Chronic Δ⁹-tetrahydrocannabinol exposure in adolescent nonhuman primates: persistent abnormalities in economic demand and brain functional connectivity

Abstract

Although chronic cannabis use during adolescence can alter brain function and impair complex behavioral processes, it is unclear whether such deficits persist into adulthood. Using a coordinated awake neuroimaging and behavioral approach in nonhuman primates, we addressed this issue by examining the impact of chronic adolescent exposure to Δ⁹-tetrahydrocannabinol (THC) on brain functional connectivity and motivational processes during early adulthood. Female and male squirrel monkeys (n = 23) were treated daily for 6 months during adolescence with vehicle or either a low (0.32 mg/kg) or high dose (3.2 mg/kg) of THC. Regional homogeneity and seed-to-whole-brain functional connectivity were analyzed prior to, during, and following discontinuation of chronic treatment to examine changes in regions implicated in reward processing. Subsequently, motivation and reward sensitivity in these subjects, now young adults, were evaluated in economic demand studies by determining the relationship between escalating response requirements and consumption of differing magnitudes of a palatable food reinforcer. Results show that adolescent THC exposure led to persistent alterations in mOFC, caudate, and ventral striatum whole-brain connectivity. Moreover, subjects treated with vehicle during adolescence displayed an orderly and expected inverse relationship between reward magnitude and demand elasticity, whereas THC-treated subjects exhibited dosage-dependent disorder in reward sensitivity and motivational deficits. Changes in neural circuitry (local connectivity in ventral striatum and whole brain connectivity in mOFC) and economic demand were correlated with indices of reward sensitivity in vehicle- but not THC-treated subjects. Taken together, these data indicate that chronic adolescent THC exposure produced long-lasting neurocognitive abnormalities in reward processing.

Introduction

Numerous studies in humans and laboratory animals suggest that cannabis use during adolescence can alter brain structure and function and impair a variety of complex behavioral processes [1–5]. Adolescent cannabis use has been associated with reductions in cortical gray matter brain volume [6] and altered brain resting-state functional connectivity (RSFC), especially within fronto-temporal and striatal networks [7–9]. Such brain changes may underlie purported deficits in cognitive domains described in adolescent cannabis users [10]. While these reports have fueled concern, the neurocognitive consequences of adolescent cannabis exposure that persist into adulthood remain incompletely understood, hampering assessments of long-term health risks. Human studies suggest some unwanted effects of habitual cannabis use, including changes in brain structure and function and in cognitive abilities, are enduring [11–14]. Considering the major neurodevelopmental changes that occur during adolescence [15–18], it is perhaps not surprising that cognitive functioning may be affected in an enduring manner [19, 20]. Such findings highlight the view of adolescent behavior as a critical determinant of adult health outcomes [21] and emphasize the need for rigorous research to gauge the impact of cannabis use during adolescence.

One longstanding concern has been the impact of adolescent cannabis exposure on motivational processes, which has been suspected of leading to a so-called cannabis amotivational syndrome, characterized by persistent deficits in reward processing that hamper goal-oriented behavior in adulthood [22]. However, evidence for linkage between chronic cannabis exposure and deficits in reward processing has been inconsistent—which may be a methodological problem. For example, a detailed review of the extant neurocognitive literature in humans [23] concluded that the vast majority of data to date, obtained in cross-sectional studies, yields only equivocal evidence of cannabinoid-selective deficits on motivational processes. On the other hand, the only two longitudinal studies [24, 25], notwithstanding variation in assessments of motivation-related constructs, identify a causal link between cannabis use and reduced motivation.

Studies in laboratory animals across reward-related paradigms have also yielded mixed findings regarding motivational deficits produced by cannabinoid exposure. For example, some studies using traditional progressive ratio procedures have shown that break points, one measure of motivation, are lower in Δ⁹-tetrahydrocannabinol (THC)-treated subjects [26]; however, see [27, 28]. Results from other operant and Pavlovian assays of motivated behavior and reward sensitivity have been similarly ambivalent. For example, THC treatment in adolescent rats has been shown to slow the acquisition of reward-predictive cues [29], whereas other reports have documented enhancement of motivation, evident by increased goal-tracking behavior and enhanced Pavlovian-to-instrumental transfer [30] or by increased response rates and palatable food-seeking behavior [27, 31, 32].

