Endocannabinoid Signaling in the Central Nervous System

Abstract

It is hard to overestimate the influence of the endocannabinoid signaling (ECS) system on CNS function. In the 40 years since cannabinoids were found to trigger specific cell signaling cascades, studies of the ECS system continue to cause amazement, surprise and confusion! CB1 cannabinoid receptors are expressed widely in the CNS and regulate cell-cell communication via effects on the release of both neurotransmitters and gliotransmitters. CB2 cannabinoid receptors are difficult to detect in the CNS but seem to “punch above their weight” as compounds targeting these receptors have significant effects on inflammatory state and behavior. Positive and negative allosteric modulators for both receptors have been identified and examined in preclinical studies. Concentrations of the endocannabinoid ligands, N-arachidonoylethanolamine (AEA) and 2-arachidonoylglycerol (2-AG), are regulated by a combination of enzymatic synthesis and degradation and inhibitors of these processes are available and making their way into clinical trials. Importantly, ECS regulates many essential brain functions, including regulation of reward, anxiety, inflammation, motor control, and cellular development. While the field is on the cusp of preclinical discoveries providing impactful clinical and therapeutic insights into many CNS disorders, there is still much to be learned about this remarkable and versatile modulatory system.

Table of contents

The endocannabinoid system in the CNS is reviewed, with an emphasis on its role in neuronal-glial communications. Recent developments in the structural biology of the cannabinoid receptors and allosteric receptor modulators are included.

Graphical Abstract

Introduction

Our understanding of the integration of the endocannabinoid signaling (ECS) system into brain function has grown exponentially over the last several decades. The ECS system is found in all brain regions, and it is intimately involved in cell-cell communication. Mobilization of ECS ligands (the endocannabinoids [eCB]) is triggered by enzymatic processes that are down-stream of many diverse signaling cascades, including G protein coupled receptors (GPCRs), tyrosine kinase linked receptors and steroid hormone receptors. Furthermore, enzymatic regulation of ligand availability allows for a wide range of eCB concentrations, enabling graded activation of ECS. The receptors of the ECS system are expressed in many cell types in the CNS, including all types of glia, projection neurons and interneurons. The widespread receptor distribution suggests a signaling paradigm that has the capacity to modulate and coordinate functions in multiple cell types in the vicinity of eCB elevation. These fundamental features of ECS are consistent with functional evidence for a pervasive influence of ECS on brain function.

Highlights of the ECS System Origin Story

The CB1 receptor is identified.

The path of scientific discovery of ECS began with structural identification of Δ⁹-tetrahydrocannabinol (THC) as the phytocannabinoid responsible for the most obvious pharmacological effects of extracts of the cannabis plant (Gaoni & Mechoulam, 1964). Within several years, structural analogs of THC were identified and examined in pharmacological assays, including derivatives with extremely high potency and efficacy (Domino, Hardman, & Seevers, 1971; Hardman, Domino, & Seevers, 1971).

One synthetic analog, levonantradol, was developed by scientists at Pfizer (Johnson et al., 1981) and studied in clinical trials for its anti-emetic properties (Diasio, Ettinger, & Satterwhite, 1981). Importantly, its high efficacy and potency allowed Howlett and colleagues to build the biochemical case that levonantradol and THC bound to a previously unidentified GPCR that signaled through the heterotrimeric G protein, Gαi (Howlett, 1984, 1985; Howlett & Fleming, 1984; Howlett, Qualy, & Khachatrian, 1986). Another analog of this series, CP55940, was tritiated and used to identify a high affinity binding site for THC and levonantradol, providing definitive evidence that THC and analogs were agonists for a brain GPCR (Devane, Dysarz, Johnson, Melvin, & Howlett, 1988). Not long after, the CB1 cannabinoid receptor was cloned and molecularly confirmed to be a GPCR with very high expression in brain (Matsuda, Lolait, Brownstein, Young, & Bonner, 1990).

[³H]-CP55940 also proved to be very useful for autoradiographic studies, which allowed for mapping of the distribution of the CB1 receptor in brain (Herkenham et al., 1990). Importantly, autoradiographic studies in the cerebellum of mutant mice provided the first evidence that the CB1 receptor is present on axon terminals (Herkenham, Groen, Lynn, De Costa, & Richfield, 1991). Shortly thereafter, two groups demonstrated that CB1 receptor activation inhibited the opening of voltage operated calcium channels (VOCC) via a mechanism consistent with activation of Gαi/o proteins (Caulfield & Brown, 1992; Mackie & Hille, 1992). These impactful discoveries, which could not have occurred without the development of high affinity, high efficacy analogs of THC, were the first steps to uncovering the ECS system.

A completely different structural class of high affinity and high efficacy CB1 receptor agonists, the aminoalkylindoles, was identified and expanded by pharmacologists at Sterling (D’Ambra et al., 1992). The most well-studied member of this series, WIN 55,212–2, has an inactive stereoisomer (WIN 55,212–3) that can serve as a negative control (see for example, (Mackie & Hille, 1992)). Unlike the levonantradol/CP series of compounds, WIN 55,212–2 is not structurally related to THC and, therefore, not subject to governmental regulations that control access to analogs of THC. This allowed for many laboratories to obtain a CB1 receptor agonist and greatly expanded the biochemical and pharmacological studies of the newly discovered receptor.

The first antagonist/inverse agonist of the CB1 receptor, SR141716(A) (later named “rimonabant”), was identified by scientists at Sanofi (Rinaldi-Carmona et al., 1994). SR141716 and its close structural analog, AM251 (Gatley et al., 1997), have been essential tools for study of endogenous CB1 receptor signaling.

Endogenous ligands for the CB1 receptor are identified.

