Cannabidiol abrogates cue-induced anxiety associated with normalization of mitochondria-specific transcripts and linoleic acid in the nucleus accumbens shell

Abstract

Anxiety disorders are one of the top contributors to psychiatric burden worldwide. Recent years have seen a dramatic rise in the potential anxiolytic properties ascribed to cannabidiol (CBD), a non-intoxicating constituent of the Cannabis Sativa plant. This has led to several clinical trials underway to examine the therapeutic potential of CBD for anxiety disorders. Yet, CBD’s anxiolytic effects are mixed with some studies reporting little to no impact on trait anxiety but significant reductions in pathological anxiety with suggestions that CBD’s effect may relate to triggered or cue-induced behavior. Here, we studied the effects of CBD on cued and non-cued behaviors and related neurobiological underpinnings. To investigate the effect of CBD on cue-induced anxiety, male rats underwent a fear conditioning protocol (odor associated with shock) followed by assessments of avoidance behavior. CBD (10 mg/kg) was administered 1hr prior to anxiety assessments. To understand molecular mechanisms associated with behavior, we investigated the transcriptome and lipid profile of the nucleus accumbens shell (NAcSh), a structure implicated in cue-mediated behaviors and aversion. Administration of CBD significantly reduced avoidance behavior, but only in animals repeatedly exposed to a shock-paired cue. CBD did not affect behavior in animals exposed to neutral cue or encoding of the cue behavioral response. RNA sequencing revealed substantial impact of the shock-paired cue in control animals, recruiting mechanisms ranging from cytoskeletal dynamics to mitochondria dysfunction. The shock-paired cue also resulted in elevated linoleic acid in vehicle animals which correlated with anxiety-like behavior. CBD either reversed or normalized these cue-induced molecular phenotypes. CBD also recruited lipid networks which correlated with transcripts involved in synaptic plasticity, signaling, and epigenetic mechanisms. These results suggest that CBD may specifically alleviate salient, conditioned anxiety and normalize related biological mechanisms in the NAcSh which may guide therapeutic interventions for anxiety disorders.

Introduction

Cannabidiol (CBD) is the second most prominent phytocannabinoid of the Cannabis Sativa plant. Approved for the treatment of intractable epilepsy conditions such as Dravet and Lennox-Gastaut syndrome^{1, 2}, there has been a boom of interest in the therapeutic potential of CBD for a variety of conditions ranging from pain to psychopathologies due to its low reinforcement properties^3–5 and limited side-effect profile^{1, 2, 6–8}. However, despite strong marketing and public interest in CBD products, there is considerable variability in reports as to its efficacy in treating conditions such as anxiety and fear. While some positive clinical effects have been observed in individuals with anxiety disorders, results are mixed for healthy individuals^{7, 9, 10}. These inconsistent outcomes are echoed by studies from animal models where CBD has been shown to be either anxiolytic, ineffective, or only effective after stress^11–15. Similarly, in studies relevant to substance use disorder, CBD’s effect on behavior is nuanced. In both animal models and placebo-controlled blinded studies, CBD reduced cue-induced seeking or craving for drug-paired cues^16–20, but has mixed effects on self-administration, extinction, or associative learning for drug contexts^{18, 21–25}. These results indicate that CBD may specifically impact cue-induced phenotypes which have common substrates involved in anxiety and addiction ²⁶. Determining CBD’s effects on cue-related behavior will help to elucidate its role in these learning-related processes and the mechanisms underlying these effects.

CBD has multiple mechanisms of action and impacts various receptor systems including GPR55, cannabinoid, serotonin, and peroxisome proliferator-activated receptor (PPAR)γ among others²⁷. Pharmacology studies show that CBD also mediates mechanisms that affect bioactive lipids including inhibiting fatty acid amide hydrolase (FAAH, the enzyme responsible for catabolism of anandamide (AEA) and other N-acylethanolamines^{28, 29}; but see also³⁰), fatty acid binding proteins³⁰, and activating N-acyl phosphatidyl ethanolamine-specific phospholipase D³¹. Endocannabinoid-related bioactive lipids are part of a broader, interconnected network and influence multiple physiological processes beyond the cannabinoid-1 receptor^{32, 33}. Given that CBD has known effects on multiple lipid species^{29, 31}, anxiogenic conditions may perturb this network and influence CBD’s molecular footprint.

In this study, we investigated the effects of CBD leveraging a rodent model of cue-induced anxiety along with neurobiological correlates. Our results revealed that CBD had lasting effects on anxiety but only in the presence of a salient shock-paired cue following multiple fear-conditioning sessions. Molecular assessment of mesocorticolimbic structures highlighted the nucleus accumbens shell (NAcSh), a region implicated in aversion learning³⁴. Both the cue and CBD induced robust effects on NAcSh mRNA expression of known CBD targets. RNA sequencing of the NAcSh revealed that CBD not only normalized behavior but also the transcriptional perturbations precipitated by the cue. In addition to transcriptional changes, the shock-paired cue increased NAcSh levels of linoleic acid in the vehicle animals which positively associated with avoidance-like behavior. CBD also significantly altered NAcSh linoleic acid and multiple other species, altering lipid networks. These results indicate that CBD is anxiolytic in very salient emotive contexts and likely has multiple molecular mechanisms through which it impacts anxiety-like behavior.

