Long-Term Outcomes of Adolescent THC Exposure on Translational Cognitive Measures in Adulthood in an Animal Model and Computational Assessment of Human Data

Key Points

Question

Is there an association between adolescent high-dose Δ⁹-tetrahydrocannabinol (THC) exposure and cognitive vulnerability in adulthood?

Findings

In this rat model, high-dose THC exposure during adolescence resulted in risky decision-making and impulsivity in adulthood, similar to results from reanalyzed data from 37 human chronic cannabis users, an association enhanced by acute THC reexposure. Adolescent THC exposure induced cell-specific and layer-specific changes in cannabinoid-1 receptor gene expression and astrocyte perturbations in the amygdala and prelimbic cortex.

Meaning

These results emphasize significant neurobiological outcomes of high-dose adolescent THC exposure and cognitive vulnerability in adulthood.

This study leverages a human data set of cannabis users and a rat model to evaluate the long-term outcomes of adolescent Δ⁹-tetrahydrocannabinol (THC) exposure on adult decision-making and impulse control.

Abstract

Importance

Although perceived as relatively harmless and nonaddictive, adolescent cannabis use significantly increases the likelihood of developing cannabis use disorder in adulthood, especially for high-potency cannabis. Risky decision-making is associated with chronic cannabis use, but given confounds of human studies, it remains unclear whether adolescent cannabis exposure and Δ⁹-tetrahydrocannabinol (THC) potency specifically predicts risky decision-making or influences cognitive response to the drug later in life.

Objective

To leverage a human data set of cannabis users and a rat model to evaluate the long-term outcomes of adolescent THC exposure on adult decision-making and impulse control.

Design, Setting, and Participants

This translational rat study tested the link between adolescent THC exposure and adulthood decision-making. A reanalysis of a previously published dataset of human chronic cannabis users was conducted to evaluate decision-making phenotypes. Computational modeling assessed the human and animal results in a single framework. Data were collected from 2017 to 2020 and analyzed from 2020 to 2022.

Main Outcomes and Measures

Decision-making was measured by the Iowa Gambling Task (IGT) and Rat Gambling Task (rGT). Impulse control was assessed in the rat model. Computational modeling was used to determine reward and punishment learning rates and learning strategy used by cannabis users and THC-exposed rats. Cell-specific molecular measures were conducted in the prefrontal cortex and amygdala.

Results

Of 37 participants, 24 (65%) were male, and the mean (SD) age was 33.0 (8.3) years. Chronic cannabis users (n = 22; mean [SE] IGT score, −5.182 [1.262]) showed disadvantageous decision-making compared with controls (n = 15; mean [SE] IGT score, 7.133 [2.687]; Cohen d = 1.436). Risky choice was associated with increased reward learning (mean [SE] IGT score: cannabis user, 0.170 [0.018]; control, 0.046 [0.008]; Cohen d = 1.895) and a strategy favoring exploration vs long-term gains (mean [SE] IGT score: cannabis user, 0.088 [0.012]; control, 0.020 [0.002]; Cohen d = 2.218). Rats exposed to high-dose THC but not low-dose THC during adolescence also showed increased risky decision-making (mean [SE] rGT score: vehicle, 46.17 [7.02]; low-dose THC, 69.45 [6.01]; high-dose THC, 21.97 [11.98]; Cohen d = 0.433) and elevated reward learning rates (mean [SE] rGT score: vehicle, 0.17 [0.01]; low-dose THC, 0.10 [0.01]; high-dose THC, 0.24 [0.06]; Cohen d = 1.541) during task acquisition. These animals were also uniquely susceptible to increased cognitive impairments after reexposure to THC in adulthood, which was correlated with even greater reward learning (r = −0.525; P < .001) and a shift in strategy (r = 0.502; P < .001), similar to results seen in human cannabis users. Molecular studies revealed that adolescent THC dose differentially affected cannabinoid-1 receptor messenger RNA expression in the prelimbic cortex and basolateral amygdala in a layer- and cell-specific manner. Further, astrocyte glial fibrillary acidic protein messenger RNA expression associated with cognitive deficits apparent with adult THC reexposure.

Conclusions and Relevance

In this translational study, high-dose adolescent THC exposure was associated with cognitive vulnerability in adulthood, especially with THC re-exposure. These data also suggest a link between astrocytes and cognition that altogether provides important insights regarding the neurobiological genesis of risky cannabis use that may help promote prevention and treatment efforts.

