Introduction

The endocannabinoid system is one of the most widely distributed neurotransmitter systems in the human brain [1, 2]. Cannabinoid receptor 1 (CB1), the main receptor involved in central endocannabinoid signaling, modulates synaptic circuits that play a prominent role in psychopathology. For example, in the basolateral amygdala, CB1 activation facilitates the extinction of aversive memories by inducing long-term depression of GABAergic synapses [3]. In the medial prefrontal cortex, CB1 receptors regulate stress reactivity by terminating stress-induced corticosterone release [4]. In the ventral striatum and midbrain, CB1 activation is thought to be responsible for cannabis’ rewarding properties [5], and blockade of the CB1 receptor leads to reductions in self-administration of cannabis and other drugs of abuse [6], possibly by decreasing drug-induced dopamine release [7, 8]. Given the diverse role of endocannabinoid signaling in these and other brain regions, the endocannabinoid system had been proposed as a pharmacological target for a range of psychiatric disorders, including mood disorders [9], anxiety disorders [10], and substance use disorders [11]. However, there is limited knowledge as to whether these disorders are associated with altered endocannabinoid signaling in humans.

To better determine whether psychiatric disorders are associated with endocannabinoid signaling abnormalities in the brain, we comprehensively reviewed human positron emission tomography (PET) studies employing radiotracers targeting the endocannabinoid system. We specifically selected studies that compared individuals with psychiatric disorders to healthy controls. If differences were present, we sought to determine whether these differences were global or regional. A literature search found relevant studies for five psychiatric disorders: cannabis use disorder, alcohol use disorder, schizophrenia, post-traumatic stress disorder (PTSD), and eating disorders (for details regarding the search strategy, see Supplemental methods). Studies investigating patients with neurological or neurodegenerative diseases, including temporal lobe epilepsy [12], migraine [13], Alzheimer’s disease [14], and Parkinson’s disease [15], were excluded from our review. Important details from each study, including the sex distribution of the sample, sample size, and method used to estimate receptor availability, are displayed in the accompanying tables.

Radiotracers targeting the endocannabinoid system

Three radiotracers have been used to probe the CB1 receptor in human PET studies of psychiatric disorders: [18F]MK-9470, [18F]FMPEP-d2, and [11C]OMAR (Table 1) [16, 17]. These tracers are used to provide an index of CB1 receptor availability, though PET cannot generally distinguish whether altered radiotracer binding is due to changes in receptor density, occupancy, or affinity [18]. Although binding potential (BPND) is commonly used as a measure of receptor availability in PET studies, its estimation often requires a reference region devoid of receptors [18]. For CB1, there is no suitable reference region; therefore, researchers have used other estimates of receptor binding, such as volume of distribution (VT) and modified standardized uptake values (mSUV). Unfortunately, these estimates can be affected by nonspecific binding (i.e., binding of the radioligand to molecules other than CB1), which represents a significant methodological limitation. Other radiotracer-related issues may also confound results. For example, radioactivity has been shown to accumulate in the skull during [18F]FMPEP-d2 scans due to in vivo defluorination and bone uptake of 18F-fluoride ions. This may, in turn, affect measurements in the adjacent cortex [19]. Despite these limitations, PET imaging using all three radiotracers has shown good test–retest reliability [19,20,21]. Furthermore, in healthy subjects, higher CB1 binding has been found in the basal ganglia, cerebral cortex, and hippocampus relative to the pons and white matter [22, 23], which is in accordance with the distribution of CB1 receptors observed in postmortem studies [2].

