Synaptic and cellular endocannabinoid signaling mechanisms regulate stress-induced plasticity of nucleus accumbens somatostatin neurons

Significance

How stress affects neuronal plasticity within brain reward circuits is important to understand given the link between stress and mood disorders and addiction. We show that endogenous cannabinoids regulate the balance of excitatory drive to the nucleus accumbens in mice and are required for stress to shift excitatory drive to this key reward center. These cannabinoid-mediated shifts in balance of excitatory drive could reveal how stress affects motivational processes and provide insight into the neural basis of cannabis reward.

Abstract

Interneuron populations within the nucleus accumbens (NAc) orchestrate excitatory-inhibitory balance, undergo experience-dependent plasticity, and gate-motivated behavior, all biobehavioral processes heavily modulated by endogenous cannabinoid (eCB) signaling. While eCBs are well known to regulate synaptic plasticity onto NAc medium spiny neurons and modulate NAc function at the behavioral level, how eCBs regulate NAc interneuron function is less well understood. Here, we show that eCB signaling differentially regulates glutamatergic and feedforward GABAergic transmission onto NAc somatostatin–expressing interneurons (NAc^SOM+) in an input-specific manner, while simultaneously increasing postsynaptic excitability of NAc^SOM+ neurons, ultimately biasing toward vHPC (ventral hippocampal), and away from BLA (basolateral amygdalalar), activation of NAc^SOM+ neurons. We further demonstrate that NAc^SOM+ are activated by stress in vivo and undergo stress-dependent plasticity, evident as a global increase in intrinsic excitability and an increase in excitation–inhibition balance specifically at vHPC, but not BLA, inputs onto NAc^SOM+ neurons. Importantly, both forms of stress-induced plasticity are dependent on eCB signaling at cannabinoid type 1 receptors. These findings reveal eCB-dependent mechanisms that sculpt afferent input and excitability of NAc^SOM+ neurons and demonstrate a key role for eCB signaling in stress-induced plasticity of NAc^SOM+-associated circuits.

The nucleus accumbens (NAc) is a critical node within the mesolimbic reward pathway (1, 2). Medium spiny neurons (MSNs) within the NAc integrate information conveyed by glutamatergic limbic inputs to guide motivated behavior (3). A key regulatory component of MSN activity, and thus NAc output, are inhibitory microcircuits composed of distinct interneuron populations (4–10). One such population are low-threshold spiking, somatostatin (SOM+), and neuropeptide Y (NPY+) coexpressing interneurons (NAc^SOM+), which play a potent role in modulating MSN activity and NAc-mediated behavior (4, 5, 11, 12). Despite this important role, the molecular mechanisms orchestrating SOM+/NPY+ activity are poorly understood but may reveal treatment targets for dysfunctional motivational states, including substance use disorders and major depression.

One system that has been identified as a potent regulator of NAc physiological and pathological function is the endogenous cannabinoid, or endocannabinoid (eCB) system (13–16). The eCB system is a lipid-derived signaling system composed of the cannabinoid receptor type 1 and 2 (CB1/2R), its endogenous ligands, N-arachidonoylethanolamine [anandamide (AEA)] and 2-arachidonoylglycerol (2-AG), and their synthetic and degradative enzymes (17, 18). It is well established that the eCB system regulates glutamatergic input onto MSNs and some interneuron populations, including fast-spiking interneurons; eCB signaling has also been shown to be recruited at interneuron synapses onto MSNs, together suggesting that the eCB system is a critical modulator of excitatory–inhibitory balance within the NAc (8, 15, 19–21). Indeed, changes in NAc eCB signaling are heavily implicated in the etiology of affective disorders, especially following stress exposure (14).

Here, we use a combination of in vivo fiber photometry and ex vivo electrophysiology and pharmacology to elucidate the role of eCB signaling in the regulation of NAc^SOM+-associated neural circuits. SOM+/NPY+ interneurons display patterns of glutamatergic connectivity like MSNs, including input from the ventral hippocampus (vHPC), basolateral amygdala (BLA), and prefrontal cortex (PFC) (22). Our data demonstrate that eCB signaling differentially regulates glutamatergic and feedforward gamma-aminobutyric acid (GABAergic) transmission onto NAc^SOM+ in an input-specific manner, while simultaneously increasing postsynaptic excitability of NAc^SOM+ neurons. These effects ultimately bias toward vHPC, and away from BLA, activation of NAc^SOM+ neurons. Furthermore, NAc^SOM+ are stress responsive, and an acute footshock exposure reorganizes the strength of glutamatergic inputs to favor vHPC, over BLA, activation, and increases the intrinsic excitability of NAc^SOM+ cells, resulting in an increased ability of both inputs to activate NAc^SOM+ neurons. Importantly, these presynaptic and postsynaptic changes can both be reversed by blocking CB1R signaling. These data reveal an eCB-dependent mechanism that modulates stress-induced plasticity of NAc^SOM+ interneuron populations and could have implications for understanding how stress dysregulates motivational processes.

Methods

Subjects.

All studies were carried out in accordance with the NIH Guide for the Care and Use of Laboratory Animals and approved by the Vanderbilt University Institutional Animal Care and Use Committee. Male and female mice were housed in a temperature and humidity-controlled housing facility under a 12-h light/dark cycle with ad libitum access to food (23). For electrophysiological recordings, SOM+-Cre (Sst-IRES-Cre; Jackson Laboratory) were crossed with Ai14 [Cg-Gt(ROSA)26Sor^{tm14(CAG-tdTomato)/Hze}/J; Jackson Laboratory] to allow for Cre-driven expression of the red fluorophore, tdTomato, in all SOM+ (Cre+) cells. For in vivo fiber photometry recordings, SOM+-Cre homozygous mice were used. Mice underwent viral injections at 5 to 10 wk and were given at least 4 wk to recover/allow for virus expression. Cells from male mice are depicted as circles, while cells from female mice are depicted as triangles.

Surgeries.

Mice were anesthetized with isoflurane and then transferred to the stereotax (Kopf Instruments, Tujunga, CA) and kept under 3% isoflurane anesthesia. Mice were injected with 10 mg/kg ketoprofen (AlliVet) as an analgesic. The hair over the incision site was shaved and the skin was prepped with alcohol and iodine. A midline sagittal incision was made to expose the skull; a hole was drilled above the target brain region, and a virus was delivered using a motorized digital software (NeuroStar, Stoelting CO., Wood Dale, IL), a 10-μL microinjection syringe (Hamilton CO., Reno, NV), and a micropump controller (World Precision Instruments, Sarasota, FL) at 0.1 μL per minute. A local, topical anesthetic, benzocaine (Medline Industries, Brentwood, TN) was applied to the incision area. After surgery, postoperative treatment with ketoprofen was administered for at least 48 h.

Ex Vivo Electrophysiology.

SOM-Ai14 mice were briefly anesthetized with isoflurane and transcardially perfused with ice-cold oxygenated (95% v/v O₂, 5% v/v CO₂) N-methyl-D-glucamine (NMDG)-based Artificial cerebrospinal fluid (ACSF) (24) comprised (in mM): 93 NMDG, 2.5 KCl, 1.2 NaH₂PO₄, 30 NaHCO₃, 20 N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid (HEPES), 25 glucose, 5 sodium-ascorbate, 3 sodium-pyruvate, 5 N-acetylcyctine, 0.5 CaCl₂·4H₂O, and 10 MgSO₄·7H₂O. The brain was quickly removed and 250-μm coronal slices containing the NAc were cut using a vibratome (Leica Biosystems, model # VT1000S) in the NMDG solution. Slices were incubated for 13 to 20 min at 32 °C in oxygenated NMDG-ACSF then stored at 24 °C until recordings were performed in HEPES-based ACSF containing (in mM): 92 NaCl, 2.5 KCl, 1.2 NaH₂PO₄, 30 NaHCO₃, 20 HEPES, 25 glucose, 5 ascorbate, 3 sodium-pyruvate, 5 N-acetylcyctine, 2 CaCl₂·4H₂O, and 2 MgSO₄·7H₂O.

