Endocannabinoids disinhibit the ventral tegmental nucleus of Gudden to dorsal premammillary nucleus pathway to enhance escape behavior following learned threat experience

Chai, Ruikai; Wang, Nawen; Nie, Jinlu; Xu, Zongyi; Zhang, Shuqian; Deng, Suixin; Wang, Rongxin; Li, Mu; Gao, Xinyi; Geng, Ruijie; Li, Haibin; Li, Lei; Wu, Hebi; Li, Zhiming; Cheng, Tian-Lin; Xu, Xiao-Hong; Shu, Yousheng; Hong, Huilin; Huang, Xiao; Wang, Weisheng

doi:10.1038/s41467-025-60080-1

Download PDF

Article
Open access
Published: 27 May 2025

Endocannabinoids disinhibit the ventral tegmental nucleus of Gudden to dorsal premammillary nucleus pathway to enhance escape behavior following learned threat experience

Ruikai Chai¹^na1,
Nawen Wang¹^na1,
Jinlu Nie¹^na1,
Zongyi Xu¹^na1,
Shuqian Zhang¹,
Suixin Deng¹,
Rongxin Wang¹,
Mu Li¹,
Xinyi Gao¹,
Ruijie Geng²,
Haibin Li ORCID: orcid.org/0000-0002-4215-1283²,
Lei Li¹,
Hebi Wu¹,
Zhiming Li¹,
Tian-Lin Cheng ORCID: orcid.org/0000-0003-0680-6710³,
Xiao-Hong Xu⁴,
Yousheng Shu ORCID: orcid.org/0000-0002-2834-2876⁵,
Huilin Hong¹,
Xiao Huang ORCID: orcid.org/0000-0002-6400-6570² &
…
Weisheng Wang ORCID: orcid.org/0000-0003-4546-3108¹

Nature Communications volume 16, Article number: 4885 (2025) Cite this article

4023 Accesses
1 Citations
5 Altmetric
Metrics details

Subjects

An Author Correction to this article was published on 18 September 2025

This article has been updated

Abstract

Innate escape behaviors, while not requiring prior learning, are shaped by an animal’s learned experiences, such as previous exposure. Here, we found that learned threat experience in mice enhances flight behaviors, which is linked to increased activation of cholecystokinin-expressing neurons in the dorsal premammillary nucleus (PMd^CCK neurons), a population that controls circa-strike escape responses. This heightened activity coincides with reduced inhibition from parvalbumin-expressing GABAergic neurons in the ventral tegmental nucleus of Gudden (VTg^PV), which typically suppress PMd^CCK activity and escape behaviors. Furthermore, threat memory prompts a prefrontal projection to stimulate the release of endocannabinoids, inhibiting the axon terminals of VTg^PV neurons. The necessity of this endocannabinoid-mediated disinhibition for the observed enhancement in flight behaviors is confirmed through genetic deletion or pharmacological blockade of endocannabinoid receptors on VTg^PV neurons. Thus, our study uncovers a neural mechanism by which experience amplifies innate escape behaviors, highlighting the crucial role of endocannabinoids.

Inhibitory neurotransmission drives endocannabinoid degradation to promote memory consolidation

Article Open access 17 December 2020

Chronic adolescent exposure to cannabis in mice leads to sex-biased changes in gene expression networks across brain regions

Article Open access 22 August 2022

Role of endocannabinoid signaling in a septohabenular pathway in the regulation of anxiety- and depressive-like behavior

Article 22 October 2020

Introduction

Escape behavior is essential for an animal’s survival¹. Due to its importance, this type of behavior is genetically hardwired in the brain without learning. Nevertheless, escape behavior undergoes dynamic adaptive changes according to the animal’s experience, to balance energy expenditure and risk¹. A particularly important aspect of such dynamic changes is that prior threat history strengthens the animal’s vigilance^2,3,4,5, leading to exaggerated flight behavior, including a lowered threshold for evading a predator, decreased flight response time, and increased escape vigor. This strengthened vigilance could reflect the animal’s recognition of increased risks in the environment due to prior encounters⁶, although the imminent threat is already out of sight. While the neural mechanism underlying this adaptive strengthening is not known, from a circuit perspective, this effect could be due to an experience-induced plastic change of the nuclei controlling the escape behavior.

The dorsal premammillary nucleus (PMd) is a central brain region that controls the escape behavior. PMd is located in the medial hypothalamus, which is the brain region that integrates threatening predator information from different sensory modalities^1,7. PMd is also the largest hypothalamic input to the dorsal periaqueductal gray (dPAG)⁸, which directly mediates the execution of escape behavior^9,10. Our previous studies demonstrated that cholecystokinin-expressing neurons in PMd (PMd^CCK neurons) exert potent control of escape vigor in multiple innate threat modalities^11,12. We hypothesized that PMd^CCK neurons could also be involved in the experience-induced strengthening of escape.

To test the hypothesis above, in this study, we established a behavior paradigm for experience-induced strengthening of escape. We found that learned experience of threat increases the activity of PMd^CCK neurons and induces exaggerated flight behavior in mice. We further dissected a tripartite synaptic mechanism underlying this effect. PMd^CCK neurons receive inputs from mPFC neurons and parvalbumin-expressing GABAergic neurons in the ventral tegmental nucleus of Gudden (VTg^PV). After repeated exposure to the threat, mPFC neurons elicited an elevated endocannabinoids (eCBs) signal on the inhibitory projections between VTg^PV neurons and PMd^CCK neurons. This eCB signal weakens the inhibition of PMd^CCK neurons, leading to strengthened escape behavior. Thus, our study revealed an mPFC/VTg to PMd circuit that controls the threat memory-driven strengthening of escape. We further demonstrated a heterosynaptic role of eCB signaling as the underlying molecular mechanism. These results have implications in understanding the neural mechanism of adaptive flexibility as well as in providing targets for modulation when such a mechanism goes awry.

Results

Learned threat experience induces exaggerated flights that depend on enhanced activation of PMd^CCK neurons

To study whether prior threat experiences modify escape, we established a paradigm using repeated exposure to the innate fear odor 2-methyl-2-thiazoline (2MT). Using a conditioned place aversion (CPA) procedure, we assessed the learning effect of this paradigm. Mice exposed to 2MT for 5 trials (referred to as the learned group) spent significantly less time in the compartment paired with 2MT (CPA score: −88.81 ± 10.43 s; p < 0.001 vs other groups). However, a single-time exposure (referred as the acute group), three-time exposure, or exposed to saline control (referred as the saline group) did not produce place aversion (Supplementary Fig. 1a, b; CPA score: acute, −16.52 ± 12.56 s; three-time, −18.27 ± 14.18 s; saline for 1 trial, 14.46 ± 11.76 s; saline for 5 trials, −7.45 ± 9.39 s). These results indicated that 2MT exposure is a negative stimulus for mice, and repeated exposure leads to measurable memory of the negative experience.

We next investigated whether this memorized negative experience could influence defensive responses to innate threats. To this end, the mice were subjected to escape-inducing assays¹² before and after the 2MT exposure paradigm (Fig. 1a). We assessed the innate flight behaviors during two different contexts: in the rat assay, mice climbed away from an awake rat for 20 min; and in the CO₂ test, mice jumped in a box filled with 15% CO₂ for 10 min. In both assays, we quantified the escape speed and number of escape attempts and found that mice in the learned group showed exaggerated flight behaviors in both metrics across two different contexts, while mice received single-time or three-time 2MT exposure or saline showed no effect (Fig. 1b, c and supplementary Fig. 1c, d). These data demonstrated a robust strengthening of mice’s innate flight responses after learned aversive experience.

**Fig. 1: Learned threat experience significantly increased the escape velocity, the number of flight behaviors, and the activity of PMd^CCK neurons.**

We recently discovered that PMd^CCK neurons play essential roles in controlling escape^11,12. Thus, we hypothesized that PMd^CCK neurons could be a control point mediating experience-driven modulation of innate flight behaviors. To directly test this hypothesis, we ablated PMd^CCK neurons with bilateral PMd injection of the AAV-DIO-taCaspase3 virus in CCK-Cre mice (Supplementary Fig. 1i–k) and tested the effect on learning-induced escape changes. Consistent with our previous reports¹², we found that the ablation of PMd^CCK neurons attenuated escape vigor and reduced the number of flights (Supplementary Fig. 1m, n), and disrupted 2MT-induced CPA behaviors (Supplementary Fig. 1l). Strikingly, PMd^CCK neuron ablation blocked learning-induced strengthening of escape after repeated exposure to 2MT (Fig. 1d, e). These results indicated that PMd^CCK neuron activity is necessary for learned experience-induced exaggeration of flight behaviors.