Introduction

Important steps have been made in recent years to decriminalize and/or legalize cannabis, which helped to diminish stigma and incarceration for its use. Unintentional consequences of these efforts also occurred, including greater cannabis availability and many teenagers believing that cannabis is without harm.^1,2 This includes a significant drop in the perception that regular or occasional use is a risk to health.¹ This belief, however, is at odds with data indicating that regular cannabis use during this critical developmental period increases the risk of developing a number of psychiatric conditions,³ including cannabis use disorder (CUD).⁴ Based on the high prevalence of cannabis use in society, approximately 10% to 30% of regular cannabis users meet the criteria for CUD,^5,6 characterized in the DSM-5 as a chronic, relapsing disorder similar to other substance use disorders.⁷ Parallel to legalization/decriminalization efforts, cannabis Δ⁹-tetrahydrocannabinol (THC) potency has also risen dramatically,^8,9 and recent data suggest that high-potency cannabis is specifically associated with greater CUD risk.^10,11 Unfortunately, limited data exist regarding the direct association between adolescent THC experience and dose with behaviors critical to the genesis of CUD. Maladaptive decision-making and poor impulse control, known risk factors for addiction and treatment dropout,^12,13 are exhibited by chronic cannabis users.^14,15,16 Decision-making impairments in cannabis users (as measured by the clinical Iowa Gambling Task [IGT]) are associated with greater reward learning and the tendency to prioritize rewarding options vs maximizing long-term gains, traits that may foster addictionlike behavior.^15,17 However, numerous factors, including THC potency, frequency of use, and environmental influences, may contribute to the cognitive sequelae associated with adolescent cannabis use and CUD risk.

Preclinical models allow for potential causal interpretations about certain risk factors and have shown that adolescent THC exposure can precipitate addictionlike behaviors and changes in reward learning.¹⁸ Although these studies have delineated effects on behaviors critical to the development of substance use disorders, few have investigated decision-making and impulsivity^19,20 and whether dose plays a role in these cognitive phenotypes. In this translational study, we assessed the association of adolescent low-dose or high-dose THC exposure with decision-making and impulsivity using a validated preclinical analogue of the IGT, the Rat Gambling Task (rGT).^21,22 To determine the translational capacity of the results, we used computational modeling to compare rGT and IGT data from human chronic cannabis users. We subsequently conducted molecular measures in the rat model of the medial prefrontal cortex (mPFC) and basolateral amygdala (BLA) to assess laminar-specific and cell-specific effects of THC and potential associations with cognitive behavior.

Methods

For a full description of the study methods, see the eMethods in the Supplement.

IGT Participants

IGT data were obtained from a previously published experiment that underwent institutional review board approval (eTable 1 in the Supplement).^15,23 All participants provided written informed consent. Chronic cannabis use was defined as more than 25 days of use per month (which correlates with cannabis dependence metrics²⁴) for a minimum of 5 years; cannabis use was initiated as adolescents (mean [SD] age, 16.4 [5.4] years). Participants abstained from cannabis use for at least 12 hours before the study (confirmed by urine toxicology). Control participants had a maximum of 100 lifetime uses of cannabis with no past-year use. Groups did not differ for years of education, sex, amount of daily cigarette use, weekly alcohol consumption, or scores of depression, anxiety, or alcohol use. However, a marginal difference in estimated full-scale IQ was detected.¹⁵

Iowa Gambling Task

For the IGT¹⁵ (eFigure 1A in the Supplement), participants started the task with $2000 in play money and were told the purpose of the game was to earn as much money as possible. To motivate play, participants were told whoever accumulated the largest sum would be awarded a bonus of $50. Participants made a series of 100 card selections from 4 decks labeled A, B, C, and D by pressing one of 4 buttons on a screen. Each deck was associated with probabilistic gains and penalties such that some cards from decks A or B resulted in a net loss of $250, whereas decks C and D produced a net gain of $250. Therefore, decks A and B had large apparent wins but also significant losses, ultimately making them disadvantageous, whereas decks C and D had small wins but small penalties, making selection of these decks more optimal. After each selection the computer presented the outcome.

Rat THC Exposure and rGT Model

An overview of the task and preclinical experiment can be found in eFigure 1 in the Supplement. As previously described,^25,26 male Long-Evans rats received intraperitoneal low-dose THC (1.5 mg/kg) or high-dose THC (5 mg/kg) or vehicle every third day throughout adolescence (postnatal day 28 to 59). This dosing protocol was meant to model recreational exposure associated with increased CUD risk (1 to 2 uses per week), as weekly use significantly predicts problematic cannabis use in adulthood.^27,28 rGT training²¹ began at postnatal day 61 (young adulthood; eFigure 1 in the Supplement). During daily 30-minute sessions, animals nose-poked into illuminated apertures associated with different amounts of reward (1 to 4 sucrose pellets), length of penalty timeout (5 to 40 seconds), and probability of winning a reward over punishment (0.9 to 0.4). Like the IGT, because of the length of punishment and low probability of winning, choice of larger reward options (P3 and P4) ultimately led to fewer sucrose pellets over the course of the session, while choice of smaller options (P1 and P2) was most advantageous. Responses made prior the presentation of the options were recorded as premature, or impulsive, responses. The number of trials completed, omitted trials (a trial where no choice was made), and latencies to choose and collect reward were also recorded. After stable responding was achieved (no effect of session, taking approximately 30 sessions), all rats were reexposed to acute doses (vehicle, 0.5, 1, and 2 mg/kg) of THC in adulthood 30 minutes prior to the rGT to assess adult cognitive response to the drug using a counterbalanced, within-subjects design (eFigure 1 in the Supplement).