Recordings were performed in a submerged recording chamber during continuous perfusion of oxygenated ACSF containing (in mM): 113 NaCl, 2.5 KCl, 1.2 MgSO₄∙7H2O, 2.5 CaCl₂∙2H2O, 1 NaH₂PO₄, 26 NaHCO₃, 1 ascorbate, 3 sodium-pyruvate and 20 glucose; at a flow rate of 2.5 to 3 mL/min. Slices were visualized using a Nikon microscope (Eclipse FN1, Nikon Instruments Inc., Melville, NY) equipped with differential interference contrast microscopy (DIC). Fluorescently labeled SOM+ interneurons in the NAc were identified using a series 120Q X-cite lamp at 40× magnification with an immersion objective with DIC. Data collection was coordinated using pClamp 10 (Molecular Devices, San Jose, CA) and cell electrical properties were monitored using a Molecular Devices 700B MultiClamp amplifier and a Digidata 1440A low-noise data acquisition digitizer. Cells with an access resistance of >30 MΩ or that exhibited greater than a 20% change over the course of recordings were not included in our datasets. For more information, see ref. 25. The drugs used are as follows: CP55,940 (5 to 10 µM); rimonabant (5 to 10 µM), JZL184 (1 µM), and PF3845 (5 µM). All drugs used were stored in DMSO aliquots and then included in the ACSF and/or the HEPES-ACSF holding solution with Bovine Serum Albumin (Fisher Scientific).

Characterization of eCB Regulation of Spontaneous Glutamatergic and GABAergic Neurotransmission.

2 to 6 MΩ borosilicate glass pipettes were filled with a Cs⁺-based internal solution (in mM): 120 CsOH, 120 D-gluconic acid, 2.8 NaCl, 20 HEPES, 5 TEA-Cl, 2.5 Mg-ATP, and 0.25 Na-GTP. Spontaneous excitatory postsynaptic currents (sEPSCs) and spontaneous inhibitory postsynaptic currents (sIPSCs) were recorded for 1 min from the same cell at −70 mV and +13 mV, respectively. Slices were incubated for 30 to 60 min in different drugs, as outlined in Fig. 1. Cells were excluded from E/I (excitation/inhibition) quantification if the access resistance changed more than 20% between sEPSC and sIPSC recordings.

Fig. 1.

eCB modulation of glutamatergic and GABAergic drive onto NAc^SOM+. (A) SOM-Ai14 mice were used for electrophysiological recordings. SOM+ cells were identified by presence of fluorophore. sEPSCs at −70 mV and sIPSCs at +13 mV were recorded from the same cell. (Scale bar represents 500 ms, 100 pA.) Slices were incubated in various drugs that target different components of the eCB system: the CBR agonist, CP55,940 (5 to 10 µM); CB1R inverse agonist, rimonabant (5 µM); monoacylglycerol lipase (MAGL) inhibitor, JZL184 (1 µM); fatty acid amide hydrolase (FAAH) inhibitor, PF3845 (5 µM). One-Way ANOVA was performed (B–G). n cells depicted above each graph from the following N mice: Veh = 32; CP = 7; Rim = 8; JZL = 6; PF = 4. Cells were recorded from male (circles) and female (triangles) mice. (B) Quantification of sEPSC frequency (Left) and cumulative probability plot (Right) after incubation with eCB-targeting drugs [F(4,112) = 7.307; P < 0.0001]. (C) Quantification of frequency of sIPSCs (Left) and cumulative probability plot (Right) following drug incubation [F(4,105) = 8.113; P < 0.0001]. (D) Analysis of E/I frequency ratio following drug incubation [F(4,98) = 1.077; P = 0.3721]. (E) Quantification of sEPSC amplitude (Left) and cumulative probability plot (Right) [F(4,133) = 1.619; P = 0.1730]. (F) Quantification of sIPSC amplitude (Left) and cumulative probability plot (Right) after incubation with different drugs [F(4,93) = 4.233; P = 0.0034]. (G) Analysis of E/I amplitude ratio following drug incubations [F(4,106) = 4.161; P = 0.0036].

Ex Vivo Optogenetics.

For circuit-specific characterization of eCB regulation, a virus expressing ChR2 (channelrhodopsin) was delivered to the vHPC (irrespective of subregion) (AP: −3.6; ML: ±3.0; DV: −4.0), BLA (AP: −1.25, ML: ± 3.30, DV: −5.00) or PFC (AP: 2.0; ML: ±0.3; DV: −2.2) of SOM-Ai14 mice. At least 4 wk after viral injection, ex vivo whole-cell patch-clamp electrophysiology recordings were performed in the NAc. For all optogenetic studies, 1-ms light exposure time was used.

2 to 6 MΩ borosilicate glass pipettes were filled with a Cs⁺-based internal solution (in mM): 120 CsOH, 120 D-gluconic acid, 2.8 NaCl, 20 HEPES, 5 TEA-Cl, 2.5 Mg-ATP, and 0.25 Na-GTP. Optically evoked excitatory postsynaptic currents (oEPSCs) and optically evoked inhibitory postsynaptic currents (oIPSCs) were recorded from the same cell at −70 mV and +13 mV, respectively. PPR (paired pulse ratio) recordings were obtained in voltage-clamp with an inter-stimulus interval of 50 ms. PPR is reported as a ratio between the amplitude of the second oEPSC divided by the first. For drug washes, a baseline of oEPSC amplitude was obtained for at least 4 min, and then drugs were washed on, as outlined in the figure legends. Data are presented as %baseline oEPSC amplitude. Cells were excluded if the baseline oEPSC amplitude was <100 pA, the access resistance was >30 MΩ, or if the access resistance fluctuated by more than 20%. oIPSC amplitude and PPR (50-ms interstimulus interval) were assessed from cells in vehicle and following drug wash-on; data are from the largest intensity stimulation where oIPSC amplitude is between 50 pA and 1,500 pA. Cells were excluded from oIPSC PPR if there was no second peak.

For optogenetically elicited action potential (AP) firing, recordings were performed in current clamp; five 1-ms light stimulations were applied at each frequency (1 to 20 Hz) for five traces, and the AP probability was calculated as the proportion of neurons firing APs (i.e., three APs out of all 5 traces = 3/25). For intrinsic excitability experiments, current-clamp recordings of somatic current injection–induced AP firing were obtained by initially injecting enough current to hold the neuron at −70 mV and then applying sequential depolarizing steps that increase by 40 pA.

For assessing stress-induced changes in neural physiology, mice were exposed to an acute footshock stress, consisting of a 7.5-min session with six 0.7 mA footshocks delivered 1 min apart using a MED Associates fear-conditioning chamber. Each shock coincided with the last 2 s of a 30-s auditory tone. For controls, mice were exposed to tone alone (no shock). Twenty four hours after shock/tone exposure, mice were killed for electrophysiological recordings as outlined in figure legends. For intrinsic excitability experiments following footshock, a K+-gluconate internal was used and picrotoxin (50 µM) was added to the ACSF. For E/I experiments, a Cs⁺-based internal solution and oEPSC amplitude was kept between 100 and 400 pA. Cells were excluded from analysis if no detectable GABA was observed (i.e., response < 10 pA).