The present studies, employing a coordinated neuroimaging and behavioral testing strategy, were designed to further examine the long-term effects of chronic THC exposure during adolescence. They were conducted in squirrel monkeys, which have been frequently used to examine in vivo effects of cannabinoids [28, 33–40]. First, the temporal consistency of blood oxygen level dependent (BOLD) signal for relevant brain regions was assessed locally and globally by measuring, respectively, regional homogeneity (ReHo) and seed-to-whole-brain functional connectivity (FC) before and after chronic treatment with vehicle or THC (low or high dosage) during adolescence. Subsequently, during early adulthood, behavioral performance in a touchscreen-based economic demand task was used to probe potential deficits in motivation and reward sensitivity. Subjects responded under escalating response requirements (price) for a highly palatable food reward (commodity) that varied in magnitude across conditions. Performance outcomes served to objectively quantify the extent to which subjects would defend consumption in the face of increasing price (motivation) and determine their ability to modify responding for consequences that differ in magnitude (reward sensitivity). Behavioral economic endpoints were then correlated with measures of RSFC to identify possible neural correlates of THC-induced changes in motivation and reward sensitivity.

Method

Subjects

Twenty-three (11 female, 12 male) adolescent squirrel monkeys (Saimiri sciureus, aged 2.0 yrs [female] and 2.5 yrs [male]; [41]) were housed in a climate-controlled vivarium with a 12-h light/dark cycle (lights on at 7 a.m.). Subjects had unlimited access to water in the home cage and were maintained at approximate free-feeding weights by post-session feedings of nutritionally-balanced high protein chow (Purina Monkey Chow) and fresh fruit. Environmental enrichment (e.g., social interactions, toys, puzzles, music) was provided daily. The study protocol was approved by the McLean Hospital IACUC in accordance with established guidelines [42].

Touchscreen apparatus

Schematics of the touch-sensitive experimental chamber for squirrel monkeys can be found in [43, 44]. Briefly, a 17” touch-sensitive screen (1739 L, ELO TouchSystems) was mounted on an inside wall of the chamber. An infusion pump (PHM-100-10, Med Associates) delivered a sweetened condensed milk (Casa Solana, Sysco Corporation, Houston, TX) solution prepared daily with room temperature tap water into a custom-designed receptacle. A speaker bar (NQ576AT, Hewlett-Packard) mounted above the touchscreen provided audible feedback. All experimental events were programmed in E-Prime 2.0 (Psychology Software Tools, Inc.).

Drug

Δ⁹-Tetrahydrocannabinol was provided by the NIDA Drug Supply Program (Rockville, MD) and prepared for intramuscular administration in a 20%:20%:60% vehicle mixture of 95% ethanol:Tween-80:0.9% saline.

Research protocols

Figure 1 presents a timeline of experimental events. Following MRI simulator training for awake scanning, whole-brain and region-specific FC were assessed on 3 occasions: immediately prior to chronic treatment (drug-naïve), during the 6th and final month of daily treatment (chronic), and 60 days following chronic treatment (discontinuation). Economic demand procedures were initiated one month following the discontinuation scan (i.e., 3 months following termination of chronic THC exposure). This schedule was specifically designed to ensure that awake scans, aimed at examining neural changes following extended treatment during adolescence (chronic), as compared to baseline (drug-naïve), were conducted in early adulthood—well beyond any potential withdrawal (discontinuation) effects, and that economic demand studies were conducted in THC-free subjects and during early adulthood [45].

Fig. 1. Twelve-month timeline of experimental events.

This was comprised of awake neuroimaging acclimation and repeated scanning, chronic THC treatment during adolescence depending on group assignment, and economic demand testing following a drug washout period during early adulthood.

Chronic dosing

Once-daily intramuscular administration—at approximately 11 am, 7 days per week—occurred for 6 months in the late adolescent period (i.e., 2.5–3.0 [females] and 3.0-3.5 [males] years old [41]). Depending on group assignment, subjects were treated with vehicle (n = 7, 3 females, 4 males), a low dose (0.32 mg/kg; n = 8, 4 females, 4 males), or a high dose (3.2 mg/kg; n = 8, 4 females, 4 males) of THC. Low and high THC chronic doses chosen were based on [28]; the once-daily injection protocol was used to provide stable plasma levels of THC and metabolites throughout chronic treatment [45]. To minimize discomfort that might have followed initial injections of the high dose, subjects in that group underwent a dose escalation protocol (0.32 mg/kg for three days, 1.0 mg/kg for three days) prior to daily treatment with 3.2 mg/kg THC.