Materials and Methods

Detailed information for all methods and results is provided in the Supplemental Information and an overview of the experiments is given in Figure 1. In Experiment 1, animals underwent the fear-conditioning protocol followed by light/dark and open field testing with CBD/vehicle given 1hr prior to the anxiety tests. For Experiment 2, CBD/vehicle was given 1hr prior to the fear conditioning protocol and animals were then tested for anxiety-like behaviors drug-free 1 day (light/dark) and 1 week (open field) after the final shock session.

Figure 1.

Experimental overview. Animals in Experiment 1 underwent fear conditioning with or without a lemon odor cue. 24 hours later, vehicle or 10 mg/kg CBD (i.p.) was given 1hr prior to the light/dark test with the cue in the present in the box. Open field was tested 1 week after the shock session with the cue situated in the center of the field. These same animals then underwent three additional fear conditioning sessions and repeated the anxiety tests with the final open field completed two weeks after the final shock. Rats were sacrificed one hour after the final open field test. In Experiment 2, rats underwent three consecutive shock sessions with or without the cue after being administered vehicle or CBD. Anxiety was tested in the light/dark box and open field 24 hrs and 1 week after the final shock session, respectively.

Animals

Male Long-Evans rats (n=48/experiment) received at postnatal day 55–60 were pair-housed under a reversed 12hr dark/light cycle (lights off at 7 am) in a temperature and humidity-controlled vivarium. Animals acclimated to the facility for 1 week and were handled 2–3 times prior to the start of the experiment. Standard rat chow and water were available ad libitum throughout the experiment.

Ethics Approval

Testing and housing were in accordance with IACUC approved protocols (Protocol # LA-12–00316).

Cue-Induced Anxiety Paradigm

All rats underwent fear conditioning adapted from a previously published paradigm¹⁵. In a six-minute conditioning session, animals were exposed to a 0.8 mA footshock once per minute. In the cued condition, a 5 ml open Eppendorf tube filled with a gauze treated with 200 µl of pure lemon essential oil (Healing Solutions Essential Oils, product- HSL2561O) and was adhered to the ceiling of the box. In the neutral condition, an open tube with a plain gauze (no oil) was present. Following the shock sessions, animals were returned to their home cages for 24hrs after which avoidance behavior testing took place.

Avoidance-like behaviors were assessed using the light/dark box and open field paradigms. In the light/dark box assay, the open-field apparatus (40”x40”x40” Plexiglas box; MedAssociates) with infrared beams was fitted with a black box that covered half of the arena. A light (500–550 lux) was positioned at the exit of the dark box to create an anxiogenic environment. During a 10-minute session, rats were initially placed into the dark compartment, and movement within the light and dark sides of the box was monitored using the infrared beams. Video cameras were also situated above the boxes to record latency to emerge into the light (all four paws in the light field of the box). In open field sessions, rats were placed in the field without the light/dark apparatus to track movement for 10-min. To determine whether the shock-paired cue influenced behavior in either paradigm, the Eppendorf tube containing either the plain gauze (neutral) or lemon oil gauze (cue) was placed on the far wall of the light side of the light/dark box or in the center of the open field. The cue used in the behavioral paradigms was the same each animal experienced in the shock session(s). This placement was used to determine whether CBD influenced both general anxiety metrics as well as approach to the cue itself in the most “anxiogenic” sections of the apparatus. In Experiment 1, rats first underwent a single shock session followed by light/dark and open field testing 24hrs and 1 week after the shock, respectively. These rats were then exposed to three successive shock sessions followed by light/dark testing 24hrs and two weeks after the final shock session. In Experiment 2, the light/dark and open field tests were completed 24hrs and one week after the final shock. Latency to enter the light compartment was recorded for the light/dark test, and number of entries and time spent in the different zones (full field, center, and outer edges for open field test, and dark/light zones) were calculated using MEDPC activity monitoring software and analyzed using RStudio.

Drug preparation and administration

CBD (RTI International) was dissolved in 100% ethanol, subsequently evaporated under nitrogen gas, and suspended in 5% cremophor EL (Calbiochem, catalog-238470–1SET) and 0.9% saline solution. Animals received by intraperitoneal injection 1ml/kg of vehicle (5% cremophor in saline) or 10 mg/kg CBD 1hr prior to anxiety assays (Experiment 1) or shock sessions (Experiment 2). This dose was chosen based on previous experiments in which CBD blocked cue-induced drug seeking¹⁸ and from pilot experiments showing maximal effect of CBD at this dose in this rat strain (data not shown).

Ex vivo

Animals were sacrificed 1hr after the final open field session (2hrs after the final CBD/vehicle injection and two weeks after the final shock for Experiment 1, one week after the final shock for Experiment 2). Animals were briefly anesthetized with CO₂, decapitated, brains extracted, and flash frozen in ice-cold isopentane. Trunk blood was collected into EDTA coated tubes and plasma separated via centrifugation (spun at 4ºC at 2000G for 15 minutes). Plasma and tissue were stored at −80ºC until analysis. Brain tissue was equilibrated to −20ºC and micropunches of medial prefrontal cortex, nucleus accumbens core and shell (NAcSh), and amygdalae nuclei were taken for subsequent analyses.

