Cannabis Use and Mental Illness: Understanding Circuit Dysfunction Through Preclinical Models

Patients with a serious mental illness often use cannabis at higher rates than the general population and are also often diagnosed with cannabis use disorder. Clinical studies reveal a strong association between the psychoactive effects of cannabis and the symptoms of serious mental illnesses. Although some studies purport that cannabis may treat mental illnesses, others have highlighted the negative consequences of use for patients with a mental illness and for otherwise healthy users. As epidemiological and clinical studies are unable to directly infer causality or examine neurobiology through circuit manipulation, preclinical animal models remain a valuable resource for examining the causal effects of cannabis. This is especially true considering the diversity of constituents in the cannabis plant contributing to its effects. In this mini-review, we provide an updated perspective on the preclinical evidence of shared neurobiological mechanisms underpinning the dual diagnosis of cannabis use disorder and a serious mental illness. We present studies of cannabinoid exposure in otherwise healthy rodents, as well as rodent models of schizophrenia, depression, and bipolar disorder, and the resulting impact on electrophysiological indices of neural circuit activity. We propose a consolidated neural circuit-based understanding of the preclinical evidence to generate new hypotheses and identify novel therapeutic targets.


INTRODUCTION
Cannabis is one of the most widely used psychoactive substances worldwide, and patients with serious mental illnesses use cannabis at rates much greater than the general population (1,2). The lifetime cannabis-use rates for patients with schizophrenia, major depressive disorder, or bipolar disorder are 80, 17, or 24%, respectively, while ∼40% of patients with schizophrenia and 20% of patients with major depressive disorder or bipolar disorder are also diagnosed with cannabis use disorder (3)(4)(5)(6). A diagnosis of mental illness increases the risk for lifetime cannabis use, while cannabis use, especially use of greater potency cannabis at an earlier age, increases the risk for developing a mental illness (1,7). Therefore, there is a need for mechanistic investigations that can then be targeted toward developing novel treatment approaches (8,9). To improve our understanding of cannabis use and serious mental illness, we herein provide an update on the preclinical evidence ( Table 1) in support of shared neurobiological mechanisms fundamental to the effects of cannabis and the symptoms of mental illness.
As described below, cannabinoids produce distinct changes in neural circuit electrophysiological activity that are similar to those observed in patients with serious mental illnesses, as well as rodent models of these illnesses (43,44). Neural circuit oscillatory activity arises from the summed electrical activity of networked neurons and is apparent in electrophysiological recordings from human subjects as well as non-human research animals, with various frequencies corresponding to certain functions (45)(46)(47)(48)(49). These frequencies and their associated functions are simplified as follows: delta oscillations (0.  are associated with signal detection and decision making; theta oscillations (4)(5)(6)(7) are associated with episodic memory and memory retrieval; alpha oscillations (8)(9)(10)(11)(12) are associated with semantic memory and attention; beta oscillations (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29) are associated with motor control as well as attention, and sensory filtering; gamma oscillations  are associated with attention, sensation, perception, memory, and conscious awareness (45)(46)(47)(48)(49). Aberrant oscillatory patterns thus correlate with different brain states, including those resulting from cannabinoid exposure or contributing to a serious mental illnessrelated symptom; dysfunctional patterns often reflect deficits in behavior and cognition. While we will ultimately identify similarities between cannabinoid-induced oscillatory changes and pathological changes associated with serious mental illnesses, cannabis exposure alone can produce phenotypes that overlap with some psychopathology, making it important to disentangle the effects of cannabis alone from its interactions with the psychopathology-associated circuit dysfunctions.