We next examined the effect of learning on the activity of PMd^CCK neurons. To do this, we performed fiber photometry to measure calcium (Ca²⁺) dynamics of PMd^CCK neurons by injecting the AAV-DIO-GCaMP7s virus into the PMd of CCK-Cre mice and implanted an optical fiber into the PMd (Fig. 1f). We were able to record GCaMP signals in PMd^CCK neurons during the rat and CO₂ assays. We found robust activation of PMd^CCK neurons around the onset of escape, consistent with our previous studies¹². Importantly, these escape-related activities significantly increased in learned mice, compared to mice with single-time or saline exposure (Fig. 1j, h). These data suggested that learning induces plastic strengthening of the PMd^CCK neurons that could be an underlying mechanism for the elevated flight behaviors.

PMd^CCK neurons receive afferent projections from both PFC and VTg

To dissect the circuit mechanism underlying the learning-induced strengthening of PMd^CCK neurons, we next examined the synaptic inputs to these neurons. To achieve this, we used a monosynaptic retrograde tracing strategy with recombinant rabies virus (RV)¹³ (Fig. 2a, b). We found a high percentage of retrogradely labeled neurons in the medial prefrontal cortex (mPFC) (Fig. 2c1), a region implicated in emotion regulation and cognitive control^14,15. The mPFC to PMd pathway has been noted in other studies, although the function is not known^16,17. In addition, PMd^CCK neurons received the densest innervations from the VTg (Fig. 2c2), a region located in the medial part of the pontine tegmental area, of which the function is poorly understood¹⁸. Additional retrograde and anterograde anatomical mapping revealed that PMd^CCK neurons are well-positioned as a control point for flexible defensive behaviors, as this nucleus forms connections with many brain regions related to emotional processes and defensive behaviors such as ventral subiculum (vSub)¹⁹ (Fig. 2C3), hypothalamic nucleus (AHN)^20,21, central amygdaloid nucleus (CeM)²², superior colliculus (SC)⁹, medial insular cortex (mIC)²³, the dorsal periaqueductal gray (dPAG)⁸ and more (Supplementary Fig. 2).

**Fig. 2: The whole brain input of PMd^CCK neurons in mice.**

VTg parvalbumin neurons directly inhibit PMd^CCK neurons and control escape behavior

Our anatomical mapping results showed that the VTg which mostly consists of parvalbumin neurons and has the most robust innervations to PMd^CCK neurons (Fig. 2d)²⁴. As several recent studies have demonstrated the involvement of inhibitory modulations on defensive behaviors^{25,26,27,28,29}, we hypothesized that VTg^PV neurons may directly suppress PMd^CCK neurons and escape behavior.

To test the hypothesis above, we first performed a more detailed characterization of the VTg^PV → PMd^CCK projections with PV-Cre and CCK-Cre mice. We verified our rabies tracing results of monosynaptic VTg→PMd^CCK projection by injecting retrobeads in PMd (Supplementary Fig. 3a) and by injecting Retro-AAV-Flpo virus in PMd and Flpo-dependent reporter virus in VTg (Fig. 3a). In addition, we found that the VTg→ PMd^CCK projectors are primarily parvalbumin (PV) neurons (Fig. 3b and d, 71%), which send robust projections that thoroughly innervate PMd (Fig. 3c). To verify whether PMd is the primarily downstream of VTg^PV neurons. We injected AAV-DIO-GFP into the VTg of PV-Cre mice. Compared to other regions, we found VTg^PV neurons projected more strongly to the PMd and mammillary body (Supplementary Fig. 3d and e). To examine the synaptic function of VTg^PV → PMd^CCK projections in a cell-type-specific manner, we took advantage of the exclusive expression of parvalbumin in VTg and cholecystokinin in PMd (Supplementary Fig. 3b and c). We were able to specifically express ChR2 in the VTg^PV neurons and a GFP reporter in PMd^CCK neurons, using transgenic mice that express Cre recombinase in both PV+ and CCK+ neurons. Again, we observed that the PMd^CCK neurons were densely surrounded by VTg^PV neurons’ axon terminals (Fig. 3g and h), suggesting potent synaptic connections. We evaluated the functions of these connections with whole-cell patch clamp recording of PMd^CCK neurons following optogenetic stimulation of ChR2-expressing VTg^PV axon terminals. We observed robust negative currents after laser stimulation, which persisted with broad blockage of excitatory channels by kynurenic acid, indicating the inhibitory nature of the synaptic transmission (Fig. 3i). This is further corroborated by the fact that these currents disappeared following tetrodotoxin application and recovered by co-treatment of potassium channel blocker 4-aminopyridine (Fig. 3i, j). Thus, these results strongly indicate that VTg^PV neurons directly project to PMd to inhibit CCK-expressing neurons.

**Fig. 3: The VTg nucleus sends monosynaptic inhibitory inputs to PMd^CCK neurons.**

Having established that VTg^PV neurons inhibit PMd^CCK neurons, we next examined the function of VTg^PV → PMd^CCK projections on escape behavior. To this end, we utilized the same rat assay and CO₂ assay described above, with selective manipulations of the VTg^PV → PMd^CCK pathway. We achieved chemogenetic manipulations by injecting retrograde AAV with Cre-inducible expression of Flpo (Retro-AAV-DIO-Flpo) into the PMd of PV-Cre mice and another AAV with Flpo-dependent expression of HM3Dq, HM4Di or mCherry in VTg (Supplementary Fig. 4a, d). For optogenetic manipulations, we expressed channelrhodopsin-2 (ChR2), anion channelrhodopsin-2 (GtACR2), or control fluorescent proteins in VTg^PV neurons and delivered blue light (473 nm) to axon terminals in the PMd (Supplementary Fig. 4b and g). While the excitation or inhibition of the VTg^PV → PMd^CCK projections did not affect locomotion (data not shown), we observed potent control of the escape behaviors in naïve mice in the two assays (Supplementary Fig. 4). Chemogenetic or optogenetic activation of VTg^PV → PMd^CCK pathway reduced the velocity and number of innate flight behaviors, while inhibition led to increases (Supplementary Fig. 4). These data demonstrated control of escape behaviors by the VTg^PV → PMd^CCK pathway. Thus, regulation of the VTg^PV neuron activities is sufficient in modulating the flight behaviors.

We next investigated whether changes in the VTg^PV → PMd^CCK neuron activities are involved in learning-induced flight exaggeration. We recorded GCaMP7s calcium signals of VTg^PV axons in PMd using fiber photometry in mice during the escape assays (Fig. 4a). In learned mice, we found a sharp suppression of calcium activities in the terminals of VTg^PV neurons that was temporally locked around the onset of flight behavior (Fig. 4b, c). Interestingly, the same pattern of activity was not observed in naïve mice, or mice with single exposure (Fig. 4b, c). To investigate the activity of VTg^PV neurons, the AAV-DIO-gCamp7s was injected into VTg of PV-Cre mice. The fluorescence signal at VTg was recorded using fiber photometr during strengthened escape behavior. We found that the calcium activities in VTg^PV neurons did not show any changes during escape behavior (Fig. 4e, f). In light of our above results that inhibition of VTg^PV → PMd^CCK pathway led to enhanced escape behaviors (Supplementary Fig. 4), these data suggest that threat memory could engage a suppression of VTg^PV → PMd^CCK pathway to achieve escape adaptation. We further tested the necessity of VTg^PV → PMd^CCK pathway suppression for threat memory-induced escape strengthening using the same virus strategies and escape assays (Fig. 4g, i). We found that activation of VTg^PV → PMd^CCK pathway, either with chemogenetic (Fig. 4h, i) or optogenetic (Fig. 4k, l) approaches, blocked elevated escape behaviors after repeated exposure to 2MT. Together, these data indicate that the VTg^PV-PMd^CCK projection is a direct control for flight behavior that is necessary for threat history induced escape adaptation.

**Fig. 4: The VTg^PV neuron → PMd projection mediates exaggerated escape behaviors.**

eCB system controls plasticity at VTg-PMd synapses and behavioral adaptation

Our above data indicated a suppression of inhibitory VTg^PV → PMd^CCK pathway controls escape adaptation with learned threat experience, thus we next tried to determine the neural mechanism underlying this plasticity. We first measured the effect of single or five-time 2MT exposure conditioning on VTg^PV neurons’ excitability (Supplementary Fig. 5a). Since we recorded the membrane potentials while holding the cell at about −45 to −60 mV with no spontaneous firing (already in the method). Although baseline membrane potentials differed among the various groups (Supplementary Fig. 5d), this only reflects variations in fundamental intrinsic and extrinsic electrophysiological properties. Crucially, when we applied simulated current injection to test excitability, the number of action potentials elicited showed no difference among groups (Supplementary Fig. 5c), indicating that the level of cellular excitability remained unchanged. These data suggest that learning-induced suppression of VTg^PV neurons does not affect their excitability at the somatic level, but could be due to changes at the VTg^PV → PMd^CCK synapses.