Computational Modeling

Maximum a posteriori estimation was used to model learning dynamics and value assignment for the 4 choices in the rGT and IGT in a trial-by-trial manner. Models had 3 free parameters: inverse temperature β, reward learning rate (η+), and punishment learning rate (η−).^15,17,29 Learning rates reflect values assigned to rewards and punishments. The inverse temperature β models the degree to which estimates of value influence choice of each option. High β values indicate a strategy that exploits current value estimates of choices (typically in favor of high-value options), whereas low β values indicate a strategy of choice exploration where the value of each option has less impact on decisions. Learning rates for reward affect the magnitude of value updates over each additional trial, based on whether a reward or punishment was obtained. Model fitting and parameter estimation were conducted on all rGT (sessions 1 to 30) and IGT trials (trials 1 to 95), along with separate blocks of acquisition trials (sessions 1 to 15 in the rGT and trials 1 to 72 in the IGT, each broken into 3 blocks) and after THC reexposure to model learning and choice dynamics at different stages of the respective tasks.

Molecular Brain Measures

Two weeks after the final acute THC reexposure session, brains were extracted and the PFC and BLA assayed for cell-specific (γ-aminobutyric acid [GABA], glutamate, and astrocyte) cannabinoid-1 receptor (Cnr1) and bulk messenger RNA (mRNA) expression of astrocyte genes using fluorescent in situ hybridization (RNAScope assay; Bio-Techne) and quantitative polymerase chain reaction (for reagent information, see eTable 2 in the Supplement).

Statistical Analysis

Repeated-measures analysis of variance (ANOVA) or mixed-effects models were used to assess behavioral and computational parameters across sessions, acute THC doses, and for cell-specific analyses (RNAScope assay). One-way ANOVAs were used to assess variables at stable performance and quantitative polymerase chain reaction data. All analyses included adolescent dosing group (vehicle, low-dose THC, and high-dose THC) or cannabis group (cannabis or controls) as between-subjects factors. For significant main effects, post hoc 1-way ANOVAs and t tests were used with false discovery rate corrections. For all analyses, results were deemed significant if 2-tailed P values were less than .05.

Results

Chronic Cannabis Use in Humans and Rat Adolescent High-Dose THC Exposure and Risky Decision-Making

Of 37 included participants, 24 (65%) were male, and the mean (SD) age was 33.0 (8.3) years. This study included 22 cannabis users and 15 controls. As reported previously,¹⁵ human cannabis users preferred the risky options of the IGT (Figure 1A; eResults and eFigure 2 in the Supplement; group, F_3,89 = 15.80; P < .001), which was maintained through stable performance (Figure 1B). In the rat model, adolescent THC dose differentially affected decision-making during task acquisition (Figure 1C; eFigure 2 in the Supplement; adolescent THC exposure, F_2,98 = 7.552; P < .001). Rats exposed to low-dose THC demonstrated greater advantageous choice and performed more optimally compared with rats exposed to high-dose THC and vehicle animals. In contrast, rats exposed to high-dose THC adopted a risky decision-making profile early in training similar to cannabis users. However, by the end of training, high-dose THC–exposed rats did not differ from control rats (Figure 1D). Adolescent THC dose also affected impulse control, where once stable, low-dose THC and high-dose THC showed marginally blunted and potentiated impulsivity, respectively (eFigure 3 in the Supplement).

Figure 1. Decision-making After Δ⁹-Tetrahydrocannabinol (THC) Exposure During Acquisition and at Stable Performance.

On the Iowa Gambling Task (IGT), cannabis users (n = 22) showed greater risky choice compared with controls (n = 15) during acquisition (A; group, F_3,89 = 15.80; P < .001; trial block × group, F_3,99 = 5.684; P = .001; Cohen d = 1.436) and at stable performance (B; t₃₅ = 4.586; P < .001; Cohen d = 1.536). Controls showed a sharp increase in advantageous choice (compared with trials 1 to 24, whereas cannabis users initially increased their score in the second trial block, which leveled off for the remainder of training (trials 1 to 24 vs trials 25 to 48, t₁₉ = 3.539; P = .01). In the Rat Gambling Task (rGT), low-dose THC–exposed rats (n = 28) and high-dose THC–exposed rats (n = 28) showed differential task acquisition (C; session, F_2,213 = 4.970; P = .006; adolescent THC exposure, F_2,98 = 7.552; P < .001), whereas low-dose THC–exposed rats showed greater optimal choice, initially showing nonsignificant differences by session 10 but then ultimately significantly differed from both vehicle rats (n = 45) and high-dose THC–exposed rats by session 20 through session 30. In contrast, high-dose THC–exposed rats showed greater risky choice during training, starting at session 10 that became significant by mid-acquisition. At stable performance (D), there was a significant effect of group (F_2,73 = 6.202; P = .003) where the 2 THC groups differed significantly (t₄₂ = 3.502; P = .001; Cohen d = 0.874) and low-dose THC–exposed rats performed better than controls (t₇₁ = 3.076; P = .002; Cohen d = 0.702); high-dose THC–exposed rats no longer differed from vehicle rats (t₅₁ = 1.076; P = .10). Graphs indicate group means with standard errors of the mean. Trials 1 to 72 and sessions 1 to 20 reflect the acquisition period. Trials 73 to 95 and session 27 to 30 reflect average stable performance.