Cannabis Use
Cannabis users inhale smoke or vapor from crudely burning cannabis flower or by vapourizing it at greater temperatures (50). Researchers modeling cannabis use employ various administration routes (i.e., injection, vaporization, oral ingestion, or inhalation) of cannabis plant components and synthetic cannabinoids, including cannabinoid-type 1 receptor (CB1R) agonists such as CP-55940 and WIN55,212-2. CB1R antagonists rimonabant and AM251 are also used to assess the involvement of the endocannabinoid system (eCB) in drug effects and psychopathology (51). Examining cannabinoids beyond their action at CB1R sites, however, is imperative, as interactions also occur via CB2R-dependent mechanisms (52)(53)(54)(55), as well as noncannabinoid receptor mediated mechanisms (56-58). Moreover, the use of CB1 receptor agonists to model the exposure to cannabis-derived cannabinoids may also have limited utility due to their limited pharmacological profile. Indeed, the effects of cannabis arise from combined constituent activity (59), not the action of a single ligand-receptor interaction, and thus future preclinical studies, unless purely pharmacological, must examine the combined effects. Also, since injections do not capture human use patterns, a recent concerted effort to establish more translationally-relevant delivery methods (e.g., vaping) for 9tetrahydrocannabinol (THC) and cannabidiol (CBD), as well as the many other constituents in the cannabis plant, has begun (28,38,60).

Schizophrenia
Schizophrenia is a complex neuropsychiatric illness characterized by severe dysfunctions including delusions, hallucinations (and other "positive" symptoms), social withdrawal (and other "negative" symptoms), and deficits in memory and sensory processing (and other "cognitive" symptoms) (61). Considering the phenotypic complexity of schizophrenia, a rodent model that singularly recapitulates the human condition does not exist. Rather, various models (i.e., genetic, neurodevelopmental, and pharmacological) produce dysfunctions that capture some parts of the disease symptoms. These models usually demonstrate positive-like (e.g., amphetamine-induced hyperlocomotion, deficits in prepulse inhibition mediated via enhanced dopamine signaling), negative-like (e.g., social withdrawal), and cognitive-like behavioral dysfunctions (e.g., deficits in attention and working memory) (62,63). As disrupted-in-schizophrenia 1 (DISC1) gene was one of the first genes implicated in schizophrenia, many transgenic models targeting this gene exist (64,65). Similarly, knocking out the NRG1 gene, which is implicated in schizophrenia, is also used to study schizophrenia-like behaviors and circuit dysfunctions (66). Neurodevelopmental models are created by altering rodent neurodevelopment, by either administering polyriboinosinicpolyribocytidilic acid (poly I:C) to pregnant dam to produce maternal immune activation (23,67), or via bilateral lesioning of neonatal ventral hippocampi (NVHL) using ibotenic acid (63,68,69). Pharmacological models involve administering a compound to modify neurotransmission; for example, phencyclidine (PCP) produces psychotomimetic effects akin to the symptoms of schizophrenia (42,70).

Major Depressive Disorder
Individuals diagnosed with major depressive disorder present with symptoms such as persistent negative affect, anhedonia, as well as disturbed sleep and appetite (71). In rodents, depressivelike symptoms are produced using genetic models or by stress exposure through chronic mild/variable stress, social defeat stress, or early life stress. Genetic models include the Flinders Sensitive Line (FSL) and Wistar Kyoto rats, which exhibit phenotypical similarities to major depressive disorder in humans (72,73). Wistar Kyoto rats are used specifically for modeling treatment-resistant depression (74,75). Chronic mild/variable stress involves daily exposure to various stressors (e.g., tail suspension, restraint, electrical shock) (20). Social defeat stress involves repeatedly exposing a submissive rodent to a dominant conspecific (76), while early life stress involves separating neonates from dams or altering the rearing environments (17,77). These modifications produce depressive-like behaviors in rodents, including reduced exploration, reduced sucrose preference (reflecting an anhedonic state), and reduced escape   (78). Interestingly, CB1R-deficient mice are used to model major depressive disorder (79)(80)(81). Social defeat stress in mice also reduces CB1R expression in the basolateral amygdala, a brain region involved in the pathophysiology of major depressive disorder, while knocking down or knocking out CB1R expression in mice enhances stress susceptibility (81,82).

Bipolar Disorder
The symptoms of bipolar disorder are characterized by cyclic changes in mood, motivation, and attention, ranging from periods of manic to depressed symptoms (83). Although producing a model comprising the complete range of symptoms has proven difficult, rodent models of psychosis, depression, and diurnal disruption are often used to model aspects of bipolar disorder (84,85). The Clock 19 transgenic mouse model shows promise as a heuristic model of bipolar disorder, having both face (behavioral cycling, hyperactivity) and predictive validity, as lithium administration decreases bipolar-like behaviors in this model (85). The dopamine transporter knock-down (DAT-KD) mouse is also used to model mania-like behaviors (enhanced motivation, hyperactivity) associated with human bipolar disorder (86)(87)(88)(89)(90). DAT knock-out (DAT-KO) mice are also sometimes used, but present with growth defects and hypoplasia. The DAT-KD mouse was subsequently developed to avoid these undesirable attributes (86,91,92).

