The amygdala NT3-TrkC pathway underlies inter-individual differences in fear extinction and related synaptic plasticity

ARTICLE OPEN The amygdala NT3-TrkC pathway underlies inter-individual differences in fear extinction and related synaptic plasticity Gianluca Masella 1,2 , Francisca Silva 1,2 , Elisa Corti 1,2 , Garikoitz Azkona 3 , Maria Francisca Madeira 1,2 , Ângelo R. Tomé 1,4 , Samira G. Ferreira 1,2 , Rodrigo A. Cunha 1,5 , Carlos B. Duarte 1,4 and Mónica Santos 1,2,6 ✉ © The Author(s) 2024 Fear-related pathologies are among the most prevalent psychiatric conditions, having inappropriate learned fear and resistance to extinction as cardinal features. Exposure therapy represents a promising therapeutic approach, the efficiency of which depends on inter-individual variation in fear extinction learning, which neurobiological basis is unknown. We characterized a model of extinction learning, whereby fear-conditioned mice were categorized as extinction (EXT)-success or EXT-failure, according to their inherent ability to extinguish fear. In the lateral amygdala, GluN2A-containing NMDAR are required for LTP and stabilization of fear memories, while GluN2B-containing NMDAR are required for LTD and fear extinction. EXT-success mice showed attenuated LTP, strong LTD and higher levels of synaptic GluN2B, while EXT-failure mice showed strong LTP, no LTD and higher levels of synaptic GluN2A. Neurotrophin 3 (NT3) infusion in the lateral amygdala was sufficient to rescue extinction deficits in EXT-failure mice. Mechanistically, activation of tropomyosin receptor kinase C (TrkC) with NT3 in EXT-failure slices attenuated lateral amygdala LTP, in a GluN2B-dependent manner. Conversely, blocking endogenous NT3-TrkC signaling with TrkC-Fc chimera in EXT-success slices strengthened lateral amygdala LTP. Our data support a key role for the NT3-TrkC system in inter-individual differences in fear extinction in rodents, through modulation of amygdalar NMDAR composition and synaptic plasticity. Molecular Psychiatry; https://doi.org/10.1038/s41380-024-02412-z INTRODUCTION Fear-related disorders gather some of the most commonly diagnosed psychiatric conditions including panic disorder (PAND), several phobias and posttraumatic stress disorder (PTSD). With a high prevalence worldwide, these disorders are highly disruptive to the professional and social life of afflicted individuals and represent considerable governmental and societal costs [reviewed in [1]. Patients in this group of disorders show impaired associative learning of contextual and sensory cues related to their ‘object’of fear [2,3], resulting in inappropriate and/or excessive fear, and inability to extinguish maladaptive fear [4]. These impairments in fear extinction negatively impact on the efficiency of exposure therapy, the first line of treatment to patients with anxiety and fear-related disorders [5,6]. Indeed, inter-individual variation in fear extinction learning is predictive of exposure therapy outcome [7,8]. We are still lacking mechanism-based therapeutic approaches for those patients with impaired extinction. Pavlovian fear conditioning and extinction represent a simple, yet robust paradigm to investigate the neural substrates and molecular machinery of acquisition and extinction of learned fear. This model has been instructive to identify the wide network of brain regions that are recruited in the processing of learned fear. The amygdala is the core brain region in the orchestration of fear response, which in turn is modulated by inputs from the hippocampus, conveying information about the context, and from the medial prefrontal cortex (mPFC), providing top-down control [9–12]. Within the amygdala, two distinct, yet interconnected, populations of glutamatergic neurons - fear neurons and extinction neurons –form opposite circuits that regulate fear states [13–16]. It is known that extinction does not erase the previously acquired fear memories, or the underlying microcircuit; instead, at every conditioned stimulus (CS) presentation, the two microcircuits compete to express or suppress fear [15]. Fear learning and extinction are associated with alterations in synaptic strength through long-term potentiation (LTP) and depression (LTD) of synaptic activity [reviewed in [17], which are considered the neurophysiological basis of learning and memory [18–20]. Accordingly, LTP and LTD at amygdalar synapses are believed to underlie acquisition versus extinction of learned fear by acting during consolidation and/or reconsolidation of these memories [21,22]. Indeed, context conditioning induces a selective increase in synaptic strength of context-responding basolateral amygdala (BLA) neurons [23]. On the other hand, extinction learning was found to induce LTD at lateral amygdala (LA) synapses [22]. Interestingly, extinction of both auditory and contextual fear memories reverses the conditioning-induced potentiation of BLA neurons [23,24]. Received: 13 June 2023 Revised: 29 December 2023 Accepted: 4 January 2024 1 CNC –Center for Neuroscience and Cell Biology, University of Coimbra, Coimbra, Portugal. 2 Institute of Interdisciplinary Research, University of Coimbra (iiiUC), Coimbra, Portugal. 