^aP < .05.

^bP < .01.

^cP < .001.

Computational modeling was used to assess learning rates and learning strategy in the IGT and rGT during task acquisition, a period of high exploratory behavior. Replicating previous studies,^15,17 cannabis users showed greater reward learning rates that sustained into stable performance (Figure 2A and E; eFigure 4 in the Supplement; t₃₃ = 5.492; P < .001). Greater reward sensitivity was strikingly mirrored in the high-dose THC–exposed rats, with increased η+ midway through training that also persisted into stable performance (Figure 2B and F; eFigure 4 in the Supplement; adolescent THC exposure, F_2,72 = 5.177; P = .008). Unlike reward learning, neither the cannabis participants nor rats with adolescent THC exposure differed for punishment learning (eFigure 5 in the Supplement). Cannabis users also consistently showed reduced β values (Figure 2C and G; eFigure 6 in the Supplement; t₃₅ = 6.625; P < .001), indicative of a decision-making profile reliant on immediate outcomes. In the rat model, high-dose THC–exposed rats had reduced β values, whereas low-dose THC–exposed rats showed elevated β values (Figure 2D; eFigure 6 in the Supplement; adolescent THC exposure, F_2,98 = 5.764; P = .004) early in training. However, midway through training and through stable performance, although low-dose THC–exposed β values remained high, it was no longer significantly different (Figure 2H; eFigure 6 in the Supplement). Therefore, in contrast to cannabis users, although high-dose THC exposure promoted reward learning, it did not robustly alter decision-making strategy, and low-dose THC exposure promoted a consistent optimal strategy. In both the IGT and rGT, β and reward-learning predicted stable choice preference, but punishment learning did not (eFigure 7 in the Supplement). In the rGT, these parameters were also associated with levels of impulsivity (eFigure 7 in the Supplement), indicating that these learning parameters were associated with cognitive control in general.

Figure 2. Computational Modeling of Decision-Making After Δ⁹-Tetrahydrocannabinol (THC) Exposure During Acquisition and at Stable Performance.

In the Iowa Gambling Task, cannabis users (n = 22) showed greater reward learning rates during training compared with controls (n = 15), which reached significance at stability (A and E; acquisition: trial block, F_2,59 = 11.09; P < .001; group, F_1,33 = 3.935; P = .06; stable performance: group, t₃₃ = 5.492; P < .001; Cohen d = 1.895). This pattern was similar in high-dose THC–exposed rats (n = 28), both in acquisition (B; block × adolescent THC exposure, F_4,255 = 7.725; P < .001; sessions 11 to 15 compared with vehicle rats [n = 45]: t₂₄ = 3.04; P = .009; Cohen d = 1.541 compared with low-dose THC–exposed rats [n = 28]: t₂₄ = 3.563; P = .005) and at stable performance (F; adolescent THC exposure, F_2,72 = 5.177; P = .008; vehicle rats: t₇₂ = 2.770; P = .004; Cohen d = 0.773; low-dose THC–exposed rats: t₇₂ = 2.895; P = .005; Cohen d = 0.894; vehicle rats vs low-dose THC–exposed rats: t₇₂ = 0.431; P = .67). Cannabis users showed significantly reduced β during acquisition (C; trial block × group, F_2,68 = 4.867; P = .01; block 1 to 24: t₃₄ = 0.185; P = .30) and stable performance (G; t₃₅ = 6.625; P < .001; Cohen d = 2.218). In the Rat Gambling Task, there was a main effect of dose for β (D; adolescent THC exposure, F_2,98 = 5.764; P = .004). Early in training, low-dose THC–exposed rats showed higher levels of β during training (compared with vehicle rats: t₄₄ = 2.503; P = .008; Cohen d = 0.716; compared with high-dose THC–exposed rats: t₄₀ = 3.811; P < .001; Cohen d = 1.069), although this became nonsignificant midway through training and by stable performance (stable: F_2,72 = 0.655; P = .52). High-dose THC–exposed rats initially had low β (sessions 1 to 5 compared with vehicle rats: t₆₅ = 1.771; P = .03; Cohen d = 0.353) but quickly matched vehicle rats by sessions 6 to 10 through to stable performance (H). Line graphs represent group means with standard errors of the mean. E-H, Distributions show relative frequencies of each group for the different trial and session blocks analyzed. Trials 73 to 95 and sessions 27 to 30 reflect average stable performance.

^aP < .05.

^bP < .01.

^cP < .001.