Since eCB signaling has been previously described as a heterosynaptic inhibitory signal^30,31,32 and a previous study demonstrated that the innervation of endocannabinoid receptor (CB₁R)-containing axons was found in the PMd using immunoelectron microscopy³³, we hypothesized that the endocannabinoid (eCB) system may be mediating the heterosynaptic regulation on VTg^PV → PMd^CCK synapses. And consistent with our hypothesis, data from Allen Brain Atlas show that VTg neurons express CB₁R mRNA at high levels (www.mouse.Brain-map.org, Supplementary Fig. 5g). By using immunostaining, we confirmed the CB₁R expression on VTg^PV neurons (Supplementary Fig. 5k–n).

To further test whether the eCB released at the VTg-PMd synapses was related to repeated threat exposure, we then measured the eCB signal during escape behavior in mice with single or repeated 2MT conditioning (Fig. 5a). Using a genetically encoded GPCR activation-based eCB biosensor (GRAB_eCB2.0)^{34,35,36,37,38,39}, we found that the eCB signal showed no change related to escape behavior in mice exposed with saline or single-time 2MT, but showed a marked activation around the escape onset in mice with repeated 2MT exposure (Fig. 5b, c). This pattern matched perfectly with the plastic changes in the VTg-PMd synapses we observed (Fig. 4).

**Fig. 5: Loss-of-function of CB₁R at VTg^PV → PMd synapses rescues the decreased eCB signal and impairs the exaggerated escape.**

The above data established spatial-temporal proximity of eCB signal elevation and VTg-PMd synaptic suppression, we next aimed to test whether there is a causal effect. To this end, we utilized a CB₁R antagonist AM251 to block the eCB signal, and recorded the calcium activity at the VTg-PMd synapses (Fig. 5d). Consistent with our previous results (Fig. 4), we observed learning-induced suppression of GCaMP signals around escape onset in both escape assays in the vehicle group. In contrast, in mice that received local injection of AM251, this learning-induced plasticity was significantly suppressed (Fig. 5e, f), indicating the necessity of an eCB signal. To further examine the impact of locally blocking eCB signal on escape adaptation, we delivered AM251 bilaterally at PMd (1 μg in 0.4 μl each side, Fig. 5g). We found that infusion of AM251 blocked escape strengthening from repeated 2MT exposure in both escape assays, and the escape behavior in naïve mice was not affected (Fig. 5h, i).

To suppress the eCB signal specifically in the VTg^PV-PMd pathway, we employed the CRISPR/Cas9 knockout strategy for local ablation of CB₁R at VTg^PV-PMd synapses (Supplementary Fig. 5h), by injecting the AAV-gRNAs-EGFP into VTg of PV^Cre::Cas9^fl mice. This strategy led to the effective reduction of CB₁R mRNA in the PMd region (Supplementary Fig. 5i), and we further confirmed the reduction of CB₁R expression specifically on PMd-projecting axons from VTg^PV neurons with immunostaining (Fig. 5j, k). Deletion of CB₁R in the VTg^PV neurons did not affect baseline escape behavior or negative emotions associated with 2MT exposure (Supplementary Fig. 5j). And strikingly, learning-induced exaggeration of flight behaviors in both rat and CO₂ assays was blocked in mice received gRNA, but not in the control group with scrambled sequences (Fig. 5k, l). Taken together, these data indicate that repeated threat exposure induces eCB release at VTg-PMd synapses to control behavioral adaptation.

mPFC releases eCB at VTg-PMd synapses with repeated threat exposure

Given that eCBs are synthesized in the dendritic spine on demand, and this postsynaptic generation involving the action of phospholipase C requires the activation of G-protein-coupled receptors (GPCRs)⁴⁰. From our anatomy results (Fig. 2), the mPFC and vSub are likely to provide excitatory inputs to PMd^CCK neurons, which have been implicated in behavioral regulation. To search for learning-induced plastic changes in the inputs of the PMd^CCK neurons, we measured c-fos expression after escape assays and compared the effect of single-time versus repeated exposure to 2MT (Fig. 6a). We found that the expression of c-fos induced by the rat escape assays in PMd was enhanced in learned mice, but not in mice with single-time or saline exposure (Fig. 6b). This is consistent with the elevated calcium responses that we recorded (Fig. 1), indicating a learning effect. Among the identified inputs of PMd^CCK neurons, we observed the same learning effect only in mPFC, but not in vSub (Fig. 6b and c). Considering previous reports on mPFC functions in top-down control over escape behaviors^41,42, the data strongly imply that the strengthening of mPFC→PMd pathway may participate in learned threat history-induced exaggeration of escape behavior.

**Fig. 6: The mPFC → PMd circuit is necessary for learning threat history-induced eCB signal increases.**

To directly test whether the activation of mPFC could increase eCB levels at VTg^PV-PMd synapses, we injected AAV-DIO-eCB2.0 virus into the VTg of PV-Cre mice and recorded fluorescence signal at PMd using fiber photometry. At the same time, we labeled PMd-projecting mPFC neurons with red-shifted excitatory opsin ChrimsonR or tdTomato (Fig. 6d). We used the same optic fiber to deliver 20 Hz red light (615 nm) to stimulate the axon terminals of mPFC neurons in the PMd, and simultaneously record the changes in the eCB signal at VTg^PV-PMd synapses. Optogenetic activation led to a robust increase in the eCB signal, but not in the control (Fig. 6e–g). These results support our hypothesis that activation of mPFC→PMd pathway could lead to the elevation of eCB at the VTg-PMd synapses. In addition, the level of eCB increase (Fig. 5) was comparable to that induced by optogenetic activation of mPFC. To clarify whether the mPFC activation is necessary for eCB production during exaggerated flight. We injected AAV-retro-Flpo into PMd, AAV-fDIO-HM4Di or -mCherry into mPFC and AAV-DIO-eCB2.0 into VTg of PV-Cre mice and implanted an optical fiber into the PMd. We found that inhibition of PMd projecting mPFC neurons significantly blocked the increased eCB signal in learned mice (Fig. 6h–j). These data indicate that learned threat experience could induce mPFC to release eCB at VTg-PMd synapses.

Inhibition of mPFC → PMd pathway impairs the escape adaptation

To directly test the causal role of mPFC→PMd pathway in learning-induced escape adaptation, we performed chemogenetic inhibition of the mPFC→PMd projectors by injecting a non-toxic rabies virus (RV)-Cre in the PMd and Cre-dependent mCherry or DREADD (designer receptors exclusively activated by designer drugs) viruses in mPFC (Fig. 7a). The DREADD constructs allowed us to control neuron activity by injecting Clozapine-N-oxide (CNO) in mice^43,44. In naive mice, chemogenetic activation of mPFC→PMd pathway did not affect the escape velocity and number of flights assessed in the rat assay or CO₂ assay (Fig. 7b, c, Supplementary Fig. 6a and b). To examine the involvement of mPFC→PMd pathway in escape adaptation, we employed repeated exposure to the 2MT paradigm described above and chemogenetically inhibited the mPFC→PMd pathway during the post-learning escape behavior tests (Fig. 7d, Exp I). We found that chemogenetic inhibition of PMd-projecting mPFC neurons significantly decreased the learning-induced exaggeration of escape behaviors in both assays (Fig. 7e, f and Supplementary Fig. 6), indicating a necessary role of mPFC→PMd pathway in the expression of learned threat history-induced escape adaption.

**Fig. 7: PMd^CCK-projecting mPFC neurons are necessary for learning-induced escape adaptation.**

We further examined the involvement of mPFC→PMd pathway in the acquisition of the learned threat, by additionally inhibiting mPFC→PMd projectors during the repeated exposure to 2MT (Fig. 7d, Exp II). We found that injection of CNO during 2MT exposure blocked learning-induced strengthening of flight behaviors in all mice, even in those receiving saline during the test assays (Fig. 7e, f), indicating that mPFC→PMd pathway is necessary for the learning of the threat history. Consistent with this view, we found that chemogenetic activation or inhibition of mPFC→PMd pathway bidirectionally controls the CPA score during repeated exposure to 2MT (Supplementary Fig. 6h, i). These data indicate that mPFC→PMd pathway is necessary for learned threat history-induced escape adaption, and the activation of this pathway is important during threat exposure as well as subsequent expression of escape flexibility. Altogether, these results demonstrated a tripartite mPFC/VTg^PV → PMd^CCK synapses that utilize eCB signal for heterosynaptic plasticity to control escape adaptation due to learned aversive memory (Fig. 7g).

Discussion

Prior threat experience heightened the animal’s sensitivity to future predator threats and enhanced the escape behavior. This capability is crucial for survival in dynamic contingencies fraught with danger, enabling animals to escape from threats at the earliest opportunity. We demonstrated here that PMd^CCK neurons orchestrate a tripartite circuit mediating prior learned threat-driven behavioral adaptation. In brief, our results suggest that activation of mPFC neurons promotes exaggerated escape behavior by inducing heterosynaptic, endocannabinoid-dependent depression of VTg synaptic transmission to the PMd (Fig. 7g).