The first endogenous ligand for the CB1 receptor was isolated from pig brain and chemically identified as N-arachidonoylethanolamine (also called anandamide and abbreviated AEA) (Devane et al., 1992). AEA binds to the CB1 receptor at nanomolar concentrations and, like THC, behaves as a partial agonist (Burkey et al., 1997). Shortly thereafter, 2-arachidonoylglycerol (2-AG), was also identified as an endogenous ligand of the CB1 receptor (Mechoulam et al., 1995; Sugiura et al., 1995). 2-AG has lower affinity for the CB1 receptor than AEA but higher efficacy (Hillard, 2000) and higher total brain concentrations (Patel et al., 2005) although microdialysis data suggest that synaptic concentrations of 2-AG and AEA are in the same order of magnitude (Buczynski & Parsons, 2010). There is ample evidence that both 2-AG and AEA can function as endogenous agonists of the CB1 receptor, and they are considered the primary eCBs.

Role for eCB/CB1 receptor signaling in regulation of synaptic transmission is discovered.

Data demonstrating a prominent location of the CB1 receptor in axon terminals and functional role in regulating calcium entry in response to depolarization suggested that CB1 receptor signaling regulated neurotransmitter release. Electrophysiological studies conducted by Alger and colleagues around the same time demonstrated that depolarization of post-synaptic neurons resulted in a transient suppression of presynaptic neurotransmitter release at GABAergic neuron-pyramidal cell synapses in the hippocampus (Pitler & Alger, 1992). This process required G protein activation in the axon terminal (Pitler & Alger, 1994). In 2001, two laboratories simultaneously published results demonstrating that this process, called “depolarization-induced suppression of inhibition” (DSI), was mediated by eCB-CB1 receptor signaling (Ohno-Shosaku, Maejima, & Kano, 2001; Wilson & Nicoll, 2001). In addition, a third group demonstrated that a parallel process, depolarization-induced suppression of excitation (DSE), was also mediated by presynaptic CB1 receptor signaling at excitatory synapses onto cerebellar Purkinje cells (Kreitzer & Regehr, 2001). DSI and DSE are forms of activity-dependent, retrograde regulation of synaptic activity. In both DSI and DSE, elevation of calcium in the post-synaptic neuron results in synthesis and release of eCB, activation of presynaptic CB1 receptor and inhibition of neurotransmitter release.

Since these seminal studies, ECS-mediated retrograde regulation of synaptic activity has been demonstrated at many synapses. Synaptic ECS can be triggered by multiple activating mechanisms; and results in both short- and long-term changes in neurotransmitter release (Freund, Katona, & Piomelli, 2003; Hillard, 2015; Winters & Vaughan, 2021). This paradigm of synaptic regulation is ideally suited to a presynaptically expressed receptor that is activated by lipid ligands that can be rapidly mobilized without the requirement for vesicular release. Thus, inhibiting neurotransmitter release is a major mechanism by which endogenous and exogenous CB1 receptor agonists alter brain function.

CB2 Receptor is Identified.

Within the time frame of the discovery of the brain CB1 receptor, studies of the effects of THC on innate immune responses to viruses (Mishkin & Cabral, 1985) and lymphocyte activation (Klein, Newton, Widen, & Friedman, 1985) led to the hypothesis that THC also targets a receptor expressed by immune cells. Indeed, a second Gαi-linked GPCR that recognizes THC was molecularly identified (Munro, Thomas, & Abu-Shaar, 1993) and ultimately named the CB2 cannabinoid receptor. The CB2 receptor was subsequently shown to be widely expressed in human lymphocytes, including B and T cells and macrophages (Galiegue et al., 1995). Like most GPCRs, the CB2 receptor signals through both G protein- and ß-arrestin-pathways and ligands can exhibit bias for these signaling pathways (Soethoudt et al., 2017). Recent studies suggest that commonly occurring variants of the human CB2 receptor exhibit differences in ß-arrestin binding to CB2 receptor, although activation of G protein signaling is the same (Turu et al., 2021). One of the variants examined in this study (rs2501432) has been associated with schizophrenia (Ishiguro et al., 2010) and autoimmune diseases (Ismail & Khawaja, 2018; Tahamtan et al., 2020), suggesting that the change in signaling balance of CB2 receptor could have pathological consequences.

Interestingly, THC binds the CB2 receptor with moderate affinity but does not inhibit adenylyl cyclase (Bayewitch et al., 1996), findings that were originally interpreted as evidence that THC is an antagonist at the CB2 receptor. However, more recent studies indicate that THC is a ß-arrestin-biased agonist at the human CB2 receptor, with little or no effect on Gαi-mediated signaling (Yuan et al., 2021).

Both CP55940 and WIN 55,212–2 are high affinity and full efficacy agonists of the CB2 receptor (Soethoudt et al., 2017). The first CB2 receptor antagonist, SR144528, was identified by researchers at Sanofi (Rinaldi-Carmona et al., 1998); and several additional antagonists/inverse agonists have been identified subsequently, including AM630 (Ross et al., 1999) and the brain penetrant compound, SMM-189 (Reiner et al., 2014).

2-AG is an unbiased, full agonist at the CB2 receptor (Mechoulam et al., 1995; Soethoudt et al., 2017; Sugiura et al., 1995) and is generally accepted as the primary endogenous ligand for the CB2 receptor. On the other hand, the affinity of AEA for the CB2 receptor is low (Felder et al., 1995; Hillard et al., 1999; Slipetz et al., 1995), so it may only activate the CB2 receptor under conditions in which AEA concentrations are very high.

Our goal in the remainder of this review is to discuss studies and developments in our understanding of this system that are recent, novel and/or controversial, with a particular focus on how the ECS system is integrated into the fabric of the brain.

Selected Recent Discoveries Regarding Components of the ECS system

CB1 cannabinoid receptor

CB1 receptor structure.