High-Dose THC Exposure and Cognitive Vulnerability to Adult THC Re-exposure

To determine whether adolescent THC exposure was associated with cognitive response to THC in adulthood, rats were reexposed to acute THC doses before the rGT. High-dose THC–exposed rats showed a nonsignificantly increased disadvantageous choice (Figure 3A; eFigure 8 in the Supplement; adolescent THC exposure × dose, F_6,159 = 2.086; P = .06) and significantly increased impulsivity (Figure 3B; eFigure 8 in the Supplement; adolescent THC exposure × dose, F_6,161 = 3.724; P = .002). In contrast, low-dose THC–exposed rats and vehicle rats showed nonsignificantly improved choice and reduced impulsivity at the highest dose (low-dose THC–exposed rats: t₁₂ = 3.059; P = .06). Acute THC exposure increased omitted responses and choice latencies in all rats but did not affect trials completed or sucrose reward collection latencies (eFigure 8 in the Supplement).

Figure 3. Outcomes of Adult Δ⁹-Tetrahydrocannabinol (THC) Reexposure on Rat Gambling Task (rGT) Performance and Computational Parameters.

Acute THC was administered using a within-subjects counterbalanced design. Therefore, all rats were exposed to all acute doses of the drug, but doses were given in counterbalanced order to prevent dose order effects. Adolescent THC exposure was associated with the response to acute THC exposure (vehicle rats [n = 31], 0.5, 1, and 2 mg/kg), with low-dose THC–exposed rats (n = 15) and high-dose THC–exposed rats (n = 14) showing opposite responses in decision-making (A; adolescent THC exposure × dose, F_6,159 = 2.086; P = .06), where high-dose THC–exposed rats had nonsignificantly increased risky decision-making. High-dose THC–exposed rats also showed greater motor impulsivity (B; dose, F_2,490 = 4.308; P = .01; adolescent THC exposure × dose, F_6,161 = 3.724; P = .002), whereas low-dose THC–exposed rats and vehicle rats had reduced impulsivity, although vehicle rats showed no differences after corrections (low-dose THC–exposed rats: t₁₂ = 3.059; P = .06). Computational modeling showed that high-dose THC–exposed rats had increased reward sensitivity (C; adolescent THC exposure, F_2,51 = 3.224; P = .048; Cohen d = 0.635) and decreased β after acute reexposure (D; adolescent THC exposure, F_2,48 = 6.672; P = .003; 2 mg/kg acute THC: high-dose THC–exposed rats vs vehicle rats: t₁₇ = 2.693; P = .03; Cohen d = 0.985). Changes in both reward sensitivity (r = −0.525; P < .001) and β (r = 0.502; P < .001) significantly correlated with changes in decision-making, with high levels of reward learning and decreased β being associated with riskier decision-making (E and F). Bar graphs represent group means with standard errors of the mean.

^aP < .01.

^bP < .05.

Computational modeling revealed that high-dose THC–exposed rats exhibit greater cognitive impairment after THC reexposure and greater reward sensitivity (Figure 3C; eFigure 8 in the Supplement; adolescent THC exposure, F_2,51 = 3.224; P = .048) and a robust drop in β value (Figure 3D; eFigure 8 in the Supplement; adolescent THC exposure, F_2,48 = 6.672; P = .003). These changes correlated with decision-making impairments after the highest THC dose (eFigure 3 in the Supplement). Therefore, high-dose THC–exposed rats readopted a strategy that was sensitive to immediate outcomes and greater reward sensitivity, which was tied to decision-making impairment, similar to human cannabis users. Punishment learning was not affected by THC reexposure (eFigure 8 in the Supplement).

Adolescent THC Experience and mRNA Expression of Cell-Specific Cnr1 and Astrocyte Gfap

Minimal insights exist regarding brain molecular signatures of human cannabis users, given the postmortem necessity of such assessments. Neuroimaging and preclinical studies have revealed reductions in cannabinoid-1 receptor (CB1R) availability in regions critical for cognition, including the amygdala and PFC after chronic cannabis and THC experience.^30,31,32 However, neuroimaging studies lack cell-type resolution. We therefore examined total and cell-specific expression (eFigure 9 in the Supplement) of Cnr1, the gene encoding CB1Rs, in relation to glutamate, GABA, and astrocyte cell populations in the rat model. We studied the BLA, an amygdala nucleus with abundant Cnr1 mRNA expression affected by adolescent THC and implicated in decision-making and impulse control.^{33,34,35,36,37} Total Cnr1 expression was reduced in high-dose THC–exposed rats (Figure 4A; adolescent THC exposure, F_2,28 = 7.332; P = .003). This was specifically recapitulated in the number of GABA cells expressing Cnr1 compared with vehicle rats (Figure 4B; t₁₀ = 2.783; P = .02) but not in the total numbers of GABAergic neurons (eFigure 9 in the Supplement). Overall, the number of Cnr1-expressing GABA cells correlated with changes in impulsivity after THC reexposure in rats with adolescent THC exposure (r = −0.668; P = .02), but not controls (r = −0.037; P = .95; eFigure 9 in the Supplement). Moreover, the number of Cnr1-expressing GABA cells correlated with change in reward sensitivity (r = −0.788; P < .001; eFigure 9 in the Supplement). Only low-dose THC–exposed rats showed altered (reduced) Cnr1 expression in GABA cells (Figure 4C; cell type × adolescent THC exposure, F_4,38 = 3.772; P = .01), indicating that in high-dose THC–exposed rats, the remaining GABA cells that express Cnr1 potentially compensate with increased expression to maintain normal levels. There were no differences of Cnr1 expression in glutamate or astrocytic cells for any group in the BLA.