Neuroimaging: MRI data acquisition

Protocol and apparatus details used for acclimation to awake MRI scanning can be found in [46]. Under terminal conditions, subjects rested comfortably in a sphinx position within a tubular enclosure for an hour-long session with their head held in place by an individualized helmet lined with padding to limit motion and optimize comfort. A platform to position a transmit/receive surface coil was designed into the helmet. Scan sounds were played within the simulator to enhance habituation. Following acclimation, MRI scan sessions were initiated. Heart rate and oxygen saturation were recorded throughout training and MRI sessions.

MRI images were acquired with a 9.4 Tesla/400 mm diameter MR system (Varian Direct Drive, Varian Inc) using an 11.9 cm gradient bore insert (Resonance Research Inc). Automated image-based shimming <75 Hz and fMRI images were acquired with a whole-brain gradient-echo planar imaging (EPI) sequence (TE = 8 ms, TR = 1500 ms, and flip angle=90°; scan matrix was 64 on a 64 mm field of view with 54 ×1 mm slices); scan time for 1200 volumes was 30 min. Distortion-matched anatomic images were acquired with a spin-echo EPI sequence (TE = 17.5 ms, TR = 1500 ms, flip angle =90°, averages=8; scan matrix=64 on a 64 mm field of view with 54 ×1 mm coronal slices matched to the fMRI slices).

Motivation and reward sensitivity: economic demand

Using previously-described methods [43], subjects were trained to respond on the touchscreen by a technician blinded to their treatment group. Next, consumption of water (0%) and 3 different concentrations of sweetened condensed milk (10%, 30%, 50%) was assessed under conditions of escalating price. Specifically, water and each concentration were tested for 3 consecutive sessions, using the following mixed-order sequence: 30%, 10%, 50%, and 0%. Reward delivery (0.15 mL of the given concentration) was contingent on touch-responses to a centered 7 × 7 cm green square on a purple background and was accompanied by an 880-ms yellow screen flash and 440-Hz tone and followed by a 10-s blackout period. During each 60-min economic demand session, the milk concentration was fixed but the response requirement increased in an ascending order (2, 5, 10, 30, 50) across five 10-min components, each separated by a 2-min inter-component blackout period.

The first session of each 3-session condition served to expose the subject to the new milk concentration and was not included in the final analyses. The number of reinforcers earned during each response requirement component during the second and third session was averaged for each subject. Next, individual subject data were used to obtain mean values for all reward magnitudes in each treatment group. Demand functions for each reward magnitude were generated by fitting the number of reinforcers earned (consumption) during each response requirement component (price) according to the exponential model of demand [47]:

$$\log Q=\log Q_0+k\left(e^{-a{Q}_{0}C}-1\right)$$

where Q is quantity consumed, Q₀ is consumption at price zero (demand intensity), k is a constant representing the range of consumption in log units, α is the rate of change in consumption as price increases (elasticity), and C is cost (response requirement). 0.1 was added to all values to allow logarithmic transforms in the event of zero consumption.

Data analysis

MRI data processing

Data processing used standard pipelines within FSL (Functional Magnetic Resonance Imaging of the Brain Software Library) and AFNI (Analysis of Functional NeuroImages). Quality control was evaluated through slice-by-slice visual inspection for artifacts in all three orthogonal directions with the MRI Quality Control tool (MRIQC; [48]). Group mean motion statistics were calculated in MRIQC. Intensity spiking was analyzed and adjusted using a 1 mm framewise displacement (FD) threshold (https://github.com/bbfrederick/spikefix). FMRIB’s Software Library (FSL, Oxford University) image preprocessing pipeline processed the data after slice-timing correction; the first 10 volumes from each scan were removed to allow for data stabilization. Motion correction by volume realignment to the middle volume was accomplished with the MCFLIRT tool. Session-averaged functional volumes were aligned to the VALiDATe [49] T1w template through a 12 DOF affine transformation followed by adjustment of nonlinear distortion fields using the jip analysis toolkit (www.nitrc.org/projects/jip) for registration. A Gaussian kernel of 2.0 mm FWHM was used for spatial smoothing for seed-based FC analysis; spatial smoothing was excluded for ReHo analyses due to potential spurious inflation of local homogeneity statistics by the averaged signal over a large area [50]. Six motion parameters (three rotational: roll, pitch, yaw; and three translational: X/Y/Z directions) and their first order derivatives were z-transformed and used as nuisance regressors. Finally, to reduce effects of low-frequency drift and high-frequency noise, temporal filtering was applied using a band pass filter (0.01 to 0.2 Hz, fslmaths-bptf).