Fiber Photometry.

For somatic recordings of NAc^SOM+ activity, AAV9-FLEX-hSyn-jGCaMP7f-WPRE (400 to 500 nL; Addgene #104492-AAV9) or AAV5-Ef1a-DIO-eYFP (Addgene, #27056-AAV5) was delivered unilaterally to the NAc (AP: 2.1; ML: ±0.9; DV: −4.33) of SOM-Cre mice. At least 4 wk following virus expression, a 400-µM mono fiberoptic cannula (Doric) was implanted above the NAc (DV: −4.29). Mice were given at least 1 wk to recover from fiber optic implantation before undergoing behavioral testing. A 465-nm LED [Lx465; Tucker Davis Technologies (TDT)] was used to detect GFP-dependent changes in fluorescence; a 415-nm LED (Lx415; TDT) was used as the isosbestic point. Light was emitted from the LEDs and passed through a minicube (Doric), that connected to the fiber implant via a 0.57 NA fiber optic patch cord (Doric). GFP-dependent changes in fluorescence were recorded by a real-time processor (RZ10x; TDT). To confirm fiber placement and virus expression, mice were perfused, and brains were sectioned for histological verification of injection sites and fiber optic placement with a fluorescent microscope (Axio Imager M2 epifluorescent microscope). Mice were excluded if the fiber optic was not above the NAc, or if there was no AAV-mediated fluorescence.

Fiber photometry was analyzed using code modified from TDT (https://www.​tdt.com/docs/sdk/offline-data-analysis/offline-data-matlab/fiber-photometry-epoch-averaging-example/). The ΔF/F was calculated via polyfit function and an algorithm sourced from Tom Davidson’s Github (https://github.com/tjd2002/tjd-shared-code/blob/master/matlab/photometry/FP_normalize.m). The z-score is calculated as the change in ΔF/F, using the seconds −3 to −1 before stimulus onset (0 s) as the baseline. The z-score was used to account for between-subject variability in signal magnitude. The change in the area under the curve (AUC) was calculated using the time before stimulus onset (0) as baseline, compared to the AUC postbaseline.

Acute Footshock Stress.

Mice were exposed to an acute footshock stress, which is a 7.5-min session consisting of six 0.7 mA foot shocks delivered 1 min apart using a MED Associates fear-conditioning chamber (St. Albans, VT). Each shock coincided with the last 2 s of a 30-s auditory tone. The onset of shock and tone was synchronized to fear conditioning software (FreezeFrame). After stress exposure, mice were returned to their home cages.

Restraint Stress.

Mice were restrained in custom made (Vanderbilt Machine Shop, Nashville, TN) restraint tubes for 10 min. Active struggle bouts were flagged by DeepLabCut as previously described (26). Briefly, R statistical software with the “tidyverse” package was used to convert X/Y position into speed of movement for the fiber and the tail during each frame. To identify whole-body struggle bouts, frames were identified in which the fiber optic and tail were moving based on speed thresholds.

2MT (2-Methyl-2-Thiazoline) Stress.

Mice were exposed to a novel cage for 5 min, or a novel cage with 2MT (25 to 30 µL) pipette in the corner on filter paper. Exposure to novel cage verse 2MT was counterbalanced to control for habituation/contextual fear expression. To calculate the frequency of calcium transients, peaks were calculated via the findpeaks function in Matlab, with the MinPeakThreshold set to 0.15. This threshold was applied to both GCaMP and YFP traces; however, changes in signal to noise between viruses likely resulted in differences in baseline event frequencies. In addition, the same duration (50 s) and threshold were used to compare the frequency in the novel cage verse 2MT exposure.

Movement Initiation.

Recordings were performed in a novel cage, and movement initiation was hand-scored by a blinded (GCaMP7f verse YFP) experimenter.

Statistics.

All statistical tests were performed using PRISM and are reported in figure legends. For analysis of two groups, an unpaired Student’s t test was used. For wash-on experiments comparing two groups, a paired Student’s t test was used. For analysis of two or more groups, a one-way ANOVA was performed. For analysis of two or more groups across two or more treatments or time points, a 2-way ANOVA with Holm–Sidak post hoc correction was used. For all datasets, a ROUT outlier test was performed, and outliers were removed. All statistical tests and results are reported in SI Appendix, Tables S1 and S2. n, N represents # of cells from # of mice. Significance was defined by a P-value of <0.05. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001

Results

eCBs Differentially Regulate Glutamatergic and GABAergic Transmission onto NAc^SOM+.

To elucidate how the eCB system regulates synaptic transmission onto NAc^SOM+, we used whole-cell patch-clamp electrophysiology in SOM-Cre × Ai14 mice. NAc^SOM+ cells were identified by the presence of red fluorophore (Fig. 1A). First, we electrically isolated spontaneous excitatory postsynaptic currents (sEPSCs) and spontaneous inhibitory postsynaptic currents (sIPSCs) from the same cell to gain insight into how the eCB system alters NAc^SOM+ synaptic dynamics (Fig. 1A). Pharmacological activation of cannabinoid receptors with CP55,940 (5 to 10 µM) decreased the frequency of sEPSCs (Fig. 1B) and sIPSCs (Fig. 1C) consistent with activation of presynaptic CB1R. Antagonism of the CB1R with the inverse agonist, rimonabant (10 µM), had no effect on sEPSC or sIPSC frequency, demonstrating that spontaneous excitatory and inhibitory input onto NAc^SOM+ cells are not regulated by tonic eCB signaling. Finally, we parsed apart the contribution of the two major eCB ligands, 2-AG and AEA, by pharmacologically inhibiting the 2-AG degradation enzyme, monoacylglycerol lipase (MAGL), or the AEA degradation enzyme, fatty acid hydrolase (FAAH). Incubation with the MAGL inhibitor, JZL184 (1 µM), or the FAAH inhibitor, PF3845 (5 µM) significantly decreased sEPSC frequency, ultimately demonstrating that both 2-AG and AEA modulate spontaneous glutamatergic input onto NAc^SOM+ (Fig. 1B). However, exclusively PF3845 decreased sIPSC frequency, suggesting that AEA, but not 2-AG, controls spontaneous GABAergic input onto these SOM+ interneurons (Fig. 1C). These data demonstrate that the primary eCBs, 2-AG and AEA, control glutamatergic and GABAergic neurotransmission differently.

We also analyzed the net effect of these pharmacological interventions on NAc^SOM+ network activity by assessing excitation-inhibition (E/I) ratios. We found no significant changes in E/I frequency ratio, suggesting that while CB1R activation via eCB augmentation alters both excitatory and inhibitory drive onto SOM+ interneurons, there is not net change in E/I balance (Fig. 1D). Finally, we also assessed postsynaptic changes by analyzing the amplitude of both sEPSCs and sIPSCs. As expected, we found no significant effect on sEPSC amplitude (Fig. 1E). Interestingly, pharmacological cannabinoid receptor activation with CP55,940 and FAAH inhibition decreased sIPSC amplitude (Fig. 1F). CP55,940 also resulted in an increase in the E/I amplitude ratio, driven by a shift in the size of sIPSC events (Fig. 1 F and G). Importantly, we found no difference between male and female mice in basal glutamatergic or GABAergic transmission (SI Appendix, Fig. S1 A–D). Together, these data establish that eCB signaling regulates synaptic glutamatergic and GABAergic transmission onto NAc^SOM+.

Cannabinoid Receptor Activation Modulates E/I Balance in a Circuit-Specific Manner.

