The use of cannabis and other cannabinergic drugs during adolescence, a developmental period when the brain undergoes maturation, is widespread and continues to grow (Bava et al., 2010; Peters et al., 2022). It is uncertain whether cannabinoid exposure during this time may lead to long-term alterations in cognition and/or brain function (Becker and Hu, 2008); however, a number of studies suggest that repeated exposure can adversely affect cognitive performance (Egerton et al., 2006; Crean et al., 2011; Zanettini et al., 2011). Of special concern, such deleterious effects on cognitive performance have been noted in a range of tasks involving fronto-cortical brain regions, which undergo extensive maturation throughout adolescence (Delevich et al., 2021). For example, deficits in the performance of cannabis users relative to age-matched controls have been observed in tasks of inhibitory control and cognitive flexibility, sustained attention, impulsivity, and memory (Gruber and Yurgelun-Todd, 2005; Fontes et al., 2011; Hanson et al., 2011; Tait et al., 2011; Wrege et al., 2014). Thus, considerable evidence supports the idea that cannabis use during adolescence can have a persistent negative impact on cognitive function.

Disruptions in cognitive performance, especially those related to memory function, previously have been observed in preclinical studies of Δ⁹-tetrahydrocannabinol (Δ⁹-THC) and other cannabinergic drugs in rodents and non-human primates. For example, daily cannabinoid exposure was shown to impair novel object recognition, short-term memory, and cognitive flexibility in adolescent rats (O’Shea et al., 2004, 2006; Cha et al., 2007; Quinn et al., 2008; Higuera-Matas et al., 2009; Zamberletti et al., 2014). Similarly, spatial memory in adolescent monkeys also was disrupted by daily Δ⁹-THC exposure over the course of 6 months (Verrico et al., 2014).

Neuroimaging studies in human subjects have provided insight into the brain regions that underlie different types of task performance (Huettel, 2012) as well as those that may be altered in adolescent cannabis users (Sager and Gruber, 2019). For example, studies of resting-state functional connectivity (RSFC) suggest a relationship between extent of cannabis use and abnormal connectivity within fronto-temporal networks (Houck et al., 2013; Cheng et al., 2014; Camchong et al., 2017). Adolescent cannabis use, in particular, has been associated with reduced interhemispheric RSFC and elevated right hemisphere connectivity that may underlie the disruption of coherent, organized cognitive and behavioral processes (Orr et al., 2013). Adolescent-onset cannabis users also may display altered patterns of brain activity (primarily in fronto-cortical regions) when it is measured during cognitive tasks, including short-term memory of several types (Padula et al., 2007; Tapert et al., 2007; Schweinsburg et al., 2008; Jacobus et al., 2009; Sagar et al., 2015). Although the results of fMRI experiments may vary depending on the particular methodology and regions of interest, the altered pattern of communication between brain areas at rest in cannabis users suggests that qualitative changes in intrinsic connectivity may be related to deficits in cognitive performance.

Research in human subjects to solidify our understanding of the consequences of adolescent exposure to cannabinoids on behavior and brain function faces considerable challenges. For example, due to ethical and practical constraints, such research (especially in adolescents) is predominantly cross-sectional rather than longitudinal. Moreover, it is difficult to control for histories that include the use or abuse of multiple substances as well as for other factors such as duration of use, age of initiation, frequency/amount of consumption, drugged versus withdrawal state (Churchwell et al., 2010; Lorenzetti et al., 2016), and polysubstance use (Moss et al., 2014). From this perspective, the extent to which results in human subjects reflect the effects of cannabis directly on behavior and brain function or, alternatively, contamination by such uncontrolled factors remains uncertain (Cheetham et al., 2012; Lubman et al., 2015). One approach to mitigating the problem of uncontrolled historical or contextual factors in preclinical research is through translational research in laboratory animals and, in particular, non-human primates. The neurological, phylogenetic, pharmacokinetic, and pharmacodynamic aspects of drug action in rhesus monkeys, for example, greatly overlap with those of humans (Weerts et al., 2007), permitting translationally-relevant longitudinal studies of the effects of cannabinoid exposure on complex behavior and brain function during adolescence.