RNA extraction, qPCR, sequencing

RNA was extracted and purified using commercially available kits. cDNA libraries were synthesized and used for real-time qPCR to assess the expression of known CBD targets (Suppl.Table1). Following the qPCR screen which revealed recruitment of the NAcSh (see results), remaining purified RNA was used for next-generation sequencing.

Transcriptome Analysis

Transcriptome analysis was completed as described previously^{35, 36}. Differential gene expression was examined with cue and drug group as between subjects variables (effect of cue group, effect of drug group, interaction of cue*drug group) for the mixed linear model using the DESeq2 package in R³⁷. Differentially expressed genes (DEGs) were defined as a nominal p<0.05 (Suppl.Table2-4). Functional annotation analysis was completed using the Enrichr database³⁸ using a False Discovery Rate (FDR) adjusted-q<0.05 and the Ingenuity Pathway Analysis platform using a Benjamini-Hochberg corrected p<0.05. To determine the effects of CBD and cue exposure on broader gene networks, we used the bioinformatics software MEGENA to create gene co-expression modules³⁹. After modules were constructed, Fisher’s exact tests identified modules significantly enriched with DEGs used for eigengene analyses. Spearman correlations were then used to associate eigengenes with behavioral traits and lipid fold-changes using the moduleEigengenes function in the WGCNA R package⁴⁰. Significant modules underwent gene ontology annotation using Enrichr using a False Discovery Rate (FDR) adjusted-q<0.05.

Lipidomics analysis

Lipid extraction and analysis via HPLC tandem mass spectrometry (MS/MS) was performed as previously described⁴¹ and in greater detail in Supplemental Data. In brief, methanolic extracts of the NAcSh underwent sonication, incubation, and centrifugation prior to partial purification via C18 solid phase extraction. Elutions of 100 percent methanol were analyzed via targeted HPLC/MS/MS to determine amounts of endogenous lipids per sample. Values of each lipid were determined by standard curve and expressed as moles per gram of tissue.

Data analyses

Behavioral, qPCR, and lipid analyses were conducted using GraphPad Prism (Version 10.1.1). All behavioral and qPCR analyses used two-way analysis of variance (ANOVA) and included cue (two levels- neutral, cue) and drug (two levels- vehicle, CBD) as between-subjects factors. Significant findings were followed up using post-hoc one-way ANOVAs or t-tests, with Tukey and Holm-Sidak corrections applied, respectively. Results were deemed to be significant if a p-value was less than an alpha of 0.05. Effect sizes are reported for main significant effects.

Data availability

Sequencing data are available on GEO under accession code GSE279163. All other relevant data supporting the findings of this study are available within the article and its Supplementary Materials or from the corresponding author upon reasonable request.

Results

CBD blocks cue-induced anxiety after repeated shock experience

To determine the effect of CBD on both non-cued and cued shock-induced anxiety, in Experiment 1, rats underwent a single shock session followed by the light/dark and open field tests 24hr and 1 week after the shock, respectively. The presence of the cue in vehicle animals significantly increased the latency to enter the light side of the chamber (Suppl.Fig1A; cue-F(1,38)=16.16, p=0.0003), reduced the number of entries (Suppl.Fig1B; cue-F(1,44)=11.55, p=0.0014) and time spent in the center of the open field (cue-F(1,42)=12.90, p=0.0009) as well as increased the overall anxiety score (Suppl.Fig1C; cue-F(1,44)=8.967, p=0.0045; d=0.87). CBD did not impact these outcomes in either cue group (drug, drug*cue- Fs<2.712, NS). Given that CBD’s therapeutic potential may be specific to pathological anxiety⁹, we hypothesized that a well-conditioned anxiety response may increase sensitivity to CBD administration. Thus, the same animals next underwent three consecutive shock sessions with the neutral or cue present followed by the same anxiety test protocol. Once again, the cue significantly increased anxiety-like behaviors, but there was a significant interaction between the cue and CBD administration for the latency to enter the light side of the box (Fig2; cue*drug-F(1,42)=9.270, p=0.004), number and duration of entries into center of the open field (entries-cue*drug, F(1,44)=5.887, p=0.0194; duration-cue*group-F(1,44)=4.540, p-0.039), and the total anxiety score (cue*group-F(1,7)=11.92, p=0.017). CBD significantly reversed the increase in avoidance-like behavior seen in the vehicle cue rats specifically for the latency to enter the light and the overall anxiety score (Fig2a,c). CBD had no effect in the neutral cue condition (veh-neutral vs CBD-neutral q’s<2.231, p>0.401).

Figure 2.