3 Department of Basic Psychological Processes and Their Development, School of Psychology, University of the Basque Country (UPV/EHU), San Sebastian, Spain. 4 Department of Life Sciences, University of Coimbra, Coimbra, Portugal. 5 Faculty of Medicine, University of Coimbra, Coimbra, Portugal. 6 Centre for Innovative Biomedicine and Biotechnology (CIBB), University of Coimbra, Coimbra, Portugal. ✉email: mjpsa[email protected] www.nature.com/mp Molecular Psychiatry 1234567890();,: Amygdalar glutamate receptors play a pivotal role in both LTP and LTD processes that underlie fear conditioning and extinction. In particular, GluN2Aand GluN2B-containing NMDA receptors (NMDAR) play opposite roles in fear processing and amygdalar synaptic plasticity. GluN2A is necessary for the induction of LTP in the amygdala and is required for the acquisition of conditioned fear [25]. Fear learning induces the recruitment of GluN2Acontaining NMDAR to BLA synapses, and their abundance is directly correlated to the strength of the fear memory [26]. Moreover, GluN2A-containing NMDAR promote the restabilization of fear memories in reconsolidation [27]. Conversely, GluN2Bcontaining NMDAR are responsible for amygdalar LTD and are required for extinction learning [25], as well as for the destabilization of fear memories made labile by retrieval [27,28]. Accumulating evidence suggests a role for neurotrophins in the regulation of fear [29–32]. In particular, the neurotrophin 3 (NT3) - tropomyosin receptor kinase C (TrkC) system has been associated to anxiety disorders in human and non-human primates [33–35], and in mouse models of disease [29,36,37]. In addition, our recent studies underscored the importance of TrkC signaling in the formation of fear memories in normal physiological conditions [32]. Neurotrophins have a critical role in synaptic plasticity [38–40] prompting them as candidates to mediate the cognitiveemotional regulation of fear [41]. In this study, taking advantage of intrinsic inter-individual variation in fear extinction performance, we have characterized a model to investigate the substrates that support differences between mice that successfully extinguish fear and those that fail. Using this model, we unveiled a role of amygdalar NT3-TrkC system in the regulation of fear extinction. Activation of the NT3TrkC system results in attenuation of learning-induced LTP, in a GluN2B-dependent mechanism, which seems paramount for successful fear extinction. RESULTS Inter-individual variation in context fear extinction Young adult C57BL/6J male mice (n=16) were trained in the contextual fear conditioning (CFC) and extinction (EXT) paradigm (Fig. 1A). A control group that did not receive any shock was also included (CTRL-no shock, n=9). With consecutive shock administrations, conditioned mice showed a progressive increase in the percentage of time spent freezing (Supplementary Fig. S1A) and, when tested 24 h later for fear retrieval showed a proper conditioned fear response (Supplementary Fig. S1B). After fear retrieval, mice were immediately trained in fear extinction acquisition. Here, extinction learning performance was defined for each individual as the ratio (expressed as percentage) between freezing levels in the last trial of extinction acquisition (E6) and those shown in fear memory retrieval/first trial of extinction acquisition (R/E1). Extinction learning performance was used to evaluate extinction and categorize mice as EXT-success (>30% reduction in freezing levels at E6 relative to R/E1) or EXT-failure (<30% or no reduction in freezing). Throughout extinction session trials (R/E1 to E6), EXT-success animals showed a reduction in freezing to levels comparable to those of CTRL-no shock animals, while EXT-failure mice increased their freezing levels (Fig. 1B, extinction trial x group interaction F (10, 110) =2.862, p=0.003; R/E1 vs. E6, EXT-success t=8.675, p=0.0048; EXT-failure t=3.215, p=0.0454). In particular, in R/E1 trial, both EXT-success and EXT-failure conditioned mice froze significantly more than CTRL-no shock mice (CTRL-no shock vs. EXT-success t=4.618, p=0.012, CTRL-no shock vs. EXT-failure t=6.400, p< 0.001, EXT-success vs. EXT-failure t=0.7485, p=0.481), showing the proper formation of a contextual fear memory. By trial E6, EXT-success mice showed freezing levels comparable to those of CTRL-no shock mice, and significantly lower than EXT-failure mice (CTRL-no shock vs. EXT-success t=1.490, p=0.177, CTRL-no shock vs. EXT-failure t=5.727, p< 0.001, EXT-success vs. EXT-failure t=2.947, p=0.039). Overall, EXT-success mice showed an average reduction in freezing levels of more than 40% relative to the levels shown in R/E1, as opposed to a 45% increase in freezing levels exhibited by EXT-failure animals (Fig. 1C, t=4.664, p< 0.001). In the extinction memory retrieval phase (EM), EXT-success animals showed statistically significant lower levels of freezing as compared to R/E1 session (Fig. 1B; extinction session x group interaction, F (2, 22) =9.566, p=0.0010; EXT-success, R/E1 vs EM t=4.537, p=0.0005), while EXT-failure animals did not show a decrease in the freezing levels from R/E1 to EM (EXT-failure, R/E1 vs EM t=0.6425, p=0.8943). Extinction memory performance (EMP) was defined as the ratio (expressed as percentage) between freezing levels in extinction