Scholar
Bender, C. L., Otamendi, A., Calfa, G. D. & Molina, V. A. Prior stress promotes the generalization of contextual fear memories: Involvement of the gabaergic signaling within the basolateral amygdala complex. Prog. Neuro-Psychopharmacol. Biol. Psychiatry 83, 18–26 (2018).
Article CAS Google Scholar
Schechtman, E., Laufer, O. & Paz, R. Negative Valence Widens Generalization of Learning. J. Neurosci. 30, 10460–10464 (2010).
Article CAS PubMed Central Google Scholar
Resnik, J., Sobel, N. & Paz, R. Auditory aversive learning increases discrimination thresholds. Nat. Neurosci. 14, 791–796 (2011).
Article CAS Google Scholar
Gao, Y. et al. Role of Cue Training, Context, and Stimulus Intensity on Fear Generalization in Humans. Behav. Sci. 13, 479 (2023).
Article PubMed Central Google Scholar
Johansen, J. P. et al. Hebbian and neuromodulatory mechanisms interact to trigger associative memory formation. Proc. Natl Acad. Sci. 111, E5584–E5592 (2014).
Article CAS PubMed Central Google Scholar
Jo, Y. S., Heymann, G. & Zweifel, L. S. Dopamine Neurons Reflect the Uncertainty in Fear Generalization. Neuron 100, 916–925.e913 (2018).
Article CAS Google Scholar
Mickleborough, M. J. et al. Effects of trauma-related cues on pain processing in posttraumatic stress disorder: an fMRI investigation. J. Psychiatry Neurosci. 36, 6–14 (2011).
Article PubMed Central Google Scholar
Tesarz, J., Baumeister, D., Andersen, T. E. & Vaegter, H. B. Pain perception and processing in individuals with posttraumatic stress disorder: a systematic review with meta-analysis. Pain. Rep. 5, e849 (2020).
Article PubMed Central Google Scholar
Asmundson, G. J. G., Coons, M. J., Taylor, S. & Katz, J. PTSD and the Experience of Pain: Research and Clinical Implications of Shared Vulnerability and Mutual Maintenance Models. Can. J. Psychiatry 47, 930–937 (2002).
Article Google Scholar
Lis, S. et al. Generalization of fear in post-traumatic stress disorder. Psychophysiology 57, e13422 (2020).
Article CAS Google Scholar
Jung, J. H. et al. Examining memory linking and generalization using scFLARE2, a temporally precise neuronal activity tagging system. Cell Rep. 42, 113592 (2023).
Article CAS PubMed Central Google Scholar
Botta, P. et al. Regulating anxiety with extrasynaptic inhibition. Nat. Neurosci. 18, 1493–1500 (2015).
Article CAS PubMed Central Google Scholar
Li, H. et al. Experience-dependent modification of a central amygdala fear circuit. Nat. Neurosci. 16, 332–339 (2013).
Article CAS PubMed Central Google Scholar
Yang, T. et al. Plastic and stimulus-specific coding of salient events in the central amygdala. Nature 616, 510–519 (2023).
Article CAS PubMed Central Google Scholar
Li, J.-N. & Sheets, P. L. Spared nerve injury differentially alters parabrachial monosynaptic excitatory inputs to molecularly specific neurons in distinct subregions of the central amygdala. PAIN 161, 166–176 (2020).
Article CAS Google Scholar
Perisse, E., Miranda, M. & Trouche, S. Modulation of aversive value coding in the vertebrate and invertebrate brain. Curr. Opin. Neurobiol. 79, 102696 (2023).
Article CAS Google Scholar
Crossley, M. et al. Proactive and retroactive interference with associative memory consolidation in the snail Lymnaea is time and circuit dependent. Commun. Biol. 2, 242 (2019).
Article PubMed Central Google Scholar
Crossley, M., Benjamin, P. R., Kemenes, G., Staras, K. & Kemenes, I. A circuit mechanism linking past and future learning through shifts in perception. Sci. Adv. 9, eadd3403 (2023).
Article CAS PubMed Central Google Scholar
Tatum Marcus, C. et al. Neuronal Serotonin Release Triggers the Heat Shock Response in C. elegans in the Absence of Temperature Increase. Curr. Biol. 25, 163–174 (2015).
Article CAS Google Scholar
Jin, X., Pokala, N. & Bargmann, C. I. Distinct circuits for the formation and retrieval of an imprinted olfactory memory. Cell 164, 632–643 (2016).
Article CAS PubMed Central Google Scholar
Vogt, M. C. & Hobert, O. Starvation-induced changes in somatic insulin/IGF-1R signaling drive metabolic programming across generations. Sci. Adv. 9, eade1817 (2023).
Article CAS PubMed Central Google Scholar
Villar, M. E. et al. Differential coding of absolute and relative aversive value in the Drosophila brain. Curr. Biol. 32, 4576–4592.e4575 (2022).
Article CAS Google Scholar
Nagashima, T. et al. Parabrachial-to-parasubthalamic nucleus pathway mediates fear-induced suppression of feeding in male mice. Nat. Commun. 13, 7913 (2022).
Article CAS PubMed Central Google Scholar
Christensen, S. L. et al. Von Frey testing revisited: Provision of an online algorithm for improved accuracy of 50% thresholds. Eur. J. Pain. 24, 783–790 (2020).
Article Google Scholar
Watabe, A. M. et al. SAD-B kinase regulates pre-synaptic vesicular dynamics at hippocampal Schaffer collateral synapses and affects contextual fear memory. J. Neurochem 136, 36–47 (2016).
Article CAS Google Scholar
Higashino, I. Aversive experiences induce valence plasticity of instructive signals to change future learning rules in mice. Zenodo, https://doi.org/10.5281/zenodo.15598908 (2025).