Figure 4. Cnr1 Messenger RNA (mRNA) Expression Levels From Bulk Quantitative Polymerase Chain Reaction (qPCR) and RNAScope Fluorescent In Situ Hybridization.

Total bulk basolateral amygdala (BLA) qPCR mRNA analysis (A) revealed that rats exposed to high-dose Δ⁹-tetrahydrocannabinol (THC) had significantly reduced Cnr1 expression (compared with vehicle: t₂₈ = 3.740; P = .002; Cohen d = 1.887). RNAScope analyses revealed that this result was recapitulated in the number of Cnr1-positive γ-aminobutyric acid (GABA) regions of interest (ROIs) (B; high-dose THC–exposed rats compared with vehicle rats: t₁₀ = 2.783; P = .02; Cohen d = 1.607; low-dose THC–exposed rats compared with vehicle rats: t₉ = 0.068), albeit not at the omnibus level (cell type, F_2,20 = 24.92; P < .001). Within-cell expression showed that only low-dose THC–exposed rats had reduced Cnr1 expression in GABA cells specifically (C; cell type × adolescent THC exposure, F_2,35 = 3.772; P = .01; compared with vehicle rats: t₈ = 3.797; P = .003; Cohen d = 2.285; compared with high-dose THC–exposed rats: t₇ = 6.174; P < .001; Cohen d = 4.07), but there were no effects of group for glutamate or astrocytic cells. In the prelimbic (PrL) cortex, there were no differences detected on the bulk level (D; adolescent THC exposure, F_2,32 = 1.443; P = .25). However, in situ hybridization showed that there were clear differences between adolescent THC groups in a laminar-specific and cell-specific manner. For the number of Cnr1-positive ROIs, high-dose THC–exposed rats had a greater number of glutamate-positive ROIs in both layers 2/3 (E). In contrast to the number of Cnr1-positive ROIs, only GABAergic cells showed reduced expression of Cnr1 in layers 2/3 (F; cell type × group, F_4,17 = 4.013; P = .02; low-dose THC–exposed rats compared with vehicle rats: t₅ = 6.342; P = .004; Cohen d = 3.279; high-dose THC–exposed rats compared with vehicle rats: t₄ = 4.30; P = .02; high-dose THC–exposed rats compared with low-dose THC–exposed rats: t₄ = 4.274; P = .02; Cohen d = 3.144). Bars represent group means with standard errors of the mean. Empty circles represent individual rats. GAT indicates GABA transporter; GAT1, GABA transporter type 1; GLAST, astrocyte-specific glutamate aspartate transporter; vGlut, vesicular glutamate transporter; vGlut1, vesicular glutamate transporter type 1.

^aP < .01.

^bP < .05.

^cP < .001.

Cnr1 expression was also examined in the prelimbic (PrL) and infralimbic (IL) cortices, mPFC subregions implicated in decision-making and impulse control.³⁸ Total Cnr1 expression did not differ in the PrL cortex (Figure 4D). Conversely, RNAScope revealed stark group differences in Cnr1 laminar and cell-type specific expression. High-dose THC–exposed rats exhibited a greater number of glutamatergic Cnr1-positive cells in both the superficial and deep layers (Figure 4E; eFigure 10 in the Supplement; layers 2/3: cell type × adolescent THC exposure, F_4,36 = 3.842; P = .02; layers 5/6: cell type × adolescent THC exposure, F_4,38 = 4.597; P = .004), with a nonsignificant reduction in astrocytes in layers 5/6. Only GABA layers 2/3 cells showed significantly reduced Cnr1 expression evident in both low-dose THC–exposed and high-dose THC–exposed rats (Figure 4F; eFigure 10 in the Supplement; cell type × group, F_4,17 = 4.013; P = .02) but was not associated with differences in total GABA cell numbers (data not shown). The IL cortex showed no group effects (eFigure 10 in the Supplement).

Astrocytes have been shown to be significantly affected by adolescent THC exposure,³⁹ including high-dose THC, as evident with perturbations in astrocyte-specific transcripts and morphology after acute stress.³⁶ Here, high-dose THC–exposed rats exhibited greater BLA Gfap expression, the transcript for the intermediate fibrillary processes of astrocytes (adolescent THC exposure, F_2,29 = 4.466; P = .02), which correlated with cognitive impairment and change in β (Figure 5A-D). An opposite association was observed in the PrL cortex, which correlated with the change in β after THC reexposure (Figure 5E and H; adolescent THC exposure, F_2,14 = 4.394; P = .03). The IL cortex showed no difference in Gfap expression nor association with any behavioral measure (eFigure 11 in the Supplement). Despite the effects evident in BLA and PrL cortex Gfap expression, minimal differences were detected for other astrocyte transcripts (eFigure 12 in the Supplement).