Schizophrenia
Patients with schizophrenia exhibit reduced resting-state and evoked theta and gamma power, as well as decreases in beta and gamma coherence spanning various brain regions (93)(94)(95)(96). Similar alterations are also evident in preclinical models (Figure 1). Single-unit and local field potential (LFP) recordings from layers II/III and IV in the barrel cortex of anesthetized NRG1 knock-out mice and wild-type (WT) controls showed that NRG1 mice exhibit reduced gamma power. This reduction was also associated with reduced gamma signal-to-noise ratio and phase-locking (for all frequencies below 50 Hz), underpinned by enhanced firing rates, possibly demonstrating mechanistic dysfunction that occurs in patients with schizophrenia (14). In DISC1 mice, in vivo LFPs from the prelimbic cortex (PrlC) and hippocampus (HIP) showed reduced theta power in the HIP and PrlC and low gamma (30-50 Hz) power only in the PrlC (33) (Figure 1). PrlC-HIP coherence remained intact as described previously (37). NVHL rats showed reduced theta and beta coherence in the dorsal HIP (dHIP) while medial prefrontal cortex (mPFC) coherence remained intact. Baseline dysfunctions in inter-spike timing, wave duration, spike to valley voltage, and wave energy were also apparent, again possibly revealing causal mechanisms of oscillatory dysfunction in human schizophrenia (24) (Figure 1). Notably, gamma power suppression is consistently demonstrated across different models, and in human subjects.

Major Depressive Disorder
Reductions in alpha, theta, and gamma oscillations appear in patients with major depressive disorder, and in genetic and stress-induced preclinical models (97,98). In anesthetized FSL rats, LFPs from the mPFC, nucleus accumbens (NAc) shell, and the subthalamic nucleus (STN) all exhibited reduced alpha and beta power compared to controls, while theta and high gamma power remained intact; low gamma power in FSL rats differed by region, with reduced power in the mPFC and NAc and enhanced power in the STN (41) (Figure 2). Although alpha, theta, and gamma suppression is apparent in patients with major depressive disorder, preclinical models demonstrate that this suppression is more varied and region-specific. Similarly, rats exposed to chronic social defeat stress (CSDS) also exhibit aberrant gamma activity. LFP recordings from the NAc of CSDS rats and unstressed controls were acquired during social interaction or free exploration. High gamma power was enhanced in controls during social interaction, whereas CSDS rats did not exhibit any change in gamma power during either activity (21). In the chronic restraint stress (CRS) mouse model, LFPs from the frontal, somatosensory, parietal, and visual cortices were captured at 7 and 21 days post-stress exposure. At the 7 day time-point, CRS mice exhibited only enhanced coherence across all frequencies compared to baseline. At the 21 day timepoint, the enhanced delta and gamma coherence disappeared. Thus, gamma was restored with remission of a depressive-like phenotype (22), which indicates that aberrant gamma power or coherence may be a viable biomarker for major depressive disorder. As is the case with schizophrenia, dysfunctional gamma is common across various rodent models and in patients with major depressive disorder.

Bipolar Disorder
Patients with bipolar disorder exhibit various aberrations in oscillatory activity, including enhanced or reduced alpha power, enhanced beta power, enhanced alpha, beta, and gamma coherence, and reduced evoked frequencies (30,(99)(100)(101)(102)(103) (Figure 3). In support of altered oscillatory activity being relevant for human bipolar disorder is preclinical evidence that lithium chlorate (600 mg/L of water) administration, the primary prescribed treatment for patients with bipolar disorder, increased low gamma phase-coupling and ameliorated phase-locking deficits in the NAc of Clock 19 mice (15). In DAT-KD mice, whole-cell patch clamp recordings of dorsal striatum (dStr) glutamatergic neurons revealed shorter half-amplitude durations and faster decay times (86). This may be contributing to the altered phase coupling reported by Dzirasa et al. (92) in DAT-KO mice. LFPs from the dHIP and the mPFC of DAT-KO mice also demonstrated enhanced dHIP-PFC gamma coherence compared to controls (92) (Figure 3).
Although the preclinical literature is sparse and does not completely reflect what is observed in clinical studies of patients with bipolar disorder, reduced oscillatory activity in the gamma frequency range is again a shared feature across models and human subjects.