The VTg is located in the medial part of the pontine tegmental area, providing a very dense input to the PMd^CCK neurons (Fig. 2) that are essential for escape behavior^11,12. The functional role of the VTg was poorly understood, despite several previous lesion studies suggesting a role in modulating vigilance states and context memory^45,46,47. Applying optogenetic control of VTg^PV neurons, we showed that this nucleus controls the escape behavior. Fiber photometry recording of the calcium activity from VTg^PV neurons indicated a learning-induced suppression, suggesting the existence of a baseline level of VTg^PV neuron activation in naïve mice. Modulating the levels of these activities provides an ideal control point to fine-tune the strength of the innate escape response. Indeed, we found that mPFC controls threat history-induced exaggerated flight behaviors by suppressing the activity of the axon terminals of VTg^PV neurons to activate the PMd^CCK neurons. Thus, our findings establish a crucial function of the VTg neurons in regulating the strength of the escape response. Whether other cell types in VTg, or other modes of modulation on VTg neuron activity, could mediate distinct forms of experience-induced flight adaptation, such as habituation²⁹ requires further investigations. Furthermore, consistent with previous reports of a top-down modulatory role of mPFC^{48,49,50,51,52,53} primarily using Pavlovian conditioning⁵⁴, we provide the evidence that mPFC modulates learning-induced innate escape behavior adaptation, and extend the understanding of the functions of mPFC projection to hypothalamic regions.

We discovered that mPFC modulates the strength of the VTg-PMd synapses via the endocannabinoids (eCBs) system. As a synaptic retrograde messenger, eCB suppresses neurotransmitter release by activating the presynaptic cannabinoid type 1 (CB₁) receptors, which has been previously described as a modulator at inhibitory synapses^30,31,32. In previous studies, the activation of eCB system is a neurobiological mechanism for attenuating anxiety⁵⁵ and learned freezing behavior^50,56,57, indicating that endocannabinoids have anxiolytic and antipsychotic effects with low-imminence threats^58,59,60. However, a recent study demonstrated that elevating the level of eCB in the amygdala strengthened innate fear-induced freezing behavior⁶¹. Similarly, pharmacological enhancement of eCB signaling increased flight behaviors elicited by predator threat^62,63. Our results indicated that increased eCB signaling exacerbated the innate flight behaviors to high-imminence threats. With the widely used genetically encoded GPCR activation-based eCB biosensor (GRAB_eCB2.0)^{34,35,36,37,38,39}, we showed that the exaggerated flight was temporally coupled with increased eCB signal at the inhibitory VTg-PMd synapses (Fig. 5). And with precise genetic deletion of CB₁R in the VTg axons, we subsequently demonstrated the necessity of the eCB signal for learning-induced escape strengthening. Together these studies suggest that the effect of eCB signaling on anxiety could potentially diverge due to different causes of the threats.

As mentioned previously, high-threat sensitivity is evolutionarily advantageous in dangerous environments for animals to quickly detect potential predators and escape from them. On the other hand, inappropriate activation of the same neural mechanism could cause panic disorder (PD) in humans⁶⁴. While the etiology of PD is not completely understood, several brain regions involved in the escape behaviors are implicated⁶⁴. Based on the anxiolytic effects of the eCBs, recent studies suggested that the eCB system could be a potential therapeutic target for treating axiety⁶⁵. However, animal studies using CB₁R knockout, particularly in the GABAergic neurons, revealed that the eCB system has anxiogenic effects⁶⁵. Similarly, pharmacological enhancement of the eCB signaling facilitated innate defensive behavior in high-threat environment^62,65. Furthermore, increased risks for panic disorder have been associated with exogenous cannabinoid use in humans^66,67. Thus, understanding the neural mechanism underlying the biphasic effect of eCB signaling on anxiety is critically important for its potential therapeutic application. Our study fills this gap of knowledge regarding the neural circuit mechanism of eCB signaling in strengthening defensive behavior. Considering that the neural circuits controlling escape overlap with those related to PD, the mechanism we discovered on eCB signaling may have implications for developing treatment for PD. Future studies are required to further dissect the neural mechanisms underlying the anxiolytic and anxiogenic effects of the eCB system under physiological escape behavior and pathological conditions.

Methods

Animals

All experiments were performed under the relevant guidelines of Fudan University and the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals and were further approved by the Animal Advisory Committee at Fudan University.

Mice

Male WT (C57BL/6), Cck-IRES-Cre (Jackson, strain name Cck^{tm1.1(cre)Zjh}/J, stock number 012706), Pvalb-IRES-Cre (Jackson, strain name B6;129P2-Pvalb^tm1(cre)Arbr/J, stock number 008069) and CAG-LSL-Cas9-tdTomato (GemPharmatech, strain name C57BL/6JGpt-Rosa26^{tm1 (CAG-LSL-Cas9-tdTomato)}/Gpt, stock number T002249) mice aged 8-14 weeks were used. Sample sizes were shown to be statistically powerful by estimating the variances of data from this study. All mice were housed under a 12 h light-dark cycle (lights on from 8:00 to 20:00) with food and water provided ad libitum. Holding and experimental room temperatures were maintained at 20–22 °C and humidity was maintained between 30% and 70%, with an average humidity of approximately 45%. All mice were randomly allocated into either the control or experimental group and were group-housed until surgery. After the surgical procedures, all experimental animals were single-housed. All behavioral tasks were performed during the light cycle.

Rats

Male SD rats (250–400 g) aged 12–16 months were obtained from GemPharmatech and were individually housed on a standard 12-h light-dark cycle and given food and water ad libitum. Rats were only used as a predatory stimulus. No attacks on mice were observed in this experiment.

Stereotaxic surgeries

Surgeries were performed as described previously¹². Eight-week-old mice were anaesthetized with 1.5–3.0% isoflurane and placed in a stereotaxic apparatus (cat. nos. 68030, 68025, RWD life science). A scalpel was used to open an incision along the midline to expose the skull. After performing a craniotomy, 40–100 nl at a titer of ~2*10¹² particles/ml virus was injected per site. Injection sites were confirmed in all animals by preparing coronal sections (40 μm) of injection sites after behavioral tests.

Viral vectors

AAV(2/9)-EF1a-DIO-eGFP, AAV(2/9)-EF1a-fDIO-mCherry, Retro-AAV(2/2)-Flpo, AAV(2/9)-EF1a-DIO-hChR2(H134R)-mCherry, AAV(2/9)-syn-fDIO-ChrimsonR-tdTomato, AAV(2/8)-hSyn-DIO-hM3D(Gq)-mCherry, AAV(2/8)-CAG-DIO-GCaMP7s-WPRE-SV40, AAV(2/8)-hSyn-DIO-hM4D(Gi)-mCherry and AAV(2/9)-hSyn-DIO-hGtACR2-eGFP were prepared by the Shanghai Taitool Bioscience Co., Ltd. AAV(2/9)-EF1a-DIO-eCB2.0 was prepared by Brain Case (Shenzhen) Biotechnology Co., Ltd. The final titer was estimated to be ~10¹² genome copies per milliliter. AAV (2/8)-CAG-DIO-TVA-eGFP (4.0 × 10¹²), AAV (2/8)-CAG-DIO-rabies glycoprotein G (4.14 × 10¹²), EnvA-G deleted rabies-DsRed and EnvA-G deleted rabies-Cre (3.0 × 10⁸) were prepared by BrainVTA (Wuhan, China) Co., Ltd.

The gRNA (gCTGGATCATGCGAACTGCG) targeting CB₁R were cloned to make AAV (2/9) - (H1-gRNA) ×3-CAG-EGFP (gRNA, or AAV-gRNAs-EGFP for short, 1.14 × 10¹³ vg/ml, WZ Bio, Inc). In addition, viruses with scrambled gRNAs (scram) targeting the same gene sites (CCGGAAGAGCGAGCTCTTCT, 1.20 × 10¹³ vg/ml, WZ Bio, Inc) were made as controls.

Viral injections and cannula implantation

Mice received unilateral 50–80 nl viral infusion using a nanoliter microinjection pump (RWD life science, R-480) and fiber optic cannula (200 mm OD, 0.22 NA, Inper Inc) implantation (0.15–0.3 mm above the viral injection coordinates). Infusion locations measured as anterior-posterior, medial-lateral, and dorsoventral coordinates from bregma were: dorsal premammillary nucleus (PMd, 80 nl, −2.46 mm; ±0.5 mm; −5.35 mm), ventral tegmental nucleus of Gudden (VTg, 50 nl, −4.9 mm; ±0.25 mm; −3.90 mm), medial prefrontal cortex (mPFC, 60 nl, 1.8 mm; ±0.5 mm; −3.20 mm). The injection glass capillary was withdrawn 10 min after the end of the infusion and the incision was closed with a suture. After implantation, all fibers were made to adhere to the skull with bone screws and dental cement. The animal was kept on a heating pad until fully recovered from anesthesia. Injection sites were confirmed in all animals by preparing coronal sections (50 or 100 mm) of injection sites after behavioral tests.