We next determined specific inputs that may be responsible for the observed decrease in sEPSC frequency following CP55,940 incubation. We used optogenetic ChR2-assisted projection targeting to assess three different glutamatergic inputs to NAc^SOM+ neurons: vHPC, BLA, and PFC. These inputs have previously been demonstrated to elicit excitatory currents onto MSNs (vHPC > BLA > PFC) that undergo synaptic depression following cannabinoid receptor activation (15, 27). Monosynaptic, optically evoked excitatory currents (oEPSCs) were recorded from NAc^SOM+ cells. Consistent with previous studies observing excitatory input onto MSNs, we found that vHPC inputs made stronger synaptic connections onto SOM+ cells, compared to PFC or BLA input (27) (SI Appendix, Fig. S2 A and B).

We assessed the presence of functional cannabinoid receptors at these three inputs by bath applying the cannabinoid receptor agonist, CP55,940 (5 µM). We found a robust depression of glutamatergic drive following CP55,940 application at vHPC-NAc^SOM+ and BLA-NAc^SOM+ synapses, but not PFC-NAc^SOM+ (Fig. 2 A–J and SI Appendix, Fig. S3 A–C). Consistent with presynaptic CB1R activation, we also observed a significant increase in the PPR, indicating decreased presynaptic release probability, at vHPC and BLA, but not PFC, inputs onto NAc^SOM+ after CP55,940 application (Fig. 2 E and J and SI Appendix, Fig. S3D). These data demonstrate that cannabinoid receptor activity modulates glutamatergic input onto NAc^SOM+ in a circuit-specific manner.

Fig. 2.

Cannabinoid receptors regulate glutamatergic feedforward GABAergic input onto NA^cSOM+. (A) Schematic of experimental setup. ChR2 (AAV-ChR2) was injected into vHPCof SOM-Ai14. Electrophysiological recordings of oEPSCs and oIPSCs onto NA^cSOM+ neurons were conducted. N = 5 mice. (B) Effect on vHPC-mediated oEPSC amplitude following bathapplication of CB1R agonist, CP55,940 (5 μM). (C) Representative traces of oEPSC before(black) and after CP55,940 wash-on (purple). (Scale bar represents 50 ms, 300 pA.) (D) Effect of CP55,940 wash on oEPSC amplitude. (E) Quantification of vHPC-NA^cSOM+ PPRfollowing CP55,940 wash-on. (F) Schematic of experimental setup. AAV-ChR2 was injectedinto the BLA of SOM-Ai14 and oEPSC and oIPSCs were assessed. N = 5 mice. (G) Effect onBLA-mediated oEPSC amplitude following bath application of CB1R agonist, CP55,940 (5μM). (H) Representative traces of BLA-NA^cSOM+ oEPSC before (black) and after CP55,940wash-on (purple). (Scale bar 50 ms, 100 pA.) (I) Effect of CP55,940 wash on oEPSC amplitude. (J) Quantification of PPR following CP55,940 wash-on. (K) Schematic offeedforward inhibition onto NAcSOM+ after optogenetic stimulation of glutamatergic input. (L) Quantification of vHPC-mediated oIPSCs with CP55,940 application. (M) Quantification ofoIPSC PPR following CP55,940. (N) Analysis of excitatory/inhibitory (E/I) ratio afterincubation with CP55,940. (O) Quantification of BLA-mediated oIPSCs with CP55,940application. (P) Quantification of BLA-NAcSOM+ feedforward oIPSC PPR followingCP55,940. (Q) Analysis of E/I ratio after CP55,940. Paired Student’s t test (D, E, I, and J), or unpaired Student’s t test (L–Q) was performed, n cells depicted above each graph from the following N mice: vHPC: N = 5; BLA: N = 4.

Activation of these glutamatergic inputs also evoked disynaptic inhibition mediated by local GABAergic neurons (7, 28) (Fig. 2K). This phenomenon ultimately shunts neuronal activation, providing a regulatory gate for excessive excitation of NAc population activity. Given that cannabinoid receptor activation induced a robust decrease in spontaneous GABAergic transmission onto NAc^SOM+, we assessed how pharmacological cannabinoid receptor activation alters feedforward inhibition elicited by optogenetic activation of distinct afferents. As expected, disynaptic optically evoked IPSCs (oIPSCs) exhibited longer onset latencies than monosynaptic excitatory inputs (SI Appendix, Fig. S2 D–G). We observed a significant reduction in the feedforward oIPSC amplitude after CP55,940 application upon vHPC, but an increase in oIPSC amplitude upon BLA, terminal stimulation (Fig. 2 L–P). We further assessed the net effect of cannabinoid receptor activation on neuronal activation patterns by quantifying the E/I ratio (29, 30). CP55,940 significantly increased the E/I ratio when vHPC input was stimulated but decreased the E/I ratio upon BLA-NAc^SOM+ stimulation, suggesting that cannabinoid receptors shift synaptic drive toward excitation of NAc^SOM+ neurons upon vHPC stimulation, but toward inhibition of NAc^SOM+ neurons upon BLA stimulation (Fig. 2 N and Q). We found no significant effect of CP55,940 on feedforward inhibition or E/I balance following optogenetic activation of the PFC (SI Appendix, Fig. S3 E–G).

We next assessed tonic eCB regulation of glutamatergic and GABAergic feedforward drive onto NAc^SOM+. Bath application of the CB1R antagonist/inverse agonist rimonabant (10 µM) significantly potentiated vHPC and BLA inputs onto NAc^SOM+ neurons, suggesting tonic eCB signaling suppresses glutamate release at these synapses; this was further supported by a significant decrease in PPR at both inputs after rimonabant wash-on (SI Appendix, Fig. S4 A–J). Feedforward inhibition elicited by glutamatergic projections supported a bidirectional net effect on NAc^SOM+ activity (SI Appendix, Fig. S4 K–P). Specifically, rimonabant significantly shifted the vHPC-activation mediated E/I ratio toward greater inhibition, on the other hand, CB1R antagonism significantly increased the BLA-evoked E/I ratio toward greater excitation (SI Appendix, Fig. S4 M and P). Taken together, and consistent with our CP55,9940 data above, these data suggest tonic eCB signaling biases toward NAc^SOM+ neuron activation by vHPC inputs, and inhibition by BLA inputs, possibly due to vHPC recruitment of more highly cannabinoid receptor sensitive disynaptic GABAergic circuits, than BLA inputs. However, how CP55,940 results in increased feedforward GABAergic transmission upon BLA stimulation is unclear.

Cannabinoid Receptors Gate Afferent-Induced Activation of NAc^SOM+ Neurons.

Our data thus far indicate that eCB signaling bidirectionally shifts the E/I balance onto NAc^SOM+ neurons in a circuit-specific manner, which is mediated primarily by differential effects of CP55,940 on feedforward GABAergic transmission. We thus hypothesized that this may translate into changes in the ability of these inputs to evoke NAc^SOM+ neuron activation. Specifically, since cannabinoid receptor activation shifts the E/I balance toward enhanced excitation upon vHPC stimulation, but enhanced inhibition upon BLA stimulation, we hypothesized that cannabinoid receptor activation could result in an increase, or decrease, in the ability of these inputs to activate NAc^SOM+, respectively (Fig. 3A).

Fig. 3.