CBD reversed cue-induced anxiety in the light/dark test (A) and open field (B) after multiple cue-shock pairings. Twenty-four hours after the last conditioning session, the cue elicited anxiety in the light/dark box as measured by increased latency to enter the light side of the box (A; cue*drug-F(1,42)=9.270, p=0.004; veh-neutral vs veh-cue, q(42)=5.885, p=0.0009; veh-cue vs CBD-cue, q(42)=5.794, p=0.001; CBD-neutral vs veh-cue, q(42)=5.460, p=0.0021), fewer entries made into the light side of the box, and less time spent on the light side (A; entries-cue*drug, F(1,44)=5.887, p=0.0194; duration-cue*group-F(1,44)=4.540, p-0.039). Two weeks later, the cue continued to increased avoidance behavior in the open field, with vehicle cue animals showing fewer entries and time spent in the center of the field (B). Vehicle cue rats exhibited an increased composite anxiety Z score which was reversed by CBD (C; cue*group-F(1,7)=11.92, p= 0.017; veh-neutral vs veh-cue, q(7)=5.702, p=0.02, d=1.63; veh-cue vs CBD-cue, q(7)=5.196, p=0.031, d=1.45; CBD-neutral vs veh-cue- q(7)=6.061, p=0.015, d=1.66). Bars and error bars represent means +/− the standard error of the mean. *p<0.05, **p<0.01, ***p<0.001.

To determine whether CBD blocked the encoding of the cue, a second cohort of animals in Experiment 2 underwent the shock paradigm with or without the presence of the cue and were administered CBD 1hr prior to each shock session. The same anxiety paradigms were conducted as above. Although the cue once again induced avoidance behaviors, CBD did not affect behavior in either the cued or neutral condition. (supplementary results, Supp.Fig.2). These data demonstrate that CBD effectively reduced cue induced anxiety-like behavior, but specifically after repeated cue-shock pairings without influencing encoding of the cue.

CBD reverses alterations of the nucleus accumbens shell transcriptome

CBD has multiple targets through which it may influence behavior²⁷. We first used a qPCR screen of select known CBD targets and cFos as a proxy for cell activation in several subregions involved in the expression of anxiety and cue-mediated behaviors including the medial prefrontal cortex, nucleus accumbens, and amygdala^{42, 43}. Although the shock-cue reduced expression of Gpr55 in the PrL and increased Fos in the NAcC, and CBD increased Cnr1 expression in the NAcC (supplementary results; Supp.Fig3-5), only in the NAcSh was there an effect of cue or cue-drug interaction for every candidate gene screened (Fig3A). This included reduced Cnr1 (CB1 receptor; cue*drug-F(1,34)=11.25, p=0.002), Faah (cue*drug- F(1,43)=3.584, p=0.0651; cue-(1,43)=8.628, p=0.0053; drug-F(1,43)=3.940, p=0.053), Gpr55 (GPR55 receptor; cue-F(1,34)=5.373. p=0.027), Htr1a (5-HT-1a receptor; cue*drug-F(1,39)=6.832, p=0.013), Gria1 (AMPA subunit 1; cue-F(1,42)=6.886, p=0.012), and Fos (cFos; cue-F(1,42)=6.886, p=0.012). These data suggested the NAcSh is a strong target for CBD’s impact on cue-induced anxiety.

Figure 3.

The anxiety-evoking cue affected mRNA expression of multiple CBD targets in the nucleus accumbens shell (NAcSh; A) and the NAcSh transcriptome (B-E). In the NAcSh, Cnr1, Faah, Gpr55, Htr1a, Gria1, and Fos were all significantly impacted by the cue (A). CBD significantly reversed the downregulation of Cnr1 and upregulation of Faah with similar effect, but not significant, for Gpr55 and Htr1a. RNAseq of the NAcSh revealed strong perturbation of the transcriptome in vehicle cue animals (B). Specific enrichment was observed for changes in synaptic plasticity related transcripts, epigenetic factors, and mitochondrial processes. In CBD animals, the effect of the salient cue was markedly diminished, with enrichment observed for LXR, prostaglandin biosynthesis, H3K27me3, and transcription factors (C). Examining the DEGs in the cue*drug interaction analysis, 1118 DEGs overlapped with significant effects of cue in the vehicle animals (D). These transcripts were almost entirely reversed in CBD animals (D, bottom panel, grey and orange genes are the same genes observed in vehicle cue animals). Ingenuity Pathway Analysis revealed the genes affected by the cue again related to plasticity, synaptic signaling, and mitochondrial dysfunction with predicted recruitment of levodopa, ACOX1, PPAR-α, RHO, and others (E). CBD reversed all of the effects observed in the vehicle cue animals. qPCR bar graphs represent means +/− the SEM.*p<0.05, **p<0.01.

To deeply interrogate effects of cue and CBD on the transcriptome, we sequenced RNA from the NAcSh. The cue resulted in 2045 differentially expressed genes (DEGs) in the vehicle group but only 432 in the CBD rats (Fig3B-C). In vehicle cue animals, upregulated transcripts enriched for pathways related to neuronal morphology and cytoskeleton as well as post-synaptic density and trans-synaptic signaling. These transcripts also enriched for the epigenetic acetylation of lysine 27 on histone H3 (H3k27ac). The cue resulted in downregulation of transcripts related to PPAR-γ activity, extracellular matrix organization, fatty acid metabolism, and mitochondria ATP synthesis and oxidative phosphorylation. In contrast, very few ontologies were enriched in the CBD cue animals suggesting the cue did not induce marked alterations in these animals. Upregulated transcripts in CBD cue rats enriched for prostaglandin synthesis and regulation, liver X receptor (LXR), a nuclear receptor involved in fatty acid homeostasis, and extracellular matrix genes. Downregulated transcripts enriched for insulin secretion, transcription factor NFE2L2 which regulates cellular response to oxidants, Suz12 (a component of the PRC2 complex that methylates histone 3), and H3K27me3. These results show a distinctly stronger impact of the cue in vehicle animals not observed in CBD rats, recruiting several plasticity, epigenetic, fatty acid, and mitochondria-related processes.