CANNABINOID EXPOSURE IN RODENTS ALTERS NEURAL CIRCUIT ACTIVITY
Cannabinoid-induced neural circuit dysfunctions in animal studies and otherwise healthy humans are similar to those in patients with a serious mental illness (18,44,104,105). THC exposure acutely suppresses gamma power and increases cortical noise in otherwise healthy individuals; changes that are associated with increased symptoms of psychosis (43,106,107). Acute THC exposure also alters oscillatory activity in patients with schizophrenia, with evidence showing it suppresses and enhances resting-state theta and low gamma power, respectively  (Figure 4). In adult anesthetized rats, Skosnik et al. (44) intravenously administered vehicle, CP-55940 (0.3 mg/kg), or CP-55940 + AM251 (3 mg/kg), while LFP and auditory evoked potentials (AEPs) were recorded from the HIP and entorhinal cortex. CP-55940 reduced AEP theta and gamma power, which was partially reversed by AM251 co-administration, suggesting THC-induced changes are partly CB1R mediated (44) (Figure 4).
We recently demonstrated that rats acutely exposed to THC vapor have oscillatory changes lasting longer than 1 week in the PFC, the orbitofrontal cortex (OFC), and the dStr after either vehicle or THC vapor (10 mg/kg) administration. Reduced gamma power was measured in all brain regions of THCtreated rats, compared to controls. Reduced dStr-OFC and OFC-PFC gamma coherence was also observed, and within-subject comparisons of rats exposed to THC in week 1 and vehicle in week 3 demonstrated persisting gamma suppression (27) (Figure 4). Infusions of THC (100 ng/0.5 µL) and CBD (100 ng/0.5 µL), either alone or in combination, directly into the vHIP of anesthetized male Sprague Dawley rats while recording VTA DA and non-DA GABAergic neurons demonstrated that THC enhanced VTA DA firing and bursting rates, compared to controls (Figure 4). Both THC and CBD enhanced firing frequencies of VTA non-DA neurons compared to controls, while only THC administration enhanced beta, gamma, and low delta power; THC also increased behavioral measures of fear responsivity, morphine conditioned place preference, and sucrose preference. CBD reversed all THC-induced changes (except for sucrose preference, which it enhanced) when coadministered (19), which supports evidence of CBD as a potential antipsychotic, antidepressant and anxiolytic (25,58,111).