retrieval and those shown in R/E1. EXT-success mice showed a better extinction memory performance as compared to EXT-failure animals when tested for extinction retrieval 24 h later. Here, EXT-success animals showed an average 45% reduction in freezing levels relative to the levels shown in R/E1, as opposed an average 9% increase in freezing exhibited by EXT-failure animals (Fig. 1D, U =0, p< 0.001). Importantly, we detected a statistically significant correlation between freezing levels in E6 and freezing levels in the extinction retrieval session (Fig. 1E; R 2 =0.5647, p< 0.001), suggesting that extinction learning by E6 is predictive of extinction memory performance. The previous correlation suggests that individual differences in extinction could be a stable trait of the animals, which could be predicted by other behavioral traits. To evaluate this hypothesis, we investigated how fear extinction and ability to learn to extinguish fear memories could be correlated with individual differences in trait anxiety. Basal anxiety-like behavior and exploratory activity were assessed in the open field (OF) and elevated plus maze (EPM) tests, before performance in the CFC and EXT paradigm. In the OF, no differences were observed among EXT-success, EXT-failure and CTRL-no shock mice in the total distance traveled nor in the percentage of time spent or distance traveled in the center of the arena (Supplementary Fig. S1C–E). Also, no differences were found among the different groups in the EPM in the total distance traveled, percentage of open arms time or percentage of open arms distance (Supplementary Fig. S1F–H). We found however that basal anxiety levels are predictive of the strength of the fear memory, but not of learning performance itself. Indeed, a statistically significant negative correlation was found between the percentage of time spent in the open arms (OAT), as well as the distance traveled in the open arms (OAD), and freezing levels in R/E1 and E6 trials of extinction acquisition (Fig. 1F, OAT vs. R/E1, R 2 =0.2916, p=0.030; OAT vs. E6, R 2 =0.2809, p=0.035; OAD vs. R/E1, R 2 =0.2809, p=0.035; OAD vs. E6, R 2 =0.2025, p=0.077), but not with extinction learning performance and extinction memory performance (Fig. 1F, OAT vs. ELP, R 2 =0, 0064, p=0.757; OAT vs. EMP, R 2 =0.0225, p=0.590; OAD vs. ELP, R 2 =0, 0009, p=0.913; OAT vs. EMP, R 2 =0.0121, p=0.679). Successful extinction learning is associated with weak LTP and strong LTD at LA synapses. Synaptic plasticity in the LA represents a cellular correlate of fear conditioning and extinction [24,42]. In particular, fear conditioning has been associated with LTP at LA synapses [43], while fear extinction is associated with LTD in the same region [22]. Here, we investigated whether differences in fear extinction learning, as those shown by EXT-success and EXTfailure mice, are supported by differences in LA synaptic plasticity. Fear-conditioned mice were trained in the EXT paradigm and categorized as EXT-success or EXT-failure, according to their extinction learning performance (Supplementary Fig. S2A, B). Horizontal brain slices including the amygdala were obtained from EXT-success (n=6 slices from 6 mice) and EXT-failure (n=5 slices G. Masella et al. 2 Molecular Psychiatry from 5 mice) mice at extinction consolidation window (two hours after E6 trial), and electrical stimulation and electrophysiological recordings were performed in the LA (Fig. 2A). No differences were observed in the input/output (I/O) profiles between EXT-success and EXT-failure slices (Supplementary Fig. S2C), demonstrating that the two groups showed a similar LA basal excitability. High-frequency stimulation (HFS) protocol (three trains of 100 Hz pulses, 1 s duration, 5 s intervals) successfully induced LTP in LA synapses of both EXT-success and EXT-failure slices, as shown by an increase in population spikes (PS) amplitude both in the first 10 min post-HFS and in the last 10 min of recordings, as compared to baseline (Fig. 2C; EXTsuccess, baseline vs. min 1–10, t=2.948 p=0.0146, baseline vs. min 36–45, t=2.513 p=0.0307; EXT-failure, baseline vs. min 1–10, Fig. 1 Inter-individual variation in context fear extinction. A Schematic representation of the Pavlovian contextual fear conditioning and extinction paradigm to which young adult (8 to 12 weeks old) C56BL/6J male mice were submitted throughout this study. BQuantification of the percentage of time spent freezing during fear extinction acquisition and extinction memory retrieval. Fear-conditioned mice were categorized as EXT-success (n=5) or EXT-failure (n=11) according to their extinction learning performance. A CTRL-no shock group was included that received no shocks (n=9). Repeated measures two-way ANOVA with either Sidak or Tukey multiple comparisons test. * CTRL-no shock vs. EXT-failure; £ CTRL-no shock vs. EXT-success; # EXT-success vs. EXT-failure; * EXT-success, R/E1 vs. E6 and R/E1 vs. EM; * EXT-failure, R/ E1 vs. E6. CExtinction learning performance, expressed as the percentage of freezing levels in E6 relative to freezing in R/E1. The dotted line marks the threshold of 30% reduction from R/E1 used to categorize mice as EXT-success or EXT-failure. *** two-tailed Student’sttest. DExtinction memory performance, expressed