For photometry experiments, mice were injected with 80 nl of AAV-DIO-GCaMP7s or AAV-DIO-eCB2.0 in the PMd or VTg, respectively of CCK-Cre or PV-Cre mice. Mice were implanted unilaterally with cannulas in the PMd. A 400 mm diameter, 0.48 NA optical fiber (Inper Inc) was used for photometry experiments.

In vivo fiber photometry recordings

Fiber photometry was applied to record Ca²⁺ signals or eCB signal¹². Briefly, microinjections of the AAV-DIO-GCaMP7s virus were performed in the PMd and VTg, or injection of AAV-DIO-eCB2.0 virus was performed in the VTg, followed by implantation of an optical fiber, with mice allowed at least 3 weeks for recovery. A monofiber optic patch cord connected to the fiber photometry system (Inper) was attached to the implanted fiber optic cannula using a ceramic sleeve with black heat-shrinkable tubes. To record the fluorescence signals from GCaMP7s or eCB2.0, light from a 470-nm LED was bandpass-filtered (470/10 nm), collimated, reflected by dichroic mirrors, focused using a 20× objective and delivered at 25–40 μW from the tip of the fiber optic cannula. The emitted fluorescence from GCaMP7s or eCB2.0 was bandpass-filtered (525/40 nm), focused on, and detected by the sensor of a complementary metal oxide semiconductor camera. The end of the fiber was imaged at a frame rate of 30 fps using the Inper Signal software; the mean value of the region of interest (ROI) at the end face of the fiber was calculated using Inper Plot software. Furthermore, a 410-nm LED light was bandpass-filtered and delivered alternately with a 470-nm LED light as an isosbestic control channel. GCaMP7s or eCB2.0 fluorescence intensity was then recorded. Fluorescence changes were quantified by ∆F/F0 = F(t) - F0(t)/F0(t) and the baseline was defined by the signals at 5 s before stimulus presentation. The Inper Plot software was applied to generate the average Ca²⁺ traces, with the shaded areas indicating the s.e.m.

Optogenetic manipulation

All mice were handled daily for 3 days before the behavioral tasks. On the day of the experiment, mice were transported to a testing room and habituated for approximately 2 h. The chronically implanted fibers (diameter 200 μm) were connected to a laser generator via optic fiber sleeves, after which mice were returned to the home cage for at least 30 min. Next, either blue light (473 nm, ChR2: 2 mw/mm², 5-ms pulses, 20 Hz; GtACR2: continuous light, 2 mw/mm², Inper) was delivered to selectively activate or inhibit the soma and circuits under the control of an intelligent optogenetic system. After completion of all behavioral tests, mice were killed to verify the virus injection and optical fiber sites. Mice with missed injections or implantations were excluded from the study.

Chemogenetics

Mice used for chemogenetic experiments were exposed to each threat and control stimuli twice, once following treatment with saline and once following treatment with CNO (LKT LABS, Cas No. 34233-69-7, 1–5 mg/kg, injected intraperitoneally) 40 min before the experiment. Only one control or threat-exposure assay was performed per day with each mouse.

Brain slice preparation

Acute brain slice preparations were performed. Mice were deeply anesthetized with pentobarbital sodium (2% w/v, intraperitoneal), then intracardially perfused with 20 ml of ice-cold oxygenated cutting solution (93 mM N-methyl-d-glucamine, 1.2 mM NaH₂PO₄, 2.5 mM KCl, 20 mM HEPES, 30 mM NaHCO₃, 2 mM thiourea, 25 mM glucose, 3 mM sodium pyruvate, 5 mM sodium ascorbate, 10 mM MgSO₄, 0.5 CaCl₂ and 3 mM glutathione (GSH)). Mice were then quickly decapitated and their brains were carefully removed from the skull. Each brain was glued on the bed plate of a vibratome and coronal slices (250 μm) containing the PMd or VTg were sliced in an ice-cold sucrose-based slicing solution, which is the normal artificial cerebrospinal fluid (ACSF) but with NaCl replaced by equimolar sucrose and sectioned at 0.1 mm s-1 (VT1200s, Leica Biosystems). Slices were incubated in a cutting solution, and then immediately transferred into HEPES artificial cerebrospinal fluid (aCSF) (126 NaCl, 2.5 KCl, 2 MgSO₄, 2 CaCl₂, 26 NaHCO₃, 1.25 NaH₂PO₄ and 25 dextrose)) at 35.5 °C for 40–60 min and then placed at room temperature until recording. After incubation, slices were placed in a recording chamber (Warner Instruments) for electrophysiological recording with continuous perfusion of oxygenated standard aCSF, with the temperature maintained at 34 °C via an in-line solution heater (TC-344B, Warner Instruments). The aCSF pH was maintained within 7.3–7.4 and osmolarity was adjusted to 315 mOsm kg^-1. All solutions were continuously boiled with 95% O₂ and 5% CO₂ throughout slice preparation and electrophysiological recordings.

In vitro electrophysiological recordings

Whole-cell patch-clamp recordings were performed on visualized PMd and VTg neurons using an infrared differential interference contrast (IR-DIC) microscope (BX-51WI, Olympus). For whole-cell recording at VTg, the glass pipettes (5–7 MΩ) were filled with intracellular solution (in mM, 140 K-Gluconate, 3 KCl, 2 MgCl₂, 0.2 EGTA, 10 HEPES, 2 Na₂ATP, (285–295 mOsm, pH 7.2)) for current-clamp recordings. Positive current pulses (10 − 100 pA per step, 500 ms in duration) and negative current pulses (−50 pA per step, 500 ms) were injected to examine the electrophysiological properties such as F-I curve when the cell was hold at about −45 to −60 mV with no spontaneous firing.

To record light-evoked postsynaptic currents, glass pipettes had an impedance of 5-7 MΩwhen filled with high-Cl- internal solution (72 K-Gluconate, 71 KCl, 2 MgCl₂, 0.2 EGTA, 10 HEPES, 2 Na₂ATP) with osmolarity of 285-295 mOsm and pH 7.20-7.25. The blue light was delivered through a LED light source (473 nm, Fluca, FL0117) positioned above the surface of the target areas. The duration of each light pulse was 1 ms, and the intensity of continuous light output measured under the 40× objective was 10 to 45 mW. The membrane potentials of neurons were held at −70 mV in voltage clamp to record IPSCs. Both recordings were immediately terminated once the series resistance reached above 10%. To eliminate polysynaptic components, TTX (1 μM, Dalian Refine Biochemical Products) was applied to the standard aCSF to block the sodium channels. At the same time, 4-AP (200 μM, Sigma-Aldrich) was added to augment light-induced postsynaptic currents. For all the recording, we employed a Multi-clamp 700B amplifier (Molecular Devices) and a Micro3 1401 analog-to-digital converter together with Spike2 software (version 8, Cambridge Electronic Design) for data acquisition.

Behavioral Tests

Rat exposure assay

The behavioral paradigm had been described in our previous study¹². Briefly, we used a corridor measuring (35 × 20 × 22 cm). There is a vertical wall with a wire mesh at the end of the corridor (with 1 cm × 1 cm square spaces in the mesh). Mice were acclimated to this environment for three days for 10 min. Mice freely explored and independently climbed up the grid ladder. The rat is restrained to the corner opposite the mesh grid by a harness attached to a string that is taped to the wall. Mice could climb to avoid the rat and hide in the burrow. After reaching the burrow mice were given 1 min of hiding time in the burrow. They were then gently placed back on the bottom floor of the context at the same level as the rat. The climb speed and number of climbs performed during 20 min were measured. This length of 20 min includes the 1-min hiding period in the burrow. The climb speed and number of climb entries into the shelter were analyzed offline using ANY-maze v7.0 (Stoelting Co.).

Carbon dioxide exposure assay

Animals were placed in a transparent box. Mice were acclimated to the environment for three days (10 min of exposure). The next day mice were exposed either to infusions of air or of 15% CO₂. The concentration of CO₂ was measured and dynamically adjusted by the ProCO₂ 120 controller (Biospherix, USA) so that CO₂ levels were maintained at 15%. This concentration of CO₂ elicited escape jumps in all mice tested. The jump speed and number of jumps in 10 minutes were recorded. The jump speed and number of jumps were analyzed offline using ANY-maze v7.0 (Stoelting Co.).

Conditioned place aversion procedures

The CPA apparatus [62 cm (length) × 24 cm (width) × 24 cm (height)] made of Plexiglas was divided into two equal-sized compartments by a removable partition with an opening (8 × 8 cm) at one end, which allowed mice ad libitum access to each compartment. Two compartments were distinguished by visual (white or black) and tactile cues (bar or slick) and provided distinct contexts that were paired with 2MT or saline. The apparatus was equipped with infrared photobeams connected to a computer that recorded the rat’s location in the chambers.