Solutions of several CB1 receptor crystal structures together with computational approaches and homology modeling have provided molecular insights into how ligands of different pharmacological and structural classes interact with the CB1 receptor (see (Shahbazi, Grandi, Banerjee, & Trant, 2020) for an excellent review). Structures have been published of the CB1 receptor in agonist bound state (Hua et al., 2017), antagonist bound state (Shao et al., 2016) and complexed with an agonist and Gαi protein (Krishna Kumar et al., 2019). These structures identified an unusual arrangement of the extracellular domain in which the N terminal and the second extracellular loop together form a lid over the orthosteric binding pocket (Shao et al., 2016). This structural feature, which isolates the ligand binding pocket from the extracellular space, suggests that the path of ligand entry into the binding pocket is through the membrane bilayer (Hurst, Schmeisser, & Reggio, 2013). This is consistent with molecular dynamic simulations by Reggio and colleagues that predict an entry path for AEA that requires its solvation into the membrane, followed by entry into the binding pocket at the interface between the bilayer and transmembrane helix (TM) 6 (Lynch & Reggio, 2006; Reggio, 2010).

Agonist binding to the CB1 receptor reduces the volume of the ligand binding domain to one-half while moving a conserved tryptophan residue that acts as a toggle switch to open the G protein binding region (Krishna Kumar et al., 2019).

A CB1 receptor crystal structure in complex with the negative allosteric modulator (NAM) ORG27569 and orthosteric agonist CP55940 has also been solved (Shao et al., 2019). The structure indicates that ORG27569 binds to a site within the inner leaflet of the membrane that overlaps with a cholesterol binding site that is conserved in GPCRs. ORG27569 binding promotes the aromatic residues at the base of the binding pocket into an inactive conformation, which blocks agonist-induced shrinking of the orthosteric ligand binding pocket and thereby reduces opening of the G protein binding pocket. This model was refined recently in a study utilizing a genetically encoded ¹⁹F NMR probe which provided evidence that ORG27569 promotes the formation of conformational state in which cholesterol, rather than G protein, occupies the G protein binding site (X. Wang et al., 2021). As a result, G protein mediated signaling is reduced, consistent with identification of ORG27569 as a CB1 receptor NAM.

Subcellular distribution pattern of the CB1 receptor in neurons.

The advent of super resolution imaging techniques, particularly STORM, together with high quality antibodies and, more recently, fluorescently labeled ligands (Prokop et al., 2021) has allowed for precise exploration of the subcellular distribution of the CB1 receptor (Dudok & Soltesz, 2022). While it is well documented that the highest density of CB1 receptor in the CNS is on neuronal axon terminals, high-resolution imaging techniques have detected the presence of CB1 receptors in other parts of neurons (Figure 1) (Dudok & Soltesz, 2022). For example, a pool of plasma membrane associated CB1 receptor along axonal processes has been identified in some neurons (Zhou, Han, Xia, & Zhuang, 2019). The axonal CB1 receptor is clustered and exhibits a 190 nm distribution period in both cultured hippocampal neuronal axons (Zhou et al., 2019) and axons of inhibitory interneurons in the cortex and hippocampus (H. Li et al., 2020). A highly ordered skeletal structure, called the membrane-associated periodic skeleton (MPS), acts as a mechanical support and organizes the distribution of proteins in the axonal membrane in a periodic distribution pattern (K. Xu, Zhong, & Zhuang, 2013). Imaging studies demonstrated that CB1 receptor clusters along the axons were colocalized with ßII-spectrin, one of the protein components of the MPS; and proximity ligation assays in cultured neurons showed that axonal CB1 receptor are tightly associated with ankyrin B, which is known to link membrane proteins with the MPS (H. Li et al., 2020). Disruption of CB1 receptor/MPS association with the cytoskeleton resulted in loss of CB1 receptor clustering and decreased CB1 receptor agonist activation of both Akt and ERK signaling. Axonal CB1 receptor could play roles in neurogenesis and axonal pathfinding, both of which are regulated by ECS (Galve-Roperh, Palazuelos, Aguado, & Guzman, 2009).

Figure 1.

CNS cellular expression of cannabinoid receptors (CB1R and CB2R) and sources of 2-arachidonoylglycerol (2-AG). CB1 cannabinoid receptors (lavender) and CB2 cannabinoid receptors (orange) are found in on plasma membranes of neurons and glia as shown. CB1 receptors are also present in neuronal and astrocyte mitochondria. Neurons, microglia and astrocytes can all synthesize and release 2-AG, and microglial-derived extracellular vesicles “microvesicles” contain N-arachidonoylethanolamine (AEA).

A series of morphological and functional studies provide evidence that the CB1 receptor is present on neuronal mitochondria and its activation results in inhibition of cellular respiration (Benard et al., 2012; Fisar, Singh, & Hroudova, 2014; Hebert-Chatelain et al., 2016). Activation of the mitochondrial CB1 receptor results in recruitment and activation of mitochondrial Gαi protein, inhibition of soluble adenylyl cyclase, and reduced mitochondrial protein kinase A (PKA) activity. Reduction of PKA-mediated phosphorylation of complex I proteins of the electron transport chain results in reduced cellular respiration (Hebert-Chatelain et al., 2016).

Studies comparing the effects of membrane-permeant and impermeant CB1 receptor ligands suggest that mitochondrial CB1 receptor contributes to DSI in hippocampal pyramidal neurons (Benard et al., 2012). Furthermore, studies in mitochondrial-specific CB1 receptor knock outs suggest that mitochondrial CB1 receptor contribute to the acute effects of THC on memory (Hebert-Chatelain et al., 2016). Most recently, activation of mitochondrial CB1 receptor in striatonigral neurons has been shown to mediate cannabinoid-induced catalepsy (Soria-Gomez et al., 2021). These novel observations indicate that the mitochondrial CB1 receptor contributes to several functions thought to be mediated by CB1 receptor at the axonal terminal. The mechanisms that link mitochondrial functional changes to the behavioral effects are not yet known.

In summary, while the CB1 receptor is expressed at high density on axonal terminals, lower abundance CB1 receptor pools have been identified at other subcellular sites (Figure 1) (Puente, Rio, Achicallende, Nahirney, & Grandes, 2019). These and other studies (reviewed in (Dudok & Soltesz, 2022)) suggest that the ECS system affects signaling beyond regulation of synaptic activity. Location/function studies enabled by super-resolution imaging will undoubtedly continue to increase understanding of the complexity of cellular and behavioral effects of CB1 receptor activation.