as the percentage of freezing levels in extinction retrieval relative to freezing in R/E1. *** Mann–Whitney U test. ECorrelation of the freezing levels during the last trial of extinction acquisition (E6) with the freezing levels in extinction retrieval. Pearson r. FCorrelation matrix of the total distance traveled, distance in open arms and time spent in the open arms in the EPM test with the percentage of time spent freezing during extinction acquisition trials and with ELP and EMP; statistics using Pearson r. CS conditioned stimulus, E1 to E6 extinction trials, EM extinction memory retrieval, ELP extinction learning performance, EMP extinction memory performance, EPM elevated plus maze, R fear retrieval, US unconditioned stimulus.; * ,£,# p≤0.05, **p≤0.01, ***p≤0.001. G. Masella et al. 3 Molecular Psychiatry t=3.141 p=0.0138, baseline vs. min 36–45, t=2.313 p=0.0495). However, slices from EXT-success mice showed a statistically significant weaker potentiation than those from EXT-failure mice in the first 10 min after HFS (Fig. 2D, t=2.383, p=0.0410). Differences in LTP were lost in the last 10 min of recordings (Fig. 2E, t=1.532, p=0.1592), suggesting that alterations in the LTP induction phase underlie differences in EXT-success and EXTfailure behavioral performance. Low-frequency stimulation (LFS, 900 stimuli at 1 Hz) successfully induced LTD in EXT-success slices, causing a decrease in PS amplitude both in the first 10 min post-LFS and in the last 10 min of recordings, as compared to baseline (Fig. 2G; EXT-success, baseline vs. min 1–10, t=7.391 p< 0.001, baseline vs. min 36–45 t=4.611 p=0.001). The same protocol did not induce LTD in EXTfailure slices (Fig. 2G; EXT-failure, baseline vs. min 1–10, t=2.764 p=0.0506, baseline vs. min 36–45 t=0.6347 p=0.6347). Brain slices from EXT-success mice showed lower PS amplitude than those from EXT-failure mice in the first 10 min post-LFS (Fig. 2H, t=2.614, p=0.0347; EXT-success n=6 slices from 5 mice, EXT-failure n=3 slices from 3 mice), which was maintained in the last 10 min of recordings (Fig. 2I, U =0, p=0.0238). Fig. 2 EXT-success mice show attenuated LTP and increased LTD in LA synapses. A Schematic representation of experimental conditions and localization of stimulating and recording electrodes in horizontal slices containing the LA. B–ELTP was induced ex-vivo in the LA of EXTsuccess (n=6) and EXT-failure (n=5) brain slices with a HFS protocol (three 1 s duration trains of 100 Hz pulses, with 5 s inter-train interval; vertical dashed line). BRepresentative traces for EXT-success and EXT-failure slices at baseline ( __ ), upon 10 min ( … ) and upon 45 min (---) after HFS. CTime course of LTP recorded for 45 min following LTP induction. PS amplitude was averaged for the (D)first 10 min and (E) last 10 min of recordings following LTP induction. F–ILTD was induced ex-vivo in the LA of EXT-success (n=6) and EXT-failure (n=3) brain slices with a low-frequency stimulation protocol (900 stimuli at 1 Hz; between the two vertical lines). FRepresentative traces for EXT-success and EXTfailure slices at baseline ( __ ), upon 10 min ( … ) and 45 min (---) of LTD induction with the LFS train. GTime course of LTD recorded for 45 min following induction. PS amplitude was averaged for the (H)first 10 min and (I) last 10 min of recordings. D,E,H,ITwo-tailed Student’sttest and Mann–Whitney U test. BLA basolateral amygdala, Ce central amygdala, E6 extinction trial 6, HFS high-frequency stimulation, LA lateral amygdala, LFS low-frequency stimulation, PS population spikes. *p≤0.05. G. Masella et al. 4 Molecular Psychiatry Taken together with the behavioral data, these results show that individuals with naturally occurring differences in fear extinction have different synaptic plasticity properties in the LA. Amygdalar GluN2Aand GluN2B-containing NMDA receptors underlie group differences in fear extinction. At the molecular level, amygdalar GluN2Aand GluN2B-containing NMDAR play separate roles in fear conditioning and extinction, and in the underlying synaptic plasticity events. In particular, GluN2A has been associated to fear learning and LTP, while GluN2B has been associated to fear extinction and LTD [25,44]. Moreover, while GluN2A/B are related with the induction phase of LTP and LTD, AMPA receptors are linked to the maintenance of basal and potentiated synaptic transmission [45,46]. We investigated whether changes in the surface density of glutamate receptors stand on the basis of behavioral and cellular differences between EXT-success and EXT-failure animals. To address this question, fear-conditioned mice were trained in the EXT paradigm and categorized as EXT-success or EXT-failure, according to their extinction learning performance (Supplementary Fig. S3A, B). Two hours after E6 trial, at extinction consolidation window, amygdalae synaptoneurosomes were isolated from EXT-success and EXTfailure mice (Fig. 3A) and stained for surface GluN2Aand GluN2Bcontaining NMDAR (Fig. 3B, E) and GluA1and GluA2-containing AMPA receptors (AMPAR) (Supplementary Fig. S3C, F). We observed that, as compared to EXT-success mice, synaptoneurosomes isolated from the amygdalae of EXT-failure animals showed a higher intensity of GluN2A signal (Fig. 3C, U =180587, p=0.0482; synaptoneurosomes: EXT-success n=631, EXT-failure n=612) and a higher percentage of GluN2A-positive synaptoneurosomes (Fig. 3D, t=3.312, p=0.0162; synaptoneurosomal preparations: EXT-success n=4, EXT-failure n=4). In turn, the intensity of GluN2B signal was higher in EXT-success than in EXTfailure amygdalae synaptoneurosomes (Fig. 3F, U =113552, Fig. 3 Differential surface density of GluN2A and GluN2B subunits of NMDA receptors in amygdala synaptoneurosomes from EXTsuccess and EXT-failure mice. A Schematic representation of experimental conditions and biologic material used. B,ERepresentative images of synaptoneurosomes isolated from the amygdalae of EXT-success and EXT-failure mice, live stained for GluN2A and GluN2B subunits of NMDA receptors. Synaptoneurosomes were identified with co-staining against the postsynaptic marker PSD95 and the presynaptic marker VGlut1 and inspection of intact membranes by phase contrast. C,FIntegrated density of GluN2A and GluN2B signal was quantified in synaptoneurosomes isolated from the amygdalae of EXT success (GluN2A n=631, GluN2B n=529) and EXT-failure (GluN2A n=612; GluN2B n=518) mice. D,GThe percentage of GluN2Aand GluN2B-positive amygdala synaptoneurosomes was calculated for EXT-success and EXTfailure animals (n=4 independent experiments). C,FMann–Whitney U test, (D,G) two-tailed Student’sttest. E6 extinction trial 6, GluN2A subunit 2A of NMDA receptor, GluN2B subunit 2B of NMDA receptor, PSD95 postsynaptic density 95, VGluT1 vesicular glutamate transporter 1. Scale bar 0.5 µm. *p≤0.05, ***p≤0.001. G. Masella et al. 5 Molecular Psychiatry p< 0.001; synaptoneurosomes: EXT-success n=529, EXT-failure n=518). No differences were observed in the percentage of GluN2B-positive synaptoneurosomes isolated from the amygdalae of EXT-success and EXT-failure animals (Fig. 3G, t=1.268, p=0.2607; synaptoneurosomal preparations: EXT-success n=3, EXT-failure n=4). For AMPA receptors, we did not observe any difference in the percentage of GluA1and GluA2-positive synaptoneurosomes, nor in the intensity of GluA1 and GluA2 signals, between EXT-success and EXT-failure mice (Supplementary Fig. S3C–H). Overall, interindividual differences in fear extinction learning evidenced using a within-session contextual fear extinction model are corroborated by cellular and molecular substrates of learning, and add to the pool of tools already available to investigate the neural and molecular mechanisms of extinction. Amygdalar TrkC activation correlates with successful fear extinction. In previous studies, we have shown the involvement of NT3-TrkC pathway in the regulation of fear memory formation and extinction in both pathological and normal physiological conditions [29,32,37]. Taking advantage of this previous knowledge, we investigated the role of the NT3-TrkC pathway in our fear extinction model. Fear-conditioned mice were trained in the EXT paradigm and categorized as EXT-success or EXT-failure, according to their fear extinction learning performance (Supplementary Fig. S4A, B). Two hours after E6 trial, within the extinction consolidation window, mice were killed and the amygdalae, hippocampi, and PFC brain regions were dissected (CTRL-no shock n=7; EXT-success n=7; EXT-failure n=14). Western blots of total protein extracts were performed to measure TrkC activation and density (Fig. 4A, B, Supplementary Fig. S4). In the amygdala, we observed an increase in relative TrkC activation, as measured by pTrkC/full-length TrkC ratio, in EXTsuccess mice as compared to CTRL-no shock and EXT-failure mice (Fig. 4C; F (2, 25) =5.121, p=0.0137; CTRL-no shock vs. EXT-success q=4.101, p=0.0202, CTRL-no shock vs. EXT-failure q=0.8053, p=0.8374, EXT-success vs. EXT-failure q=3.931, p=0.0266). Although not statistically significant, the same trend was observed in the total levels of phosphorylated TrkC (Fig. 4D; F (2, 25) =3.685, p=0.040; CTRL-no shock vs. EXT-success q=3.350, p=0.065, CTRL-no shock vs. EXT-failure q=0.4027, p=0.956, EXT-success vs. EXT-failure q=3.466, p=0.054). No differences were observed in the total levels of full-length TrkC (Fig. 4E; F (2, 25) =1.917, p=0.1680). In the PFC, we observed a decrease in pTrkC/full-length TrkC ratio in EXT-success mice as compared with EXT-failure Fig. 4 Amygdalar TrkC activation is associated with a successful fear extinction performance. A Schematic representation of experimental conditions and dissected brain region. BRepresentative images of western blots for phosphorylated TrkC and total TrkC performed in brain extracts from the amygdala. The panel shows non-contiguous lanes from the same membrane. Quantification of (C) relative TrkC activation as measured by pTrkC/full-length TrkC ratio, (D) total pTrkC levels, (E) total full-length TrkC levels, (F) full-length TrkC/truncated TrkC ratio and (G) truncated TrkC levels. CTRL-no shock (n=7), EXT-success (n=7) and EXT-failure (n=14) mice. β-actin was used as a loading control. C–GOneway ANOVA with Tukey’s multiple comparisons test. CTRL control group, E6 extinction trial 6, pTrkC phosphorylated TrkC, TrkC tropomyosin receptor kinase C. *p≤0.05. G. Masella et al. 6 Molecular Psychiatry (Supplementary Fig. S4D). Differences were not due to decreased levels of pTrkC, as no changes were observed in