Download references

Acknowledgements

We thank all the members of the Watabe laboratory for their helpful discussions and assistance. We are particularly indebted to Kazuko Shibahara, Aimi Yuasa, Mariko Komatsu, Kana Morikyu, and Emiko Suzuki for their technical support. We also thank Dr. Toshihisa Ohtsuka and Shun Hamada for providing AAVDJ-Syn-eYFP, AAVDJ-Syn-eGFP, and AAVDJ-Syn-mCherry. This work was supported by Japan Agency for Medical Research and Development (AMED) Brain Mapping by Integrated Neurotechnologies for Disease Studies (Brain/MINDS) (JP19dm0207081 to A.M.W.), AMED Brain/MINDS 2.0 (JP24wm0625208 to A.M.W.), JSPS Grant-in-Aid for Scientific Research (B) (JP19H04062 and JP22H03542 to A.M.W.), Challenging Research (Exploratory) (JP21K18564 and JP24K21504 to A.M.W.), Scientific Research (C) (JP23K06005 to M.N.), Early-Career Scientists (JP21K16374 and JP24K19295 to T.N.; JP22K15231 to S.T.; JP23K16994 to Y.Y.), Research Activity Start-up (JP20K22694 to S.T.), JST (Moonshot R and D) (JPMJMS2024 to A.M.W. and H.N.), and Cooperative Study Program of Exploratory Research Center on Life and Living Systems (ExCELLS; program number 19-102 to H.N.).

Author information

Authors and Affiliations

Institute of Clinical Medicine and Research, Research Center for Medical Sciences, The Jikei University School of Medicine, Chiba, Japan
Suguru Tohyama, Takashi Nagashima, Fumiko Arima-Yoshida, Kanae Hiyoshi, Masashi Nagase & Ayako M. Watabe
Laboratory for Data-driven Biology, Nagoya University Graduate School of Medicine, Nagoya, Aichi, Japan
Iori Higashino, Yuichiro Yada & Naoki Honda
Laboratory for Data-driven Biology, Graduate School of Integrated Sciences for Life, Hiroshima University, Higashihiroshima, Hiroshima, Japan
Iori Higashino & Naoki Honda
Center for One Medicine Innovative Translational Research (COMIT), Nagoya University, Nagoya, Aichi, Japan
Naoki Honda
Theoretical Biology Research Group, Exploratory Research Center on Life and Living Systems (ExCELLS), National Institutes of Natural Sciences, Okazaki, Aichi, Japan
Naoki Honda

Authors

Suguru Tohyama
View author publications
Search author on:PubMed Google Scholar
Takashi Nagashima
View author publications
Search author on:PubMed Google Scholar
Iori Higashino
View author publications
Search author on:PubMed Google Scholar
Fumiko Arima-Yoshida
View author publications
Search author on:PubMed Google Scholar
Kanae Hiyoshi
View author publications
Search author on:PubMed Google Scholar
Masashi Nagase
View author publications
Search author on:PubMed Google Scholar
Yuichiro Yada
View author publications
Search author on:PubMed Google Scholar
Naoki Honda
View author publications
Search author on:PubMed Google Scholar
Ayako M. Watabe
View author publications
Search author on:PubMed Google Scholar

Contributions

A.M.W. conceived and supervised the research project. S.T. and A.M.W. designed the experiments. S.T. and K.H. performed and analyzed the behavioral experiments. T.N., A.Y.F., and M.N. performed and analyzed the electrophysiological experiments. I.H., Y.Y., and N.H. performed mathematical analyses. S.T., T.N., and A.M.W. wrote the manuscript with feedback and revisions from all the co-authors.

Corresponding author

Correspondence to Ayako M. Watabe.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Communications Biology thanks the anonymous reviewers for their contribution to the peer review of this work.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Description of Additional Supplementary Files

Supplemental Data 1

reporting summary

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Tohyama, S., Nagashima, T., Higashino, I. et al. Aversive experiences induce valence plasticity of instructive signals to change future learning rules in mice. Commun Biol 8, 1002 (2025). https://doi.org/10.1038/s42003-025-08367-3

Download citation

Received: 17 July 2024
Accepted: 10 June 2025
Published: 04 July 2025
DOI: https://doi.org/10.1038/s42003-025-08367-3