Brain functional connectivity analyses

Seed-to-whole-brain FC: To investigate whether THC exposure altered FC of demand-related regions to the rest of the brain, regions with documented roles in reward and motivational processing—mOFC, dlPFC, bilateral amygdala, caudate, ventral striatum—were selected as seeds. mOFC and dlPFC regions of interest (ROI) masks were constructed based on functional parcellation of the squirrel monkey prefrontal cortex [51] while subcortical ROIs (bilateral amygdala, caudate, and ventral striatum) were based on previously described anatomical landmarks [52]. AFNI’s 3dNetCorr calculated seed-to-whole-brain Z-scored FC values of the 8 ROIs. Subject-specific time series for each ROI were extracted (AFNI: 3dNetCorr) and analyzed for Fisher Z-transformed Pearson’s correlation connectivity estimates. The Z-scored spatial maps for all subjects across 3 sessions (baseline, chronic THC, discontinuation) were then used in a model treating subjects, session, and THC group as main effects and subject sex, age, and motion as covariates (AFNI: 3dLME) for each ROI independently. Group*session interaction maps were calculated using an uncorrected threshold of p = 0.005. Clusters in the interaction maps were used to extract the average Z-scored FC (AFNI: 3dROIstats).

ReHo analysis

ReHo is a data-driven method to quantify temporal synchronization of BOLD activity within ROIs. ReHo was calculated as the Kendall’s Correlation Coefficient (KCC) metric representing the similarity between a single voxel and the surrounding 26 voxels using AFNI’s 3dReHo. KCC is a coherence measure with high test-retest reliability to establish strong brain-behavior correlations [53]. ROI masks used for seed-analysis were also applied for ReHo (mOFC, dlPFC, amygdala, R/L caudate, and R/L ventral striatum). ROI volumes were then normalized to individual mean whole-brain KCC values. ReHo values were extracted (AFNI: 3dROIstats) for the 8 ROIs.

Exponential demand modeling

The well validated exponential model of demand (detailed above and see [47]) was used to fit all demand functions describing consumption (rewards earned) as a function of price (touchscreen-based response requirement). Goodness of fit using the exponential model was quantified for each demand function in each treatment group using R². In addition, a global fit across all demand functions in all groups was evaluated with an adjusted R² that was weighted for the number of independent predictor variables in its calculation. To assess quality of the model’s predictions (i.e., distance between predicted and actual values), Root Mean Squared Error (RMSE) was also calculated using the global fit across demand functions, in which the value can range from 0 (precise prediction) to positive infinity as more error (less precise prediction) is observed. Finally, a two-way ANOVA with a Geisser-Greenhouse correction, using THC group and milk concentration as factors, was conducted to statistically evaluate the α and Q₀ exponential demand parameters across reward magnitudes tested.

Behavioral and neuroimaging correlations

Pearson correlation analysis in a multiple linear regression model (R:lm) was applied to examine associations between demand metrics—R² for α or Q₀ for various milk concentrations – and change in averaged FC value (discontinuation—baseline scan) in cluster maps from seed-based analysis and ReHo (i.e., KCC) value. Subject age at the discontinuation scan time point, sex, and FD averaged over the scan duration were used as nuisance regressors.