any of the groups (Supplementary Fig. S4E), but to differences in full-length TrkC density, increased in EXT-success as compared with EXT-failure mice (Supplementary Fig. S4F). In the hippocampus, we did not observe any difference among groups in TrkC activation as measured by pTrkC/full-length TrkC ratio (Supplementary Fig. S4J), total pTrkC levels (Supplementary Fig. S4K) or total TrkC levels (Supplementary Fig. S4L). The truncated isoform of TrkC lacking the phosphorylation domain can act as a dominant negative, dimerizing with fulllength TrkC and preventing its activation [47]. Overall, no differences were observed in the full-length TrkC/truncated TrkC ratio, nor in the density of truncated TrkC in the amygdala (Supplementary Fig. S4F, G), PFC (Supplementary Fig. S4G, H) or hippocampus (Supplementary Fig. S4M, N). Amygdala NT3-TrkC signaling rescues fear extinction deficits. Next, we tested whether there is a causal link between TrkC activation in the amygdala and fear extinction performance. To this end, mice were bilaterally implanted with cannulas positioned above the BLA and trained in the CFC and EXT paradigm. At the extinction consolidation window, EXT-failure mice were infused with NT3 and tested for extinction memory (Fig. 5E). To confirm that NT3 infusion at a dose of 1 µg/µL selectively activates TrkC, NT3 was infused unilaterally (contralateral side was sham-manipulated) in a first batch of unconditioned animals (n=4) and amygdalae were collected 15 min later. Western blot analysis showed an increase in the levels of pTrkC in the NT3infused amygdalae as compared to the contralateral not-infused amygdalae (Fig. 5A, B, t=6.486, p< 0.001). Considering that NT3 can also bind to TrkB receptors, even though with low affinity [48], we measured TrkB activation levels. We did not observe Fig. 5 NT3 infusion in the BLA rescues fear extinction deficits in EXT-failure mice. Representative images of western blots for (A) pTrkC and (C) pTrkB performed using protein extracts from NT3-infused amygdalae versus contralateral not-infused amygdalae (n=4 per condition). Quantification of total (B) pTrkC and (D) pTrkB levels. β-actin was used as a loading control. Two-tailed Student’sttest. ESchematic representation of experimental conditions and treatments. FSchematic representation of cannulas placement in the BLA. Circles represent the tip of the internal cannula. Mice with misplaced cannulas were excluded from the analysis (EXT-success n=2, EXT-failure n=1). GQuantification of the percentage of time spent freezing during extinction acquisition session. Fear-conditioned mice were categorized as EXT-success (n=6) or EXT-failure (n=12), according to their extinction learning performance. Repeated measures two-way ANOVA with Tukey multiple comparisons test. * EXT-success, R/E1 vs. E6. HExtinction learning performance, expressed as the percentage of freezing levels in E6 relative to freezing in R/E1. Mann–Whitney U test. IExtinction memory performance, expressed as the percentage of freezing levels in extinction memory retrieval relative to freezing in R/E1. The dotted line marks the threshold of 30% reduction from R/E1 used to categorize mice as EXT-success or EXT-failure. One-way ANOVA with Tukey multiple comparisons test. E1-E6, extinction trial 1 to 6; ELP, extinction learning performance; EMP, extinction memory performance; NT3, neurotrophin 3; pTrkC, phosphorylated tropomyosin receptor kinase C; pTrkB, phosphorylated tropomyosin receptor kinase B; R/E1, fear retrieval/extinction trial 1. *p≤0.05, **p≤0.01, ***p≤0.001. G. Masella et al. 7 Molecular Psychiatry alterations in the levels of pTrkB in the NT3-infused versus contralateral not-infused amygdalae (Fig. 5C, D, t=0. 07351, p=0.9887). Taken together, data confirm that at a dose of 1 µg/µL NT3 infusion in the BLA selectively activates TrkC receptor. Next, a second batch of cannula-implanted mice was fear conditioned and trained in the EXT paradigm, and categorized as EXT-success or EXT-failure, according to their extinction learning performance. Throughout extinction acquisition trials, EXT-success, but not EXT-failure mice, showed a reduction in their freezing levels (Fig. 5G, extinction trial x group interaction F (5, 80) =7021, p< 0.001, R/E1 vs. E6, EXT-success t=4.185, p< 0.001; EXT-failure t=2.052, p=0.1989). In particular, in R/E1 trial both EXT-success and EXT-failure mice showed similar robust levels of freezing (EXTsuccess vs. EXT-failure, t=0.7927, p=0.9696), showing the proper formation of a contextual fear memory. By E6, EXT-success mice showed an average 44% reduction in freezing levels relative to the levels shown in R/E1, as opposed to average 35% increase in freezing levels by EXT-failure mice (Fig. 5H, U =0, p< 0.001). Two hours after E6, coinciding with the extinction consolidation window, a subset of EXT-failure animals was bilaterally infused with NT3 into the BLA. EXT-failure not-infused and EXT-success animals were sham manipulated. Importantly, when tested for extinction retrieval 24 h later, EXT-failure NT3-infused animals showed an improved extinction memory performance (60% reduction in freezing levels) as compared to EXT-failure sham animals (1% reduction in freezing levels), and comparable to that of EXT-success