Table 1 Radioligands targeting the endocannabinoid system used in human PET studies

Endocannabinoid signaling is primarily terminated by the enzymes fatty acid amide hydrolase (FAAH) and monoacylglycerol lipase (MAGL), which metabolize the endogenous cannabinoids anandamide and 2-arachidonoylglycerol (2-AG), respectively. [11C]CURB is the only radiotracer that has been used to probe FAAH activity in human PET studies of individuals with psychiatric disorders (Table 1). [11C]CURB is an irreversible FAAH inhibitor that shows high brain uptake in rodents, with highest uptake levels in the cortex and cerebellum [24]. As with radiotracers targeting CB1, imaging with [11C]CURB has been demonstrated to have good test–retest reliability [25]. There are several lines of evidence suggesting that [11C]CURB provides an index of FAAH activity. First, individuals with the FAAH C385A polymorphism (rs324420), which reduces FAAH expression and activity [26, 27], have reduced [11C]CURB binding [28]. Second, oral administration of a FAAH inhibitor greatly reduces [11C]CURB binding in humans [25]. At the present time, no radiotracers targeting MAGL have been used in human studies.

PET studies of substance use disorders

Cannabis use disorder

Cannabis contains multiple compounds that bind directly to central CB1 receptors; therefore, cannabis users are an ideal sample for PET studies of the endocannabinoid system. The endocannabinoid system is also a promising target for the treatment of cannabis use disorder. Agonists of the CB1 receptor, such as dronabinol, nabilone, and nabiximols, have been used to successfully treat cannabis withdrawal [29,30,31,32]. Whether CB1 agonists or antagonists are useful in relapse prevention is less clear. Clinical trials testing dronabinol and nabiximols for cannabis use disorder have failed to show increased abstinence rates [33,34,35]. There is some evidence from the non-human primate literature that CB1 antagonists could be useful for this indication [36, 37], although these compounds have never been tested for relapse prevention in humans. Given the lack of effective pharmacotherapies for this disorder [38], a better understanding of cannnabis’ effect on central endocannabinoid signaling could help guide the development of novel pharmacological treatment strategies.

Human PET studies have found that chronic cannabis use is associated with decreased CB1 availability (Table 2) [39,40,41]. The preponderance of evidence indicates that cannabis-induced reductions in CB1 availability are global [40, 41], although one study found that reductions were mostly restricted to cortical regions [39]. These changes may be responsible for the symptoms of cannabis withdrawal, as CB1 availability on day 2 of abstinence was negatively correlated with withdrawal severity (Spearman’s Rho = −0.67) [40]. CB1 availability appears to normalize rapidly after a period of abstinence [39, 40], indicating that cannabis-induced alterations in endocannabinoid signaling may be reversible. However, these findings require replication in female samples, as there were only two female cannabis-using subjects across all three studies.

Table 2 PET studies investigating the endocannabinoid system in substance use disorders

The effect of chronic cannabis use on FAAH binding has also been examined in humans [42]. In accordance with the CB1 receptor findings, chronic cannabis users demonstrated global reductions in FAAH binding. Strong negative correlations were found between FAAH binding and delta-9-tetrahydrocannabinol (Δ9-THC) metabolite concentrations in blood and urine, suggesting that these reductions are directly related to Δ9-THC exposure. The observed findings were contrary to the authors’ hypothesis; FAAH was expected to be upregulated to compensate for high levels of exogenous CB1 stimulation. However, chronic Δ9-THC use may also lead to reduced anandamide synthesis which, in turn, could lead to decreased FAAH synthesis or activity. Indeed, there is evidence that cerebrospinal fluid levels of anandamide are reduced following heavy cannabis exposure [43], but whether these reductions are secondary to decreased anandamide synthesis or other mechanisms remains unclear.

Alcohol use disorder

There is evidence that altered endocannabinoid signaling affects human alcohol consumption. Individuals of European ancestry with the FAAH C385A polymorphism have a higher probability of being diagnosed with alcohol dependence [44]. Furthermore, dependent individuals with this polymorphism consume more alcohol, have higher Alcohol Use Disorders Identification Test (AUDIT) scores, and are more likely to meet criteria for severe dependence [44]. This accords with extensive rodent literature showing that increasing CB1 signaling induces alcohol consumption [45], whereas CB1 blockade reduces alcohol intake [46,47,48]. On the other hand, clinical studies testing rimonabant, an inverse agonist of the CB1 receptor, have not shown any efficacy in reducing alcohol consumption or relapse [49, 50]. Human PET studies could help clarify whether individuals at risk for alcoholism have aberrant endocannabinoid signaling and could also prove useful to elucidate whether acute and chronic consumption alter endocannabinoid metabolism and CB1 receptor density.

