Cannabinoid modulation of corticolimbic activation during extinction learning and fear renewal in adults with posttraumatic stress disorder

Abstract

Failure to successfully extinguish fear is a hallmark of trauma-related disorders, like posttraumatic stress disorder (PTSD). PTSD is also characterized by dysfunctional corticolimbic activation and connectivity. The endocannabinoid system is a putative system to target for rescuing these behavioral and neural deficits. In healthy adults, acute, low-dose delta-9-tetrahydrocannabinol (THC) facilitates fear extinction and increases cortico-limbic activation and connectivity in response to threat. The present study determines the effect of acute, low-dose THC on fear-related brain activation and connectivity during fear extinction in trauma-exposed adults with (PTSD = 19) and without PTSD [trauma-exposed controls (TEC) = 26] and non-trauma-exposed [healthy controls (HC) = 26]. We used a Pavlovian fear conditioning and extinction paradigm, where we measured concurrent functional magnetic resonance imaging (fMRI) and behavioral responses (i.e., skin conductance responding and expectancy ratings). Using a randomized, double-blind, placebo-controlled design, N = 71 subjects were randomized to receive placebo (PBO, n = 37) or THC (n = 34) prior to fear extinction learning. During early extinction learning, individuals with PTSD given THC had greater vmPFC activation than their TEC counterparts. During a test of the return of fear (i.e., renewal), HC and individuals with PTSD given THC had greater vmPFC activation compared to TEC. Individuals with PTSD given THC also had greater amygdala activation compared to those given PBO. We found no effects of trauma group or THC on behavioral fear indices during extinction learning, recall, and fear renewal. These data suggest that low dose, oral THC can affect neural indices of fear learning and memory in adults with trauma-exposure; this may be beneficial for future therapeutic interventions seeking to improve fear extinction learning and memory.

1. Introduction

Posttraumatic stress disorder (PTSD) is a debilitating mental health condition that can occur following a traumatic event and is characterized by flashbacks, avoidance, hyperarousal, and negative changes in cognition and mood (American Psychiatric Association: Diagnostic and Statistical Manual of Mental Disorders, 2013a). PTSD symptoms stem from an impairment in fear extinction, as memories of an aversive experience are pervasive and continue to trigger a fear response long after the event has occurred (Rauch et al., 2006; Rosen & Schulkin, 1998). Prolonged Exposure (PE) therapy is a first-line psychotherapy for PTSD that utilizes extinction learning principles by repeatedly exposing the patient to fearful memories and trauma reminders in a safe environment (Foa, 2011). This re-exposure integrates corrective information, therefore diminishing fear responses to traumatic memories (Foa et al., 2007). Although PE produces clinically meaningful improvements for many PTSD patients, there is still substantial room for improvement. Approximately 20-30% of treatment completers retain their PTSD diagnosis, slightly more (30-40%) fail to achieve a stringent criterion for good end-state functioning (i.e., they are still symptomatic), and others fail to complete treatment (20.5%) (Foa et al., 1999; Hembree et al., 2003; Rothbaum et al., 2005).

This exposure-based learning can be modeled in the laboratory, in both animals and humans, using Pavlovian fear conditioning models. Fear is acquired when an aversive stimulus (unconditioned stimulus; US) is linked to a previously innocuous cue (conditioned stimulus; CS) and then extinguished by presenting the CS alone in the absence of the aversive stimulus (producing extinction). Successful extinction is exhibited by a decrease in fear responding. Unfortunately, a major limitation of extinction is that it is a temporary phenomenon -- extinguished fears can re-emerge over time (Bouton, 2002). Fear extinction and its later recall have become the prime translational neuroscience target for the treatment of PTSD, as these mechanisms are thought to underly PE treatment response and long-term maintenance of treatment gains.

Extinction learning processes are dependent on corticolimbic brain regions, namely the amygdala, hippocampus, and prefrontal cortex (for review see Milad & Quirk, 2012). The amygdala and dorsal anterior cingulate cortex (dACC) are involved in forming fear associations and producing fear responses (LeDoux, 2000; Milad, Quirk, et al., 2007). Prefrontal brain regions that interconnect with the amygdala, specifically the ventromedial prefrontal cortex (vmPFC), are important for the recall of extinction learning. The vmPFC attenuates fear responses by inhibiting amygdala output neurons (Hartley et al., 2011; Milad, Quinn, et al., 2005; Milad, Wright, et al., 2007; Milad & Quirk, 2002; Paré & Smith, 1993). Similarly, hippocampus activation is associated with successful recall of extinction learning and is positively correlated with vmPFC activation during extinction recall in healthy humans (Milad, Quirk, et al., 2007). Individuals with PTSD not only have poor extinction recall, but display aberrant corticolimbic activation; vmPFC and hippocampus activation is lesser than controls, while amygdala and dACC activation is greater (Milad et al., 2009). Poor extinction recall and corticolimbic dysfunction displayed by patients with PTSD may undermine the therapeutic efficacy of exposure-based interventions, like PE (Foa, 2000; Milad et al., 2008, 2009; Orr et al., 2000; Pitman et al., 2001; Rougemont-Bücking et al., 2011; van Minnen & Hagenaars, 2002). Enhancing neural and neurochemical substrates of inhibitory fear learning could target this dysfunction, thereby improving PTSD treatment outcomes.

Several decades of research highlight the endocannabinoid (eCB) system as an essential moderator of our responses to and recovery from fear and stress (for extensive review see Morena et al., 2015). Notably, rodent models of fear suggest that activating the eCB system within brain structures important for extinction (i.e., amygdala, vmPFC, hippocampus) regulates extinction learning and retention. For instance, blocking or genetically deleting type-1 cannabinoid receptors within these brain structures prevents fear extinction, whereas activation of them, via an agonist such as Δ⁹-tetrahydrocannibinol (THC), can facilitate extinction (de Oliveira Alvares et al., 2008; Lin et al., 2009; Marsicano et al., 2002; Pamplona et al., 2008). In addition, drugs that increase the eCB levels during extinction not only enhance extinction retention, but also impair the return of extinguished fear in rats (Chhatwal et al., 2005; Morena et al., 2018; Pamplona et al., 2008). In healthy humans, eCB system modulation exerts similar effects on fear extinction learning and recall as reported in animal models. Using a double-blind, placebo-controlled, between subjects design, we have previously demonstrated that an acute dose of THC, administered prior to extinction learning, facilitates later recall of extinction learning (Rabinak et al., 2013). We have also previously demonstrated that THC administered to healthy adults prior to extinction learning: decreases amygdala activation during extinction learning, increases vmPFC and hippocampus activation during extinction recall, and increase prefrontal-amygdala functional connectivity (Rabinak et al., 2014, 2018; see also Hammoud et al., 2019).