animals (67% reduction in freezing levels) (Fig. 5I, F (2, 15) =12.47, p< 0.001; EXT-success vs. EXT-failure q=6.404, p=0.001, EXT-success vs. EXT-failure-NT3 q=0.6208, p=0.900, EXT-failure vs. EXT-failure-NT3 q=5.783, p=0.003; n=6 per group). This set of experiments shows that activation of the NT3-TrkC pathway in the BLA during extinction consolidation is sufficient to rescue fear extinction deficits. NT3-TrkC signaling modulates LTP strength in a GluN2B-dependent mechanism. Finally, we sought to investigate the cellular and molecular mechanisms underlying amygdalar NT3-induced rescue of fear extinction deficits. Here, fear conditioned mice trained in the EXT paradigm were categorized as EXT-success or EXT-failure, according to extinction learning performance (Supplementary Fig. S5A, B). After E6, at extinction consolidation window, mice were killed to prepare horizontal brain slices including the amygdala and ex vivo electrophysiological recordings were performed (Fig. 6A). First, recapitulating in vivo experiments, we treated EXT-failure slices with NT3 to test whether neurotrophin could also decrease LA LTP (Fig. 6B–E). In the first 10 min following HFS, EXT-failure slices treated with NT3 showed lower potentiation than untreated EXT-failure slices (Fig. 6D; EXT-failure vs. EXT-failure-NT3 q=4.101, p=0.0447; EXT-failure n=6 slices from 6 mice; EXT-failure-NT3 n=6 slices from 3 mice), an effect that was maintained in the last 10 min of recordings (Fig. 6E; NT3 effect F (1, 17) =18.59, p=0.0005). We hypothesized that the effects of NT3 on LA LTP could be mediated by GluN2B-containing NMDAR. LTP recordings were performed in the presence of NT3 and ifenprodil, an antagonist of GluN2B-containing NMDAR. We observed that in the first 10 min following HFS, ifenprodil administration prevented the effects of NT3 on LTP (Fig. 6D; NT3 x ifenprodil interaction F (1, 17) =5.171, p=0.0362; EXT-failure vs. EXT-failure-NT3+ifenprodil, q=0.1917, p=0.9991, EXT-failure-NT3 vs. EXT-failure-NT3+ifenprodil, q=4.101, p=0.0447; EXT-failure-NT3+ifenprodil n=5 slices from 4 mice). However, in the last 10 min of recordings ifenprodil was no longer able to prevent the effects of NT3 (Fig. 6E; NT3 x ifenprodil interaction F (1, 17) =0.04139, p=0.8412), fitting with the attributed role of NMDAR in the induction phase of LTP [46]. Despite blocking the effects of NT3 on LTP induction, when administered alone ifenprodil did not show an effect neither in the first 10 min after HFS (Fig. 6D; EXT-failure vs. EXT-failure-ifenprodil, q=0.5621, p=0.9780; EXT-failure-ifenprodil n=4 slices from 3 mice) nor in the last 10 min of recordings (Fig. 6E; ifenprodil effect F (1, 17) =0.4615, p=0.5061), demonstrating that GluN2Bcontaining NMDAR inhibition selectively affects NT3 effects on LA LTP. Finally, we treated EXT-success slices with TrkC-Fc chimera to scavenge endogenous NT3 and in this way block NT3 signaling (Fig. 6F–I). In the first 10 min following HFS, EXT-success slices treated with TrkC-Fc showed a stronger potentiation than EXTsuccess untreated slices (Fig. 6H; t=3.084, p=0.0150; EXTsuccess n=6 slices from 5 mice, EXT-success-TrkC-Fc n=4 slices from 2 mice). However, the difference in LTP was lost in the last 10 min of recordings (Fig. 6I; t=1.884, p=0.0964). The observed effects were specific to LTP since none of the drugs affected the basal synaptic transmission, as no differences were observed in the I/O curves before and after treatments with NT3, NT3+ifenprodil or TrkC-Fc (Supplementary Fig. S5A–C). Taken together, our results show that NT3 leads to an attenuation of LA LTP induction in EXT-failure slices, through a GluN2B-dependent mechanism. Conversely, blocking endogenous NT3 signaling with TrkC-Fc in EXT-success slices potentiated LTP, demonstrating that endogenous NT3 is necessary for LA LTP weakening and thereby fear extinction. Inter-individual variation in fear extinction relies on a win or lose competition between the fear and extinction microcircuits [49]. Our current findings support a theoretical model where NT3-TrkC system is a key player in mediating this balance (Fig. 7). DISCUSSION Inter-individual differences in the ability to extinguish fear have a dual outcome: first on setting the vulnerability to develop anxiety and fear-related disorders, and second on determining the effectiveness of exposure therapy towards patients in this group of disorders. Indeed, fear extinction mechanisms that support exposure therapy principles [8] are often impaired in patients with fear-related disorders [50]. Here, taking advantage of the individual variation in fear extinction performance, we characterized (at the behavioral, cellular, and molecular level) a model in rodents to study intrinsic group differences in contextual fear extinction. Moreover, we found that NT3-TrkC signaling in the amygdala is sufficient to rescue fear extinction deficits and the underlying synaptic plasticity, in a GluN2B-dependent mechanism. The behavioral model herein proposed represents a powerful tool to study group differences in fear extinction. Using this model, 40% of fear-conditioned mice were able to extinguish fear (EXT-success) when trained in a contextual extinction paradigm, as opposed to EXT-failure who failed to show a