Results

Economic demand results

Figure 2 presents economic demand relationships of consumption (rewards earned) to price (response requirement) during sessions when either 0, 10, 30, and 50% milk was available. The exponential demand equation provided an excellent fit to the findings in all three treatment groups (vehicle: R² = 0.93–1.0; low dose THC: R² = 0.86–0.99; high dose THC: R² = 0.98–1.0). The adjusted R² in the global fit was also high (0.96) and, additionally, the quality of the model’s predictions (i.e., distance between predicted and actual values) in the global fit across demand functions yielded a low RMSE (0.12), indicating very small error between actual data points and the regression line. However, analysis of the orderliness of the relationships between consumption and price across reward magnitudes revealed prominent differences among treatment groups. Subjects treated during adolescence with vehicle displayed orderly demand functions (Fig. 2A). Consumption of 50% milk as a function of price yielded the most inelastic curve, followed by curves for 30%, 10%, and 0%. However, in subjects treated chronically during adolescence either with low dose (Fig. 2B) or high dose THC (Fig. 2C), demand functions were disordered, exhibiting overlap and crossover among the milk concentrations. For example, greater consumption of 30% than 50% milk was observed at lower prices in the low dose THC group and at the higher prices for the high dose THC group. Likewise, 10% milk consumption approximated or exceeded that of 30% milk in the low dose THC group and approximated that of 50% milk consumption at higher response requirements in the high dose THC group.

Fig. 2. Economic demand functions that describe the exponential relationship between price (response requirement) and consumption (number of rewards earned) as a function of percent sweetened condensed milk (reward magnitude).

Mean data were obtained from young adult subjects previously treated daily for 6 months during their adolescence period with either vehicle (Panel A; n = 7), low dose THC (0.32 mg/kg; Panel B; n = 8), or high dose THC (3.2 mg/kg; Panel C; n = 8).

Figure 3 provides a quantitative characterization of order and disorder in demand functions, displaying the relationships between elasticity (α, left panels) or demand intensity (Q₀, right panels) and reward magnitude (top panels) or adolescent THC exposure (middle panels). (To probe the parametric space where order and disorder was observed, demand data from 0% milk concentration [i.e., water] conditions were excluded from these analyses.) Differences in α were statistically significant by milk concentration (F[2,60] = 7.61, p = 0.001), with a trend observed by chronic treatment group (F[2,60] = 2.30, p = 0.11); however, there was no interaction between factors (F[4,60] = 0.31, p = 0.87). As Fig. 3A shows, the orderly and expected negative relationship between milk concentration and α in the vehicle group was not observed in either the low or high dose THC group. Analysis by linear regression (Fig. 3B) indicates an excellent fit of the data for the vehicle group but not the low or high dose THC groups. As the R² values plotted in Fig. 3C show, reductions in goodness of fit between reward magnitude (milk concentration) and reinforcing strength (α) were dosage-related. Finally, Fig. 3A also reveals higher α values under all three milk concentration conditions in both THC treatment groups than in the vehicle group, suggesting THC-related deficits in motivation.

Fig. 3. Economic demand parameter analysis.

Summarized α (elasticity, Panels A–C) and Q₀ (demand intensity, Panels D–F) economic demand parameters derived from exponential functions presented in Fig. 2. Panels (A, D) present mean (±SEM) parameter values by treatment group and reward magnitude; Panels (B, E) present parameters by treatment group with a linear regression fit across reward magnitudes tested; Panels (C, F) present the group-level mean goodness of fit (R²) of linear regressions by treatment group, with data points representing values from individual subjects.

Order (vehicle group) and disorder (THC groups) across differing reward magnitudes were also evident in the demand intensity parameter (Q₀). Statistically significant difference in Q₀ emerged by milk concentration (F[2,60] = 22.72, p < 0.0001) but not treatment group (F[2,60] = 0.66, p = 0.52); however, there was a significant milk concentration*THC group interaction (F[4,60]) = 3.75, (p = 0.009). As expected, concentration-dependent increases in Q₀ in the vehicle group and, to a slightly lesser extent, the high dose THC group were orderly (Fig. 3D). However, this was clearly not the case for the low dose THC group (Fig. 3E). R² values from these linear regressions are depicted in Fig. 3F and confirm a modest reduction in fit of the high dose THC group relative to vehicle, and a substantially poorer fit in the low dose THC group.