Given that extinction retention deficits and corticolimbic dysfunction are observed in patients with PTSD, and that enhancing cannabinoid transmission facilitates extinction recall, the eCB system is a promising target for improving these deficits. Specifically, modulating the eCB system could enhance the learning that occurs during therapy, therefore also increasing the efficacy and durability of PE in treating PTSD (i.e., shortening treatment while strengthening and prolonging gains). Despite the promising evidence, direct tests of cannabinoid effects on extinction recall and associated neural circuits have not been conducted in PTSD patients. To address this critical gap in knowledge, we report on a randomized, double-blind, placebo-controlled study to determine the effects of THC on neural and behavioral indices of fear extinction, extinction recall, and fear renewal in trauma-exposed adults with PTSD compared to trauma-exposed adults without PTSD and healthy controls. In this study, we aimed to determine the following: (1) Among those that received PBO, are there significant neural and/or behavioral differences between healthy controls, trauma-exposed controls, and trauma-exposed adults with PTSD during fear extinction learning, recall, and fear renewal?; (2) When given THC, do trauma-exposed adults with PTSD have similar neural and/or behavioral fear indices compared to both control groups during extinction learning, recall, and fear renewal?; and (3) Compared to PBO, do trauma-exposed adults with PTSD given THC have enhanced fear extinction?

2. Methods

2.1. Participants

Participants were recruited from the community via flyers, online postings, etc. (Detroit, MI). Eligible participants were right-handed, between the ages of 21-45, fluent in English, had at least a high school diploma or equivalent, and were not currently taking any concomitant medications that may interact with dronabinol. Participants were free of any MRI contraindications (e.g., metal in the body) and any current medical condition, neurocognitive impairment, or treatment with medication that would interfere with task performance (e.g., benzodiazepines). Participants who were pregnant or breastfeeding, posed a risk of harm to themselves or others, or had participated in another fear conditioning study in the past 30 days were excluded. Participants with a current or past allergic/ adverse reaction or known sensitivity to cannabinoid-like substances (dronabinol/marijuana/cannabis/THC, cannabinoid oil, sesame oil, gelatin, glycerin, and titanium dioxide) were also excluded from the study. Additional exclusionary criteria for those with PTSD are described in the Supplemental Information Section S1.2.

All participants had to produce a negative result on a urine drug screen and alcohol breathalyzer test before MRI scanning. Control participants (HC, TEC) were free of any major psychiatric diagnoses based on a computerized version (NetSCID-4, TeleSage) of the Structured Clinical Interview for DSM-IV (SCID-IV; First et al., 2002) or the Mini International Neuropsychiatric Interview (MINI version 6.0.0; Sheehan et al., 1998). Eighty-six individuals met prescreening eligibility and were enrolled in the study. Of the N = 86, n = 15 were excluded for the following reasons: ineligible based on exclusion criteria (n = 6), lost to follow-up (n = 1), brain abnormality (n = 1), and incomplete fMRI data (i.e., missing a run or did not complete task in scanner; n = 7) (See Figure 1). Thus, N = 71 participants are included in the final analyses.

Fig. 1. Screening and randomization of participants.

Eighty-six individuals met inclusion/exclusion criteria for the study and were randomized to THC or PBO. Of the 86, n = 15 were excluded for the following reasons: 6 were ineligible based on participation criteria (e.g., recently diagnosed mood disorder after enrollment), 1 was lost to follow-up, 1 had a brain abnormality, and 7 had incomplete imaging data. Seventy-one participants had reliable data for analyses. PTSD = posttraumatic stress disorder, TEC = trauma-exposed control, HC = healthy control, THC = delta-9-tetrahydrocannabinol, PBO = placebo.

The study protocols were approved by the Wayne State Institutional Review Board, and all participants provided written informed consent prior to completing study procedures. Participants were compensated for their time. Participants varied in racial/ethnic makeup, which is consistent with the racial/ethnic composition of the recruitment area (Detroit, Michigan, USA; See Table 1 for details). This study is a registered clinical trial investigating the effects of THC on fear extinction in participants with PTSD (NCT02069366)

Table 1.

Participant demographics.

	THC (n = 34)				PBO (n = 37)				p-value
	PTSD (n = 9)	TEC (n = 12)	HC (n = 13)	Total	PTSD (n = 10)	TEC (n = 14)	HC (n = 13)	Total
Age (mean years, SD)	26.7 (5.3)	28.7 (7.6)	24.3 (2.5)	26.5 (5.6)	23.8 (3.2)	26.4 (7.0)	26.9 (6.6)	25.8 (6.1)	0.636
Gender (n, female)	8	3	6	17	5	9	4	18	0.909
Race (n)									0.802
African American	2	3	0	5	1	3	2	6
Asian	1	3	5	8	1	4	4	9
American Indian/Alaskan Native	0	1	0	1	0	0	0	0
Caucasian	5	5	8	18	7	5	7	19
More than 1 race	1	0	0	1	0	1	0	1
Other	0	0	0	0	0	1	0	1
Hispanic or latino (n)	1	0	0	1	1	1	2	4	0.195
PTSD Severity (CAPS-5)
Total severity score (mean, SD)	34.4 (11.2)	2.8 (6.6)	--	16.3 (18.2)	34.2 (6.4)	4.2 (5.9)	--	16.7 (16.2)	0.942
Number of clinically significant symptoms (mean, SD)	11.3 (2.7)	0.9 (2.4)	--	5.4 (5.8)	12.4 (2.7)	1.2 (2.0)	--	5.9 (6.1)	0.783

2.2. Trauma Ratings and Interview

All participants completed the Life Events Checklist for DSM-5 to determine exposure to traumatic events in the participant’s lifetime (Weathers et al., 2013). A total of n = 45 participants endorsed a traumatic event that met Criterion A per DSM-5 criteria for PTSD (American Psychiatric Association: Diagnostic and Statistical Manual of Mental Disorders, 2013b). Of those that endorsed a Criterion A event, n = 19 met criteria for PTSD, by either meeting CAPS-5 diagnosis or having a total CAPS-5 severity score ≥ 25 and were classified as the PTSD group (Weathers et al., 2018). The n = 26 who did not meet these criteria were classified as trauma-exposed controls (TEC) and the n = 26 who did not endorse a Criterion A event were classified as healthy controls (HC).