DIFFERENTIAL EFFECTS OF CANNABINOID EXPOSURE IN RODENT MODELS OF SERIOUS MENTAL ILLNESS Schizophrenia
Cannabinoid exposure differentially affects neural circuit activity in rodent models of schizophrenia, in a dose-and regiondependent manner, and when comparing model animals to controls. Seillier et al. (36) examined whether THC exposure dose-dependently alters baseline VTA DA activity in PCP-treated or control rats. In PCP-treated rats, a reduced number of active VTA DA neurons at baseline was evident when compared to controls [a result that contradicts existing evidence of augmented DA population activity (11)]. The low dose (0.1 mg/kg, i.p.) of THC enhanced VTA DA population activity in PCP-treated rats, while the high dose (1 mg/kg, i.p.) of THC had no effect. In control rats, the low dose of THC reduced the number of active VTA DA neurons to baseline levels observed in PCPtreated rats. Extending this result to examine the involvement of the eCB in additional brain regions, Aguilar et al. (12) used the subchronic PCP-treated rat model to examine changes in neural circuit activity after exposure to THC (1 mg/kg, i.p.) or URB597 (0.3 mg/kg, i.p.), a fatty-acid amide hydrolase inhibitor, administered to awake, behaving rats. Single-unit and LFP recordings from the PFC and vHIP were taken prior to injection and at 30 min intervals post-administration. THC reduced mPFC firing rates in controls, without producing any effect in rats treated with PCP. URB597 increased mPFC firing rates in rats treated with PCP, compared to controls (Figure 4). THC also did not produce any effect on firing rates in the vHIP. PFC and vHIP baseline oscillatory activity was similar between PCPtreated rats and controls. THC administration also did not alter oscillatory activity in either group, which the authors attribute to the low dose of THC used. URB597 administration increased gamma power in the mPFC and decreased delta power in the vHIP for both groups (Figure 4). THC and URB597 produced opposing effects on mPFC-vHIP coherence, increasing, and decreasing delta coherence, respectively, for both groups; thus, also revealing that eCB modulation is differentially impacted by various modulations (12). Cannabinoid-induced alterations to neural activity may also be age-dependent; Lecca et al. (23) used the poly I:C rat model to examine whether adolescent cannabis exposure alters adult VTA DA neuron sensitivity. Adolescent poly I:C rats and controls were administered either vehicle or escalating doses of THC (2.5-10 mg/kg, i.p.) for 11 days. Once in adulthood, recordings were captured from the VTA of anesthetized rats. In Poly I:C rats, the number of spikes per burst and the intra-burst frequency of VTA DA neurons were reduced, compared to controls, which was ameliorated by adolescent THC exposure. Taken together, THC exposure may enhance DA population activity in rodent models of schizophrenia only at higher doses and in particular brain regions, whereas THC suppresses DA population activity in control animals. This possibly reflects an underlying mechanism supporting data from patients with schizophrenia that demonstrates THC acutely enhances oscillatory power in patients and suppresses power in controls (43,108).

Major Depressive Disorder
As was observed in the rodent models of schizophrenia described above, cannabinoid exposure enhances neural circuit activity in rodent models of major depressive disorder, and this effect seems more pronounced compared to control animals. Abush and Akirav (10) examined the longevity of cannabinoid-induced neural changes in a rat model of major depressive disorder. Male, Sprague-Dawley rats underwent daily chronic restraint stress (CRS) and administration of vehicle or WIN55,212-2 (1.2 mg/kg, i.p.).Thirty days after stress exposure, LTP recordings were captured from the NAc shell of anesthetized rats. Stress-exposed rats exhibited reduced LTP in the NAc while stress-exposed rats administered WIN55,212-2 were like unstressed controls, suggesting WIN55,212-2 rescued stress-induced LTP deficits. Within-subject comparisons of the unstressed, WIN55,212-2-treated rats revealed that WIN55,212-2 enhanced LTP compared to baseline. WIN55,212-2 + AM251 co-administration ameliorated WIN55,212-2 enhancements of LTP (10). These results were also reproduced using the chronic mild/variable stress model of major depressive disorder, as WIN55,212-2 administration (0.5 mg/kg, i.p.) prevented stressinduced reductions in LTP in the NAc and this was reduced by AM251 (0.3 mg/kg, i.p.) co-administration (35). Unlike the rodent models of schizophrenia mentioned above, WIN55,212-2 exposure produced a similar effect in control rats, albeit to a lesser extent than CRS rats (10). Another study, using restraint and tail-shock stress in Sprague Dawley rats to induce a depressivelike phenotype, demonstrated using slice electrophysiology that WIN55,212-2 (1 µM bath) induces long-term depression (LTD) in the lateral habenula of both stress-exposed rats and controls, a brain region implicated in the pathogenesis of major depressive disorder (29). Taken together, these results indicate that WIN55,212-2 exposure alters LTP and LTD mechanisms in both model and control rats.

Bipolar Disorder
Studies using validated models of bipolar disorder to examine cannabinoid-induced functional changes in neural circuit electrophysiological activity are very limited, so we have included an alternate measure of neural circuit alteration below. One study investigated the impact of CBD on manialike behavior and neurobiology using amphetamine-induced hyperlocomotion in male Wistar Kyoto rats, to model mania in bipolar disorder. Rats were treated with escalating doses of CBD (15-60 mg/kg, i.p.) before, or after, exposure to Damphetamine (2 mg/kg) to produce an acute manic episode. CBD did not affect amphetamine-induced hyperlocomotion in this rat model, which contradicts existing literature and may reflect innate differences in Wistar Kyoto rats (31). The moderate dose of CBD (30 mg/kg) increased HIP BDNF expression and rescued amphetamine-induced damage (possibly due to its effects as an anti-oxidant) when given after amphetamine exposure, whereas CBD pre-treatment had no effect on BDNF expression in the HIP, showing a protective effect of CBD on the biochemical changes associated with this model of bipolar disorder, only when administered posttreatment (40). This exemplifies how the effect of cannabinoid exposure depends on the chosen administration protocol. Moreover, given the involvement of BDNF signaling in coordinating gamma activity (112), we expect that future studies examining oscillatory activity in this rodent model of bipolar disorder may reveal reductions in gamma activity induced by cannabinoid exposure.