Three PET studies have examined whether alcohol dependence is associated with CB1 availability in males (Table 2). A small preliminary study comparing alcohol-dependent males to healthy male controls found increased CB1 receptor availability in dependent individuals [51]. Subsequent larger studies have not supported this finding. These studies found global reductions in CB1 availability in alcohol-dependent individuals versus controls, which persisted after a period of abstinence [52, 53]. The second of these studies also tested the effects of acute intravenous ethanol administration on CB1 availability in healthy males [53]. Interestingly, as opposed to the decreased CB1 availability observed with chronic ethanol use, acute administration was associated with global increases in CB1 availability. The reason that acute and chronic ethanol consumption appear to have opposing effects remains to be determined. Follow-up studies should also evaluate whether alcohol use affects endocannabinoid signaling in females.

PET studies of other psychiatric disorders

Schizophrenia

The endocannabinoid system may be involved in the etiology of psychotic disorders. For example, there is evidence that cannabis use has dose-dependent psychotogenic effects, as both higher levels of cannabis use [54] and use of cannabis with higher concentrations of Δ9-THC [55] are associated with a greater likelihood of developing psychosis. In individuals with schizophrenia, cannabis abuse has been associated with relapse of psychotic symptoms [56] and higher rates of hospitalization [57]. The link between Δ9-THC use and psychosis has also been demonstrated more directly in human laboratory studies. Intravenous administration of Δ9-THC in healthy subjects dose-dependently increases both psychotic symptoms [58] and an electroencephalography measure of cortical noise associated with psychosis [59]. Interestingly, there is growing evidence that cannabidiol, a negative allosteric modulator of the CB1 receptor [60], reduces psychotic symptoms in patients with schizophrenia, both as a monotherapy [61] and as an adjunctive treatment [62]. On the other hand, a clinical trial found that blockade of CB1 was ineffective at reducing psychotic symptoms in individuals with schizophrenia and schizoaffective disorder [63], suggesting the possibility that cannabidiol may exert its antipsychotic effects through other molecular targets.

Three human PET studies have assessed CB1 receptor binding in individuals with schizophrenia. The first study found higher [11C]OMAR binding in individuals with schizophrenia than in healthy controls, which only reached statistical significance in the pons [22]. This study had several limitations. First, all subjects with schizophrenia were undergoing treatment with atypical antipsychotics (either risperidone or olanzapine); therefore, any differences in radiotracer binding could have been secondary to medication effects rather than underlying pathophysiology. Second, the schizophrenia group’s mean age was 8.6 years older than the control group, and CB1 binding was negatively correlated with age in this study. A much larger study overcame these issues by examining [18F]MK-9470 binding in medicated and unmedicated individuals with schizophrenia and age-matched controls [64]. Global increases in CB1 receptor availability were observed in both schizophrenia groups. However, in contrast to earlier findings, a third study employing [11C]OMAR found global reductions in CB1 receptor availability in both medicated and unmedicated patients with schizophrenia [65], raising questions about whether endocannabinoid signaling is increased or decreased in this disorder.