2.3. Fear Conditioning and Extinction Paradigm

Participants completed an adapted version of a well-validated Pavlovian fear-extinction paradigm originally developed by Milad and colleagues (2007a), which manipulates context using an ABBA design. In this paradigm, the fear acquisition context (‘danger’ context; CXT+ or A) was separate from the ‘safety’ context where fear was extinguished (CXT− or B) (described previously in Rabinak et al., 2017). Fear conditioning was conducted outside of the MRI scanner environment in a standard lab office with the task displayed on a 24” LCD computer monitor. Fear extinction learning, extinction recall, and fear renewal occurred during fMRI scanning. The task was displayed using Presentation (Presentation Version 18.1, Neurobehavioral Systems, Berkeley, CA, USA) on a screen viewed behind the scanner using a mirror affixed to the headcoil. Two different outdoor environments constituted the contexts (Fig. 2). The CSs were lamp posts lit up in three colors, chosen for maximum discriminability by color vision deficient individuals (Fig. 2), and the US was a white noise burst (500 ms, 95 dB).

Fig. 2. Pavlovian fear conditioning paradigm.

The paradigm occurs over three days: Day 1 contains fear conditioning, Day 2 contains fear extinction learning, and Day 3 contains extinction recall and fear renewal. Participants received a capsule of either PBO or 7.5 mg of THC 120 minutes prior to extinction learning. Days 2 – 3 occur in the functional imaging scanner. Icons sourced from Flaticon.com.

During fear conditioning (~16 min), participants were presented with a CS on a screen that co-terminated with an aversive white noise burst (US) at a partial reinforcement rate of 60% (CS+; two different colored lights, i.e., blue light and yellow light). A third CS (e.g., pink light) was presented during fear conditioning but was never paired with the US (CS−). CS duration was 4 seconds total, regardless of if paired or unpaired with the US; the US co-terminated with the last 0.5 seconds of the CS. Fear conditioning consisted of 8 non-reinforced presentations of each of the CS+s, intermixed with an additional 12 presentations of each CS+ that co-terminated with the US, and 20 presentations of the CS−. All stimuli were presented within the conditioning context (CXT+; Fig. 2 Day 1). Twenty-four hours later, participants completed an extinction learning session (~11 min) during fMRI scanning wherein one of the CS+s was subsequently extinguished, i.e., presented in the absence of the US (CS+E). Extinction learning consisted of 20 non-reinforced CS+E and CS− trials presented in the extinction context (CXT-; Fig. 2 Day 2). Of note, the other CS+ was not presented during extinction learning and remained unextinguished (CS+U). Twenty-four hours later, participants returned to the MRI scanner for an extinction recall test phase (~16 min), which consisted of 20 non-reinforced presentations of each of the CSs (CS+E, CS+U, and CS−) in the CXT- (Fig. 2 Day 3). Approximately two minutes later, participants completed a fear renewal phase (~16 min), which consisted of 20 non-reinforced presentations of each of the CSs (CS+E, CS+U, and CS−), this time within the CXT+. The designation of colored lights (blue, yellow, or pink) as CS+E, CS+U, or CS−, and the context (building or forest) as CXT+ or CXT− was randomized across the participants. In each experimental session (fear acquisition, extinction learning, extinction recall, fear renewal) the context picture was presented for 7–12 s: 3–8 s alone followed by 4 s in combination with the CS+ or CS−. The mean intertrial interval was 6 s (range: 4–9 s) determined from the offset of the CS to the onset of the context. In addition, the order of trials was pseudo-randomized, such that no more than 2 presentations of the same color light (pink, yellow, or blue) occurred in a row.

2.4. Drug Administration

Participants were randomized to either receive an acute oral dose of THC (dronabinol, 7.5 mg capsule; Ascend Laboratories, LLC, Parsippany, NJ) or dextrose (PBO). This dose of THC was previously shown to be the lowest dose to still induce a subjective effect (Kirk & de Wit, 1999) and reduce amygdala activation to social threat in healthy adults (Phan et al., 2008).

2.5. Study Procedures

Approximately 120 minutes prior to extinction learning, participants ingested a capsule of either THC (n = 34) or PBO (n = 37). This administration corresponds to peak drug effects of THC (Wachtel et al., 2002). The principal investigator (CAR) randomized participants, wherein participants and study personnel were blinded to the capsule content and drug group assignments.

2.6. US Expectancy Ratings

Before each phase, participants were instructed to submit a rating immediately when they saw the colored lights to indicate whether they believed they would hear a loud noise. Possible responses were “Yes, I will hear a loud noise”, “No, I will not hear a loud noise”, and “I don’t know”. All responses were registered using a MR-compatible 4-button response device. For analysis of US expectancy ratings, early fear acquisition was defined as the second trial of each stimulus. For extinction learning, extinction recall, and fear renewal, early was defined as the first trial for each stimulus in each phase, following our prior work (Zabik et al., 2022). Late fear acquisition, extinction learning, extinction recall, and fear renewal were defined as the last trial (20th) of each stimulus per phase. If 10% or more of responses were missing for a given trial (HC/TEC > 1; PTSD ≥ 1), responses were replaced with the adjacent trial (3^rd trial for early fear acquisition, 2^nd trial for early response in subsequent phases, and 19^th trial for late response). Several stimuli had zero responses for a given category, likely due to small drug x trauma sample sizes. To address this issue for statistical analyses, we collapsed responses into two categories: ‘Yes/I don’t know’ and ‘No’. Previous work from Rabinak et al. (2017) report that individuals with PTSD are more likely to expect an US to occur, which is reflected in their increased “Yes” and “I am not sure” responses. During each phase, binary logistic regressions were used to assess the probability of participants responding ‘Yes/I don’t know’ vs. ‘No’. Group (drug or trauma) and time (early, late) were included in the regression for each phase, as well as stimulus (fear acquisition = CS+, CS−; extinction learning = CS+E, CS−; extinction recall = CS+E, CS+U, CS−; extinction renewal = CS+E, CS+U, CS−). Results are considered significant with the omnibus test (X²), where direction of the effect is determined by either B-weights or estimated marginal means with 95% confidence intervals that do not include (1). Data were analyzed using SPSS software (IBM SPSS Statistics 27) and considered significant at α = 0.05; Bonferroni correction was applied to main effects and interactions for extinction learning, extinction recall, and fear renewal, where the corrected α-level = 0.01. Bonferroni correction was also applied to estimated marginal means within the SPSS software.