To determine the specific interaction between effects of the cue and drug, we extracted DEGs from the drug*cue interaction DESeq2 analysis. There were a total of 1868 genes in the interaction term, and 1118 were significantly impacted by cue exposure in vehicle animals (Fig3D). In the CBD group, only 60 of these transcripts had a significant effect of cue, the majority of which showed opposite expression compared to veh-cue rats (i.e., upregulated genes in veh-cue animals were down regulated in CBD-cue rats; Fig3D). The remaining 1058 genes significantly impacted by the cue in the vehicle group were not significant in CBD rats, suggesting that CBD blocked changes in, or normalized, expression of these transcripts precipitated by the cue in vehicle animals. Top DEGs reversed by CBD (Suppl.Fig6) included matrix metallopeptidase (Mmp12; vehicle, log2 fold change (log2FC): −6.405, p<0.0001; CBD, log2FC=3.78, p=0.019), ATPase proton transporting subunit Atp6v0d2 (vehicle, log2FC=−4.942, p=0.0003; CBD, log2FC=4.113, p=0.005), nerve growth factor receptor Ngfr (vehicle, log2FC=−2.596, p=0.006; CBD, log2FC=2.981, p=0.0024), and Annexin 1 Anxa1 (vehicle, log2FC=−2.010, p=0.01; CBD, log2FC =2.307, p=0.004). Other transcripts found to be significantly impacted by the cue in vehicle rats but normalized or significantly expressed in the opposite direction in CBD animals involved in the extracellular matrix, phospholipid binding, enzymes, transporters, and epigenetic factors (Suppl.Fig.6). These results indicate that CBD may be normalizing diverse mechanisms that regulate synaptic plasticity and signaling as well as phospholipid dynamics. We then used Ingenuity Pathway Analysis to determine enrichment for biological processes and whether these processes were activated or inhibited by the cue or CBD. In vehicle animals, there was significant predicted activation of mitochondrial dysfunction as well as signaling pathways including calcium, opioid, estrogen receptor, synaptogenesis, cholecystokinin, G-protein beta gamma, corticotropin releasing hormone, but inhibition of S100 family signaling (Fig3E). The opposite effect was observed for all pathways in CBD cue animals. In evaluating putative upstream regulators of these processes (Fig3E), of the top fifteen significantly enriched regulators eleven were predicted to be inhibited in vehicle cue animals. These related to modulation of transcription (beta-estradiol, p38 MAPK, CD38, SREBF1, PPAR-α), plasticity (PI3K complex, PDGF, VEGFA), and angiogenesis (PDGF, EGF). Interaction term genes also predicted activation of levodopa, ACOX1, and RHO in vehicle cue animals, which regulate dopamine synthesis, lipid metabolism, and cytoskeleton plasticity, respectively. In CBD animals, these upstream regulators were collectively predicted to be in the opposite direction (Fig3E). Secondary analysis using Enrichr confirmed several of these results (Suppl.Fig7). These data suggest that the cue recruited several mechanisms related to mitochondria, synaptic plasticity and signaling, cellular activity and metabolism in the NAcSh. Normalization of these mechanisms by CBD suggests diverse mechanisms through which it may mediate its effects.

Concurrent CBD and cue experience alters lipid networks in the NAcSh

It is well known that CBD modulates several bioactive lipids including the canonical endocannabinoids and their congeners^{29, 31, 44}. However, little is known as to whether behavioral context influences changes in lipid profiles. Therefore, we analyzed select lipids in plasma and several bioactive species in the NAcSh from the rats in Experiment 1 (supplementary methods)^{31, 32, 41}. Administration of CBD in both cue conditions produced similar plasma and brain concentrations of CBD (Suppl.Fig8A). CBD also increased plasma endocannabinoids AEA and 2-AG as well AEA congener OEA and arachidonic acid (Suppl.Fig.8B) regardless of cue group. In the NAcSh, there was no significant effect of CBD for the endocannabinoids 2-AG and AEA (Fig4A). CBD administration did, however, increase levels of several other lipids (Fig4A; Supplementary Results, Suppl.Fig9) including arachidonic acid, SEA and OEA (congeners of AEA), 2-LG and 2-OG (congeners of 2-AG), P-alanine, and p-glycine. Despite broad effects of CBD administration on these lipids in the NAcSh, only linoleic acid showed an interaction between drug and cue (F(1,36)=4.865, p=0.034) where CBD reversed elevated linoleic acid observed in vehicle-cue rats (vehicle-neutral vs vehicle-cue-q(36)=4.803, p=0.009; vehicle-neutral vs CBD-cue- q(36)=1.995, p=0.501; Fig4A). Linoleic acid levels also positively correlated with the overall anxiety score (Fig4B; R²=0.248, p=0.0006), suggesting a potential relationship between normalizing linoleic acid and mitigation of cue-induced anxiety by CBD administration.