In the present research, a touchscreen-based titrating-delay matching-to-sample task of short-term memory function (TDMTS; Kangas et al., 2010) was paired with longitudinal neuroimaging (fMRI and magnetic resonance spectroscopy) measures to investigate the effects of acute and chronic exposure to the cannabinoid receptor type 1 (CB₁) agonist AM2389 in adolescent female rhesus macaques. Female macaques were studied in the present work because (1) cannabinergic sex differences have been documented, including a greater CB₁ density across several brain regions in male than female laboratory animals (e.g., Rubino et al., 2008; Reich et al., 2009; Burston et al., 2010; Mateos et al., 2011), (2) the role of density in regions of interest related to memorial processes, including the orbitofrontal cortex (Lamar et al., 2004), are unknown, and (3) there continues to be a need to address the well-publicized underrepresentation of female subjects in biomedical research (Woitowich et al., 2020). Data also were obtained following chronic treatment to examine the immediate and longer-term effects of discontinuing daily injections of AM2389. Lastly, the effects of the CB₁ antagonist rimonabant were determined during chronic AM2389 to assess the presence of behavioral or neural signs of antagonist-precipitated withdrawal. AM2389 was selected for study because its long duration of behavioral action (∼17 h in vivo half-life; Järbe et al., 2012) assures continuous CB₁ receptor occupancy with once-daily treatment. In addition, as a CB₁ receptor full agonist, its agonist effects may be more definitive than those of partial agonists like Δ⁹-THC (Paronis et al., 2012; Kangas et al., 2020). TDMTS task performance was selected as the cognitive endpoint for the present studies in view of the well-known ability of cannabinoids to disrupt short-term memory, an adverse effect that has been repeatedly documented in chronic cannabis users (Crean et al., 2011; Gruber et al., 2012; Crane et al., 2013) and preclinical studies in rodents and non-human primates (e.g., Cadoni et al., 2001; Verrico et al., 2012; Kangas et al., 2016). Neuroimaging procedures have focused on the medial orbitofrontal cortex, which has been shown to be involved in DMTS performance in an age-dependent manner (Lamar et al., 2004) and is also thought to play a critical role in substance abuse (London et al., 2000; Winstanley, 2007; Schoenbaum and Shaham, 2008).

In the present study, acute administration of the CB₁ full agonist AM2389 produced dose-related disruptions in behavioral and neurochemical measures (tNAA/tCrw ratio) but, notwithstanding a trend toward decreased local frontal connectivity, did not significantly alter connectivity in the mOFC network. During chronic exposure, tolerance developed to the effects of AM2389 on both motoric performance (response rates under the FR20 schedule of food reinforcement) and memorial function (titrated delay in the matching-to-sample procedure; TDMTS) but not to its neurochemical effects in the mOFC. Finally, whereas the administration of the CB₁ antagonist rimonabant during chronic treatment disrupted motoric performance but not memorial function, discontinuation of the chronic AM2389 regimen disrupted both types of behavior. Overall, the present findings are consistent with the results of previous studies in humans and laboratory animals showing that, first, cannabinoid agonists may produce adverse effects on short-term memory processes (e.g., Abush and Akirav, 2012; Crane et al., 2013; Verrico et al., 2014; Stringfield and Torregrossa, 2021) and second, tolerance develops to such behavioral effects of cannabinoid agonists over the course of chronic exposure (Ramaekers et al., 2011). The magnitude of tolerance observed here was substantial (1.5 log unit rightward shift) and is comparable in magnitude to that reported in previous behavioral studies of cannabinoid tolerance and cross-tolerance in non-human primates (e.g., McMahon, 2011; Desai et al., 2013). It is interesting that tolerance to the effects of the daily dosage of AM2389 appeared more complete on memorial function than on motoric performance. Thus, whereas TDMTS recovered fully during the chronic AM2389 regimen, response rates recovered to only 70% of pre-chronic values and, during dose-response re-determinations, were markedly disrupted by doses of AM2389 that had only moderate effects on titrated delay values. Previous studies have shown that the extent to which acutely administered cannabinoids disrupt performance on cognitive tasks may vary with the difficulty of the task, i.e., cognitive load (Kangas et al., 2016). The present results indicate that tolerance to the disruptive effects of chronically administered cannabinoid agonists also may vary with the class of behavior being measured, i.e., cognitive vs. motoric task performance.

Although the precise defining features and severity of cannabis withdrawal continue to be a subject of inquiry (e.g., Budney and Hughes, 2006; Schlienz et al., 2017), it is now commonly accepted that withdrawal symptoms can emerge following the abrupt discontinuation of cannabis products, especially in high-intake chronic users. This view was codified in the DSM-5 (American Psychiatric Association, 2013) with the inclusion of Cannabis Use Disorder which, for the first time, included diagnostic criteria for cannabis dependence. Moreover, it is widely recognized that such withdrawal discomfort can include pronounced disruptions in cognitive function (Cousijn and van Duijvenvoorde, 2018). The present results showing that discontinuation of AM2389 treatment in the present studies was followed by decrements in short-term memory is consistent with this understanding of cannabis withdrawal. Such effects were most severe on the second day after treatment was discontinued, which is in line with the long duration of action reported for AM2389 (Järbe et al., 2012). The recovery of task performance over the course of approximately 5–10 days also comports well with findings from a previous study of human memory processes in adolescent and young adults showing that 1 month (but as brief as 1 week) of abstinence from cannabis use was associated with improvements in memory task performance (Schuster et al., 2018; see also Scott et al., 2018).