2.7. Skin Conductance Responding (SCR)

We recorded skin conductance responses (SCRs) using two disposable carbon fiber electrodes (EL509, BIOPAC Systems, Inc., Goleta, CA) attached between the first and second phalanges of the second and third digits of whichever hand showed greater SCRs during a screening task, or the left hand if there was no difference in responding between hands. The electrodes were connected to a BIOPAC Systems skin conductance module (GSR100C) and skin conductance was continuously sampled at a rate of 1000 samples per second. SCR data were processed using AcqKnowledge 5.0 software (BIOPAC Systems, Inc.) by applying a low pass Blackman filter (cutoff frequency = 25 Hz) and mean value smoothed over 100 adjacent data points. Following prior work, SCR for each CS trial was calculated by subtracting the mean skin conductance level during the 2 sec before stimulus onset from the highest skin conductance level that occurred in the 0.5 to 4.5 sec window after stimulus onset (Rabinak et al., 2014). Values below 0.02 μS were excluded as non-responses (Boucsein, 2012), and raw SCRs from each phase were square root transformed to normalize the distributions (LaBar et al., 1998; Milad, Orr, et al., 2005; Schiller et al., 2010). To ascertain the level of conditioned anticipation of the US separate from responses to the noise bursts, we included only non-reinforced trials of the CS+ in our analyses.

Early trials were defined as the average of the first four trials, while late trials were defined as the average of the last four trials. For acquisition, the four early (1 – 4) and late (5 – 8) trials correspond to the non-reinforced trials of the CS+ (i.e., no US). For extinction learning, extinction recall, and fear renewal, early trials correspond to trials 1 – 4, while late corresponds to trials 17 – 20. Due to limited poor data quality, SCR data were not analyzed; valid n for groups ranged from 0 – 17. However, plots of means can be found in Supplemental Information (Section S2.2, Fig. S1).

2.8. Subjective Units of Distress (SUDs)

Participants reported on their subjective distress before, during the middle of, and after extinction learning, extinction recall, and fear renewal. In the MRI scanner, participants used the MR-compatible response device to report how they were feeling at that moment on a scale from 0 to 100 (increments of 10). 0 = “No Anxiety”, 20 = “Mild anxiety, Alert, able to cope”, 50 = “Moderate anxiety, Some trouble concentrating”, 80 = “Severe anxiety, Thoughts of leaving”, and 100 = “Very severe anxiety, Worst ever experienced” (Wolphe, 1969).

2.9. Functional Imaging

2.9.1. Data collection

Blood-oxygen level-dependent (BOLD) fMRI data were collected from 66 axial, 2-mm-thick slices (set to the manufacturer-minimum gap size) using a T2*-sensitive gradient echo EPI acquisition sequence (repetition time = 2000 ms; echo time = 30 ms; 128 x 128 x 66 matrix; field of view = 256 mm; flip angle = 73 degrees; 2.0 x 2.0 x 2.0 mm voxels; multiband acceleration factor = 3). A T1-weighted anatomical image (repetition time = 1820 ms; echo time = 3.52 ms; 256 x 256 x 120 matrix; field of view = 240 mm; flip angle = 8 degrees; 0.94 x 0.94 x 1.5 mm voxels) and field map image (repetition time = 2412 ms; echo time = 51 ms; 128 x 128 x 66 matrix; field of view = 256 mm; flip angle = 90 degrees; multiband acceleration factor = 3) were collected within the same scan session for coregistration and distortion correction.

2.9.2. Preprocessing

fMRI data were analyzed using SPM12 software (Wellcome Department of Cognitive Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm). The following preprocessing steps were applied, in order: (1) distortion correction, (2) realignment to the first image, (3) slice timing correction, (4) coregistration, (5) normalization to MNI space, (6) reslicing, and (7) spatial smoothing (6-mm FWHM Gaussian kernel).

2.9.3. Motion-related artifact and quality control

Preprocessed data were submitted to quality control analyses, and data from N = 71 participants that met criteria for high quality and scan stability with minimum motion correction were included in the analyses. For images identified as containing high movement (≥3 mm displacement in any one direction), frames with excessive motion (framewise displacement > 0.8 mm) were censored, as well as one frame before and two frames after the high-motion frame. These frames were censored from first-level analyses to reduce the impact of spurious participant movement. See Supplemental Information Section S1.2 for more details regarding subject head motion.

2.9.4. First-level model

Following preprocessing, a general linear model was applied to the time series, convolved with the canonical hemodynamic response function (HRF) and with a 128 s high-pass filter (Worsley & Friston, 1995). Individual statistical parametric maps were calculated using this general linear model for each condition type (CS+E, CS− for fear extinction learning; CS+E, CS+U, CS− for extinction recall and fear renewal) for each participant in all phases. To reduce the impact of motion-related artifact, the six movement parameters obtained during realignment were included in the first-level models as nuisance regressors, along with their derivatives and the quadratic terms of both the motion and motion derivatives. Volumes with excess motion (i.e., > 0.8 mm framewise displacement) were also added as nuisance regressors in the models.

2.9.5. Second-level model

First-level contrasts were subsequently submitted to second-level analyses in a random-effects statistical model in SPM8 (Friston et al., 1998), for every phase. Previous work suggests that activation of fear-related brain regions is strongest during early phases of learning (i.e., early extinction learning and extinction recall; see Milad, Wright, et al., 2007; Milad & Quirk, 2002). Therefore, each phase was split into early and late, defined as the first and second halves of each phase, respectively (i.e., 10 trials/stimuli for early, 10 trials/stimuli for late). We tested for effects of drug or trauma and time and stimulus for each a priori region of interest (ROI; described below), separately. To achieve this, during fear extinction learning, a group [drug (THC, PBO) or trauma (PTSD, TEC, HC)] × stimulus (CS+E, CS−) × time (early, late) RM-ANOVA was performed. For extinction recall and fear renewal, group [drug (THC, PBO) or trauma (PTSD, TEC, HC)] × stimulus (CS+E, CS+U, CS−) × time (early, late) RM-ANOVAs were performed. All significant F-tests were followed up by t-tests. Results were considered significant within each ROI using small volume familywise error correction p_FWE < 0.05, applied via Gaussian random field theory for small volumes, as implemented in SPM12 (Worsley et al., 1996). To avoid spurious results, we used a contiguous voxel threshold of k = 10 (Lieberman & Cunningham, 2009). Variables that contained outliers (z-scores > +/− 3.29) were winsorized.