Figure 5. Gfap Messenger RNA Expression in the Basolateral Amygdala (BLA) and Prelimbic (PrL) Cortex and Its Correlation With the Most Pronounced Behavioral Changes and Computational Parameters After Adult Δ⁹-Tetrahydrocannabinol (THC) Reexposure.

In the BLA (A), Gfap expression was increased in high-dose THC–exposed rats (n = 6; adolescent THC exposure, F_2,29 = 4.466; P = .02; compared with vehicle rats [n = 21]: t₂₉ = 2.973; P = .01; Cohen d = 1.163) but not in low-dose THC–exposed rats (n = 6). BLA Gfap expression significantly correlated with change in score (B; r = −0.402; P = .03), impulsivity (C; r = 0.482; P = .006), and β (D; r = −0.469; P = .01). The PrL cortex showed the opposite pattern, where high-dose THC–exposed rats had reduced Gfap expression (E; adolescent THC exposure, F_2,14 = 4.394; P = .03; high-dose THC–exposed rats compared with vehicle rats: t₂₅ = 2.789; P = .03; Cohen d = 0.975). PrL cortex Gfap expression did not correlate with the change in score or impulsivity (F and G) but did correlate with change in β (H; r = 0.553; P = .001). Empty circles represent individual rats.

^aP < .05.

Discussion

This translational study provides evidence that THC dose is associated with the protracted outcomes of adolescent THC exposure on decision-making and impulsivity. Similar to chronic cannabis users, high-dose THC–exposed rats demonstrated heightened reward learning and disadvantageous choice early in training. However, only high-dose THC–exposed rats showed exacerbated cognitive deficits, reward sensitivity, and change in strategy after THC reexposure, consistent with phenotypes seen in human chronic users with cannabis exposure initiated during adolescence. These results highlight the long-term cognitive outcomes of THC dose associated with adolescent cannabis use and the enhanced sensitivity when reexposed to the drug later in life.

Consistent with studies of chronic cannabis users,¹⁷ high-dose THC–exposed rats had elevated reward learning rates. Importantly, reward learning predicted risk preference and impulsivity as well as increased risky choice under the influence of THC, indicating a potential contribution to these cognitive deficits. Reduced β values seen in cannabis users is a consistent feature in substance use populations¹⁷ and may reflect sensitivity to explore options based on immediate outcomes vs long-term gains, a phenotype strongly associated with drug-seeking behavior.^40,41 The low β values and risky choice also seen in high-dose THC–exposed rats during baseline IGT conditions did normalize after extensive training. Nevertheless, on reexposure to THC, these rats uniquely exhibited a remarkable drop in β values that correlated with risky choice. Unlike human cannabis users who had a short abstinent period (approximately 12 hours since last cannabis use) before testing, high-dose THC–exposed rats exhibited the risky behavioral phenotypes after acute reexposure. Further investigation is needed to examine the duration of the cognitive phenotype following acute THC reexposure as well as the outcomes of repeated THC reexposure.

Interestingly, low-dose THC–exposed rats adopted a value-based decision-making strategy early in training accompanied by greater advantageous choice. Improvement of value-based decision-making is consistent with our previous study showing that low-dose THC–exposed rats are sensitive to changes in reward value.³⁶ Low-dose THC exposure has also been shown to increase the propensity to self-administer heroin in multiple studies.^25,36 Thus, low-dose THC exposure may increase reward value sensitivity that might promote opioid addiction liability^42,43 but leave cognition intact or even improved. In contrast, high-dose THC–exposed rats do not differ from controls for heroin or sucrose self-administration.³⁶ Therefore, adolescent THC dose produces unique, protracted effects on decision-making and impulsivity but does not increase vulnerability to drugs in general and instead may depend on the type of reward.

Repeated THC exposure is well-demonstrated to alter CB1 receptors, and consistent with previous results,^30,31,32 adolescent THC exposure also had protracted effects on Cnr1. We now also reveal cell-type and layer-specific relevance of these perturbations. Interestingly, significant alterations were only evident in the PrL cortex and not the IL cortex. The PrL and IL cortices have been shown to play unique and sometimes oppositional effects on motivated and affective behavior, including cognition, where the PrL cortex mediates acquisition and expression of goal-directed behavior and the IL cortex promotes extinction and updating of previously learned contingencies.^38,44 In the PrL cortex, both low-dose and high-dose THC exposure in rats had reduced GABA Cnr1 expression in layers 2/3, which shares reciprocal connections with the amygdala.^45,46 High-dose THC exposure had greater numbers of glutamate Cnr1-positive cells in both layers 2/3 and layers 5/6. Increased Cnr1 and the subsequent potential increase of CB1Rs in layers 2/3 on GABAergic interneurons would normally disinhibit glutamatergic pyramidal cells,^47,48 whereas increased CB1Rs in layers 5/6 (localized on glutamatergic terminals⁴⁹) would reduce excitatory activity.⁵⁰ Therefore, if CB1R is reduced in layer 2/3 GABAergic cells but increased on glutamatergic terminals, one would expect reduced excitatory output from the PrL cortex. In the BLA, GABAergic CB1R on interneurons gates projection excitability.^51,52 Therefore, reduced GABA Cnr1 seen in THC–exposed rats would likely suggest greater BLA activity. Although functional activity and protein expression were not assessed in a cell-specific manner in this study, a substantial body of literature supports dysfunction in corticolimbic regions of human cannabis users. For example, individuals who initiate cannabis use during adolescence show accelerated PFC thinning in areas of high CB1 expression associated with increased impulsivity.⁵³ Furthermore, individuals diagnosed with CUD have blunted mPFC activity while performing the IGT.^7,54 Additionally, a number of studies have demonstrated that animals with low-dose THC exposure have reduced PrL cortex layer 2/3 dendritic branching,⁵⁵ and chronic THC exposure in adolescence or adulthood reduces activity in the mPFC^56,57 but increases BLA activity.⁵⁶ Altogether, the molecular patterns we observed are consistent with morphological and functional changes documented in the literature but highlights cell-specific effects within the PrL cortex and BLA induced by adolescent THC exposure.

