Dysregulations of appetite, eating behavior and body weight are hallmark symptoms of a wide range of chronic and disabling illnesses that can collectively be referred to as food intake disorders (FID).¹ In this sense, FID include obesity (OB) and eating disorders such as anorexia nervosa (AN) and bulimia nervosa (BN).^{2, 3} The core behavioral features of these disorders include either food avoidance or excessive food intake,² which may be accompanied by compensatory behavior (that is, fasting, excessive physical exercise, vomiting and/or laxative/diuretic use) intended to control weight (especially in AN and BN). This, in turn, might underlie abnormalities in body mass index (BMI), ranging from extreme underweight to morbid OB. Moreover, functional dyspepsia (FD), a prevalent functional gastrointestinal disorder characterized by meal-induced epigastric symptoms, is often accompanied by disturbed appetite and food intake as well as unintentional weight loss.⁴ Together, these disorders represent major global health problems that put an enormous demand on health-care services, not at least because of their high medical comorbidity.

During the last decade, the endocannabinoid system (ECS) emerged as one of the most important neuromodulatory systems involved in both the central and peripheral regulation of food intake and body weight.^{5, 6} The cerebral type 1 cannabinoid receptor (CB₁R) is the most abundant G-protein-coupled receptor in the central nervous system, where it resides predominantly at presynaptic nerve terminals to directly or indirectly modulate glutamatergic and GABAergic neurotransmission.⁷ It is now well accepted that stimulation of cannabinoid receptors by synthetic or plant-derived ligands such as Δ⁹-tetrahydrocannabinol (Δ⁹-THC) increases appetite and food intake in both humans and laboratory animals, especially toward foods with high palatability.^{8, 9} Conversely, pharmacological blockade of CB₁R reduces hunger, food intake and body weight of patients with OB.^{10, 11} Animal experiments indicate that these effects result in large part from targeting CB₁R in several interconnected brain circuits linking homeostatic centers in the brainstem and hypothalamus with the mesolimbic reward system that encompasses ventral tegmental area, striatum, amygdala, hippocampus and orbitofrontal cortex.^{5, 12} Together, these neural structures represent the major integration centers for the regulation of appetite and food intake, where the ECS is believed to modulate energy homeostasis, reward sensitivity and motivated behavior.^{13, 14} Specifically, it appears that endocannabinoids not only regulate the expression and release of hypothalamic orexigenic and anorexigenic signals, but also modulate activity in mesolimbic dopaminergic incentive pathways and opioidergic hedonic circuits, hence facilitating appetitive motivation as well as the pleasure of food during ingestion.^{15, 16, 17, 18}

Given the direct involvement of the ECS in the central neurocircuitry mediating energy homeostasis and food reward, it is not surprising that increasing evidence points toward disturbed endocannabinoid signaling in FID. Several reports indicate differences in plasma and/or tissue endocannabinoid levels as well as altered central CB₁R availability in both obese and anorectic conditions.^{9, 19, 20, 21, 22, 23, 24} Moreover, specific genetic variants of several ECS components have been associated with AN, BN and OB.^{25, 26, 27, 28, 29} However, despite mounting evidence supporting disturbed ECS signaling in several separate pathological eating- and weight-related conditions, so far there are no in vivo human studies linking endocannabinoid function in the key food intake-related brain areas to body weight along the BMI spectrum.

In this study, we used positron emission tomography (PET) imaging with the selective CB₁R radioligand [¹⁸F]MK-9470³⁰ to investigate for the first time whether in vivo cerebral CB₁R availability in the key homeostatic and reward-related brain areas is associated with BMI in patients with FID covering a wide BMI range (AN, BN, FD with severe weight loss and OB), and in healthy subjects within the normal BMI range. As a second objective, conjunction and interaction analyses were performed to investigate whether potential CB₁R–BMI associations would differ between both groups.

The study was approved by the local ethics committee of the University Hospital and KU Leuven and was performed according to the latest version of the World Medical Association Declaration of Helsinki. All subjects provided written informed consent after receiving a full explanation of the study procedures.

The psychobiological processes involved in (disordered) eating behavior and body weight regulation are complex and incompletely understood,⁴¹ but converging evidence points toward an important role of the neural circuits involved in the homeostatic and reward-related aspects of food intake, where the ECS has a vital role as neuromodulatory system.⁴² As a result, ECS dysfunction has become an auspicious pathophysiological mechanism and treatment target for several disorders of food intake, especially OB; however, the exact nature of this dysfunction remains unclear.