To determine baseline place preference, the mice were initially placed in the apparatus with the doors removed for 15 min (preconditioning test). Conditioning was performed using an unbiased, counterbalanced protocol, as described in our previous work⁶⁸. Conditioning took place over the next 2 d. On the first day, the mice were confined to either compartment with saline in a counterbalanced manner for 20 min. On the second day, the mice were confined to the compartment with 2MT opposite to that on the first day for 20 min. This compartment will be referred to as the 2MT treatment-paired compartment. In the post-training testing phase (24 h after the conditioning trial), all mice were allowed to explore the entire apparatus for 10 min, and the amount of time spent in each compartment was recorded. CPA score represents the time in the drug treatment-paired compartment during the testing phase minus that during the preconditioning phase. Time spent in the compartment was analyzed offline using ANY-maze v7.0.

Fluorescent retrobead retrograde tracing

Mice were injected unilaterally with red fluorescent retrobeads (100 nl; LumaFluor Inc. #78R170) into the PMd (AP: −2.46 mm; ML: ±0.5 mm; DV: −5.35 mm) The incision was closed with a suture and a tissue adhesive and the animal was kept on a heating pad until it recovered from anesthesia.

Rabies tracing

The strategy builds on the targeting of Cre-dependent co-expression of the avian receptor TVA and the rabies glycoprotein G (RVG) on principle neurons through injection of helper viruses (AAV-CAG-DIO-TVA-eGFP and AAV-CAG-DIO-RVG) into the PMd (AP: −2.46 mm; ML: 0.5 mm; DV: −5.35 mm) of CCK-IRES-Cre mice on day 1. A genetically modified rabies virus (RV) pseudotyped with the avian envelope protein was delivered to the same coordinates on day 21 after the injection of helper viruses. Given that RV lacks an endogenous gene of RVG and has been genetically modified to express the fluorescent protein dsRed, the RV can only infect cells expressing the TVA while further spread of the rabies virus is limited due to the deletion of RVG. 7 days after RV injection, mice were perfused with 4% PFA in PBS. Brains were stored in 30% sucrose in PBS at 4 °C overnight and then 40 μm slices were collected for imaging by using an Olympus microscope (VS200).

Immunostaining for NeuN, Fos and CB₁R

Animals were perfused, fixed brains were kept in 30% sucrose, and the brains were sectioned on a sliding microtome in the frontal plane (Leica CM 1950). Sections were rinsed in 0.01 M PBS (pH = 7.2) with 10 min × 3 times and transferred sections into 0.3% (v/v) Triton-X 100 in PBS and incubated for 30 min (room temperature, RT). After PBS rinsing, sections were blocked with 5% (w/v) normal bovine serum for 1 h at RT or 4 °C overnight and then were incubated with primary antibodies overnight in 4 °C: Anti-cfos (1:1000, rabbit, abcam, ab190289), Anti-NeuN antibody (1:500, rabbit, Novusbio, NBP1- 77686SS) or Anti-CB₁R (1:500, rabbit, Frontier Institute, MSFR100580). The next day, brain sections were stained for 2 h in secondary antibodies (1:1000, abcam, ab150083, Alexa Fluor 594 goat anti-rabbit). Antibodies were diluted in phosphate-buffered saline which contained 2% (w/v) BSA and 0.3% (v/v) Triton X-100. Sections were washed in PBS 3 times for 10 min, incubated with DAPI (1/50000 dilution in PBS, Invitrogen), washed again in PBS, and mounted under coverslips with Fluoromount Aqueous Mounting Medium (Sigma-Aldrich). Image acquisition was performed with Olympus VS200 laser scanning microscope or Evident FV3000 using 20× or 40× objectives. Confocal images were analyzed using ImageJ.

RT-PCR

RT-PCR was performed as described previously. Cytoplasm from ventral tegmental nucleus of Gudden of PV^Cre::Cas9^fl mouse with control or gRNA virus injection after behavioral test. Harvested mRNA was first reverse transcribed to cDNA using a Super Script III kit (Invitrogen, USA), incubated at 50 °C for 50 min, and then terminated at 65 °C for 20 min. The entire cDNA was then amplified by multiplex PCR for 25 cycles (94 °C for 40 s, 57 °C for 90 s, 72 °C for 60 s, and a final extension at 72 °C for 10 min) using a QIAGEN Multiplex PCR Master Mix kit (QIAGEN, Germany) in a 100 mL volume. The transcripts of each tested gene were detected by a second round of PCR using an individual primer set: 1 mL of the multiplex PCR mixture was applied as a template in a 10 mL reaction with HotStarTaq polymerase (TianGen, China) and cycled again (35 cycles of 94 °C for 40 s, 57 °C for 40 s, 72 °C for 60 s, and a final extension at 72 °C for 10 min). The PCR products were run on a 2% agarose gel stained with GelRed (Invitrogen, USA) and visualized with ultraviolet light. The identities of the amplicons were confirmed by DNA sequencing (Sangon Biotech, China). Primers for GAPDH and CB1Rs were designed according to NCBI published sequences, as follows: CB₁Rs (295 bp): Forward 50-AGGAGAACGAGGACAACA-30, Reverse 30-ACATTGGGACTATCTTTGC-50; GAPDH (572 bp): Forward 50-ACTCCACTCACGGCAAATTC-30, Reverse 30-CACATTGGGGGTAGGAACAC-50; To avoid false negatives, only PCR detections of GAPDH were counted.

Imaging and analysis

Whole-brain slides were imaged using a ×20 magnification on an Olympus VS 200 virtual slide microscope system or Evident FV3000 and processed with ImageJ. The placement of the optical fibers was assessed based on the lesions in the tissues produced by the fiber tips. The expression levels of virus-infected cell bodies and axons were outlined in the corresponding sections in the mouse brain stereotaxic coordinates. We use Simple Neurite Tracer’s moniker (SNT)⁶⁹, which is very well integrated with the ImageJ platform to trace and reconstruct the expression of CB₁R along dendrites (Fig. 7).

Statistical analysis

Two-tailed student’s paired t-tests or unpaired t-tests, and one- or two-way ANOVA tests were used to determine statistical differences for anatomical, behavioral, and electrophysiological data using GraphPad Prism 10 (GraphPad Software). Post hoc analysis was applied when ANOVA showed a significant main effect. Raw data as shown in Supplementary Dataset, related to Figs. 1, 2, 3, 4, 5, 6, 7, S1, S2, S3, S4, S5, and S6. *p < 0.05, **p < 0.01, ***p < 0.001. All data are presented as means ± SEM.

Reporting summary

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

Data availability

The datasets generated and/or analyzed during the current study can be found in the paper and the supplementary materials. Any additional research materials will be made available on a request at weishengw@fudan.edu.cn. Source data are provided with this paper.

Code availability

No original code was generated in this study.