Cannabinoid receptors differentially modulate afferent-induced excitation of NAc^SOM+ neurons. (A) Schematic of experimental design. ChR2 was virally delivered to vHPC or BLA. Afferents were optogenetically activated for 5 pulses at different frequencies, and the probability of an AP firing was assessed with or without CP55,940 incubation (10 µM). (B) Representative traces of oEPSCs originating from vHPC (green) (Top) and BLA (blue) (Bottom). (Scale bar represents 10 ms, 100 pA.) (C) Average vHPC- (N = 5) and BLA- (N = 3) mediated oEPSC amplitude. (D) AP probability following optogenetic stimulation of vHPC or BLA afferents across different frequencies: 1, 6, 10, and 20 Hz. (E) vHPC-mediated oEPSC amplitude (N = 4). (F) Representative traces of vHPC-mediated APs in vehicle (black) or CP55,940 (purple) across different frequencies. (G) Average vHPC-induced AP probability across frequencies after slices were incubated in CP55,940 (10 μM). (H) Average amplitude of BLA-mediated oEPSC (N =3). (I) Representative traces of BLA-mediated APs in vehicle (blue) or CP55,940 (purple) across different frequencies. (J) Average BLA-induced AP probability following CP55,940 incubation. (K) Current step recordings from NAcSOM+ neurons (N = 5). (L) Representative trace of AP firing frequency at 300 pA current step injection in slices incubated in vehicle (black) or CP55,940 (purple). (Scale bar represents 50 mV, 200 ms.) (M) Average number of APs at the highest current step injection, 300 pA. (N) Quantification of the ½ width of the first AP fired at 300 pA, comparing vehicle to CP55,940 incubated slices. (O) Average latency for the first AP to fire following CP55,940 incubation. (P) Quantification of the first AP amplitude. (Q) Quantification of the afterhyperpolarization potential (AHP) after the first AP. Unpaired Student’s t test (C, E, H, and M–Q) or 2-Way ANOVA (D, G, J, and K) were performed. n cells depicted above each graph from N mice.

To directly test this hypothesis, we optogenetically stimulated vHPC or BLA terminals at different frequencies to generate postsynaptic action potentials (APs) in NAc^SOM+ neurons. The amplitude of oEPSCs was kept comparable between inputs, and drug conditions, to control for variability in virus expression and isolate differential effects of CP55,940 on feedforward GABAergic transmission (Fig. 3 B, C, E, and H). At baseline, we observed that BLA inputs were more likely to trigger APs in NAc^SOM+, compared to vHPC inputs (Fig. 3D). However, following incubation with CP55,940 (10 µM), NAc^SOM+ exhibited a frequency-dependent increase in vHPC-stimulated AP probability (Fig. 4 E–G). Importantly, cannabinoid receptor activation had no effect on AP probability at BLA-NAc^SOM+ synapses (Fig. 4 H–J).

Fig. 4.

NAc^SOM+ cells respond to footshock stress. (A) Schematic of experimental design. SOM-Cre mice were injected with a Cre-dependent virus expressing GCaMP7f (FLEX-GCaMP7f) or YFP (DIO-YFP) into the NAc. At least 4 wks later, mice were implanted with a fiberoptic directly above the NAc. Following recovery, mice were exposed to a single fearconditioning session. (B) Fear conditioning paradigm. Mice were exposed to 6 tone (30 s)-shock (0.7 mA, 2 s) pairings. (C) Representative traces of resulting changes in 465 (blue) or 405 (purple) in GCaMP7f (Top) or YFP (Bottom) injected mice; (scale bar represents 5 mV, 50 s.) (D) Representative z-score of ΔF/F from GCaMP7f-injected mouse across six shock exposures. (E) Average z-score of change in ΔF/F following footshock (blue bar) in GCaMP7f (green)and YFP (yellow) injected mice. (F) Quantification of AUC following footshock. (G) Representative z-score of ΔF/F from GCaMP7f-injected mouse across six tone exposures. (H) Average z-score of change in fluorescence following tone presentation (gray) inGCaMP7f- (green) and YFP (yellow)-injected mice. (I) Quantification of AUC followingtone onset. 2-Way ANOVA (F and I) was performed for statistical analysis.

In addition to effects on neurotransmitter release, cannabinoid receptors can affect postsynaptic excitability under some conditions (31–33), which could contribute to our findings above. Thus, we assessed intrinsic excitability of NAc^SOM+ neurons following CP55,940 incubation. Surprisingly, we found that pharmacological cannabinoid receptor activation significantly increased AP firing frequency at high current injections (>180 pA) (Fig. 3 K–M). This was not associated with any changes in passive membrane properties, including the resting membrane potential, input resistance, capacitance, or membrane constant (SI Appendix, Fig. S5 A–D). However, further analysis into the kinetics of AP firing revealed that cannabinoid receptor activation significantly decreased the ½ width of elicited APs, with no effect on the latency, amplitude, or afterhyperpolarization (Fig. 3 N–Q). Preincubation with the CB1R inverse agonist, rimonabant, prevented the CP55,940-induced increase in excitability, confirming CB1R specificity (SI Appendix, Fig. S5 E–G). Furthermore, rimonabant incubation on its own significantly decreased NAc^SOM+ excitability, supporting bidirectional effects of CB1R on NAc^SOM+ neuron excitability and the possible existence of tonic CB1 signaling that maintains high levels of intrinsic excitability in NAc^SOM+ neurons (SI Appendix, Fig. S5 H–J). Taken together, these data indicate eCB signaling decreases excitatory input at both vHPC- and BLA-NAc^SOM+ synapses, preferentially suppresses feedforward inhibition triggered by vHPC afferents, and increases the intrinsic excitability of NAc^SOM+ neurons. Ultimately these pleotropic eCB signaling effects result in a sculpting of afferent drive to favor NAc^SOM+ neuron activation by vHPC, over BLA, afferents.

NAc^SOM+ Interneurons Respond to Stress In Vivo.

Thus far we have characterized eCB-dependent modulation of NAc^SOM+ interneurons, demonstrating that eCB/CB1R signaling ultimately shifts afferent excitation balance toward activation by vHPC inputs and away from BLA inputs. The vHPC–NAc pathway is a critical locus of stress-induced changes in plasticity and anxiety-like behavior following stress exposure (34–36). Given that NAc^SOM+ cells receive such strong input from the vHPC, we explored how NAc^SOM+ respond to stress using in vivo fiber photometry (Fig. 4). SOM-Cre mice were injected with a Cre-dependent virus that either expresses the calcium indicator, GCaMP7f [FLEX-GCaMP7f (AAV9-FLEX-hSyn-jGCaMP7f-WPRE)], or control, YFP [DIO-YFP (AAV5-Ef1a-DIO-eYFP)]; mice were then implanted with a fiber optic directly above the NAc to allow for somatic recordings of NAc^SOM+ activity (Fig. 4A). Following surgery recovery, mice were exposed to six tone (CS+)-shock (US) pairings (Fig. 4B). We observed that footshock significantly increased Ca²⁺ influx in SOM+ cells in the NAc in mice injected with GCaMP7f, but not YFP controls (Fig. 4 C–E). The shock-paired tone did not elicit an effect on fluorescent signal across learning (Fig. 4 G–I).

We confirmed that stress exposure more broadly can activate NAc^SOM+ by exposing mice to a 10-min restraint stress session; during active bouts of struggle behavior, we observed a significant increase in NAc^SOM+ Ca²⁺ activity (SI Appendix, Fig. S6 A–C). To validate that this was specific to stress, and not general movement, we also exposed mice to an acute predator odor stress (2MT) which induces significant freezing in mice (SI Appendix, Fig. S6D). 2MT significantly increased the frequency of Ca²⁺ transients only in mice injected with GCaMP7f, and not YFP control (SI Appendix, Fig. S6 E–G). Furthermore, movement-initiated increases in fluorescent signal did not significantly differ in GCaMP7f-injected mice compared to YFP controls, confirming that NAc^SOM+ activity is due to stress exposure and not movement (SI Appendix, Fig. S6 H–J). Together these data demonstrate that NAc^SOM+ cells respond to diverse modalities of stress with increases in activity.