Another question raised by these studies is whether antipsychotic treatment alters CB1 availability. Ceccarini et al. [64] compared CB1 availability in medicated and unmedicated patients with schizophrenia and found that the medicated group had CB1 availability closer to that of controls, suggesting that antipsychotic treatment may normalize CB1 availability. In concordance with this finding, Ranganathan et al. [65] also found that antipsychotic treatment tended to normalize CB1 availability. Normalization of CB1 availability could be secondary to the pharmacological effects of antipsychotics, such as D2 receptor antagonism, or may be a sign of clinical response to treatment. Several lines of evidence suggest that the latter may be the case. First, Ranganathan et al. [65] found that antipsychotic dose was not correlated with CB1 availability, suggesting that increased D2 antagonism does not reduce CB1 availability. Second, all three studies demonstrated correlations between CB1 availability and symptom severity measures [22, 64, 65], although there was disagreement as to which symptoms correlated best with CB1 availability. Third, individuals with schizophrenia have higher circulating and cerebrospinal levels of anandamide than healthy controls [66, 67], but circulating levels return to normal following clinical remission [66]. Thus, these findings provide preliminary evidence that CB1 receptor availability and circulating levels of endocannabinoids may have some utility as a biomarker for clinical response in individuals with schizophrenia.

Post-traumatic stress disorder

Endocannabinoid signaling may modulate anxiety and stress responses in humans. For example, subjecting healthy subjects to experimental stress induction paradigms leads to increases in circulating cannabinoid levels [68, 69], which may buffer human stress responses. In support of this, individuals carrying a FAAH polymorphism that decreases anandamide metabolism demonstrate more rapid fear extinction [26, 70] and a reduced amygdala response to fearful faces [71]. Interestingly, acute stress-induced increases in circulating endocannabinoid concentrations have not been observed in subjects with comorbid PTSD and alcohol dependence [72], indicating potential aberrancies in endocannabinoid-related stress regulation in these subjects.

Several studies have found reductions in circulating endocannabinoid levels in individuals with PTSD [73, 74], although one study demonstrated lower circulating levels of anandamide but not 2-AG [74], whereas another demonstrated lower levels of 2-AG but not anandamide [73]. Hair samples derived from individuals with PTSD also contain lower levels of endocannabinoids compared to controls [75]. Only one human PET study has examined central CB1 binding in PTSD. This study compared untreated individuals with PTSD to trauma-exposed and healthy controls and found global increases in CB1 binding in the PTSD group compared to both control groups [74] (Table 3). This increase in CB1 binding could represent upregulated receptor levels secondary to depressed endocannabinoid levels. In support of this, [11C]OMAR binding was found to correlate negatively with circulating anandamide levels, although the strength of this correlation was modest (r = −0.27). In a subsample of this study (n = 20), both decreased circulating anandamide levels and elevated CB1 receptor availability in the amygdala were associated with increased attentional bias to threat using a dot-probe task [76]. These studies provide preliminary evidence that the endocannabinoid system is involved in PTSD symptomatology.

Table 3 PET studies investigating the endocannabinoid system in other psychiatric disorders

Since endocannabinoids appear to influence stress and anxiety, cannabinoid receptor agonists have been proposed as a potential treatment for PTSD. One small randomized placebo-controlled crossover trial indicated that nabilone, a synthetic analog of Δ9-THC, reduced PTSD symptoms in male military personnel compared to placebo [77]. Despite the lack of randomized trials investigating cannabis use for this indication, some states allow the use of medical marijuana to treat PTSD [78, 79]. This is concerning given that individuals with PTSD have a substantially increased likelihood of being diagnosed with cannabis use disorder [80]. Further research is needed to clarify whether medical marijuana use for PTSD treatment is causing more benefit than harm.

Eating disorders

Exogenous cannabinoids have long been known to influence appetite and food intake. Whereas cannabis intoxication increases appetite [81], cannabis withdrawal is associated with decreased caloric intake and weight loss [31]. Dronabinol, an encapsulated form of oral Δ9-THC which was the first Food and Drug Administration (FDA)-approved cannabinoid, received its initial indication as an appetite stimulant for individuals with AIDS who had experienced anorexia and weight loss [82]. Despite the appetite-stimulating effects of acute Δ9-THC intoxication, chronic Δ9-THC use may actually be protective against obesity [83]. These putative anti-obesity effects remain poorly understood, although several central and peripheral mechanisms have been proposed [84, 85]. Like their exogenous counterparts, endogenous cannabinoids may also regulate appetite and food intake. Presenting food to normal weight and obese individuals has been shown to increase circulating anandamide [86]; levels of anandamide gradually decrease following food consumption or intravenous glucose administration [86, 87]. Multiple studies also suggest that exercise increases circulating anandamide levels [88, 89], suggesting that anandamide may stimulate appetite following energy utilization. The endocannabinoid system therefore represents a viable pharmacological target for disordered eating.