Change history

18 September 2025
A Correction to this paper has been published: https://doi.org/10.1038/s41467-025-64121-7

References

Tseng, Y. T., Schaefke, B., Wei, P. & Wang, L. Defensive responses: behaviour, the brain and the body. Nat. Rev. Neurosci. 24, 655–671 (2023).
Article CAS PubMed Google Scholar
Cooper, W. E. Jr. Optimal escape theory predicts escape behaviors beyond flight initiation distance: risk assessment and escape by striped plateau lizards Sceloporus virgatus. Curr. Zool. 55, 123–131 (2009).
Article Google Scholar
Hawlena, D., Kress, H., Dufresne, E. R. & Schmitz, O. J. Grasshoppers alter jumping biomechanics to enhance escape performance under chronic risk of spider predation. Funct. Ecol. 25, 279–288 (2011).
Article Google Scholar
Freymiller, G. A., Whitford, M. D., Higham, T. E. & Clark, R. W. Recent interactions with snakes enhance escape performance of desert kangaroo rats (Rodentia: Heteromyidae) during simulated attacks. Biol. J. Linn. Soc. 122, 651–660 (2017).
Article Google Scholar
Bateman, P. W. & Fleming, P. A. Switching to Plan B: changes in the escape tactics of two grasshopper species (Acrididae: Orthoptera) in response to repeated predatory approaches. Behav. Ecol. Sociobiol. 68, 457–465 (2014).
Article Google Scholar
Wooster, E. I. F. et al. Animal cognition and culture mediate predator-prey interactions. Trends Ecol. Evol. 39, 52–64 (2024).
Article PubMed Google Scholar
Gross, C. T. & Canteras, N. S. The many paths to fear. Nat. Rev. Neurosci. 13, 651–658 (2012).
Article CAS PubMed Google Scholar
Tovote, P. et al. Midbrain circuits for defensive behaviour. Nature 534, 206–212 (2016).
Article ADS CAS PubMed Google Scholar
Evans, D. A. et al. A synaptic threshold mechanism for computing escape decisions. Nature 558, 590–594 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Deng, H., Xiao, X. & Wang, Z. Periaqueductal gray neuronal activities underlie different aspects of defensive behaviors. J. Neurosci. 36, 7580–7588 (2016).
Article CAS PubMed PubMed Central Google Scholar
Wang, W. et al. Dorsal premammillary projection to periaqueductal gray controls escape vigor from innate and conditioned threats. Elife 10 (2021).
Wang, W. et al. Coordination of escape and spatial navigation circuits orchestrates versatile flight from threats. Neuron 109, 1848–1860 e1848 (2021).
Article CAS PubMed PubMed Central Google Scholar
Wickersham, I. R., Finke, S., Conzelmann, K. K. & Callaway, E. M. Retrograde neuronal tracing with a deletion-mutant rabies virus. Nat. Methods 4, 47–49 (2007).
Article CAS PubMed Google Scholar
Qi, S. et al. How cognitive and reactive fear circuits optimize escape decisions in humans. Proc. Natl. Acad. Sci. USA 115, 3186–3191(2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Tomsic, D., de Astrada, M. B., Sztarker, J. & Maldonado, H. Behavioral and neuronal attributes of short- and long-term habituation in the crab Chasmagnathus. Neurobiol. Learn Mem. 92, 176–182(2009).
Article PubMed Google Scholar
Comoli, E., Ribeiro-Barbosa, E. R. & Canteras, N. S. Afferent connections of the dorsal premammillary nucleus. J. Comp. Neurol. 423, 83–98 (2000).
Article CAS PubMed Google Scholar
Gongwer, M. W. et al. Brain-wide projections and differential encoding of prefrontal neuronal classes underlying learned and innate threat avoidance. J. Neurosci. 43, 5810–5830 (2023).
Article CAS PubMed PubMed Central Google Scholar
McNaughton, N. & Vann, S. D. Construction of complex memories via parallel distributed cortical-subcortical iterative integration. Trends Neurosci. 45, 550–562 (2022).
Article CAS PubMed PubMed Central Google Scholar
Yan, J. J. et al. A circuit from the ventral subiculum to anterior hypothalamic nucleus GABAergic neurons essential for anxiety-like behavioral avoidance. Nat. Commun. 13, 7464 (2022).
Article ADS CAS PubMed PubMed Central Google Scholar
Xie, Z. et al. Mechanically evoked defensive attack is controlled by GABAergic neurons in the anterior hypothalamic nucleus. Nat. Neurosci. 25, 72–85 (2022).
Article CAS PubMed Google Scholar
Wang, L., Chen, I. Z. & Lin, D. Collateral pathways from the ventromedial hypothalamus mediate defensive behaviors. Neuron 85, 1344–1358 (2015).
Article CAS PubMed PubMed Central Google Scholar
Whittle, N. et al. Central amygdala micro-circuits mediate fear extinction. Nat. Commun. 12, 4156 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Klein, A. S., Dolensek, N., Weiand, C. & Gogolla, N. Fear balance is maintained by bodily feedback to the insular cortex in mice. Science 374, 1010–1015 (2021).
Article ADS CAS PubMed Google Scholar
Dillingham, C. M. et al. Calcium-binding protein immunoreactivity in Gudden’s tegmental nuclei and the hippocampal formation: differential co-localization in neurons projecting to the mammillary bodies. Front Neuroanat. 9, 103 (2015).
Article PubMed PubMed Central Google Scholar
Zhou, Z. et al. A VTA GABAergic neural circuit mediates visually evoked innate defensive responses. Neuron 103, 473–488 e476 (2019).
Article CAS PubMed Google Scholar
Venkataraman, A. et al. Modulation of fear generalization by the zona incerta. Proc. Natl. Acad. Sci. USA 116, 9072–9077 (2019).
Article ADS CAS PubMed PubMed Central Google Scholar
Ahmadlou, M. et al. A cell type-specific cortico-subcortical brain circuit for investigatory and novelty-seeking behavior. Science 372 (2021).
Fratzl, A. et al. Flexible inhibitory control of visually evoked defensive behavior by the ventral lateral geniculate nucleus. Neuron 109, 3810–3822 e3819 (2021).
Article CAS PubMed PubMed Central Google Scholar
Evans, D. A., Stempel, A. V., Vale, R. & Branco, T. Cognitive control of escape behaviour. Trends Cogn. Sci. 23, 334–348 (2019).
Article PubMed PubMed Central Google Scholar
Dudok, B. et al. Retrograde endocannabinoid signaling at inhibitory synapses in vivo. Science 383, 967–970 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
Freund, T. F. & Hajos, N. Excitement reduces inhibition via endocannabinoids. Neuron 38, 362–365 (2003).
Article CAS PubMed Google Scholar
Castillo, P. E., Younts, T. J., Chavez, A. E. & Hashimotodani, Y. Endocannabinoid signaling and synaptic function. Neuron 76, 70–81 (2012).
Article CAS PubMed PubMed Central Google Scholar
Wittmann, G. et al. Distribution of type 1 cannabinoid receptor (CB1)-immunoreactive axons in the mouse hypothalamus. J. Comp. Neurol. 503, 270–279 (2007).
Article CAS PubMed Google Scholar
Farrell, J. S. et al. In vivo endocannabinoid dynamics at the timescale of physiological and pathological neural activity. Neuron 109, 2398–2403 e2394 (2021).
Article CAS PubMed PubMed Central Google Scholar
Gunduz-Cinar, O. et al. A cortico-amygdala neural substrate for endocannabinoid modulation of fear extinction. Neuron 111, 3053–3067.e3010 (2023).
Article CAS PubMed PubMed Central Google Scholar
Liput, D. J. et al. 2-Arachidonoylglycerol mobilization following brief synaptic stimulation in the dorsal lateral striatum requires glutamatergic and cholinergic neurotransmission. Neuropharmacology 205, 108916 (2022).
Article CAS PubMed Google Scholar
Kondev, V. et al. Endocannabinoid release at ventral hippocampal-amygdala synapses regulates stress-induced behavioral adaptation. Cell Rep. 42, 113027 (2023).
Article CAS PubMed PubMed Central Google Scholar
Dong, A. et al. A fluorescent sensor for spatiotemporally resolved imaging of endocannabinoid dynamics in vivo. Nat. Biotechnol. 40, 787–798 (2022).
Article CAS PubMed Google Scholar
Liu, Z. et al. Deficiency in endocannabinoid synthase DAGLB contributes to early onset Parkinsonism and murine nigral dopaminergic neuron dysfunction. Nat. Commun. 13, 3490 (2022).
Article ADS CAS PubMed PubMed Central Google Scholar
Hashimotodani, Y., Ohno-Shosaku, T. & Kano, M. Ca(2+)-assisted receptor-driven endocannabinoid release: mechanisms that associate presynaptic and postsynaptic activities. Curr. Opin. Neurobiol. 17, 360–365 (2007).
Article CAS PubMed Google Scholar
Ritter, A., Habusha, S., Givon, L., Edut, S. & Klavir, O. Prefrontal control of superior colliculus modulates innate escape behavior following adversity. Nat. Commun. 15, 2158 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
Miller, E. K. & Cohen, J. D. An integrative theory of prefrontal cortex function. Annu Rev. Neurosci. 24, 167–202 (2001).
Article CAS PubMed Google Scholar
Armbruster, B. N., Li, X., Pausch, M. H., Herlitze, S. & Roth, B. L. Evolving the lock to fit the key to create a family of G protein-coupled receptors potently activated by an inert ligand. Proc. Natl Acad. Sci. USA 104, 5163–5168 (2007).
Article ADS PubMed PubMed Central Google Scholar
Zhang, S. et al. Molecular basis for selective activation of DREADD-based chemogenetics. Nature 612, 354–362 (2022).
Article ADS CAS PubMed Google Scholar
Vann, S. D. Gudden’s ventral tegmental nucleus is vital for memory: re-evaluating diencephalic inputs for amnesia. Brain 132, 2372–2384 (2009).
Article PubMed Google Scholar
Bassant, M. H. & Poindessous-Jazat, F. Ventral tegmental nucleus of Gudden: a pontine hippocampal theta generator?. Hippocampus 11, 809–813 (2001).
Article CAS PubMed Google Scholar
Vann, S. D. Dismantling the Papez circuit for memory in rats. Elife 2, e00736 (2013).
Article PubMed PubMed Central Google Scholar
Friedman, N. P. & Robbins, T. W. The role of prefrontal cortex in cognitive control and executive function. Neuropsychopharmacology 47, 72–89 (2022).
Article PubMed Google Scholar
Koechlin, E. Prefrontal executive function and adaptive behavior in complex environments. Curr. Opin. Neurobiol. 37, 1–6 (2016).
Article CAS PubMed Google Scholar
Sangha, S., Diehl, M. M., Bergstrom, H. C. & Drew, M. R. Know safety, no fear. Neurosci. Biobehav Rev. 108, 218–230 (2020).
Article PubMed Google Scholar
Soltani, A. & Koechlin, E. Computational models of adaptive behavior and prefrontal cortex. Neuropsychopharmacology 47, 58–71 (2022).
Article PubMed Google Scholar
Borkar, C. D. et al. Top-down control of flight by a non-canonical cortico-amygdala pathway. Nature 625, 743–749 (2024).
Article ADS CAS PubMed PubMed Central Google Scholar
Chou, X. L. et al. Inhibitory gain modulation of defense behaviors by zona incerta. Nat. Commun. 9, 1151 (2018).
Article ADS PubMed PubMed Central Google Scholar
Hanganu-Opatz, I. L. et al. Resolving the prefrontal mechanisms of adaptive cognitive behaviors: A cross-species perspective. Neuron 111, 1020–1036 (2023).
Article CAS PubMed 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 PubMed Google Scholar
Izquierdo, I., Furini, C. R. & Myskiw, J. C. Fear Memory. Physiol. Rev. 96, 695–750 (2016).
Article PubMed Google Scholar
Moscarello, J. M. & Maren, S. Flexibility in the face of fear: Hippocampal-prefrontal regulation of fear and avoidance. Curr. Opin. Behav. Sci. 19, 44–49 (2018).
Article PubMed Google Scholar
Patel, S., Hill, M. N., Cheer, J. F., Wotjak, C. T. & Holmes, A. The endocannabinoid system as a target for novel anxiolytic drugs. Neurosci. Biobehav Rev. 76, 56–66 (2017).
Article CAS PubMed PubMed Central Google Scholar
Lutz, B., Marsicano, G., Maldonado, R. & Hillard, C. J. The endocannabinoid system in guarding against fear, anxiety and stress. Nat. Rev. Neurosci. 16, 705–718 (2015).
Article CAS PubMed PubMed Central Google Scholar
Chadwick, V. L., Rohleder, C., Koethe, D. & Leweke, F. M. Cannabinoids and the endocannabinoid system in anxiety, depression, and dysregulation of emotion in humans. Curr. Opin. Psychiatry 33, 20–42 (2020).
Article PubMed Google Scholar
Kondev, V. et al. The endocannabinoid 2-Arachidonoylglycerol Bidirectionally Modulates Acute and Protracted Effects of Predator Odor Exposure. Biol. Psychiatry 92, 739–749 (2022).
Article CAS PubMed PubMed Central Google Scholar
Heinz, D. E., Genewsky, A. & Wotjak, C. T. Enhanced anandamide signaling reduces flight behavior elicited by an approaching robo-beetle. Neuropharmacology 126, 233–241 (2017).
Article CAS PubMed Google Scholar
Uribe-Marino, A. et al. Anti-aversive effects of cannabidiol on innate fear-induced behaviors evoked by an ethological model of panic attacks based on a prey vs the wild snake Epicrates cenchria crassus confrontation paradigm. Neuropsychopharmacology 37, 412–421 (2012).
Article CAS PubMed Google Scholar
Guan, X. & Cao, P. Brain mechanisms underlying panic attack and panic disorder. Neurosci. Bull. 40, 795–814 (2024).
Article PubMed Google Scholar
Petrie, G. N., Nastase, A. S., Aukema, R. J. & Hill, M. N. Endocannabinoids, cannabinoids and the regulation of anxiety. Neuropharmacology 195, 108626 (2021).
Article CAS PubMed Google Scholar
Patton, G. C. et al. Cannabis use and mental health in young people: cohort study. BMJ 325, 1195–1198 (2002).
Article PubMed PubMed Central Google Scholar
Gorelick, D. A. Cannabis-related disorders and toxic effects. N. Engl. J. Med. 389, 2267–2275 (2023).
Article CAS PubMed Google Scholar
Wang, W. S. et al. Extinction of aversive memories associated with morphine withdrawal requires ERK-mediated epigenetic regulation of brain-derived neurotrophic factor transcription in the rat ventromedial prefrontal cortex. J. Neurosci. 32, 13763–13775 (2012).
Article CAS PubMed PubMed Central Google Scholar
Arshadi, C., Gunther, U., Eddison, M., Harrington, K. I. S. & Ferreira, T. A. SNT: a unifying toolbox for quantification of neuronal anatomy. Nat. Methods 18, 374–377 (2021).
Article CAS PubMed Google Scholar