2.9.5.1. Regions of interest (ROIs).

ROIs included anatomically defined amygdala, hippocampus, and vmPFC, based on the AAL atlas (Tzourio-Mazoyer et al., 2002). An anatomically defined dACC ROI based on Brodmann area 32 was also used and created with the WFU PickAtlas toolbox version 3.0.5 (Maldjian et al., 2003). ROIs were assessed bilaterally, for a total of four ROIs.

2.9.5.2. Functional connectivity.

Functional connectivity of the ROIs described above was assessed using a generalized psychophysiological interaction analysis (gPPI; McLaren et al., 2012) in SPM12. First, we created a seed region using the peak coordinates of a significant activation result, and deconvolved the time series of that seed region with the HRF to put the seed time series in “neuronal space”. Then, we created interaction terms (PPI) by multiplying the deconvolved time series from the coordinate with the onset times for each stimulus (CS+E, CS+U, CS−) during extinction learning, extinction recall, and fear renewal. GPPI estimates were determined by small-volume correction (in the contrast of interest) using the anatomically defined ROIs (described above). Significance was considered at p_FWE < 0.05 (two-tailed).

3. Results

3.1. US Expectancy Ratings

3.1.1. Fear Acquisition

All participants acquired differential fear to the CS+ vs. CS−. No differences were detected between trauma groups (PTSD, TEC, HC). There was a main effect of stimulus [X²(1) = 49.082, p < 0.001] and time [χ²(1) = 4.346, p = 0.037]. Participants were 309% more likely to respond ‘Yes/I don’t know’ during early acquisition, compared to late [B = 1.409, p = 0.035, 95% CI (1.104, 15.156)] (See Fig. 3). Likewise, participants were 1468% more likely to respond ‘Yes/I don’t know’ to CS+, compared to CS− [B = 2.753, p = 0.003, 95% CI (2.567, 95.805)]. There was also a stimulus × time interaction [χ²(1) = 6.681, p = 0.010], where participants were more likely to respond ‘Yes/I don’t know’ to early CS+, compared to early [x = 0.37, p < 0.001, 95% CI (0.15, 0.59)] or late CS− [x = 0.58, p < 0.001, 95% CI (0.42, 0.75)]. They were also more likely to respond ‘Yes/I don’t know’ to late CS+, compared to early [x = 0.41, p < 0.001, 95% CI (0.21, 0.62)] or late CS− [x = 0.63, p < 0.001, 95% CI (0.45, 0.81)]. Likewise, they were more likely to respond ‘Yes/I don’t know’ to early CS−, compared to late [x = 0.21, p = 0.017, 95% CI (0.02, 0.40)]. No significant differences were detected between drug groups prior to administration; See Supplemental Data Section S2.1 for full results.

Fig. 3. US expectancy ratings across trauma groups during fear acquisition.

Y-axis indicates the percentage of responses for a given category. The first column shows ratings from individuals with PTSD, the second column from TEC, and the third column from HC. Stacked bars represent responses, from top to bottom: ‘No’, ‘I don’t know’, and ‘Yes’. No differences were detected between trauma groups. PTSD = posttraumatic stress disorder; TEC = trauma-exposed control; HC = healthy control; CS+ = conditioned stimulus paired with unconditioned stimulus; CS− = conditioned stimulus never paired with unconditioned stimulus.

3.1.2. Fear Extinction Learning

In adults who received PBO, there was a main effect of time [χ²(1) = 6.624, p = 0.010]. Participants were 2000% more likely to respond ‘Yes/I don’t know’ during early extinction learning, compared to late [B = 3.045, p = 0.007, 95% CI (2.307, 191.168)]. See Figure 4 for details.

Fig. 4. US expectancy ratings across trauma groups given (a) PBO or (B) THC during fear extinction learning.

Y-axis indicates the percentage of responses for a given category. The first column shows ratings from individuals with PTSD, the second column from TEC, and the third column from HC. Stacked bars represent responses, from top to bottom: ‘No’, ‘I don’t know’, and ‘Yes’. PBO = placebo; THC = delta-9-tetrahydrocannabinol; PTSD = posttraumatic stress disorder; TEC = trauma-exposed control; HC = healthy control; CS+E = conditioned stimulus that is extinguished (i.e., no longer paired with unconditioned stimulus); CS− = conditioned stimulus never paired with the unconditioned stimulus.

In adults who received THC, time (early, late) was analyzed separately due to the matrix being singular. Early extinction was still singular and is not reported. During late extinction learning, no main effects or interactions were detected.

In adults with PTSD, time (early, late) was analyzed separately due to the matrix being singular. Early extinction was still singular and is not reported. During late extinction learning, no main effects or interactions were detected.

See Supplemental Data Section S2.1.2 for analyses regarding differences between drug groups (THC vs. PBO) within HC or TEC, as well as a full trauma × drug × stimulus × time model. Briefly, no effects of trauma or drug group were detected.

3.1.3. Extinction Recall

In adults who received PBO, there was a main effect of time [χ²(1) = 10.291, p = 0.001]. Participants more likely to have a ‘Yes/I don’t know’ response during early recall, compared to late [x = 0.24, p = 0.001, 95% CI (0.10, 0.38)] (See Fig. 5a).

Fig. 5. US expectancy ratings across trauma groups given (a) PBO or (b) THC during extinction recall.

Y-axis indicates the percentage of responses for a given category. The first column shows ratings from individuals with PTSD, the second column from TEC, and the third column from HC. Stacked bars represent responses, from top to bottom: ‘No’, ‘I don’t know’, and ‘Yes’. PBO = placebo; THC = delta-9-tetrahydrocannabinol; PTSD = posttraumatic stress disorder; TEC = trauma-exposed control; HC = healthy control; CS+E = conditioned stimulus that is extinguished (i.e., no longer paired with unconditioned stimulus during extinction); CS+U = conditioned stimulus that is not extinguished (i.e., never presented during extinction); CS− = conditioned stimulus never paired with unconditioned stimulus.

