Preparations from the hemp plant Cannabis sativa L. have been used medicinally for millennia. Among their unique active components (the cannabinoids), Δ⁹‐tetrahydrocannabinol (THC) is the most relevant owing to its high potency and abundance. THC exerts a wide variety of biological effects by mimicking endogenous substances [the endocannabinoids anandamide (N‐arachidonoylethanolamine, AEA) and 2‐arachidonoylglycerol (2‐AG)] that engage specific cellular cannabinoid receptors (Mechoulam et al., 2014; Pertwee et al., 2010). Endocannabinoids, together with their receptors and the proteins responsible for their synthesis, transport, degradation, and bioconversion, constitute the so‐called “endocannabinoid system” (ECS), a pivotal neuromodulatory network that controls a plethora of biological functions. So far, two major cannabinoid‐specific receptors ‐ CB₁R and CB₂R ‐ have been cloned and characterized from mammalian tissues. Most of the effects produced by cannabinoids in the central and peripheral nervous system rely on the activation of CB₁R molecules located largely on neurons. In contrast, CB₂R is more highly abundant in immune cells, and may also be present in cells ‐including cancer cells‐ from other origins. The cannabinoid receptor ligands AEA and 2‐AG are synthesized from membrane lipids primarily by the enzymes N‐acyl‐phosphatidylethanolamine‐phospholipase D (NAPE‐PLD) and diacylglycerol lipase α/β (DAGLα/β; DAGLα accounting for most 2‐AG production in the adult brain), respectively. Subsequently, AEA and 2‐AG are deactivated mainly by the enzymes fatty acid amide hydrolase (FAAH) and monoacylglycerol lipase (MAGL), respectively (Piomelli, 2003; Katona & Freund, 2008; Castillo et al., 2012).

It has been long known that glial cells express functional CB₁R and CB₂R, the former especially in neuroglia and the latter especially in microglia (Stella, 2010). Likewise, many glioma cell lines express detectable amounts of CB₁R and CB₂R mRNA and protein (Galve‐Roperh et al., 2000; Lorente et al., 2011; Sánchez et al., 1998; Vaccani et al., 2005). Moreover, a series of studies has analyzed the expression of CB₁R and CB₂R, as well as other ECS elements, in specimens from human gliomas (Table 1). However, several inconsistencies between these studies have come out, likely because the limited availability of these samples precludes the achievement of appropriate patient‐population sizes and matched (particularly in age and sex) control specimens. Which sample represents the best internal control is a recurrent question in glioma research. So far, the most accepted procedure is to use adjacent, non‐tumoral tissue if resection is safe for the patient during surgery (Lemée et al., 2013), but this has not been applied to all the studies conducted to date (Table 1). Other inherent limitations, such as demographic bias and sample handling, might contribute as well to the observed differences. Overall, the most consistent finding of these studies seems to be the up‐regulation of CB₂R expression in high‐grade glioma samples (Calatozzolo et al., 2007; De Jesús et al., 2010; Ellert‐Miklaszewska et al., 2007; Hashemi et al., 2020; Held‐Feindt et al., 2006; Sánchez et al., 2001; Schley et al., 2009; Wu et al., 2012). This is believed to occur mainly in endothelial cells of blood vessels and in infiltrated immune cells, although tumor cells also express this receptor (Hashemi et al., 2020; Held‐Feindt et al., 2006). On the contrary, increases, decreases or no changes of CB₁R expression in high‐grade glioma biopsies have been reported. When comparing to adjacent, non‐tumoral tissue, CB₁R was found up‐regulated in two studies (Hashemi et al., 2020; Wu et al., 2012), and the report that showed a decreased CB₁R density (De Jesús et al., 2010) used membrane preparations instead of whole‐tissue homogenates, thus ruling out intracellular receptors a priori. Unlike CB₂R, CB₁R seems to reside mainly on glioma cells, as assessed by co‐localization with the astrocytic marker GFAP (Hashemi et al., 2020; Wu et al., 2012).

Fewer studies have analyzed the expression of enzymes involved in endocannabinoid metabolism, and the levels of endocannabinoids themselves, in human gliomas (Petersen et al., 2005; Wu et al., 2012) (Table 1). While enzymes responsible for the synthesis and degradation of AEA were down‐regulated in samples from glioma patients in both studies, the levels of AEA were inconsistent, with the initial report showing increased AEA and the subsequent one finding decreased AEA in the tumor specimens. Another study found reduced levels of AEA in meningioma samples, as well as in a sole glioblastoma biopsy (Maccarrone et al., 2001). As AEA levels are well known to increase over time in postmortem, anoxic brains (Schmid et al., 1995), these discrepancies might reflect differential handling of the samples between studies. So far, to the best of our knowledge, only one study has analyzed the expression of enzymes involved in 2‐AG metabolism. The authors found that while DAGLα levels were similar to those of matched controls, MAGL expression was significantly reduced, in concert with the enhanced levels of 2‐AG in the glioblastoma samples (Wu et al., 2012). This is in line with a previous report showing increased levels of 2‐monoacylglycerols in glioma specimens, although in this study a significant 2‐AG boost was not evident (Petersen et al., 2005). Taken together, these data suggest that human gliomas may have an overactive 2‐AG‐CB₂R signaling axis, although information regarding most components of the system is scarce and, in some cases, inconsistent. Hence, further research is necessary to clarify which precise cell types within the tumor express the different elements of the ECS, as well as how the dynamic regulation of these proteins and lipids along disease malignancy occurs.

The recent development of next‐generation sequencing procedures has produced vast amounts of genomic data of samples of basically every disease, including glioblastoma. Public and user‐friendly data‐mining search engines currently empower researchers to easily interrogate freely‐accessible genomic datasets. As a mere proof of concept, here we used the Xena platform (http://xena.ucsc.edu) (Goldman et al., 2020) to compare the mRNA expression of the main components of the ECS between healthy brain, low/medium‐grade gliomas (grades 2–3) (TCGA Research Network, 2015), and glioblastoma (grade 4) (Brennan et al., 2013) by using a combined cohort of The Cancer Genome Atlas (TCGA) (https://www.cancer.gov/tcga) and Genotype Tissue Expression (GTEx) (https://gtexportal.org/home) samples (Figure 1). This approach allowed a very large sample size, which provides sufficient statistical inference power to unveil minor differences between groups. In fact, significant changes between the three groups were detected for every element of the ECS examined, with the sole exception of CB₁R expression between healthy brain and glioblastoma samples. These differences, both regarding CB₂R expression and broadly speaking, are similar to the published data discussed above and shown in Table 1. The same trends were observed when comparing with adjacent, non‐tumoral tissue, although in this case the control sample size was too small (n = 5) for statistical analyses to be performed with enough confidence. Moreover, according to the Chinese Glioma Genome Atlas (http://www.cgga.org.cn), CB₁R expression is significantly up‐regulated in IDH‐wild type grade 4 gliomas versus IDH‐mutant grade 4 gliomas (p = 2 × 10⁻⁵), while CB₂R expression does not change (p = .94). We also conducted survival analysis on the 166 samples of the aforementioned TCGA glioblastoma dataset (Brennan et al., 2013), which were stratified for high or low mRNA expression of the different ECS components, as assessed by the Cutoff Finder bioinformatic tool (Budczies et al., 2012). Log‐rank tests on the respective Kaplan–Meier plots only showed a subtle difference (p = .0418) in the case of MAGL mRNA, with individuals bearing tumors with low MAGL expression displaying slightly less overall survival than those with high MAGL expression.