CB1 receptor interacting proteins.

Like other GPCRs, it is well established that the CB1 receptor associates with G proteins (Mukhopadhyay, McIntosh, Houston, & Howlett, 2000), ß-arrestins (Jin et al., 1999) and GPCR kinases (GRKs) (Morgan et al., 2014). Like many GPCRs, the CB1 receptor is regulated by clathrin-mediated endocytosis following phosphorylation by G-protein receptor kinases and binding to ß-arrestin (Daigle, Kwok, & Mackie, 2008). The process of endocytosis involves the interaction of the CB1 receptor with several proteins, including the adaptor protein 3 (Rozenfeld & Devi, 2008) and G protein-associated sorting protein 1 (GASP1) (Martini et al., 2007). However, CB1 receptor interactions with another protein, SGIP1, prevent internalization and are likely responsible for stabilizing some pools of CB1 receptor at the axon terminal (Hajkova et al., 2016). Interactions between CB1 receptor and SGIP1 also reduce CB1 receptor-activation of ERK, as this signaling occurs after receptor internalization and requires ß-arrestin (Hajkova et al., 2016). Recent behavioral data using SGIP1 knock out mice indicate that this interaction contributes to the effects of CB1 receptor activation on anxiety-driven behaviors and nociception, among other behaviors (Dvorakova et al., 2021). These kinds of studies begin to elucidate the specific signaling pathways responsible for the behavioral effects of CB1 receptor activation and could contribute to the development of specific activation approaches.

A proteomic screen of mouse brain neurons identified proteins of the WAVE1/SCAR1 complex as interacting with the CB1 receptor (Njoo, Agarwal, Lutz, & Kuner, 2015). CB1 receptor activation regulates the activity of this complex, resulting in the collapse of growth cones of developing neurons and attenuated inflammation-induced remodeling of dendritic spines in spinal neurons. An important potential functional role for this interaction could be in the suppression of morphological changes that contribute to hyperalgesia (Njoo et al., 2015). Interestingly, CB1 receptor-mediated long-term depression of inhibition (iLTD) in the hippocampus is accompanied by a reduction in expression of WAVE1 (Monday, Bourdenx, Jordan, & Castillo, 2020), suggesting that there are multiple mechanisms of interaction between CB1 receptor and WAVE1 and support an important role for CB1 receptor in the regulation of the actin cytoskeleton.

Several other proteins have been identified that can interact with the CB1 receptor. One of these, cannabinoid receptor interacting protein 1a (CRIP1a) (Niehaus et al., 2007) binds to the C terminal tail of the CB1 receptor and can interfere with both ß-arrestin binding and G protein-mediated signaling (Booth, Walker, Lowther, & Howlett, 2019). The crystal structure of CRIP1a has been solved and homology studies of its structure predict that it binds lipidated proteins (Booth et al., 2021). In accord with this prediction, CRIP1a was demonstrated to bind myristolated Gαi with micromolar affinity (Booth et al., 2021), suggesting that CRIP1a might function to sequester Gαi proteins away from CB1 receptor and thereby interfere with signaling.

Guzman and colleagues have recently identified and characterized another CB1 receptor interacting protein, binding immunoglobulin protein (BiP) (Costas-Insua et al., 2021). BiP is an Hsp70 molecular chaperone that is enriched within the endoplasmic reticulum and contributes to proper folding of newly synthesized proteins (Gething, 1999). BiP, which is also present in the cytosol, binds to the CB1 receptor with micromolar affinity (Costas-Insua et al., 2021). Immunohistochemical data indicate that BiP is preferably co-expressed with CB1 receptor in GABAergic terminals and co-expression of BiP and the CB1 receptor results in a very significant suppression of CB1 receptor signaling through Gαq/11, without affecting Gαi/o signaling. Partial deletion of BiP enhanced the anxiogenic response to THC, which previous studies have shown is dependent upon CB1 receptor signaling in GABAergic neurons (Rey, Purrio, Viveros, & Lutz, 2012). Thus, these findings are consistent with BiP functioning to selectively suppress CB1 receptor signaling in GABAergic neurons.

CRIP1a, which is preferentially expressed in glutamatergic neurons (Booth et al., 2019), and BiP, which is preferentially expressed in GABAergic neurons (Costas-Insua et al., 2021), could provide important therapeutic targets that would differentiate between the simultaneous effects of CB1 receptor agonists to inhibit both excitation and inhibition (Figure 2).

Figure 2.

Pharmacological approaches to increase or decrease CB1 receptor signaling. A. Orthosteric agonist binding to the endocannabinoid (eCB) binding site of the receptor mimic the effects of eCBs and activate signaling. B. Positive allosteric modulators (PAMs) bind to allosteric sites and increase/stabilize eCB binding or potentiate signaling; while inhibitors of CRIP1A or BiP would remove inhibitory protein interactions and enhance signaling. C. Regulation of eCB concentrations available to bind to the CB1 receptor is a combination of enzymatic synthesis (top funnel) and degradation (bottom funnel). Inhibitors of the degradative enzymes fatty acid amide hydrolase (FAAH) and monoacylglycerol lipase (MAGL) elevate eCB concentrations and increase CB1 receptor signaling. D. Orthosteric antagonists occupy the eCB binding site and inhibit CB1 receptor signaling. E. Negative allosteric modulators (NAMs) bind to allosteric sites and inhibit eCB-induced molecular rearrangement of the CB1 receptor that enables G protein signaling. F. Inhibition of the synthetic enzymes diacylglycerol lipase (DAGL) and N-acyl-phosphatidylethanolamine phospholipase D (NAPE-PLD) reduce eCB available to the receptor and reduce signaling.

Allosteric ligands for the CB1 receptor.