In adults who received THC, there was a main effect of time [χ²(1) = 11.155, p < 0.001], such that early trials were more likely to have a ‘Yes/I don’t know’ response, compared to late [x = 0.32, p < 0.001, 95% CI (0.15, 0.49)] (See Fig. 5b).

In individuals with PTSD, there was a main effect of stimulus [χ²(2) = 11.561, p = 0.003] and time [χ²(1) = 8.195, p = 0.004]. Follow-up tests reveal that individuals with PTSD were more likely to respond ‘Yes/I don’t know’ to CS+U, compared to CS+E [x = 0.25, p = 0.005, 95% CI (0.08, 0.43)] or CS− [x = 0.31, p = 0.005, 95% CI (0.09, 0.53)]. For the main effect of time, early trials were more likely to have a ‘Yes/I don’t know’ response, compared to late [x = 0.36, p = 0.002, 95% CI (0.13, 0.58)].

See Supplemental Data Section S2.1.3 for analyses regarding differences between drug groups (THC vs. PBO) within HC or TEC, as well as a full trauma × drug × stimulus × time model; also see Supplemental Data Section S2.1.2 for analyses regarding the transition between extinction learning and extinction recall (late extinction learning, early extinction recall as ‘time’ factors). Briefly, no effects of trauma or drug group were detected.

3.1.4. Fear Renewal

In adults who received PBO, there was a main effect of stimulus [χ²(2) = 14.058, p < 0.001] and time [χ²(1) = 9.827, p = 0.002]. Participants were more likely to respond ‘Yes/I don’t know’ to CS+E [x = 0.28, p = 0.004, 95% CI (0.07, 0.48)] or CS+U [x = 0.34, p < 0.001, 95% CI (0.13, 0.55)], compared to CS−. For the main effect of time, early trials were more likely to receive a ‘Yes/I don’t know’ response compared to late [x = 0.24, p = 0.001, 95% CI (0.09, 0.38)] (See Fig. 6a).

Fig. 6. US expectancy ratings across trauma groups given (a) PBO or (b) THC during fear renewal.

Y-axis indicates the percentage of responses for a given category. The first column shows ratings from individuals with PTSD, the second column from TEC, and the third column from HC. Stacked bars represent responses, from top to bottom: ‘No’, ‘I don’t know’, and ‘Yes’. PBO = placebo; THC = delta-9-tetrahydrocannabinol; PTSD = posttraumatic stress disorder; TEC = trauma-exposed control; HC = healthy control; CS+E = conditioned stimulus that is extinguished (i.e., no longer paired with unconditioned stimulus during extinction); CS+U = conditioned stimulus that was not extinguished (i.e., never presented during extinction); CS− = conditioned stimulus never paired with unconditioned stimulus.

In adults who received THC, there was a main effect of time [χ²(1) = 11.335, p < 0.001]. Participants were more likely to respond ‘Yes/I don’t know’ during early fear renewal, compared to late [x = 0.23, p = 0.001, 95% CI (0.10, 0.37)] (See Fig. 6b).

In individuals with PTSD, there was a main effect of stimulus [χ²(2) = 11.259, p = 0.004]. Participants were more likely to respond ‘Yes/I don’t know’ to CS+E [x = 0.28, p = 0.025, 95% CI (0.03, 0.52)] or CS+U [x = 0.43, p < 0.001, 95% CI (0.21, 0.64)], compared to CS−.

See Supplemental Data Section S2.1.4 for analyses regarding differences between drug groups (THC vs. PBO) within HC or TEC, as well as a full trauma × drug × stimulus × time model; also see Supplemental Data Section S2.1.4 for analyses regarding the transition between extinction learning and fear renewal (late extinction learning, early fear renewal as ‘time’ factors) and extinction recall and fear renewal (late extinction recall, early fear renewal as ‘time’ factors). Briefly, no effects of trauma or drug group were detected.

3.2. SUDs

A trauma (PTSD, TEC, HC) × drug (PBO, THC) × time (early, mid, late) RM-ANOVA was performed to test for distress related to extinction, recall, fear renewal. No main effects or interactions were detected during extinction learning, recall, or fear renewal. See Figure S2 in Supplemental Data Section S2.3 for average distress across all participants.

3.3. Functional Activation

3.3.1. Extinction Learning

In adults who received PBO, there was a main effect of time in the hippocampus (37 voxels, F = 16.56, Z = 3.74, p_FWE = 0.041). However, follow-up t-tests did not detect a difference between activation in early vs. late extinction learning.

In adults who received THC, there was a trauma × time interaction in the vmPFC (16 voxels, F = 10.96, Z = 3.88, p_FWE = 0.040). Follow-up t-test revealed that individuals with PTSD given THC had greater vmPFC activation during early (vs. late) extinction learning, compared to TEC given THC (50 voxels, t = 5.47, Z = 4.19, xyz = 10, 32, −20, p_FWE = 0.017; see Fig. 7).

Fig. 7. BOLD fMRI activation of the vmPFC showing a trauma x time interaction in adults given THC during extinction learning.

Individuals with PTSD given THC exhibited greater vmPFC activation during early (> late) extinction learning, compared to TEC given THC. Resulfs significanf at pFWE < 0.05; image shown at p < 0.005 from fhe significanf contrast.

No main effects or interactions were detected between individuals with PTSD given PBO vs. THC.

3.3.2. Extinction Recall

In adults who received PBO, there was a main effect of time in the amygdala (13 voxels, F = 13.12, Z = 3.36, p_FWE = 0.035) and vmPFC (149 voxels, F = 18.36, Z = 4.01, p_FWE = 0.019). Follow-up t-tests in the amygdala did not reveal any significant differences between early and late activation. The follow-up t-tests in the vmPFC revealed that activation was greater during late extinction recall, compared to early (165 voxels, t = 4.01, Z = 3.65, xyz = 4, 48, −24, p_FWE = 0.034; see Figure S3 in the Supplemental Information).