Figure 4.

Lipid profiling of the NAcSh. CBD upregulated several bioactive lipids including arachidonic acid, SEA, OEA, 2-LG, 2-OG, PGLy, and S-Alanine (A). However, only linoleic acid was significantly increased in the vehicle cue animals and reversed in CBD cue rats (A). Linoleic acid levels positively correlated with the composite anxiety score (B). Open circles represent animals in the neutral group, closed circles animals in the cue group. Vehicle and CBD rats are depicted by grey and orange markers, respectively. Partial correlation analyses were used to estimate lipid networks. In neutral animals (C), arachidonic acid positively correlated with linoleic acid (for which it is a precursor). In cue animals, this correlation was lost, but CBD strengthened relationships between multiple lipid species especially arachidonic acid and OEA. Asterisks indicate significant relationship p<0.05. L-acid, linoleic acid; A-acid, arachidonic acid; 2-LG, 2-linoleoylglycerol; 2-OG, 2-oleoylglycerol; SEA, N-Stearoylethanolamine; OEA, N-Oleoylethanolamine.

Linoleic acid is utilized in several metabolic pathways to synthesize arachidonic acid and oxylipins, products of fatty acid oxidation⁴⁵. Leveraging the RNAseq dataset, we queried whether the cue or CBD influenced transcripts associated with linoleic acid metabolism. In the vehicle cue animals, fatty acid desaturases were consistently downregulated, as was Cyp4b1 and Ephx2 cytochrome 450 subunit and soluble epoxide hydrolase which may be involved in the conversion of linoleic acid to linoleic epoxides (Suppl.Fig10). These effects were typically unchanged in CBD cue animals, suggesting the cue decreased conversion of linoleic acid to arachidonic acid and oxylipins which may be normalized by CBD.

Several of the lipid species studied are precursors, products, or byproducts of one another via various metabolic pathways³³. To explore how CBD may impact the relationships between these lipids, we completed partial correlation analyses on the species significantly impacted by CBD. In the neutral groups, linoleic acid positively correlated with arachidonic acid (for which it is a precursor) and negatively correlated with OEA, whereas OEA positively correlated SEA and arachidonic acid. In contrast, in vehicle-cued animals, none of these relationships were maintained, suggesting a substantial shift in normal lipid networks. Surprisingly, in CBD cued animals, the partial correlation between linoleic and arachidonic acid was not restored. However, several significant correlations emerged throughout the network in CBD cue rats, with strong relationships between arachidonic acid and the diacylglycerols, OEA, and PGly, as well as between OEA and the diacylglcerols. These relationships were not observed in the CBD neutral group, indicating a critical interaction between drug and cue experience and bioactive lipid networks.

DEG-enriched gene networks recruited in cue animals distinctly correlate with behavior or NAcSh lipids

Using MEGENA co-expression analysis, gene networks were constructed for all cued animals. We then identified modules significantly enriched with DEGs from the interaction term and that correlated with behavior and lipids (Fig5A). Eight networks were primarily correlated with behavior, and five correlated with lipids, all of which were derived from two parent networks--- c1_5 and c1_2. We also assessed whether these networks contained known CBD targets. Between both parent modules, 21 of the 26 expressed CBD candidate genes were present (Fig5B; 80.8%; Odds Ratio for enrichment: 3.25, p=0.016), indicating these networks may represent the broader effects of CBD activity in the NAcSh. Distinct c1_5 child networks either primarily correlated with anxiety behavior or bioactive lipid levels. The modules that distinctly correlated with behavior contained genes involved in mitochondria function, including the electron transport chain, ATP synthase complex, oxidative phosphorylation, and the inner membrane of the mitochondrial matrix (Fig5C). In contrast, the c1_5 child modules that inversely correlated with lipid levels enriched for genes related to pre- and post-synaptic plasticity and H3K27me3 (Fig5D). The c1_2 child modules most robustly correlated with the lipid network and contained transcripts that enriched for endocannabinoid signaling, phospholipase D, calcium activity, morphine addiction, PPAR-γ, and H3K27me3 (Fig5E). These results recapitulated several results from the DESeq2 analysis but show broader recruitment of these processes which may underlie different effects of the cue and CBD.

Figure 5.

Results from MEGENA gene network analysis. Several modules significantly enriched with DEGs correlated with either behavior or NAcSh lipids (A). Approximately 80% of known CBD targets were present in the two parent networks (B). Gene modules from the c1_5 parent network that correlated primarily with behavior enriched for mitochondria-specific ontologies (C). In contrast, genes from the same parent network that correlated with lipids enriched for ontologies related to pre-and post-synaptic membranes, ion channel complex, and H3K27me3 (D). Modules from the parent network c1_2, which also correlated with the NAcSh lipid network, enriched for retrograde endocannabinoid signaling, phospholipase D, calcium signaling, morphine addiction, PPAR- γ, and H3K27me3.