Footshock Exposure Enhances NAc^SOM+ Excitability and Promotes Plasticity at vHPC-NAc^SOM+ Synapses.

Given that stress exposure increases NAc^SOM+ activity, we hypothesized that this may cause lasting changes in neural physiology. To assess how stress exposure alters NAc^SOM+ physiology, SOM-Ai14 mice were exposed to a single session of auditory fear conditioning, where six CS (tone)—US (shock) associations were administered. Control mice were exposed to just tone (Fig. 5A). Twenty four hours after conditioning, electrophysiology recordings were conducted. Mice exposed to footshock showed a significant increase in NAc^SOM+ excitability, specifically at higher postsynaptic current injections (300 pA) (Fig. 5 B–D). This increase in excitability was not associated with any significant changes in passive membrane properties, including membrane resistance, resting membrane potential, capacitance, or membrane time constant (SI Appendix, Fig. S7 A–D). We further assessed spontaneous synaptic transmission by assessing sEPSCs and sIPSCs following tone alone or paired with footshock. We observed a significant increase in sEPSC frequency and amplitude, with no change in sIPSC frequency or amplitude, consistent with previous literature demonstrating stress-effects on glutamatergic plasticity (SI Appendix, Fig. S7 E–J) (37, 38). Interestingly, this increase in glutamatergic signaling ultimately resulted in a significant increase in the E/I ratio, evident as an increase in the E/I frequency ratio.

Fig. 5.

Stress enhances NAc^SOM+ excitability and increases the ability of glutamatergic input to promote NAc^SOM+ neuron firing. (A) Schematic of experimental design. Mice were exposed to 6 tones (T) or 6 tone-shock pairings (S), and then 24 h later, electrophysiological recordings were conducted from NA^cSOM+neurons. (B) Current step injections and resulting AP firing frequency in tone (gray; N = 6) and shock (blue; N = 6) exposed mice. (C) Representative images of APs elicited at 300 pA in tone (black) and shock (blue) exposed mice; (scale bar represents 50 mV.) (D) Quantification of the number of APs at 300 pA current step injection. (E) Schematic of methodology to assess stress effects on vHPC-NA^cSOM+ circuitry. (F) Representative traces of vHPC-evoked oEPSCs after tone (black) and shock (blue); (scale bar represents 50 ms, 100 pA.) (G) Average oEPSC amplitude in tone- (N = 3) and shock- (N = 3) exposed mice. (H) vHPC-mediated AP probability following tone (gray) and shock (blue). (I) Average oEPSC PPR. (J) Schematic of methodology to assess BLA-NA^cSOM+ circuitry following stress. (K) Representative traces of BLA-evoked oEPSCs after tone (black) and shock (blue); (scale bar represents 50 ms, 100 pA.) (L) BLA-mediated oEPSC amplitude in tone- (N =4) and shock- (N = 3) exposed mice. (M) BLA-evoked AP probability. (N) Average BLA-mediated oEPSC PPR after shock. 2-Way ANOVA (B, H, and M) or unpaired Student’s t test (D, G, I, L, and N) were performed. n cells depicted above each graph from N mice as described above.

Given that cellular excitability and glutamatergic signaling were both significantly increased following footshock, we hypothesized that stress may drive increased activation of NAc^SOM+ by glutamatergic inputs. To directly test this, we probed circuit-specific effects of footshock by assessing AP probability following vHPC or BLA stimulation (Fig. 5 E and J). As before, the amplitude of vHPC or BLA-mediated oEPSCs was kept consistent between conditions to control for differences in virus expression (Fig. 5 G and L). Footshock promoted a significant increase in the ability of both vHPC and BLA stimulation to evoke NAc^SOM+ APs (Fig. 5 H and M), potentially due to an overall increase in postsynaptic SOM+ excitability. However, we found that presynaptic release probability was significantly decreased in a circuit-specific manner, evident by a significant increase in PPR exclusively at vHPC-NAc^SOM+ synapses, but not BLA-NAc^SOM+ (Fig. 5 I and N). These data suggest that while postsynaptic NAc^SOM+ excitability is enhanced to increase activation of NAc^SOM+, stress-induced circuit-specific synaptic plasticity drives input bias mediated via a reduction in release probability selectively at vHPC-NAc^SOM+ synapses.

We thus assessed net synaptic effects of shock exposure by exploring direct excitation and feedforward inhibition following vHPC or BLA activation (Fig. 6). In a separate experiment, direct oEPSC and feedforward oIPSC were recorded from the same cell following optogenetic stimulation of either vHPC or BLA input; oEPSC amplitude was again kept consistent between conditions (Fig. 6 A, B, G, and H). Twenty four hours after footshock exposure, we observed the characteristic increase in oEPSC PPR at vHPC-NAc^SOM+ synapses, which was absent at BLA-NAc^SOM+, confirming circuit-specific modification of presynaptic strength following acute stress (Fig. 6 C and I). Furthermore, footshock promoted a significant decrease in vHPC-mediated feedforward oIPSC amplitude (Fig. 6 D and J) which was absent at BLA-NAc^SOM+ synapses. This decrease in GABAergic transmission ultimately resulted in a significant increase in E/I ratio driven by vHPC inputs onto NAc^SOM+, with no changes at BLA-NAc^SOM+ synapses (Fig. 6 D–F and J–L and SI Appendix, Fig. S8).

Fig. 6.

Stress increases E/I balance at vHPC, but not BLA, input onto NAc^SOM+ neurons. (A) Schematic of methodology to assess vHPC-NA^cSOM+ changes following stress. Mice were exposed to tone or tone+footshock. The next day direct oEPSCs and feedforward oIPSCs were recorded from the same cell using a Cs⁺-gluconate-based internal solution. (B) vHPC-mediated oEPSC amplitude following tone (N =3) or footshock (N = 5). (C) Average vHPC-mediated oEPSC PPR. (D) Feedforward oIPSC amplitude following vHPC stimulation. (E) Representative trace of oEPSC and oIPSC from same cell from tone (black) and shock (blue)-exposed mice. (Scale bar represents 40 ms, 500 pA.) (F) Average E/I ratio. (G) Methodology to assess stress-induced changes in BLA-NA^cSOM+ circuitry. (H) BLA-mediated oEPSC amplitudefollowing tone (N = 4) or shock (N = 3). (I) Average oEPSC PPR following stress. (J) Feedforward oIPSC amplitude. (K) Representative traces of oEPSC and oIPSC from samecell in tone (black) and shock (blue)-exposed mice. (Scale bar represents 40 ms, 200 pA.) (L) Average E/I ratio. Unpaired t test used for all analysis. n of cells depicted above each graph from N mice.

Together, these data demonstrate that footshock increases NAc^SOM+ excitability, which seems to result in an enhanced activation of these cells irrespective of the source of excitatory input. However, shock exposure restructures vHPC, but not BLA, input via competing synaptic mechanisms: decreasing presynaptic release probability at glutamatergic synapses and robustly decreasing feedforward GABAergic inhibition, thus increasing net E/I balance.

Blocking CB1R Reverses Stress-Induced Plasticity within NAc^SOM+-Associated Circuits.