In adults who received THC, there was a main effect of time in the hippocampus (34 voxels, F = 16.90, Z = 3.83, p_FWE = 0.027). Follow-up t-test revealed that activation was greater during late (vs. early) extinction recall in the right hippocampus (62 voxels, t = 4.70, Z = 4.08, xyz = 26, −10, −14, p_FWE = 0.013; see Figure S4 in the Supplemental Information).

In individuals with PTSD, there was a main effect of time in the vmPFC (56 voxels, F = 21.79, Z = 4.24, p_FWE = 0.008), such that activation was greater during late (vs. early) extinction recall (57 voxels, t = 5.98, Z = 4.38, xyz = −2, 42, −18, p_FWE = 0.007; see Figure S5 in the Supplemental Information). A drug × time interaction was detected in the hippocampus (15 voxels, F = 16.22, Z = 3.67, p_FWE = 0.046). However, follow-up t-tests did not reveal a significant difference between drug groups.

3.3.3. Fear Renewal

In adults given PBO, no main effects or interactions were detected.

In those given THC, a trauma × stimulus interaction was detected in the vmPFC (19 voxels, F = 6.99, Z = 3.98, p_FWE = 0.025). HC had greater activation to CS+E (vs. CS+U) compared to TEC (14 voxels, t = 5.05, Z = 4.10, xyz = 16, 4, −10, p_FWE = 0.018; see Fig. 8a). Likewise, individuals with PTSD given THC had greater vmPFC activation to CS+E (vs. CS+U) compared to TEC (177 voxels, t = 5.85, Z = 4.37, xyz = 10, 38, −18, p_FWE = 0.008; see Fig. 8b).

Fig. 8. BOLD fMRI activation of the vmPFC showing a trauma x stimulus interaction in adults given THC during fear renewal.

(a) HC given THC exhibited greater vmPFC/pregenual ACC activation to CS+E (> CS+U) during fear renewal compared to TEC given THC. Likewise, (b) individuals with PTSD given THC exhibited greater vmPFC activation to CS+E (> CS+U) during fear renewal compared to TEC given THC. Results significant at pFWE < 0.05; image (a) shown at p < 0.05 and (b) at p < 0.005 for display purposes only and images represent the significant contrast of interest.

In individuals with PTSD, a drug × stimulus × time interaction was detected in the amygdala (14 voxels, F = 10.45, Z = 3.75, p_FWE = 0.010). Follow-up t-tests revealed that individuals with PTSD given THC had greater left amygdala activation during early presentations of CS+E (vs. CS−) compared to PBO (24 voxels, t = 4.24, Z = 3.45, xyz = −24, −4, −12, p_FWE = 0.031; See Fig. 9).

Fig. 9. BOLD fMRI activation of the left amygdala showing a drug x stimulus x time interaction in adults with PTSD during early fear renewal.

Individuals with PTSD given THC exhibited greater left amygdala activation to CS+E (> CS−) during early fear renewal, compared to PBO. Results significant at pFWE < 0.05; image shown at p < 0.005 and displays the significant contrast. Left amygdala shown for display purposes only, region was analyzed bilaterally.

3.4. Functional Connectivity

To follow-up our functional activation analyses, we also investigated functional connectivity of our primary results: vmPFC activation during extinction learning in adults given THC, vmPFC activation during fear renewal in adults given THC, and left amygdala activation in adults with PTSD during fear renewal. The primary activation result coordinate served as the seed region, with target regions being the other a priori ROIs (amygdala, hippocampus, vmPFC, and dACC). We found no significant effects in these follow-up analyses.

4. Discussion

Successful fear extinction learning is a cornerstone of first-line behavioral treatments for trauma-related disorders, like PTSD. The eCB system is a promising target for improving extinction learning and recall, as it is critically involved in fear learning and emotional memory processes that occur within corticolimbic brain regions (Mayo et al., 2021). While the effect of THC on extinction learning in healthy adults was previously demonstrated, this effect has not yet been extended to trauma-exposed adults. To our knowledge, we are the first to test the effects of THC on neurobehavioral indices of fear extinction memory recall and fear renewal in trauma-exposed adults with and without PTSD. We aimed to determine if: (1) the population we recruited had differential neural and/or behavioral indices of fear between healthy individuals with and without trauma exposure (TEC, HC) and those with PTSD, (2) individuals with PTSD given THC performed similarly to their healthy counterparts during a Pavlovian fear conditioning task, and (3) individuals with PTSD given THC displayed altered neural and/or behavioral indices of fear compared to their healthy counterparts given PBO. We found no effect of drug condition or trauma on US expectancy ratings. However, we found an effect of drug condition on functional brain activation, such that individuals with PTSD given THC had greater vmPFC activation during early extinction learning, compared to those in the TEC group. Both HC and PTSD given THC had greater vmPFC activation to the CS+E compared to TEC during fear renewal. Additionally, individuals with PTSD given THC had greater left amygdala activation to the CS+E during fear renewal compared to those given PBO.

During early extinction learning, individuals with PTSD given THC had greater vmPFC activation compared to their TEC counterparts who were also given THC. VMPFC activation is often associated with successful extinction recall, as it prevents fear responding by inhibiting output from the centromedial amygdala (Berretta et al., 2005; Milad & Quirk, 2002; Phelps et al., 2004). However, there is also evidence to support its role during extinction learning, as activating infralimbic (vmPFC) neurons in rodents during extinction learning both (1) increases the rate of extinction and (2) improves successful extinction recall the next day, as evidenced by lesser freezing behavior (Milad & Quirk, 2002). Prior work demonstrates that adults with PTSD exhibit lesser vmPFC activation during extinction learning, compared to healthy adults (Milad, Wright, et al., 2007; Suarez-Jimenez et al., 2020). Extinction learning and recall success is also affected by intra-infralimbic (i.e., vmPFC) eCB modulation. Specifically, administration of a CB1R antagonist into the infralimbic cortex impairs within-session extinction learning, while CB1R agonism enhances extinction learning by decreasing fear-potentiated startle in rodents the next day (Lin et al., 2009; Pamplona et al., 2008). Therefore, both vmPFC activation and eCB function within the vmPFC are necessary for within-session extinction, and likely influences later recall. Our findings suggest that THC may help to rescue dampened vmPFC activation during fear extinction in adults with PTSD, which may play a role in improving extinction learning and retention. However, our ability to discern whether THC-related increased vmPFC activation in individuals with PTSD relates to behavioral measures of extinction learning is limited, as we found no effect of THC on US expectancy ratings, within or across trauma or drug groups. This is not entirely unexpected, as Rabinak et al. (2013b) do not report differences in fear indices between healthy adults who receive THC or PBO during extinction learning.