Discussion

Our results demonstrate that CBD has anxiolytic properties, but only in the presence of a cue previously associated with multiple cue-shock pairings and thus greater emotive salience. CBD also produced two clear molecular outcomes. CBD either reversed or normalized the transcriptional impact of the shock-cue in the NAcSh, including normalizing downregulation of mitochondria transcripts that correlated with behavior. Additionally, CBD altered the NAcSh lipid profile, where it normalized elevated levels of linoleic acid levels observed in vehicle cue animals but also substantially recruited lipid networks. These results were unique to the shock-cue group, as CBD in the neutral condition did not affect anxiety behavior and had no distinct impact on the molecular landscape. Collectively these results support nuanced effects of CBD in cue-induced behaviors likely via interactions with mitochondria and modulation of lipids in the NAcSh.

The finding that CBD reversed avoidance-like behavior only in the presence of a shock-paired cue after multiple conditioning sessions suggests a relevance to heightened anxiety conditions. Several preclinical studies have reported significant effects of CBD using anxiety assays^11–13 including after uncued footshock in Sprague-Dawley rats¹⁵. However, a recent meta-analysis revealed that unconditioned anxiety showed only a moderate response to CBD, whereas conditioned anxiety and pre-existing anxiety phenotypes were associated with stronger drug effects¹². These results are similar to those seen in humans, with mixed results obtained in healthy controls but more robust effects in those diagnosed with anxiety disorders under anxiogenic task conditions^{7, 9}. In substance use disorder, CBD has been shown to reduce trait and cue-elicited anxiety^{16, 17}. In animal models of addiction, although CBD did not decrease drug-taking or extinction it significantly reduced relapse-like behavior to a drug-paired cue^{18, 19}. Our results complement these findings and suggest that behavioral-state and cue associations are likely important for potential therapeutic effects of CBD on psychopathology-like phenotypes. However, the effects observed may differ based on strain (e.g., Long-Evans rats have been shown to be more exploratory in approach-avoidance tasks¹⁸), type of anxiety trait examined, or dose and timing of CBD administration. Future experiments incorporating multiple assays of anxiety behavior with stress-paired cues can help to elucidate whether these effects extend to other paradigms.

The current molecular findings also provided significant insights including the initial qPCR panel that explored known CBD targets in multiple nuclei recruited by anxiety and cues. Of the mesocorticolimbic structures (accumbens, amygdala and prefrontal cortex) studied, the candidate genes were predominantly altered in the NAcSh. Although mainly recognized for its role in reward, the NAcSh is also known for its contribution to aversion behaviors, especially in response to cues. For example, the immediate early gene cFos is elevated in the NAcSh during active avoidance⁴⁶ and after stressful stimuli⁴⁷. Lesioning the NAcSh impairs conditioned fear discrimination⁴⁸ and dopamine release in the NAcSh (but not the NAc core subregion) is elevated in response to aversion learning, especially to cues that predict negative outcomes^{49, 50}. Interrogating the transcriptomic profile of the NAcSh revealed changes in transcripts related to synaptic plasticity, trans-synaptic signaling (including opioid signaling), and calcium signaling, all of which were reversed by CBD. Of these transcriptional changes, downregulation of mitochondria-related transcripts, especially those part of complex 1, showed the strongest association with anxiety-like phenotypes. These results suggest that reduced oxidative phosphorylation (and potentially ATP production) may contribute to the expression of anxiety elicited by the shock-cue two weeks after the final shock exposure. It is known that stress induces mitochondrial dysfunction associated with anxiety^{51, 52}. More recent studies show that highly anxious rats have abrogated complex 1 and 2 protein activity as well as reduced ATP production and oxygen consumption specifically within the NAc⁵³ and blunting cell-specific mitochondria fusion in the NAc can induce social subordination⁵⁴. In vitro models have also shown that CBD decreases mitochondria respiration and induces mitochondrial release of calcium into the cytosol^{55, 56}. Other data suggest the effect of CBD on mitochondria may depend on excitability of the cell as evident in cultured hippocampal cells where CBD increased mitochondria calcium function, but reduced calcium activity in hyperexcitable cells and blocked epileptiform activity⁵⁷. These data indicate that changes in cell calcium activity likely produce molecular changes that are effectively blocked by CBD. The exact mechanism underlying the impact of CBD on mitochondria remains unknown but has multiple candidates including a novel unnamed mitochondrial chloride channel⁵⁸, mitochondria-localized CB1 receptors, voltage-dependent anion channel proteins, and through components of the mitochondria permeability transition pore⁵⁵. The causal impact modulating these mechanisms on physiology or behavior after CBD has yet to be studied.