Ligands that bind to the orthosteric binding site of the CB1 receptor engage signaling (agonists and partial agonists); occupy the site without changing it (antagonists); or stabilize the inactive conformation of the receptor and thereby reduce constitutive activity (inverse agonists). Allosteric molecules bind to regions outside of the orthosteric binding pocket and modulate orthosteric ligand binding, signaling or both. An allosteric modulator can inhibit (NAM) or enhance (positive allosteric modulator or PAM) orthosteric agonist signaling. Like orthosteric ligands, allosteric modulators can exhibit bias (Slosky, Caron, & Barak, 2021). Thus, this ligand class can fine-tune signaling through GPCRs. Not surprisingly, one of the most active areas of development of drugs that affect ECS has been the identification and optimization of CB receptor allosteric modulators (Figure 2).

ORG27569 is the first allosteric modulator identified for the CB1 receptor (Price et al., 2005). The effects of ORG27569 on CB1 receptor signaling are complex-it reduces G-protein activation and inhibits constitutive activity, which are characteristics of a NAM; on the other hand, ORG27569 slows orthosteric agonist dissociation and increases recruitment of ß-arrestin1, which are characteristics of a PAM (Ahn, Mahmoud, & Kendall, 2012; Baillie et al., 2013; Cawston et al., 2013; Price et al., 2005). Functionally, ORG27569 has the characteristics of a NAM. Some very interesting novel drugs that are structural hybrids of ORG27569 and another NAM, PSNCBAM-1 (Horswill et al., 2007) were recently designed based upon the ORG27569/CP55940/CB1 receptor crystal structure described above (T. Nguyen et al., 2021). One of these compounds, RTICBM-74, exhibits favorable pharmacokinetics and is brain penetrant (Lovelock, Nguyen, Van Voorhies, Zhang, & Besheer, 2021). Another compound developed using this approach, ABM300, docks to the same site as ORG27569 and increases orthosteric agonist binding, but completely reduces both ß-arrestin and G-protein signaling in response to CB1 receptor agonist, thus acts as a NAM (Mielnik et al., 2021). Importantly, ABM300 profoundly reduced the behavioral consequences of hyperdopaminergia in two different mouse models, without effects in the cannabinoid tetrad assay (Mielnik et al., 2021). These data suggest that this approach could be useful for treatment of schizophrenia, for example, without the significant adverse effects of direct antagonism of dopamine receptors.

Another putative CB1 receptor NAM is the neurosteroid, pregnenolone (Raux, Drutel, Revest, & Vallee, 2021; Vallee et al., 2014). Pregnenolone was originally reported to be a biased NAM, with a bias toward inhibiting THC-mediated ß-arrestin recruitment over G protein activation. In accord with this, pregnenolone has been shown to be ineffective in altering G protein activation (Gamage et al., 2017) or cannabinoid agonist effects on synaptic transmission, which are mediated by G protein activation (Krohmer, Brehm, Auwarter, & Szabo, 2017; Straiker, Mitjavila, Yin, Gibson, & Mackie, 2015). On the other hand, pregnenolone blocked the effects of CB1 receptor agonists to inhibit transmitter release in Schaffer-commissural afferents to the CA1 region of the hippocampus, an effect that suggests inhibition of G protein signaling (Wang et al., 2018). Recent in vivo studies indicate that pregnenolone can block some negative consequences of THC exposure, including reversing the hyperdopaminergic effects of prenatal THC exposure (Frau et al., 2019) and the psychotic effects of THC (Busquets-Garcia et al., 2017).

Peptides derived from hemopressin can function as CB1 receptor NAMs (Bauer et al., 2012; Straiker et al., 2015). Pepcan-12, a member of this family that is found in brain, is a potent NAM of CB1 receptor agonist mediated G protein activation in cellular assays (Bauer et al., 2012). Circular dichroism and NMR studies suggest that pepcan-12 is disordered in an aqueous environment but becomes more helical in a hydrophobic environment (Emendato et al., 2018). Docking of a rigidized pepcan-12 into previously identified allosteric sites of the CB1 receptor suggests pepcan-12 occupies a site on the extracellular side of the receptor, in a region that overlaps with orthosteric antagonist binding (Shore et al., 2014).

The cannabis constituent, cannabidiol (CBD), can reduce or inhibit several pharmacological effects of THC and other CB1 receptor agonists (Russo & Guy, 2006). However, CBD has very low affinity for the orthosteric binding site of the CB1 receptor (Devane et al., 1988), making it unlikely that it is an orthosteric antagonist at moderate doses. Studies employing cell-based signaling assays have found that CBD functions as a NAM of both CB1 receptor agonist ß-arrestin recruitment and G protein-mediated signaling (Laprairie, Bagher, Kelly, & Denovan-Wright, 2015). In agreement with these findings, low micromolar concentrations of CBD inhibit DSE in hippocampal neuronal cultures in a manner consistent with a NAM mechanism of action (Straiker, Dvorakova, Zimmowitch, & Mackie, 2018). Mutagenesis studies (Laprairie et al., 2015) and recent molecular dynamics simulations of CBD and THC docking to the CB1 receptor (Jakowiecki et al., 2021) both indicate that the allosteric binding site of CBD involves residues from the N terminal domain and the second extracellular loop. This is a different region of the receptor than occupied by ORG27569 discussed above, suggesting that there are multiple mechanisms of CB1 receptor allosterism.

Several CB1 receptor PAMs have also been identified. The first was the anti-inflammatory lipid, lipoxin A4 (Pamplona et al., 2012), although later studies did not find a PAM effect in DSE at cultured hippocampal neurons (Khajehali et al., 2015).

A second CB1 receptor PAM, ZCZ011, both increases binding of agonists to the receptor and enhances signaling in a nonbiased manner (Ignatowska-Jankowska et al., 2015). ZCZ011 has been shown to alleviate the effects of THC withdrawal (Trexler, Eckard, & Kinsey, 2019) and the effects of naloxone-precipitated withdrawal in oxycodone-dependent mice (Dodu et al., 2022).