Fear renewal (i.e., return of previously extinguished fear responses) is a limitation of exposure-based therapies, as extinction learning is highly context dependent. We found that individuals with PTSD given THC had greater amygdala activation to CS+E (> CS−) during fear renewal compared to individuals with PTSD who received PBO. Fear renewal requires hippocampus activation in concert with the amygdala and prefrontal regions, like the dACC (Hermann et al., 2016; Jin & Maren, 2015; for review see Ji & Maren, 2007). Despite the elevated expression of CB1R in these regions, as well as the role of the eCB system in fear responding, limited work determines the role of this system during fear renewal. In healthy adults, genetic variation in eCB function and inhibition of eCB degradation, both producing elevated eCB levels, have no effect on fear renewal in a fear-potentiated startle paradigm (Mayo, Asratian, Lindé, Holm, et al., 2020). In addition to greater amygdala activation, we also found that HC and individuals with PTSD given THC exhibited greater vmPFC activation to CS+E (> CS+U) during fear renewal, compared to TEC. Translational evidence suggests that dorsal areas of the PFC are largely involved in fear renewal processes, while ventral areas usually contribute to extinction recall (Åhs et al., 2015; Wang et al., 2016). Considering the increased amygdala activation during fear renewal, administering THC during extinction learning in those with PTSD may prevent or lessen contextual return of fear. However, whether greater vmPFC and amygdala activation translates to recalling extinction is unclear, as we did not detect any differences in US expectancy ratings within or across trauma or drugs groups.

During early extinction recall, we expected to detect differences in neural and/or subjective ratings of fear between healthy controls (HC, TEC) and PTSD given PBO, as well as between those with PTSD given THC or PBO. However, we did not detect these differences. There is some evidence that vmPFC and hippocampus activation is lesser in those with PTSD compared to controls during extinction recall, but a recent meta-analysis suggests the opposite (Milad et al., 2009; Suarez-Jimenez et al., 2020). There may be more nuances to aberrant corticolimbic activation in those with PTSD that we are unable to capture in this sample, especially considering the complexity of the disorder and its (numerous) clinical presentations (Hickling et al., 2013). With respect to the lack of differences between those with PTSD given THC or PBO, a majority of prior eCB modulation work is done exclusively in healthy adults (i.e., (Mayo, Asratian, Lindé, Morena, et al., 2020; Rabinak et al., 2013, 2014). It may be that, with the added complexity of significant trauma exposure, modulation with THC may only be beneficial for extinction recall in select individuals with a specific clinical presentation and/or eCB deficit (Mayo et al., 2021).

Several limitations of this study warrant mention. First, we were unable to collect enough SCR data for our hypotheses. SCR is a gold-standard measure of non-conscious threat processing in humans, and is used throughout the PTSD and fear conditioning field as a marker of extinction recall (Hermann et al., 2016; Milad, Wright, et al., 2007). Lack of usable data limited our ability to fully interpret neural activation findings. However, we still report interesting and worthwhile data regarding THC’s effects in trauma-exposed adults that will greatly contribute to understanding the role of the eCB system in emotional learning and memory. Second, we report on a relatively small sample given the number of groups in this study. Our sample size may have limited our ability to report small, but meaningful, effects, especially as it relates to our behavioral data. Future studies may benefit from fewer groups in an experimental design (e.g., eliminating a control group). Third, we did not collect blood samples to determine individual levels of THC or its metabolites. Genetic variation in several genes (e.g., CNR1, FAAH) may contribute to individual differences in THC metabolism, which may also affect how THC influences behavior or neural activation (Hryhorowicz et al., 2018). However, subjective drug effects are correlated with blood concentrations of THC and all participants who received THC report experiencing similar drug effects (Vandrey et al., 2017; see our previous publications using different tasks with this dataset: Pacitto et al., 2022; Rabinak et al., 2020). Future studies should collect blood samples from participants throughout the extinction learning visit, as well as the following visit (extinction recall and fear renewal), to determine (1) levels of THC and its metabolites in each participant, (2) if there are ‘lingering’ metabolites of THC the following day, and (3) if these concentrations of THC/metabolites contribute to or predict extinction retrieval. Future studies should also consider administering smoked/vaped THC to circumvent first-pass metabolism and metabolic variability that occurs when taking an oral capsule. Lastly, we report on a single dose of THC and how it affects fear indices and neural activation. In clinical application, THC may be administered several times (i.e., prior to therapy sessions) to improve extinction learning and recall, similar to prior work with other novel, adjunctive pharmacotherapies for PTSD (i.e., d-cycloserine; Rothbaum et al., 2014). Thus, whether the effects we report also reflect what would occur after multiple doses of THC, or if multiple doses of THC would be tolerable to patients, is unclear. Despite this, our study is an important step towards understanding the role of the eCB system in emotional learning and memory in those with and without significant trauma exposure.

5. Conclusion

Taken together, these findings suggest that an acute, low dose of oral THC may improve corticolimbic activation during extinction learning and fear renewal in adults with PTSD. Our results extend prior preclinical and human data that the eCB system plays a pivotal role in learning and memory and may be a viable avenue for therapeutic intervention. We also add to the limited literature determining the neural and behavioral indices of fear renewal in trauma-exposed adults. Overall, these data add to the literature that the use of cannabinoid-like compounds as adjunctive pharmacotherapies for exposure-based therapies may be beneficial for those with PTSD.

Highlights:

• PTSD is characterized by poor fear extinction and aberrant corticolimbic activity.

• Endocannabinoid system activation in corticolimbic regions facilitates extinction.

• Acute THC increased corticolimbic activation in PTSD during fear extinction.

• Acute THC also increased corticolimbic activation in PTSD during renewal of fear.

• Acute THC did not significantly affect subjective ratings of fear.

Abbreviations:

THC

delta-9-tetrahydrocannabinol

PTSD

posttraumatic stress disorder

TEC

trauma-exposed control

HC

healthy control

vmPFC

ventromedial prefrontal cortex

dACC

dorsal anterior cingulate cortex

CS

conditioned stimulus

US

unconditioned stimulus

ROI

region of interest