In the present study, we demonstrate for we believe the first time that CB₁R availability in homeostatic and mesolimbic reward regions is inversely related to BMI in health and FID along the BMI continuum (that is, AN, BN, FD with weight loss, and OB). Specifically, our PET results show that lower CB₁R levels in homeostatic brain areas such as the hypothalamus and brainstem are significantly associated with higher BMI in both healthy subjects and patients with FID. These findings indicate that variations in the endocannabinoid neurocircuitry in brain regions essential for energy balance regulation are inextricably linked to body weight, possibly reflecting a compensatory mechanism aimed at restoring energy homeostasis. However, in patients with FID along the BMI spectrum, additional negative correlations between CB₁R availability and BMI were found throughout the mesolimbic reward system, including the midbrain, striatum and orbitofrontal cortex. This suggests that CB₁R level deviations in brain areas implicated in encoding the incentive and hedonic value of food may have a role in the disordered hedonic eating behavior and body weight as observed in these patients.

Our data do not provide evidence for causality of the observed changes in CB₁R levels. In our opinion, two possible interpretations are conceivable. First, these ECS changes may predispose subjects to aberrant body weight by interfering with the central regulation of appetite, food intake and energy balance. Alternatively, changes in CB₁R availability might be a consequence of abnormal BMI and hence, indirectly, disturbed food intake. However, as there is currently no human evidence advocating one assumption over the other, future follow-up studies in subjects along the BMI spectrum should address the state- or trait-related nature of our findings. Furthermore, in both these explanations, the altered CB₁R receptor availability may be a primary phenomenon or secondary to abnormal central endocannabinoid levels.

To date, only a limited amount of human data exist on the association between ECS function and BMI. Some peripheral components, including circulating plasma endocannabinoid levels,⁴³ activity of the endocannabinoid-degrading enzyme fatty acid amide hydrolase in subcutaneous adipocytes,⁴⁴ and perirenal visceral adipose tissue CB₁R expression levels⁴⁵ have been found to correlate positively with BMI in subjects ranging from normal weight to OB. In addition, a CB₁R gene polymorphism was associated with lower BMI in healthy subjects with a wide BMI spread.²⁶ However, most research has focused on ECS alterations within separate FID subgroups, especially AN and OB, which has led to the hypothesis of a (chronic) hypo- and hyperactivity of the (peripheral) ECS in, respectively, AN and OB conditions.^{9, 21, 22, 23, 46, 47} For example, upregulation of peripheral endocannabinoid signaling in overweight and OB individuals with and without binge-eating disorder has been demonstrated.^{9, 21, 48} Moreover, both animal and human studies have demonstrated the efficacy of CB₁R antagonists/inverse agonists such as Rimonabant in reducing food intake and body weight in OB.^{10, 11} Preclinical studies have also shown increased hypothalamic endocannabinoid levels in diet-induced OB as well as several genetic models of OB.^{19, 49} It is suggested that this ECS overactivity in OB might originate from a high-fat diet and subsequent increased availability of polyunsaturated fatty-acid precursors for endocannabinoid biosynthesis. The hypothesis of ECS hypoactivity in anorectic conditions mostly originates from indirect evidence and animal work. Cannabinoid agonists such as dronabinol are used as therapeutic agents to treat AN and cachexia in cancer and AIDS patients.^{50, 51} Preclinical studies, addressing the effects of short-term starvation, have reported increased endocannabinoid levels in the limbic forebrain and hypothalamus of rats. However, in a context of prolonged starvation, reduced rather than increased brain endocannabinoid levels were observed throughout the entire mouse brain.^{52, 53} These apparent discrepancies can be interpreted as homeostatic endocannabinoid adaptations. In the short term, elevated endocannabinoid levels may be beneficial to trigger eating behavior, whereas in conditions of prolonged starvation (as in AN) this orexigenic mediator might be downregulated as an adaptive response to better cope with lack of food.^{47, 53}

The presumed ECS hypoactivity in AN and hyperactivity in conditions of hyperphagia and OB is thought to be accompanied by, respectively, compensatory CB₁R up- and downregulation. Using the same radiotracer, our group recently demonstrated increased CB₁R binding in the AN and FD subgroups of this FID cohort^{23, 31} as well as the activity-based rodent model of AN.²⁴ Conversely, CB₁R downregulation in OB has been showed preclinically in forebrain and hindbrain regions.^{20, 54} These data clearly support our finding of an inverse association between CB₁R availability and BMI in subjects across the BMI spectrum.