A third PAM, GAT211, exhibits both PAM and allosteric agonist activity (i.e. activation of signaling in the absence of an orthosteric agonist) in multiple assays (Laprairie et al., 2017). GAT211 is a racemic mixture of a CB1 receptor allosteric agonist, GAT228, and a CB1 receptor PAM, GAT229, suggesting that the dual modulatory effects of the racemic GAT211 result from two separate binding interactions. GAT211 has been shown to mimic the effects of CB1 receptor agonists to reduce pain, symptoms of psychosis, and symptoms of Huntington’s Disease (Garai, Leo, et al., 2021; Laprairie et al., 2019; McElroy et al., 2021; Slivicki et al., 2018). Analogs of GAT211 have recently been developed that have improved bioavailability and potential usefulness as therapeutic agents (Garai, Schaffer, et al., 2021).

In summary, allosteric modulators have been identified that can alter CB1 receptor signaling in complex, but potentially very important ways (Figure 2). While CB1 receptor allosteric modulators have promise as potential therapeutic modifiers of ECS, there are important outstanding questions. For example, it is unknown whether the GAT analogs or ZCZ011 produce the “high”, dependence and other effects of THC that would complicate their usefulness as therapeutic agents. Additionally, it is unknown whether CB1 receptor NAMs can produce beneficial effects without also inducing dysphoria, anxiety and depression as was seen with the orthosteric antagonist, rimonabant (T. Nguyen, Thomas, & Zhang, 2019).

CB2 cannabinoid receptor

CB2 receptor structure.

The crystal structure for the CB2 receptor bound to an antagonist (AM10257) has been solved (X. Li et al., 2019). There are several similarities with the CB1 receptor structure, which is not surprising since there are several agonists that bind both receptors with similar affinities. In particular, the residues with which AM10257 interacts in the CB2 receptor structure overlap considerably with the residues interacting with agonists in the CB1 receptor structure (Hua et al., 2017). On the other hand, there are distinct differences between antagonist interactions with binding site residues for CB1 and CB2 receptors which is in accord with the high selectivity of antagonists for their respective receptors. The crystal structure of the CB2 receptor bound to an agonist has been recently solved (Hua et al., 2020). As expected, the CB2 receptor crystal structure shows that agonist binding is associated with the canonical movements of the transmembrane domains and toggle switch that allows for the G protein binding site to open. Like the CB1 receptor, the agonist binding site of the CB2 receptor is capped by the N terminal domain, suggesting that ligands enter the binding pocket from the phospholipid bilayer. Indeed, molecular dynamics simulations demonstrate that 2-AG enters the CB2 receptor binding pocket from the bulk lipid of the membrane by passing through the interface between transmembrane helices 6 and 7 (Hurst et al., 2010).

Cryo-EM structures of CB2 receptor/Gαi complexes bound to WIN 55,212–2 (Xing et al., 2020) and AM12033 (Hua et al., 2020) have been determined. These data also reveal several similarities between agonist binding to CB1 and CB2 receptor, including ligand-receptor binding interactions that are primarily hydrophobic in nature and a similar a toggle switch mechanism of receptor activation. The structural basis for selectivity of CB2 receptor for Gαi over Gαs was also hypothesized from these solutions.

Allosteric modulators of the CB2 receptor.

In vitro and receptor modeling studies support the hypothesis that CBD is NAM of the CB2 receptor (Martinez-Pinilla et al., 2017; Navarro et al., 2021). Modeling and mutagenesis studies find that CBD binds near the interface between TM1 and TM7; and suggest two potential mechanisms for its NAM activity (Navarro et al., 2021). CBD occupancy of the site could, first, prevent agonist entry and, second, interfere with agonist-mediated activation of the toggle switch. Interestingly, when the pentyl chain of CBD was shortened to methyl, ethyl or propyl groups, the resulting compounds were PAMs of agonist activation of G protein signaling; lengthening the chain resulted in increasing efficacy as NAMs. Modeling studies suggest that all the CBD derivatives bind in the same pocket and can contribute to conformational changes that synergize with those of the orthosteric agonist to induce activation. However, the longer chain analogs also interfere with the toggle switch and thereby interfere with movement into the active state.

In addition to its ability to act as a NAM at the CB1 receptor, pepcan-12 is a very effective PAM at CB2 receptor, potentiating 2-AG/CB2 receptor-mediated inhibition of adenylyl cyclase activity 5–10 fold (Petrucci et al., 2017). Given that pepcan-12 is present in brain (Bauer et al., 2012) and is increased in response to injury and ischemia (Petrucci et al., 2017), perhaps pepcan-12 amplifies brain CB2 receptor signaling under pathological conditions and contributes to the seemingly outsized functional influence of CB2 receptor in the CNS compared to its very low expression.

A series of rationally designed, synthetic compounds has been found to function as CB2 receptor PAMs (Gado et al., 2019). One of these, designated C2 in (Gado et al., 2019), significantly enhanced CB2 receptor agonist-induced G protein signaling at concentrations that had little effect on agonist activation of the CB1 receptor. This effect was due to a significant slowing of the dissociation of orthosteric ligand from the receptor.

Thus, allosteric modulators of the CB2 receptor are available and show interesting pharmacological properties. These allosteric modulators provide tools for the study of CB2 receptor mechanisms and could eventually also be a source of novel therapeutics.

New tools for the study of CB2 receptor.

As is the case for many GPCRs, antibodies against the CB2 receptor display a large degree of non-selectivity (Atwood & Mackie, 2010; Baek, Darlington, Smith, & Ashton, 2013; H. Y. Zhang et al., 2019), making definitive identification of CB2 receptor protein difficult, particularly in the brain where expression of the receptor is very low. A group, led by chemists and pharmacologists from Roche, reported a series of fluorescently tagged CB2 receptor ligands that will be useful probes for the CB2 receptor protein in binding studies and flow cytometry (Sarott et al., 2020). These probes demonstrate nM K_D values for the CB2 receptor and provide an important advance towards unambiguous detection of the CB2 receptor.