The competing synaptic and intrinsic excitability changes induced by stress parallel the effects observed after cannabinoid receptor activation. Specifically, both footshock exposure and incubation with CP55,940 enhance NAc^SOM+ excitability, decrease vHPC-NAc^SOM+ release probability and increase net E/I ratio, and ultimately increase NAc^SOM+ activation. We thus tested the hypothesis that recruitment of eCB signaling by stress mediates these synaptic and cellular adaptations. A separate cohort of mice were exposed to 6 footshock-tone pairings and NAc slices were obtained 24 h later, as before. Importantly, slices were incubated in either vehicle or the CB1R inverse agonist, rimonabant (10 µM) (Fig. 6A). Footshock exposure elicited an increase in NAc^SOM+ excitability, which was significantly restored back to control levels by pharmacological blockade of CB1R signaling with rimonabant (Fig. 6 B–D). We hypothesized that rimonabant-induced decreases in NAc^SOM+ excitability would be sufficient to reverse stress-induced increases in vHPC- and BLA-driven AP probability back to control levels. Indeed, following rimonabant incubation, stress-induced increases in vHPC- and BLA-elicited AP probability were significantly decreased back to levels of tone only-exposed mice, further supporting the notion that increased postsynaptic excitability enhances afferent-induced firing irrespective of input source.

Furthermore, we replicated that shock exposure promotes a significant increase in vHPC-NAc^SOM+ oEPSC PPR and found that this increase could also be reversed by rimonabant incubation back to levels seen following tone-only exposure (Fig. 6 E–H). Finally, the footshock-induced increase in vHPC-evoked E/I ratio was also significantly decreased following rimonabant incubation (Fig. 6 I–K). Together, these data demonstrate that blocking CB1R signaling acutely reverses stress-induced changes in NAc^SOM+ intrinsic excitability that promote vHPC-NAc^SOM+-driven firing of NAc^SOM+ neurons, as well as reversing circuit-level remodeling of vHPC-NAc^SOM+ synapses. These data support the hypothesis that stress recruits tonic eCB signaling to facilitate firing of NAc^SOM+ neurons and reorganization of afferent drive to favor increased excitation from vHPC, over BLA, afferents (Fig. 7).

Fig. 7.

CB1R blockade reverses stress-induced increase in excitability and changes in vHPC-NA^cSOM+ circuitry. (A) Schematic of experimental design. Mice were exposed to six tone-shock pairings or tone alone (control). The next day, NAc slices were obtained and electrophysiological recordings of NA^cSOM+ performed. Slices were incubated in vehicle (DMSO) or the CB1R inverse agonist, rimonabant (10 μM) to block CB1Rsignaling. (B) Quantification of AP firing frequency at different current step injections in mice exposed to tone (gray; N = 4), shock and vehicle-incubated (blue; N = 4) or rimonabant-incubated (orange; N = 4). (C) Representative traces at 300 pA current injection in tone (black), shock (blue), or shock+rimonabant (orange) slices; (scale bar represents 200 ms, 50 mV.) (D) Quantification of the number of APs fired at 300 pA current step injection. (E) Schematic of BLA-NA^cSOM+ approach and average BLA-mediated oEPSC amplitude (Tone: N = 3; Shock: N = 3; Shock+Rim: N = 3). (F) Average AP probability at different frequencies following optogenetic BLA activation (Left) and data again with individual values (Right). (G) Schematic of vHPC-NA^cSOM+ approach and average vHPC-mediated oEPSC amplitude (Tone: N = 5; Shock: N = 3; Shock+Rim: N = 3). (H) Average vHPC-mediated AP probability at different frequencies (Left) and data again showing individual values (Right). (I) Representative trace of oEPSC PPR following tone exposure, shock exposure, or shock exposure + rimonabant incubation before AP probability recordings using K+-gluconate internal. (Scale bar represents 50 ms, 200 pA.) (J) Average oEPSC PPR from tone-exposed (N = 5), shock-exposed (N = 4), or shock-exposed +rimonabant incubation (N = 4) using K+-gluconate internal. (K) Representative traces of oEPSC and oIPSC from the same cell with Cs+-gluconate internal. (Scale bar represents 40 ms, 200 pA.) (L) Average oEPSC amplitude for E/I experiments (Tone: N = 4; Shock: N = 3; Shock+Rim: N =3). (M) Feedforward oIPSC amplitude following treatments. (N) Net vHPC-mediated E/I ratio. 2-Way ANOVA (B, F, and H) or one-way ANOVA (D, E, G, J, and L–N) were performed for analysis. n of cells depicted above each graph from N mice.

Discussion

The NAc is a central limbic node contributing critically to goal-directed behavior, reinforcement learning and stress-adaptation, and dysregulation of NAc function has been implicated in a variety of neuropsychiatric disorders ranging from substance use to major depression (9, 39). Excitatory afferent drive to the NAc supports goal-directed behavior, which is sculpted by local GABAergic inhibition and neuromodulators including dopamine, opioids, and eCBs (40). Recent characterization of different NAc interneuron subtypes has revealed a role for NAc^SOM+ neurons in potentiating locomotor and rewarding effects of cocaine suggesting these interneurons may play an important role in the regulation of NAc function (11), and SOM+ neurons in the dorsal striatum have been demonstrated to be a major target of chronic stress (41). While eCBs are known to regulate excitatory drive to NAc MSNs and fast-spiking interneurons, as well as feedforward GABA release from local parvalbumin-expressing interneurons onto MSNs, the role of eCBs in the regulation of NAc^SOM+ neuron function is not known (8, 15, 19, 20). Here, we reveal a role for eCB signaling in regulating the balance between vHPC and BLA activation of NAc^SOM+ neurons through multiple mechanisms including direct presynaptic inhibition, increases in cellular excitability, and differential effects on E/I balance that ultimately mediate changes in AP probability triggered by vHPC and BLA afferents. Importantly, acute stress exposure activates NAc^SOM+ neurons and results in reorganization of glutamatergic afferents and intrinsic excitability of NAc^SOM+ neurons via recruitment of these eCB signaling mechanisms. These data reveal mechanisms regulating NAc^SOM+ neuron activation by distinct long-range afferents and point to recruitment of eCB signaling as a key mechanism linking acute stress exposure to reorganization of NAc^SOM+-associated neural circuits. These findings also provide synaptic and cellular insight into how eCB signaling may regulate goal-directed behavior and stress adaptation relevant to substance use disorders and stress-related disorders.

First, we show that the eCB system differentially regulates glutamatergic and GABAergic input onto NAc^SOM+ cells in a circuit-specific manner. Previous literature has reported strong afferent projections from the vHPC, BLA, and PFC onto MSNs in the NAc (15); our data here reveal that NAc^SOM+ interneurons receive similar inputs. This also supports functional connectivity between these brain regions and SOM+ interneurons, further validating previous tracing data (22, 42). We reveal that functional cannabinoid receptors are present at vHPC and BLA, but not PFC afferents onto NAc^SOM+ interneurons, supporting the notion that the eCB system controls glutamatergic input in a circuit-specific manner (15). Cannabinoid receptor activation ultimately shifts the E/I balance onto NAc^SOM+ toward enhanced excitation by vHPC inputs, but enhanced inhibition by BLA inputs. Specifically, eCB signaling significantly augments the ability of vHPC, but not BLA, input to elicit NAc^SOM+ firing. This effect is the net result of several, sometimes opposing, signaling mechanisms including presynaptic inhibition of glutamate release, changes in the magnitude of feedforward inhibition, and increases in postsynaptic excitability. Indeed, E/I balance is shifted toward excitation for vHPC inputs by eCB signaling mediated by a relatively greater reduction in feedforward GABAergic transmission onto NAc^SOM+ neurons, coupled with increased postsynaptic excitability, ultimately biasing toward activation of NAc^SOM+ neurons by vHPC inputs. In contrast, eCB signaling shifts the E/I balance toward inhibition upon BLA stimulation. It is possible that vHPC afferents recruit a greater number of fast-spiking interneurons that are known to express CB1R to mediate feedforward inhibition onto NAc^SOM+ neurons than BLA inputs, thus explaining suppression of feedforward inhibition upon cannabinoid receptor activation only upon vHPC activation (4, 20). This hypothesis remains to be tested experimentally. Overall, since glutamatergic inputs to the NAc have been suggested to provide information about environmental context and cues to guide goal-directed behavior, shifting the strength of glutamatergic inputs onto NAc^SOM+ cells could modulate the relative contribution, or importance, of vHPC information in orchestrating behavior (43–45).