reduction, and in some instances even showed an increase, in freezing levels. Both EXT-success and EXT-failure mice acquire and form equally strong fear memories, suggesting that the observed behavioral differences are specific to the extinction process. Importantly, behavioral differences in EXT-success and EXT-failure mice were maintained 24 h after extinction learning, when mice were tested for extinction retrieval, highlighting the predictive value of learning performance to extinction outcome. Group differences in fear extinction have been studied before, particularly using models of stress-enhanced fear learning [51–53]. In these models, a prior ‘traumatic’event results in the enhancement of fear learning and deficits in fear mitigation upon extinction [54]. A modified version of such models to study fear extinction differences in stress-resistant and stress-susceptible mice was previously reported, with the latter showing fear extinction deficits [53]. In these stress-enhanced fear learning models, the different response of the animals to a traumatic event sets the basis for the different extinction performances. Instead, in G. Masella et al. 8 Molecular Psychiatry our model, animals with equally strong fear memories show individual differences in extinction that are inherent to the animals and observed in the absence of any genetic or environmental manipulation. A model of approach/avoidance conflict task, more closely related to our model, was established to assess individual differences in response to conflicting stimuli [55]. The authors found that animals categorized a priori as ‘avoiding’,in comparison to ‘balancing’, did not extinguish fear [55,56]. Importantly, behavioral differences were supported by morphological, functional and transcriptomic signatures in the mPFC and amygdala pyramidal neurons, that predispose to maladaptive fear extinction [56]. The formation of fear memories and their extinction is dependent on synaptic plasticity events occurring at amygdalar fear and extinction microcircuits [42]. In particular, the formation Fig. 6 NT3 attenuates LTP at LA synapses in a GluN2B-dependent mechanism. A Schematic representation of experimental conditions and localization of stimulating and recording electrode in horizontal brain slices containing the LA. A HFS protocol (three 1 s duration trains of 100 Hz pulses, with 5 s inter-train interval; dashed vertical line) was used to induce ex-vivo LTP in the LA of (B–E) EXT-failure slices untreated (n=6), treated with NT3 (n=6, 50 ng/mL), treated with ifenprodil (n=4, 3.25 µg/mL), treated with NT3+ifenprodil (n=5, NT3 50 ng/mL and ifenprodil 3.25 µg/mL) and of (F–I) EXT-success slices untreated (n=6) and treated with TrkC-Fc chimera (n=4, 0.5 µg/mL). Tested drugs were added 45 min before HFS and were maintained in the medium until the end of the experiment. BRepresentative traces for EXT-failure slices untreated and treated with NT3, ifenprodil and NT3+ifenprodil at baseline ( __ ), upon 10 min ( … ) and 45 min (---) of HFS. CLTP time course recorded for 45 min following induction. Amplitude of PS was averaged for the (D)first 10 min and (E) last 10 min of recordings following LTP induction. Repeated measures two-way ANOVA with Tukey multiple comparisons test. FRepresentative traces for EXT-success slices untreated and treated with TrkC-Fc at baseline ( __ ), upon 10 min ( … ) and 45 min (---) of HFS. GLTP time course recorded for 45 min following induction. PS amplitude was averaged for the (H)first 10 min and (I) last 10 min of recordings following LTP induction. Two-tailed Student’sttest. BLA basolateral amygdala, Ce central amygdala, E6 extinction trial 6, HFS high frequency stimulation, LA lateral amygdala, NT3 neurotrophin 3, PS population spikes. TrkC-Fc tropomyosin receptor kinase C–Fc chimera. *p≤0.05. G. Masella et al. 9 Molecular Psychiatry results. EC performed the electrophysiology experiments. GA helped with the stereotaxic surgeries and in vivo drug infusions. MFM performed western blots and histology, and prepared the graphical abstract and the figure of the proposed model. AT and SF assisted with the electrophysiology recordings and discussed the data. RAC and CBD provided resources and contributed to the discussion of the results. MS performed and analyzed behavioral tests and assisted in stereotaxic surgeries and in vivo drug infusions, designed the study, supervised research and data analysis, and was a major contributor in writing the manuscript and funding acquisition. COMPETING INTERESTS The authors declare no competing interests. ETHICS APPROVAL All procedures were performed in accordance with the Directive 2010/63/EU regarding the protection of animals used for scientific purposes. The Animal Welfare Ethics Committee of the University of Coimbra controlled and approved all procedures (ORBEA_209_2018/08102018). ADDITIONAL INFORMATION Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41380-024-02412-z. Correspondence and requests for materials should be addressed to Mónica Santos. Reprints and permission information is available at http://www.nature.com/ reprints Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creativecommons.org/licenses/by/4.0/. © The Author(s) 2024 G. Masella et al. 16 Molecular Psychiatry