The failure of the CB₁ receptor antagonist rimonabant to produce withdrawal effects comparable to or, depending on dose, more intense than those observed following discontinuation of the AM2389 regimen was surprising. A defining feature of drug dependence resulting from chronic exposure to other classes of drugs, for example, opioids or benzodiazepines (e.g., Gellert and Holtzman, 1979; Lukas and Griffiths, 1982) is the ability of receptor antagonists to precipitate overt, pronounced signs of withdrawal, and previous studies have shown that the stimulus properties of 24 h abstinence after chronic treatment with Δ⁹-THC or AM2389 overlap with those of rimonabant in non-human primates (Stewart and McMahon, 2010; Kangas et al., 2020). Studies in rodents and non-human primates that are chronically treated with Δ⁹-THC or other CB₁ agonists have further suggested that rimonabant-precipitated withdrawal can be characterized by overt signs including paw tremor and head shakes (Verberne et al., 1981; Cook et al., 1998; Stewart and McMahon, 2010). However, this remains controversial, as such behavioral disturbances can be produced by rimonabant alone (including psychiatric side effects that led to its withdrawal from market; Nguyen et al., 2019) and are not always apparent in subjects treated with cannabinoids like Δ⁹-THC or AM2389 (Kangas et al., 2020). Interestingly, studies in cannabinoid-treated rodents also have revealed time- and dose-dependent changes in motor activity consequent to rimonabant injection that could serve as a more reliable indicator of cannabinoid withdrawal (Huang et al., 2009; Missig et al., 2022; Paronis et al., 2022). With the latter findings in mind, it is noteworthy that, in the present study, lower doses of rimonabant produced considerable increases in fixed-ratio responding in AM2389-treated monkeys, likely reflecting a mild agitation or motor stimulation. Possibly, assays that are particularly sensitive to such changes in motor activity or assessment of daily patterns of locomotor activity during cannabinoid treatment and discontinuation will provide more consistent demonstrations of antagonist-precipitated cannabinoid withdrawal than are currently available.

Neuroimaging studies in human subjects suggest that use of cannabis during adolescence may lead to abnormalities in brain function, structure, and chemistry in several regions related to cognitive processing (reviewed by Lichenstein et al., 2022; Ma et al., 2022). Studies conducted in adolescent cannabis users have often focused on mOFC, a brain region that undergoes extensive pruning and reorganization during adolescence (Sturman and Moghaddam, 2011) and that is thought to play a role in maintaining drug use in addiction (London et al., 2000; Winstanley, 2007; Schoenbaum and Shaham, 2008). The mOFC also has been linked to performance on cognitive tasks including delayed match-to-sample performance (Lamar et al., 2004). Accordingly, ICA–which revealed a fronto-cortical functional network that primarily encompasses the mOFC, prefrontal cortex, and anterior cingulate—and an MRS voxel in the mOFC were utilized in the present study to probe changes in brain function and chemistry during and after chronic AM2389 treatment. The present results, illustrating greater functional connectivity within a fronto-cortical network during chronic AM2389 treatment, are consistent with human studies showing a relationship between the extent of cannabis use and frontal connectivity (Houck et al., 2013; Orr et al., 2013; Lopez-Larson et al., 2015). For example, increased OFC functional connectivity with other frontal and motor regions (Lopez-Larson et al., 2015) as well as with dorsal striatal regions (Zimmermann et al., 2018) has been demonstrated previously in heavy cannabis users when compared to healthy controls. Importantly, greater OFC connectivity with frontal and subcortical regions also has been associated with greater resistance to behavioral change in high-risk adolescent cannabis users (Feldstein Ewing et al., 2017), consistent with the view that alterations in this network may be a key component of compromised executive control in adolescent cannabis users. In the present study, changes in RSFC within this network could be related to changes in performance on the TDMTS task. For example, RSFC within the fronto-cortical network showed a non-significant trend for decreased connectivity following acute treatment with a behaviorally disruptive dose of AM2389 (10 ug/kg). In contrast, RSFC increased within the frontal network following a regimen of daily AM2389 administration that induced tolerance to its performance–disrupting effects. It is tempting to speculate that increased connectivity among frontal regions during chronic AM2389 treatment represents a recruitment of brain regions that helped to overcome the disruptive behavioral effects of AM2389 that were evident after acute administration. These effects were not persistent, however, as functional connectivity following treatment discontinuation was able to recover to pre-treatment levels.