DISCUSSION
Understanding the pathophysiology of co-occurring cannabis use and serious mental illnesses remains a challenge in neuropsychiatry. Aberrant theta, alpha, beta, and gamma oscillations are observed after cannabinoid exposure in humans and in mental illness, while cannabinoid exposure differentially impacts the symptoms of the afore-mentioned mental illnesses. Similarly, preclinical models exhibit suppressed baseline neural circuit oscillatory activity, while cannabinoid exposure produces differential effects on electrophysiological neural circuit activity in otherwise untreated animals and has mixed effects in preclinical models on oscillatory activity, neuroplasticity, and neuronal firing rates. That said, our mini-review (albeit not exhaustive) shows that aberrant gamma activity is consistently observed. Whether aberrant gamma is etiological or only a consequence remains contested; however, the rescuing effect of gamma modulation implies that it is involved in mediating pathophysiological mechanisms (16,113). Gamma oscillations arise from competing excitatory and inhibitory control, involving extensive coordination between glutamatergic and GABAergic transmission, and the pathogenic loss of inhibitory control via impaired interneurons may contribute to aberrant gamma oscillations (13,33,114). Why then, do we observe both enhancements and reductions in gamma oscillatory power and coherence after cannabinoid exposure? Gamma signal is involved in maintaining local and global circuits, and a local disruption could lead to global changes in distal brain areas (13). Also, eCB tone is critical for coordinating neurodevelopment and early-life disruptions of the eCB, like adolescent cannabinoid exposure (115,116), can lead to long-term baseline oscillatory changes implicated in the etiology of mental illnesses (26).
Although the review focused on schizophrenia, major depressive disorder, and bipolar disorder, it is important to consider which other serious mental illnesses may involve, or be precipitated or ameliorated by, cannabinoids and eCB modulation. Patients with attention-deficit/hyperactivity disorder, obsessive-compulsive disorder, anxiety-related disorders, and post-traumatic stress disorder all report greater rates of cannabis use and worsened outcomes (117,118). Some of these patients may also benefit from therapeutic targeting of the eCB (119)(120)(121)(122)(123), highlighting the need for further research into this area.
Future studies should focus on comparing the effects of varying administration routes, toward a set of standardized methodologies used to examine cannabinoid exposure in models of mental illness, as these studies often use different routes and doses of cannabinoid administration which may contribute to inter-study variability. The other cannabinoids in the cannabis plant must also be investigated, as they demonstrate partial agonist activity for both CB1R and CB2R, and produce signature CB1R-dependent behavioral responses in mice tested using the cannabis tetrad (59). Additionally, whole-brain examinations must be completed to connect disparate observations of regional differences in brain activity both before, and after, acute and chronic cannabinoid exposure. Recent studies have used preclinical magnetic resonance imaging to examine the mechanisms underlying serious behavioral dysfunctions in some models of mental illness (34). Furthermore, differences in circuit activity may be related to different medications taken by patients, or inherent differences in brain functional connectivity and/or the measured state (124), which might make the reverse translation of these findings to animal models difficult. Thus, a holistic approach to studying and translating circuit dysfunctions in animal models is imperative. Finally, although the therapeutic relevance of disrupted oscillatory activity in cannabis use and serious mental illnesses is contested, modulation of gamma oscillations should be investigated to characterize the pathogenic nature of aberrant gamma signal, and to hopefully reveal a circuit-based mechanism that can be targeted for intervention (16,39,125).

AUTHOR CONTRIBUTIONS
BJ drafted the review and formatted Figures 1-4. JK formulated the idea for the review and guided the research and writing process. Both authors contributed to the article and approved the submitted version.