Expression of the astrocyte plasticity mark Gfap correlated with cognitive deficits after THC reexposure in high-dose THC–exposed rats. Astrocytes express CB1 and endocannabinoid signaling modulates astrocyte synaptic function.^58,59,60 A growing body of evidence implicates astrocytes in the effects of THC³⁹ and addictionlike behaviors.^61,62 Changes in astrocyte plasticity have also been detected weeks after final drug administration^39,61; therefore, alterations detected in high-dose THC–exposed rats may also reflect protracted effects of the adolescent experience. Determining the time course of astrocyte perturbations in structure and function during training and after THC reexposure will help elucidate the contribution of these cells to cognitive vulnerability.

Limitations

This study has limitations. Compared with the IGT, acquisition of the rGT requires several sessions, and we did not assess rGT performance in relation to repeated THC exposure in adulthood or during acute withdrawal (a period more comparable with human cannabis users). Furthermore, parenteral acute THC exposure, which also lacks the diverse phytoconstituents found in the cannabis plant, is not a direct proxy for volitional intake of smoked cannabis. Also, our THC administration protocol extended from early to late adolescence, while most of cannabis users reported initiating use in mid to late adolescence.⁶³ Despite these limitations and evident differences between humans and animal models, the results replicated behavioral vulnerability that is similar to cannabis users and thus provides a valuable model to assess the long-term outcomes of THC exposure. We also did not assess sex differences. Although more male users use cannabis and develop CUD, female users rapidly develop CUD,⁷ and there are known sex differences in relation to metabolism and behavioral responses to adolescent THC exposure.^64,65 It is important to examine sex effects in future studies.

Conclusions

In conclusion, this study demonstrates that adolescent THC exposure was associated with dose-differential long-term effects on decision-making, impulse control, and cognitive response to THC reexposure in adulthood. Critically, high-dose THC–exposed rats closely resembled several phenotypes exhibited by human cannabis users, supporting the suggestion that potent THC affects cognition, which may influence CUD risk. Furthermore, cell-specific and laminar-specific perturbation in Cnr1 expression and astrocytes within the PrL cortex–BLA circuit likely underlie behavioral consequences of high-dose THC exposure. Overall, these translational data raise concern regarding the increased THC potency today experienced by adolescents that may affect the genesis of risky cannabis use and increase cognitive vulnerability particularly on cannabis reexposure later in life.

Supplement.

eMethods.

eResults.

eFigure 1. Overview of the IGT and rGT Task Structure and Timeline of the Preclinical rGT Experiment

eFigure 2. Decision-making Performance Broken Down by Individual Choice Selection and Score in Cannabis (CB) Individuals and Rat Model

eFigure 3. Additional Behavioral Measures During Acquisition and at Stable Performance

eFigure 4. Reward Learning Rates During Specific Trial Blocks of the IGT and Session Blocks of the rGT

eFigure 5. Punishment Learning Rates During Specific Trial Blocks and Stable Performance of the IGT and Session Blocks and Stable Performance of the rGT

eFigure 6. Decision-making Strategy Metric β During Specific Trial Blocks of the IGT and Session Blocks of the rGT

eFigure 7. Correlations Between Computational Parameters and Task Performance

eFigure 8. Effects of Adult THC Reexposure on All rGT Measures and Computational Parameters

eFigure 9. Representative RNAScope Image and BLA GABA Cell Counts and Correlations Between Cnr1-Positive GABA Cell Numbers and Behavior

eFigure 10. Cnr1 RNAscope Results From Layer 5/6 of the PrL and the IL

eFigure 11. IL Gfap mRNA Expression and Correlations With Behavior After THC Reexposure

eFigure 12. Results from Bulk qPCR Analyses on Astrocyte-Specific Genes

eTable 1. Demographic Characteristics of Controls and Chronic Cannabis Users

eTable 2. RNAScope In Situ Hybridization and TaqMan Probe Information

eReferences.