Functional connectivity results

Changes in seed-to-whole-brain FC for the 8 seed ROIs were compared using subject-level FC maps for the three imaging sessions (drug-naïve, chronic, discontinuation). These analyses found that FC for three seeds, mOFC, L Caudate, and R vStr (Fig. 4A, shown in green), showed a significant group*session interaction (uncorrected voxel-wise p = 0.005, minimum cluster threshold of 20 voxels) between the seed and target regions (shown in red). FC extracts from target regions for the three treatment groups and three sessions are shown in Fig. 4B. FC between the mOFC and the R visual cortex passed the above threshold. Group*session interactions were noted in multiple regions (see table in Fig. 4) and were especially prominent in a cluster in the R visual cortex (L Caudate seed) and in clusters in the bilateral parietal and visual areas (R vStr seed). For complete statistics regarding main effects and interactions for the analyses conducted, please refer to Supplement Table 1.

Fig. 4. Seed-to-whole brain FC in regions of interest for vehicle, low dose THC (0.32 mg/kg), and high dose THC (3.2 mg/kg) groups at the three timepoints (drug-naïve, chronic, discontinuation).

Panel (A) shows spatial maps representing significant group*session interactions for FC in mOFC, L caudate, and R ventral striatum between seed (green) and target (red) regions. Panel (B) shows changes in FC for each treatment group across the three scan timepoints, with data points representing values from individual subjects. Table (C) lists regions showing a group*session interaction for the seed-to-whole-brain FC analysis for the mOFC, L Caudate, and R vStr seeds (uncorrected voxel-wise p = 0.005, minimum cluster threshold of 20 voxels).

The 8 ROIs were then used as independent clusters for ReHo analyses, using subject-level normalized ReHo maps for the three imaging sessions. ReHo values averaged over the ROI were extracted for each subject and indicated that THC group did not influence local connectivity in any of the a priori selected ROIs (all p’s > 0.05).

To examine the relationship between behavioral and neuroimaging endpoints, R² α (reward magnitude vs. reinforcing strength) or R² Q₀ (reward magnitude vs. demand intensity) were used as independent factors and the brain metric (extracted FC strength in the seed analysis averaged over target regions or KCC from ReHo averaged over ROIs) served as the dependent factor, with subject age, sex, and motion (during discontinuation scan) as covariates for correlational analyses. Two significant relationships emerged. First, R² α and seed-to-whole-brain FC change between the mOFC and the R visual cortex were positively associated in the vehicle group but not in either THC group (vehicle, p = 0.04; low dose THC, p = 0.91; p = 0.52, high dose THC; Fig. 5A). Second, R² α was positively associated with ReHo within the L vStr ROI in vehicle-treated subjects but not in either THC group (vehicle, p = 0.01; low dose THC, p = 0.66, high dose THC, p = 0.12; Fig. 5B). Both results indicate the positive relationship between R² α and FC (global or local) in the vehicle group (notwithstanding the expected truncation in R² range in vehicle-treated subjects) was absent in the THC-treated groups. When modeling with R² Q₀ as the independent factor, no association with seed-to-whole-brain FC or ReHo change was evident in any of the three groups. Thus, THC treatment did not meaningfully alter the relationship between R² Q₀ and either neuroimaging endpoint.

Fig. 5. Relationship between behavioral and neuroimaging endpoints. Data points represent relations for individual subjects.

Top panels (A) show extracted FC strength in the seed analysis averaged over target regions for mOFC as a function of R² α (reward magnitude vs. reinforcing strength) for the vehicle (left), low dose THC (0.32 mg/kg; middle), and high dose THC (3.2 mg/kg; right) groups. Bottom panels (B) show Kenall’s Correlation Coefficient (KCC) from Regional Homogeneity (ReHo) analysis averaged over (L) ventral striatum as a function of R² α (reward magnitude vs. reinforcing strength) for the vehicle (left), low dose THC (0.32 mg/kg; middle), and high dose THC (3.2 mg/kg; right).