The combined use of fMRI and MRS in the same subjects provided the opportunity to gain insights into the neurochemical mechanisms that may contribute to drug-induced changes in functional connectivity (Kohut and Kaufman, 2021). Here, we showed that the tNAA/tCrw ratio in mOFC decreased in a dose-dependent manner during AM2389 treatment but returned to baseline levels following discontinuation. We focused on NAA levels because they are thought to be a marker of neuronal health and integrity and have been shown to be altered in substance use disorders (Magalhaes, 2005), including cannabis use disorder. For example, increased striatal NAA has been positively correlated to subjective reports of “high” following acute THC administration to occasional cannabis users (Mason et al., 2021). On the other hand, decreased NAA concentration has been reported in frontal regions such as anterior cingulate cortex (Prescot et al., 2011) and dorsolateral prefrontal cortex (Hermann et al., 2007) in chronic cannabis users suggesting that the effects of cannabis on brain NAA may be region-dependent or change with chronic exposure. Further, Sung et al. (2013) showed decreased frontal NAA levels in adolescents that use methamphetamine + marijuana compared to levels in methamphetamine only users or healthy controls; NAA ratios were also inversely related to lifetime and duration of marijuana use. The relationship between neurochemical environment in a given brain region and its functional connectivity is likely to be complex and influenced by multiple factors, Lord et al. (2017) described a quadratic relationship between NAA concentration and functional connectivity in posterior cingulate cortex of healthy volunteers in which functional connectivity and NAA decreased linearly before increasing again. While a causative influence of NAA on functional connectivity (or vice versa) cannot be proven from data in the present study, it is noteworthy that the effects of AM2389 administration generally followed this same pattern. Further studies that target other metabolites in frontal regions may shed greater light on how cannabinoids alter the relationship between neurochemical indices and functional connectivity.

There are several caveats and limitations of these studies that deserve mention. First, all adolescent rhesus macaques examined were female. As greater CB₁ density in several brain regions has been consistently reported in males than females in laboratory animal studies (e.g., Rubino et al., 2008; Reich et al., 2009; Burston et al., 2010; Mateos et al., 2011), it will be important for future studies to determine whether the present findings can be replicated in adolescent male monkeys. Second, relating hormonal phases to behavioral and neuroimaging endpoints was beyond the scope of the present studies and, therefore, not systematically assessed. Interestingly, although these subjects had not begun menses prior to the initiation of these studies, first menses and subsequent cycling were observed in all subjects at various points during these studies, confirming that these experiments indeed were occurring during the intended developmental stage. Also, magnetic resonance spectroscopy conducted at 3T field strength was not sensitive enough in non-human primates to distinguish resonance peaks for other neurochemicals of interest such as glutamate, y-aminobutyric acid (GABA), and glutamine. Future studies using other acquisition techniques or ultra-high field scanners to enhance the signal of such metabolites will help further characterize the neurochemical environment in the mOFC and related brain regions during chronic drug administration (Ma et al., 2022). Next, although accuracy in the TDMTS task was adversely affected by acute AM2389 administration as well as discontinuation of chronic treatment, it was noteworthy that all rewards earned by subjects were indeed consumed. Despite this, a decrease in motivation for the milk reinforcer cannot be ruled out as the behavioral mechanism of observed task disruption and, as such, the effects of chronic AM2389 treatment and discontinuation upon TDMTS performance maintained by a non-food consequence will be necessary to examine this possibility. Finally, it is important to note that inasmuch as these studies employed a within-subject behavioral and neuroimaging design during pre-drug exposure, an untreated control group was not included during or following chronic exposure. While a drug-free control group would have allowed for interesting comparisons of behavioral and neuroimaging endpoints during these phases of this research, the longitudinal design of the present study nonetheless clearly reveals modulation of both cognitive and MRS/fMRI metrics during acute and chronic cannabinoid exposure and, as well, their normalization following its discontinuation.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by McLean Hospital Animal Care and Use Committee.

SK, JB, BF, JJ, and BK participated in research design. SK, CZ, JJ, and BK conducted the experiments. SJ, SN, and AM contributed the new reagents or analytic tools. DM, LC, BF, SK, and BK performed the data analysis. SK, LC, DM, JB, and BK contributed the writing of the manuscript. All authors contributed to the article and approved the submitted version.