Download references

Acknowledgements

We thank Profs. Peng Yuan (Fudan University) and Hongbin Yang (Zhejiang University) for helpful discussions. We were supported by grants from the STI2030-Major Projects (2021ZD0202703 to W.W.) and National Natural Science Foundation of China (NNSF32271067 to W.W. NNSF 82371527 to X.H.,); We thank Profs. B. Song for providing rats.

Author information

These authors contributed equally: Ruikai Chai, Nawen Wang, Jinlu Nie, Zongyi Xu.

Authors and Affiliations

Department of Neurology, Jinshan Hospital, Institute for Translational Brain Research, State Key Laboratory of Brain Function and Disorders, MOE Frontiers Center for Brain Science, Fudan University, Shanghai, 201508, China
Ruikai Chai, Nawen Wang, Jinlu Nie, Zongyi Xu, Shuqian Zhang, Suixin Deng, Rongxin Wang, Mu Li, Xinyi Gao, Lei Li, Hebi Wu, Zhiming Li, Huilin Hong & Weisheng Wang
Department of Psychological Medicine, Zhongshan Hospital, Fudan University, Shanghai, 200032, China
Ruijie Geng, Haibin Li & Xiao Huang
Institute of Pediatrics, National Children’s Medical Center, Children’s Hospital, Institute for Translational Brain Research, State Key Laboratory of Brain Function and Disorders, MOE Frontiers Center for Brain Science, Fudan University, Shanghai, 200032, China
Tian-Lin Cheng
Institute for Brain Science, State Key Laboratory of Brain Function and Disorders, MOE Frontiers Center for Brain Science, Fudan University, Shanghai, 200032, China
Xiao-Hong Xu
Institute for Translational Brain Research, State Key Laboratory of Brain Function and Disorders, MOE Frontiers Center for Brain Science, Fudan University, Shanghai, 200032, China
Yousheng Shu

Authors

Ruikai Chai
View author publications
Search author on:PubMed Google Scholar
Nawen Wang
View author publications
Search author on:PubMed Google Scholar
Jinlu Nie
View author publications
Search author on:PubMed Google Scholar
Zongyi Xu
View author publications
Search author on:PubMed Google Scholar
Shuqian Zhang
View author publications
Search author on:PubMed Google Scholar
Suixin Deng
View author publications
Search author on:PubMed Google Scholar
Rongxin Wang
View author publications
Search author on:PubMed Google Scholar
Mu Li
View author publications
Search author on:PubMed Google Scholar
Xinyi Gao
View author publications
Search author on:PubMed Google Scholar
Ruijie Geng
View author publications
Search author on:PubMed Google Scholar
Haibin Li
View author publications
Search author on:PubMed Google Scholar
Lei Li
View author publications
Search author on:PubMed Google Scholar
Hebi Wu
View author publications
Search author on:PubMed Google Scholar
Zhiming Li
View author publications
Search author on:PubMed Google Scholar
Tian-Lin Cheng
View author publications
Search author on:PubMed Google Scholar
Xiao-Hong Xu
View author publications
Search author on:PubMed Google Scholar
Yousheng Shu
View author publications
Search author on:PubMed Google Scholar
Huilin Hong
View author publications
Search author on:PubMed Google Scholar
Xiao Huang
View author publications
Search author on:PubMed Google Scholar
Weisheng Wang
View author publications
Search author on:PubMed Google Scholar

Contributions

Conceptualization, R.C. and W.W.; Methodology, R.C. and W.W.; behavioral tests and viral injections, R.C., N.W., J.N.,L.L., Z.X., S.D.; immunohistochemistry studies, H.W., M.L., R.G., H.L., X. G; Electrophysiological studies, Z.X.; Perfusion, R.W., Z.L.; Virus generation, S.Z.; Writing—Original Draft, H.H. and W.W.; Writing—Review & Editing, W.W. and X.-H.X.; Funding Acquisition, X.H. and W.W.; Resources, T.-L.C. and Y.S.; Supervision, W.W.

Corresponding authors

Correspondence to Yousheng Shu, Huilin Hong, Xiao Huang or Weisheng Wang.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. [A peer review file is available.]

Additional information

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

Supplementary information

Supplementary Information

Reporting Summary

Transparent Peer Review file

Source data

Source Data

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

Chai, R., Wang, N., Nie, J. et al. Endocannabinoids disinhibit the ventral tegmental nucleus of Gudden to dorsal premammillary nucleus pathway to enhance escape behavior following learned threat experience. Nat Commun 16, 4885 (2025). https://doi.org/10.1038/s41467-025-60080-1

Download citation

Received: 06 September 2024
Accepted: 13 May 2025
Published: 27 May 2025
DOI: https://doi.org/10.1038/s41467-025-60080-1