Discussion

Behavioral outcomes

The present behavioral economic studies uncovered several abnormalities in motivation and reward processing in young adult monkeys that had been exposed to chronic THC treatment during adolescence. First, indices of elasticity (α) across all reward magnitudes were lower in the vehicle-treated subjects than in either low- or high-dose THC-treated subjects. Higher α values signify a reduced tendency to defend consumption of reward in the face of increasing price (i.e., lessened motivation). Second, highly orderly relationships between concentration-dependent decreases in elasticity (α) and increases in demand intensity (Q₀) in the vehicle group contrasted sharply with disorder in these relationships in both THC groups. Behavioral assessments began 3-months post-discontinuation of chronic adolescent THC exposure to ensure (1) observations occurred during early adulthood, (2) negligible THC concentrations in plasma and brain [45], and (3) that any potential symptoms of cannabinoid withdrawal, including disruptions in cognitive performance, had fully resolved [54]. Thus, these findings reflect consequences of chronic adolescent THC exposure rather than effects of residual THC or withdrawal. The deficits and disorder in these well-validated quantitative metrics are consistent with the view that an amotivational syndrome may persist into adulthood following chronic THC exposure during adolescence [22].

Economic demand procedures have long provided highly translational characterizations of drug effects on motivation in humans and laboratory animals [55]. Here, we have expanded this analytic approach to identify a putative behavioral biomarker of an amotivational syndrome. Specifically, the parametric examination of a reinforcer that can vary in magnitude allows the determination of a family of exponential functions that, as observed under vehicle conditions, differ in a rational, ordered manner. In turn, goodness of fit values that effectively characterize order or disorder can be used without excessive inferential baggage to numerically quantify motivational consequences of drug treatment. Importantly, linearity between demand parameters (α and Q₀) and reward magnitude is assumed by these regressions. While this assumption was thoroughly supported here, that is, in drug-free control subjects displaying excellent fits (R₂ = 0.94-0.96), additional studies employing this analytic tactic are needed to establish the generality of this summary metric.

Neuroimaging outcomes

Nonhuman primate neuroimaging techniques have become increasingly feasible in awake (non-anesthetized) subjects, enhancing their translational relevance. Using this approach here, alterations in local (ReHo) and whole-brain (seed-based) FC of brain regions involved in reward processing were identified as long-term neural consequences of adolescent THC exposure. While local connectivity in any of the a priori selected ROIs was not substantively altered, adolescent THC exposure led to persistent alterations in whole-brain connectivity with mOFC, caudate, and ventral striatum. These regions play specific roles in reward valuation [56] and are implicated in neurobiological effects of cannabis in adolescent and adult users [4, 57, 58]. For example, marijuana-dependent users exhibit greater resting state mOFC-dorsal striatal FC and stronger mOFC-dorsal striatal coupling in response to negative emotional stimuli [59]. Further, mOFC FC with anterior cingulate, prefrontal cortex, and caudate in adolescent female macaques was previously shown to significantly increase during chronic treatment with a high efficacy CB₁ receptor agonist [54]. Thus, brain areas in which regional FC was altered by THC exposure in the present study are consistent with those identified in previous human and nonhuman primate studies.

In addition to changes in canonical brain reward circuitry, the observed changes in FC within visual and sensorimotor processing regions (e.g., parietal areas) across multiple ROIs in THC-exposed groups and the relationship of such changes to behavioral endpoints are major findings in the present study. While mOFC and striatal regions are well-known to be involved in reward valuation [56, 60–62], their connection with visual and or sensory processing is less well investigated. However, visual regions are known to contain a high density of cannabinoid receptors [63, 64] which undergo several changes during early development [65] and, unsurprisingly, THC administration has well characterized effects on the processing of sensory information. For example, acute THC administration has been shown to increase FC in the sensorimotor network, including visual regions [66], whereas chronic cannabis use has been associated with long-term changes in visual processing, perhaps due to alterations in cannabinoid receptor signaling [67]. Indeed, abstinent marijuana users have been shown to have greater activation of frontal, parietal and visual (occipital) brain regions during a visual-attention task, suggesting altered brain development as a consequence of chronic marijuana exposure [68]. Consistent with these observations, metanalysis of available data has shown significantly greater activation in components of the central executive (e.g., parietal) and default mode networks (e.g., visual) in adolescent cannabis users compared to nonusers [69]. Finally, two key findings from our previous work have suggested that mOFC FC with attentional regions may be especially vulnerable to THC exposure during adolescence. First, in a comparison of rsFC networks in adolescent and adult squirrel monkeys, adolescents showed greater FC between an mOFC network and visual cortex, suggesting that FC of these regions is developmentally sensitive [46] and, in turn, likely vulnerable to the influence of repeated cannabinoid exposure during development. Second, a similar mOFC-visual network has been identified in adolescent female rhesus macaques [54]. In conjunction with these earlier observations, our current results strongly support the view that THC exposure during adolescence may produce long-term changes in visual attentional sensory processes that interact with reward-related regions to influence behavioral responses to food-based reinforcement.