Linoleic acid was also an interesting target since potentiated levels evident in cue-shocked animals were normalized by CBD. Moreover, changes in linoleic acid both correlated with anxiety and mitochondrial gene networks, suggesting a link between these outcomes. Linoleic acid is an essential long-chain fatty acid used for synthesis of other lipids including arachidonic acid⁴⁵. In the neutral animals, linoleic acid positively correlated with arachidonic acid levels, but this relationship was abolished in cued-shocked rats, even after CBD administration. This suggests that CBD did not necessarily restore normal linoleic acid conversion to arachidonic acid but may have recruited other metabolic mechanisms, such as oxidation or esterification of linoleic acid. Esterified linoleic acid produces reactive oxygen species from mitochondria and lipid peroxidation⁵⁹. When oxidized, linoleic acid produces metabolites⁴⁵ which have been shown to promote axon outgrowth in development⁶⁰ and when blocked reduce anxiety in the open field assay⁶¹. Downregulation of fatty acid desaturases (reducing metabolism of linoleic acid to arachidonic acid) and upregulated cytochrome p450 and epoxide hydrolases which influence fatty acid oxidation were observed in vehicle-cue animals. These effects were either not observed or reversed by CBD, which may either help to restore the normal lipid network or block conversion of linoleic acid to oxylipins which may influence behavior. Alternatively, normalization of linoleic acid may have impacted signaling through PPAR nuclear receptors. Both PPAR-α and PPAR-γ contribute to lipid metabolism by upregulating ACOX1 expression, the transcript responsible for the first step of beta-oxidation of lipids in peroxisomes^{62, 63}. These receptors also affect mitochondria by mitigating reactive oxygen species (PPAR-α) or through stimulating mitochondrial biogenesis and oxygen consumption (PPAR-γ)^63–65. PPAR-α agonism has also been shown to modulate glutamatergic and seizure activity⁶⁶. The cue downregulated DEGs that enriched for PPAR-γ in the vehicle animals, and IPA analyses indicated that PPAR-α was inhibited in vehicle cue animals but activated in CBD rats. CBD is a known PPAR-γ agonist⁶⁷ and also significantly elevated arachidonic acid and OEA which are potent agonists of PPAR-α ⁶³. Therefore, CBD’s actions at PPAR-γ, and potentially indirect activity via elevated lipids at PPAR-α, may impact both acute cell function, lipid homeostasis and mitochondrial function. What is unclear is whether the cue suppresses activity of these receptors that CBD reverses or if their activation is sufficient to impact behavior alone. Studies are required to investigate cue-induced changes in PPAR protein expression, lipid levels, and the functional role of these ligands/receptors with CBD in relation to behavior.

In both the neutral and cue condition, CBD had broad impact on the lipid landscape increasing levels of arachidonic acid, N-acylethanolamines, diacylglycerols, P-glycine, and S-Alanine. Many of these lipids were tightly correlated in the cue animals, suggesting recruitment of these lipids as a network in response to the cue. Surprisingly, this network did not have a strong relationship with anxiety behavior, but instead was inversely correlated with gene modules that enrich for changes in synaptic membranes and signaling. Furthermore, elevated lipids also negatively correlated with H3K27me3-related genes, and CBD cue animals showed downregulation of genes enriched for H3K27me3. These results suggest that CBD’s impact on the lipid network may influence epigenetic regulation of gene expression of synaptic plasticity. Emerging data shows that endogenous lipids influence chromatin architecture and remodeling^{68, 69}. CBD has been shown to impact several epigenetic mechanisms including inducing H3K27me3 and induces DNA methylation in the cortex^{57, 70, 71}. Such epigenetic changes may contribute to the protracted effects of CBD on behavior including reduced anxiety in the open field we observed two weeks after the final shock or reduced cue-induced drug-seeking weeks or even months after the final CBD administration^{18, 19}. Disentangling whether CBD or the elevated lipid network drives these changes is an exciting future direction.

Despite significant novel findings from this study, there are some limitations. We assessed the role single vs. repeated shock pairings using a within-subjects design which does not test extinction to the cue or incubation of fear. CBD has been shown to facilitate fear extinction⁷² (but not extinction for drug reward¹⁸), therefore it is possible that CBD could influence extinction or sensitization of fear-related behavior. We also did not test the effect of CBD on avoidance behaviors without fear conditioning, and several other anxiety studies have indicated anti-anxiety effects without cues^11–13. However, as stated above, CBD may have limited effects on anxiety without stress experience, including in the light/dark test^{14, 15}, and our results show no effect of CBD without the shock-paired cue. Testing the effects of CBD at all stages of cue learning, extinction, and incubation with manipulations of potential mechanisms described above (oxidative phosphorylation, linoleic acid, PPAR modulation) will be critical to further disentangling its anxiolytic potential. Furthermore, the results reported are exclusively in males. This was done to be consistent with previous work^{15, 18}. However, recent studies indicate that CBD has sex specific effects in pain and drug metabolism^{73, 74}, and emerging results showing influence of the estrus cycle in CBD efficiacy^{75, 76}. Future studies will incorporate females into our now validated model to determine whether sex influences the behavioral and neurobiological responses to CBD.

In conclusion, these findings support that CBD does have anti-anxiety potential, but mainly related to emotive salient cues. Of the neurobiological signature induced by CBD, reversal of both elevated levels of linoleic acid and mitochondrial dysfunction in the NAcSh suggest they may drive cue-induced anxiety. These novel results expand upon our understanding as to how CBD may improve avoidance-like behavior and further our understanding on novel therapeutic targets for treating anxiety disorders.