In addition, two transgenic mouse lines have been developed that function as cellular reporters of CB2 receptor expression. The first mouse line was generated by insertion of a BAC transgene expressing green fluorescent protein (GFP) in place of the open reading frame of the CB2 receptor (Schmole et al., 2015). The second mouse line was generated by recombination; in this line, a transgene encoding an IRES and eGFP was added to the 3’ end of the CB2 receptor gene (Lopez et al., 2018). CNS GFP expression has been examined carefully in both of these models and only is detected in microglia.

Cellular distribution of CB2 receptor in the CNS.

CB2 receptor was originally characterized as the “peripheral cannabinoid receptor” based upon failure to detect CB2 receptor mRNA or protein in brain tissue (Galiegue et al., 1995; Schatz, Lee, Condie, Pulaski, & Kaminski, 1997). However, the application of more sensitive techniques, including quantitative PCR and the reporter mice described above, have detected low CB2 receptor expression in healthy brain tissue (Figure 1) (Y. Li & Kim, 2015; Van Sickle et al., 2005; H. Y. Zhang et al., 2014). CB2 receptor expression in brain is 1/100^th of the expression level in spleen (Y. Li & Kim, 2015; Q. R. Liu et al., 2009; H. Y. Zhang et al., 2014), and this very low level of expression, together with a lack of highly specific antibodies (Atwood & Mackie, 2010; Baek et al., 2013) have made unequivocal evidence for brain CB2 receptor expression at the cellular level difficult to obtain.

The first evidence that CB2 receptor is expressed in brain-resident cells was by Cabral and colleagues who found that microglia from neonatal rat cerebral cortex expressed CB2 receptor mRNA (Carlisle, Marciano-Cabral, Staab, Ludwick, & Cabral, 2002). They found that CB2 receptor expression was very low in resting microglia, increased significantly in proliferating microglia, but was significantly reduced in LPS-activated microglia. Similar regulation patterns were seen in mouse primary microglia, in which GM-CSF and interferon gamma increased CB2 receptor mRNA, while LPS reduced expression (Maresz, Carrier, Ponomarev, Hillard, & Dittel, 2005). These and other studies suggest that expression of CB2 receptor in microglia is variable, and dependent upon the degree of microglial activation (Duffy, Hayes, Fiore, & Moalem-Taylor, 2021). Microglia can assume a continuum of activation and morphological states; and single cell RNA sequencing studies will be very helpful in identifying which microglial populations express the CB2 receptor.

Recent studies have found that the antioxidant response element (ARE) is present in the promoter region of the CB2 receptor (Galan-Ganga, Del Rio, Jimenez-Moreno, Diaz-Guerra, & Lastres-Becker, 2020). An important activator of the ARE is nuclear factor erythroid 2-related factor 2 (Nrf2) (Itoh et al., 1997). Dimethyl fumarate increases both Nrf2 and CB2 receptor mRNA expression, the latter shows a time delay of several hours, in mouse microglia (Galan-Ganga et al., 2020). Furthermore, primary microglial cells prepared from neonatal Nrf2 deficient mice had a 90% reduction in CB2 receptor mRNA expression compared to wild-type mice, consistent with CB2R expression being Nrf2-depedent (Galan-Ganga et al., 2020) (Figure 3). Nrf-2 is redox sensitive and upregulates the expression of hundreds of genes that are associated with reduced inflammation and oxidative stress (Jayaram & Krishnamurthy, 2021). Although there are no studies to date, an open and interesting question is whether Nrf-2 similarly regulates CB2 receptor expression in neurons and other brain cell types.

Figure 3.

Mechanism of CB2R upregulation and effects of CB2R activation in microglia. Upregulation of CB2R in microglia is dependent on nuclear factor erythroid 2-related factor 2 (NRF2), a pleiotropic transcription factor. CB2R activation induces a microglial phenotype that produces high levels of anti-inflammatory markers (IL-10, BDNF), low levels of pro-inflammatory markers (TNFα, IL-1β, IL-6), increases cell proliferation, and decreases phagocytic activity via an IL1RL1-dependent process.

CB2 receptor activation in microglial cells is associated with increased expression of IL-10 (anti-inflammatory factor) and BDNF (neurotrophic factor) and reduced expression of TNF-α, IL-1β, and IL-6 (pro-inflammatory factors) (L. Ma et al., 2015), evidence that it contributes to an anti-inflammatory milieu in microglia. There is evidence that microglial CB2 receptor expression also affects the activation state of microglia; for example, loss of CB2 receptor expression is associated with reduced amyloid plaque burden (Lopez et al., 2018) and improved cognition (Schmole et al., 2018) in mouse models of Alzheimer’s Disease. These findings suggest the hypothesis that CB2 receptor “restrains” microglial phagocytic activity, which results in reduced clearance of amyloid but in other situations could have negative consequences (Figure 3).

Although the reporter mice described above do not exhibit CB2 receptor expression in astrocytes (Lopez et al., 2018), a splice variant of the human CB2 receptor (CB2RA) has been detected in substantia nigral astrocytes in patients who died with Parkinson’s disease (F. Navarrete, Garcia-Gutierrez, Aracil-Fernandez, Lanciego, & Manzanares, 2018). Immunohistochemical studies using in situ proximity ligation approaches identified CB2 receptor in astrocytes, and not microglia or neurons in the substantia nigra (F. Navarrete et al., 2018).