Correlational analyses of changes in neural circuitry and in economic demand revealed significant associations between both local connectivity (ReHo in ventral striatum) and whole brain FC (seed: mOFC) and R² α in vehicle-, but not THC-treated, subjects. Several points are noteworthy. First, as with the behavioral economic analyses themselves, the orderly relationship present in vehicle-treated subjects was absent in THC-treated subjects. Second, consistent with human studies, ventral striatum and mOFC were identified as core brain regions playing distinct motivational roles, i.e., anticipation and consumption of reward and processing reward magnitude, respectively [70]. Similarly, BOLD activity in ventral striatum and mOFC has been positively associated with overall value in effort-related reward [71]. Next, significant correlations were found with R² α, but not R² Q0 (see also [72]). While α is thought to measure reinforcing effectiveness of a reward in the context of increasing response cost (i.e., effort), Q0 is thought to be a metric of unconstrained consumption of a reward. Though motivational processes are likely involved, it may be that Q0 is governed by brain regions related to regulation of satiety or of general consumption that maintains physiological homeostasis. Finally, reward processing changes dramatically during adolescence and may differ from processing in adults [73]. For example, sweet liquid reward processing in adolescents involves a stronger response in ventral striatum but a lesser response in frontal executive control regions than in adults [73, 74]. Our previous FC comparisons between adolescent and adult squirrel monkeys also show mOFC whole-brain FC [46] and right-left ventral striatum FC [75] differ significantly between adolescent and adult squirrel monkeys, suggesting that these circuits are vulnerable to disruption during critical developmental periods. In conjunction, these lines of evidence suggest that chronic cannabis use during adolescence produces persistent abnormalities in neural circuitry and motivational processes, revealing risks to neurocognitive health.

There are caveats that deserve discussion. First, although deficits observed in reward processing and FC following chronic adolescent THC exposure were evident in early adulthood, it is not known whether an identical chronic regimen during adulthood would yield similar outcomes. We have reported that FC in squirrel monkeys is highly dynamic during adolescence and relatively stable thereafter [46, 75]. The behavioral and neuroimaging abnormalities reported here may be a unique consequence of THC exposure during this dynamic developmental period. However, studies employing the same chronic treatment and testing regimens in adult monkeys are necessary to confirm this possibility. Further, while the present work focused on changes in rsFC and its relation to behavioral economic demand, the investigation of other neurobiological endpoints, such as changes in brain structure [76] or activation during stimulus presentation (i.e., task- or event-related fMRI in which subjects are responding for and receiving various magnitude of rewards) is needed to fully understand the effects of adolescent THC exposure on reward sensitivity. A second caveat lies in the selectivity of the present effects of chronic THC exposure on reward processing. Thus, it is unknown whether the THC-related deficits in reward processing for a palatable food reinforcer would extend to other types of reinforcement (e.g., drugs).

In summary, the present studies utilized a coordinated behavioral and neuroimaging approach in nonhuman primates to explore long lasting consequences of chronic adolescent THC exposure. Persistent abnormalities in reward processing were identified using behavioral economics, fMRI FC, and ReHo analyses. Such motivational and neural abnormalities have clear translational implications regarding health risks of chronic cannabis use during adolescence.

Author contributions

BDK, SLW, RDS, JB, and SJK designed the research, BDK, SLW, and SJK performed the research, BDK, HUD, and SJK analyzed the data, and BDK, HUD, SLW, RDS, JB, SJK wrote the paper.

Funding

This project was supported by R01-DA047575 from the National Institute on Drug Abuse. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Competing interests

Over the past 3 years, BDK has had sponsored research agreements with BlackThorn Therapeutics, Compass Pathways, Delix Therapeutics, Engrail Therapeutics, Neurocrine Biosciences, and Takeda Pharmaceuticals. No funding from these entities was used to support the current work. All other authors have no conflicts of interest or relevant disclosures.

Supplementary information

The online version contains supplementary material available at 10.1038/s41386-024-02024-9.