Although measurements of central endocannabinoid levels are impossible in humans in vivo, it is plausible that the negative CB₁R–BMI correlation in our study represents the statistical embodiment of compensatory changes in CB₁R availability, aimed at counteracting the above-mentioned aberrant endocannabinoid levels along the BMI continuum/FID spectrum. However, it has to be noted that deviations in CB₁R levels could also occur independently from endocannabinoid content⁵⁵ or follow changes of endocannabinoid tone in the same direction,⁵⁶ as has been observed in AN²¹ and several other pathological conditions.⁵⁷ Although differences in experimental methods can partly explain opposite findings within the same (food intake) disorder, this may well reflect the complexity of ECS regulation under pathological as well as physiological conditions.¹³ For example, it is uncertain whether peripheral endocannabinoid levels reflect the CNS status, as endocannabinoids are released on demand and rapidly metabolized in tissues.²³ Measured (brain) tissue levels also do not necessarily reflect extracellular, and hence CB₁R-active, content. However, whereas several explanations for the aberrant CB₁R availability along the BMI spectrum are conceivable, we speculate that a compensatory mechanism (receptor desensitization and/or downregulation) is plausible from a large intracellular CB₁R reserve.⁵⁸ In support, other G-protein-coupled receptors (for example, serotonin 5-HT_1A and dopamine D₂ receptors) are also inversely regulated by ligand availability.^{59, 60}

Despite these interesting PET data, some caution is warranted when interpreting our results. Although we did find regional differences in the negative CB₁R–BMI correlation between FID and CON, the group-by-BMI interaction analysis did not identify any areas where the negative correlation with BMI was significantly different between both groups. However, this could be because of the smaller sample size of the CON group compared with the FID, which lowers the power to detect such interaction effect. In addition, exploratory voxel-wise correlation analyses within the FID subgroups showed similar cluster patterns for AN, BN and FD but not OB. However, the latter might well be due to a ‘floor’ effect in CB₁R availability within the OB group, as the variability in mSUV (reflecting CB₁R availability) within this group was substantially smaller than in the other FID subgroups, leaving little variance to be explained by BMI. Moreover, confounding effects of weight differences on [¹⁸F]MK-9470 quantification are unlikely. Although the large differences in body weight between AN and OB may represent differences in [¹⁸F]MK-9470 distribution volume (V_T), we have previously validated the use of mSUV in both patient groups and the activity-based rat model of AN, where full kinetic modeling showed a strong positive correlation (R²=0.9) of mSUV with V_T.²⁴ Moreover, the mSUV parameter is normalized by a reference weight to exclude possible confounding effects due to large weight differences between groups. In this way, the systematical underweight of AN patients would imply an underestimation of CB₁R availability, whereas overweight would result in an overestimation. Hence, weight differences are not expected to explain the negative CB₁R–BMI correlation observed in this study. Indeed, similar findings have recently been reported in both cannabis users and controls using a different CB₁R tracer and quantification method, where a negative correlation between V_T and BMI was not driven by a peripheral confound.⁶¹ Finally, OB and part of the healthy control subjects were scanned using a different camera and, hence, also different acquisition protocols. However, our results remained unchanged when including camera/protocol as additional covariate of no interest in our analyses, thus excluding potential intercamera/protocol differences in CB₁R assessment.

In conclusion, to the best of our knowledge, we demonstrate for the first time that CB₁R availability in homeostatic brain regions is inversely related to BMI in both healthy subjects and patients with FID covering a wide BMI range (AN, BN, FD and OB). However, in FID, CB₁R availability is also negatively correlated with BMI throughout the mesolimbic reward system. These results indicate that the cerebral homeostatic CB₁R system is inextricably linked to BMI, with additional involvement of reward areas under conditions of disordered body weight. Thus, combined with (pre)clinical findings concerning peripheral ECS functioning, our results corroborate a key role for the ECS in body weight regulation and support the idea of pharmacological manipulation of the central ECS as a beneficial therapeutic approach for FID.

We thank the subjects who participated in this study. We also thank the Leuven PET radiopharmacy and nuclear medicine team for their collaboration, as well as Merck & Co for the availability of the [¹⁸F]MK-9470 precursor. This work was financially supported by the Research Council KU Leuven (OT/05/58). KVL is Senior Clinical Investigator and JC is a postdoctoral fellow of the Research Foundation - Flanders (FWO). LVO is a research professor of the KU Leuven Special Research Fund (BOF, Bijzonder Onderzoeksfonds). NW is funded by a research grant from the FWO. KVL received a research grant for the study in obese patients by Merck. JT and LVO did not receive financial support related to this work.