Clinical trials have investigated medications that block the CB1 receptor as a treatment for obesity. Four large randomized trials found that rimonabant, an inverse agonist of the CB1 receptor, led to weight loss and improvement of cardiometabolic risk factors when compared to placebo [90,91,92,93]. Rimonabant was approved by the European Medicine Agency as a treatment for obesity in 2006 [94]. However, regulatory approval was subsequently withdrawn when rimonabant was discovered to increase anxious and depressive symptomatology [95, 96], which may have led to the higher rate of suicide attempts and completions observed in a large trial investigating the use of rimonabant to prevent cardiovascular death in high-risk patients [97]. Given that blocking CB1 receptors decreases food intake, it would stand to reason that decreased CB1 signaling may be present in individuals who restrict their caloric intake. Indeed, there is evidence that individuals with anorexia nervosa do not exhibit normal endocannabinoid responses to food consumption [98].

Human PET studies have examined CB1 availability in individuals with disordered eating (Table 3). One study, using [18F]MK-9470, found that CB1 availability was globally increased in subjects with anorexia nervosa compared to both healthy controls and subjects with bulimia nervosa [99]. The bulimia nervosa group also had global increases in CB1 availability compared to healthy controls, although the difference did not reach statistical significance. Another study examined CB1 availability in functional dyspepsia, a disorder that is not included in the Diagnostic and Statistical Manual of Mental Disorders (DSM), in which patients have abdominal pain with no obvious organic etiology on diagnostic testing [100]. This study also found increased CB1 availability in cases versus controls, which persisted when subjects were re-scanned 3 years later [101]. An analysis pooling data from both studies, which also included an additional group of individuals with obesity, found a negative association between radiotracer binding and body mass index (BMI) [102], replicating an association observed in some [39] but not all [40, 51, 74] CB1 PET studies. This association raises the possibility that the fluctuations in BMI associated with disordered eating may explain the observed alterations in radiotracer binding, suggesting that altered CB1 availability may be a result of disordered eating rather than a causal factor. Prospective studies of subjects with anorexia and bulimia may help resolve this issue.

Discussion

Human PET studies have examined CB1 availability in several psychiatric conditions. Of these, studies comparing cannabis users to healthy controls have the most consistent findings. All three published studies demonstrated reduced CB1 availability in cannabis users, likely due to receptor downregulation following heavy cannabis exposure. This parallels studies of stimulant users, which suggest that chronic dopamine exposure leads to reductions in D2/D3 receptor availability [103]. As with stimulant users [104], a period of abstinence from cannabis use seems to normalize receptor availability [39, 40]. Studies investigating CB1 availability in individuals with alcohol use disorder and schizophrenia have been less consistent, with some studies reporting increased availability and others reporting decreased availability. Preliminary results in samples with PTSD and anorexia nervosa indicate altered endocannabinoid signaling, but other common psychiatric disorders such as major depressive disorder have yet to be studied. Across disorders, differences in CB1 availability tend to be global rather than regional, with radioligand binding showing strong correlations across brain regions.