We further report an eCB-dependent phenomenon whereby activation of CB1R increases NAc^SOM+ excitability. While the molecular mechanism subserving this effect is unknown, one possibility involves direct or indirect effects of postsynaptic CB1R on ion channel conductance; indeed, CB1R is expressed postsynaptically in other NAc cell types (31, 32). Given that pharmacological cannabinoid receptor activation reduces the ½ width of APs fired, this enhanced firing frequency may rely on inhibition of A-type potassium channels; in support of this hypothesis, 2-AG has been previously reported to inhibit A-type potassium channels, increasing AP firing rates (46). This enhanced intrinsic excitability was most evident at large depolarizing steps (300 pA), suggesting that changes in depolarization-induced inhibition, which has been shown to rely on voltage-gated sodium channels (VGSC), could be mediating changes in excitability; indeed, AEA has been previously shown to inhibit VGSCs, offering another mechanism by which eCB/CB1R signaling alters intrinsic excitability (47). Finally, it remains possible that these effects involve CB1R-dependent regulation of other neuromodulatory systems, such as dopamine (48, 49). For example, augmentation of eCB signaling, and subsequent CB1R activity, have been demonstrated to increase AP firing elicited by low doses of D1/D2 agonists (48). One important consideration regarding these changes in excitability is the variability in AP firing and subsequent depolarization-induced inhibition; the NAc is an incredibly heterogeneous region and inherent differences in excitability in the core verse shell could underlie the heterogeneity in AP firing. Future studies will be required to explore how anatomical differences in NAc^SOM+ cells could alter basal and eCB-dependent changes in excitability and the mechanism mediating this effect.

Using in vivo fiber photometry, we reveal that NAc^SOM+ neurons respond to stress. Distinct populations of MSNs (D1-MSNs vs. D2-MSNs) have been demonstrated to differentially encode positive verse negative valence (50–53). Yet, others have shown that both cell types can convey information related positive and negative valence stimuli depending on their stimulation patterns and downstream targets (54, 55). It remains to be determined whether NAc^SOM+ also show bivalent responses to positive verse negative stimuli, or preferentially respond to aversive stimuli. Data thus far suggest that NAc^SOM+ interneurons may have opposing responses to reinforcing verse aversive experiences. For example, Ribeiro, et al. have shown that repeated exposure to a reinforcer, cocaine, can decrease the excitability of these cells, while we demonstrate here that an acute footshock session can increase the excitability of these cells (11). Ultimately, future work will need to characterize the role of NAc^SOM+ in encoding valence. Finally, while we did not observe a strong effect of sex on basal neurotransmission or stress-induced effects on excitability, our in vivo fiber photometry recordings here were done in exclusively male mice. Thus, it remains possible that NAc^SOM+ activity and regulation could show sexual dimorphism, but this will have to be explicitly tested in the future.

Importantly, we also show that an acute footshock exposure enhances NAc^SOM+ excitability, which can augment the ability of excitatory inputs to induce NAc^SOM+ activity, evident as an increase in vHPC- and BLA-evoked AP probability. As mentioned above, this increased excitability may be related to depolarization-induced inhibition when NAc^SOM+ cells begin to fire at higher frequencies. Indeed, the increase in AP probability is still present at high-frequency optogenetic stimulation of BLA input (20 Hz), suggesting that the increase in intrinsic excitability (decrease in depolarization block) could play a physiological role in modulating NAc^SOM+ excitability and subsequent excitation by glutamatergic drive. Furthermore, while the enhancement of intrinsic excitability follows activation by both vHPC and BLA inputs, we observed a decrease in glutamatergic transmission (increase in PPR), yet increased net E/I balance exclusively at vHPC-NAc^SOM+ synapses, suggesting that stress exposure may bias toward net excitation by vHPC input. These effects parallel results seen following pharmacological cannabinoid receptor activation with CP55,940 (increased NAc^SOM+ excitability, vHPC-NAc^SOM+ PPR, and net E/I balance). In support of this parallel, these stress-induced reorganizations in NAc^SOM+-associated circuitry were completely reversed by CB1R blockade, indicating that stress-induced circuit reorganization, including changes in vHPC release probability, E/I balance, and postsynaptic excitability, are mediated via recruitment of eCB signaling mechanisms.

Augmentation of eCB/CB1R activity within the NAc prevents the induction of anxiogenesis following chronic social defeat stress, suggesting eCB plasticity observed here could serve an adaptive role aimed at counteracting the adverse effects of stress (14). In support of this hypothesis, chronic stress exposure increases vHPC-NAc PPR (reduces release probability) onto MSNs exclusively in stress-exposed mice that do not develop maladaptive coping behaviors, suggesting that synaptic reorganization of this circuit may contribute to stress resiliency (36). Additionally, enhanced glutamatergic signaling has been associated with stress susceptible phenotypes, including generalized anxiety- and depressive-like states (36, 38). It remains possible that eCB signaling the NAc can produce bidirectional, and opposing, effects on behavior depending on the cell-types or circuits activated (56–59), suggesting further studies will be required to determine the function of stress-induced plasticity within NAc^SOM+ -associated circuits.

While we demonstrate eCB regulation and stress-dependent plasticity of NAc^SOM+, one major question that remains is the role of these cells in guiding NAc output. Interneuron populations canonically regulate MSN activity via feedforward inhibition (4, 20). Interestingly, one study found that vHPC-elicited feedforward inhibition is driven primarily by parvalbumin-expressing interneurons, suggesting that this phenomenon does not significantly depend on SOM+ activity (4). Rather, SOM/NPY+ cells may modulate behavior via direct projections to the lateral hypothalamus (LH) (42). The LH plays a critical role in mediating innate avoidance and maladaptive anxiety-like behavior following stress (60, 61). NAc-LH circuitry has been demonstrated to play a role in the motivation for natural rewards, including food, which can be modulated by stress exposure (62–64). Ultimately, future studies will be required to characterize the role of NAc^SOM+ neurons in modulating NAc function, either via feedforward inhibition onto MSNs, or directly via projections to the LH.

In conclusion, here we elucidate synaptic and cellular eCB mechanisms regulating NAc^SOM+ neuron activity and define a key role for eCB signaling in biasing activation of NAc^SOM+ neurons in response to vHPC, over BLA, afferent stimulation. We find that NAc^SOM+ neurons are activated by diverse stressors and show persistent adaptations favoring activation by glutamatergic inputs, but ultimately biasing toward net excitation via vHPC drive. These adaptations were found to be dependent on eCB signaling and thus reveal neuromodulatory mechanisms capable of sculpting afferent drive to a relatively understudied, but potentially important, class of NAc neurons that express SOM+. These data also reveal mechanisms by which eCB signaling and exogenous cannabinoids could regulate NAc function. Specifically, since vHPC afferents to the NAc support instrumental responding and place preference (27), cannabinoid receptor–induced bias toward vHPC activation of NAc^SOM+ neurons could represent a synaptic substrate for the reinforcing effects of cannabis, especially following stress exposure.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.