Several factors could explain the discrepant results observed in the schizophrenia and alcohol use disorder studies. One possibility is that the use of different radiotracers influenced findings [105]. Indeed, several PET studies using [11C]OMAR, a reversible radiotracer, have yielded different results from those using the irreversible radiotracers [18F]MK-9470 and [18F]FMPEP-d2 [51,52,53, 64, 65]. On the other hand, two studies using [11C]OMAR to compare patients with schizophrenia to healthy controls reported results in the opposite direction [22, 65], which makes this hypothesis seem somewhat less plausible. Genetic factors may also affect CB1 availability. There is evidence that a variant of the CB1 gene (rs2023239) alters radiotracer binding, but only one study controlled for this polymorphism [52]. Furthermore, there is preliminary evidence of lower CB1 availability in healthy subjects of Indian ancestry compared to those of European and African ancestry [39]. A recent analysis of raclopride studies demonstrated significant differences in binding based on ancestry [106]; thus, ancestry may deserve greater consideration in human PET studies than previously assumed. Other confounding factors could include length of illness and symptom severity, which were correlated with CB1 availability in some studies (Supplemental Tables 1 and 2). Future studies should attempt to parse out the effects of each of these variables.

Tobacco use represents another potential confound, as subjects with psychiatric disorders tended to smoke at much higher rates than controls across studies (Supplemental Tables 1 and 2). A published abstract directly comparing CB1 availability in smokers and non-smokers using [18F]FMPEP-d2 found that smokers had significantly lower CB1 availability throughout the brain [107]. However, this contrasts with supplementary analyses performed in several published studies, which found that smoking status was not associated with CB1 availability in individuals with comorbid psychopathology [39, 41, 52, 53, 64, 74]. Only one supplementary analysis suggested a main effect of comorbid tobacco use, indicating that smokers with schizophrenia had higher CB1 availability than non-smokers with schizophrenia, although this was only significant at a trend level (P = 0.096) [65]. Thus, the preponderance of evidence suggests that comorbid tobacco use has no effect on CB1 availability in patients with psychiatric disorders. However, these supplementary analyses were often limited to a subgroup of cases who smoked and were potentially underpowered to detect main effects of tobacco use. Additional studies should be conducted to assess CB1 availability in smokers and non-smokers with and without psychiatric comorbidities in order to resolve this issue.

Most human PET studies evaluating the endocannabinoid system have focused on CB1. Less is known about whether psychiatric disorders are associated with alterations in other molecules involved in endocannabinoid signaling, such as FAAH, MAGL, and CB2. To date, only one published PET study has probed for abnormalities in endocannabinoid degradation in a clinical sample, employing [11C]CURB to measure FAAH activity in cannabis users [42]. This tracer should be used in other clinical populations with hypothesized aberrancies in endocannabinoid signaling, including samples with schizophrenia and PTSD. The development of radiotracers targeting MAGL [108] ensures that there will also be new studies evaluating the role of 2-AG in these disorders. CB2 receptors represent another target of potential interest. Although they are mainly expressed on peripheral immune cells, there is preclinical evidence that they are also expressed in the brain [109, 110], where they may affect addictive and schizophrenia-related behaviors [110,111,112]. Several attempts have been made to develop CB2 radiotracers, but these ligands have been unsuitable for clinical imaging studies due to high non-specific binding and unfavorable pharmacokinetic properties [113]. Hopefully, the synthesis of new tracers for these targets will allow for a greater understanding of all aspects of endocannabinoid signaling rather than focusing exclusively on CB1.

At the present time, we only possess a rudimentary understanding of endocannabinoid signaling in humans. The PET studies summarized in this review represent an initial attempt at characterizing alterations in endocannabinoid signaling in psychiatric disorders. Subsequent PET studies should be complemented by studies measuring circulating and cerebrospinal endocannabinoid levels and postmortem studies employing autoradiography and immunohistochemistry in individuals with and without disease. Furthermore, both in vitro and animal studies will be necessary to clarify how receptor downregulation or upregulation and changes in enzymatic activity affect cellular signaling and contribute to the behavioral phenotypes observed in these disorders. This work will hopefully lead to a greater understanding of endocannabinoid signaling in human psychopathology.