CB2 receptor expression in neurons remains a controversial topic. Early studies failed to show conclusive evidence of neuronal expression, which requires a combination of functional (pharmacological approach with agonists and antagonists), biochemical (western blotting/PCR-based approaches), and anatomical (ISH/IHC) studies with appropriate positive (spleen tissue) and negative (knockout tissue) controls (Atwood & Mackie, 2010). However, recent studies are more convincing. For example, Xi and colleagues demonstrated that 1) CB2 receptor agonism inhibited ventral tegmental area (VTA) dopaminergic neuronal firing; an effect blocked by CB2 receptor antagonists and absent in CB2 receptor knockout mice; 2) CB2 mRNA expression in the midbrain by qPCR; and 3) used in situ hybridization to confirm CB2 receptor mRNA in VTA dopaminergic neurons (H. Y. Zhang et al., 2014). Although this group also showed CB2 receptor protein using immunohistochemistry, the poor selectivity of antibodies makes those data difficult to rely on. Recent studies from this group have used multiple approaches to study CB2 receptor mRNA expression at the cellular level and have found that repeated cocaine treatment upregulates CB2 receptor expression in D1 receptor-expressing medium spiny neurons by RNA scope and the use of RiboTag-expressing mice (H. Y. Zhang et al., 2021). Together, these data provide high quality evidence of CB2 receptor mRNA expression in this neuronal population and provide a roadmap for future studies. The problem of unambiguous identification of CB2 receptor protein in neurons is still unsolved, however, as is a clear understanding of the mechanisms of regulation of CB2 receptor expression in neurons.

Several studies have demonstrated that CB2 receptor activation can affect neuronal activity. The CB2 receptor agonist, JWH133, decreased VTA dopaminergic neuron action potential firing through a signaling cascade that involved Gαi, inhibition of PKA, enhancement of M-type K channel current, resulting in hyperpolarization (Z. Ma et al., 2019). Both elevation of 2-AG concentrations and exogenous CB2 receptor agonists generated a long-lasting membrane hyperpolarization in CA3 pyramidal neurons that required activation of the Na/bicarbonate co-transporter (NBC) (Stempel et al., 2016). NBCs are expressed in hippocampal pyramidal neurons (Majumdar et al., 2008) and tend to increase intracellular pH that consequently reduces neuronal excitability through effects on channels and other membrane proteins (Ruffin, Salameh, Boron, & Parker, 2014). Stempel and colleagues showed G protein-coupled inwardly rectifying potassium (GIRK) antagonism failed to repolarize CA3 pyramidal neurons post hyperpolarization induction (Stempel et al., 2016), data that argue against a role for CB1 receptor in the effect. In addition, CA3 pyramidal neuron hyperpolarization did not occur in experiments blocking NBC and replacing sodium in the media with N-methyl-D-glucamine, suggesting for the first time that CB2 receptor agonism can induce changes in long term plasticity through interaction with NBC in hippocampal neurons (Stempel et al., 2016).

Mechanisms of Endocannabinoid Regulation: Therapeutic Potential

The two main eCB, 2-AG, and AEA, are both synthesized enzymatically from membrane phospholipids and their metabolic pathways of synthesis and degradation are quite distinct. Since eCB signaling concentrations are regulated enzymatically, this provides an opportunity to manipulate ECS through inhibition or activation of these enzymatic processes (Figure 2). In this section, we discuss the current understanding of the major mechanisms by which receptor-available concentrations of the eCBs are regulated and the status for therapeutic development of drugs targeting these proteins. Importantly, the recent development of the “eCB-GRAB” constructs, which function as in vivo biosensors for the eCBs, will greatly enhance our ability to determine the pathways and processes that regulate changes in eCB concentrations (Dong et al., 2022).

Regulation of 2-AG Synthesis.

The formation of the 2-AG can occur via multiple enzymatic pathways from distinct membrane lipid precursors (reviewed in (Baggelaar, Maccarrone, & van der Stelt, 2018)). The most common synthetic route involves phospholipase C (PLC)-mediated release of diacylglycerol (DAG) from the phosphoinositide, phosphoinositide 4,5, bisphosphate (PIP₂). DAG, in which the sn-2 lipid is usually an unsaturated, long chain fatty acid such as arachidonic acid, is hydrolyzed by diacylglycerol lipase (DAGL), releasing 2-acylglycerols, including 2-AG. Less common, alternative biosynthetic routes have also been described that may make a minor contribution to the formation of 2-AG (Hillard, 2015).

Two isoforms of DAGL have been characterized: DAGLα and DAGLβ. Although both serine hydrolases are involved in the synthesis of 2-AG, DAGLα is the predominant isoform in the CNS and is highly expressed in neurons (Bisogno et al., 2003). DAGLα activity is required for most forms of eCB/CB1 receptor-mediated retrograde regulation of synaptic transmission (Tanimura et al., 2010; Yoshino et al., 2011). DAGLβ, primarily expressed in the periphery in adult animals, is also expressed in the developing CNS (Bisogno et al., 2003; C. S. Wu et al., 2010) and is involved in neurite outgrowth (Jung, Astarita, Thongkham, & Piomelli, 2011). DAGLß is expressed in microglia (Viader et al., 2016), which is consistent with data that microglia can synthesize 2-AG (Carrier et al., 2004).

Given its position in the synthetic pathway of 2-AG, modulation of DAGLα activity provides a viable mechanism to affect ECS. In support of this notion, studies have demonstrated changes in DAGLα expression in response to environmental conditions. For example, socially dominant zebrafish exhibit increased expression compared to subordinates and controls (Orr et al., 2021) while male and female rats exposed to early life stress exhibit reduced DAGLα expression (Marco et al., 2014). DAGLα mRNA expression is regulated by the inflammation-responsive transcription factor Sp1 in neural stem cells (Walker, Suetterlin, Reisenberg, Williams, & Doherty, 2010) and by miR-223 in hepatocytes (D. P. Wang et al., 2021). However, whether these or other regulatory processes contribute to DAGLα expression in the brain is not known.

DAGLα enzymatic activity can be regulated by post-translational modification. Phosphorylation of DAGLα by calcium/calmodulin-dependent protein kinase II (CaMKII) can reduce enzymatic activity and suppress the synthesis of 2-AG (Murray et al., 2021; Shonesy et al., 2013). On the other hand, cAMP-dependent protein kinase (PKA), phosphorylates and enhances DAGLα activity and increases 2-AG/CB1 receptor signaling (Shonesy, Stephenson, Marks, & Colbran, 2020). The latter mechanism has been shown link D1 dopamine receptor activation to alteration ECS-mediated regulation of the direct output nigrostriatal pathway (Shonesy et al., 2020). Interestingly, PKA-mediated phosphorylation of DAGLα has also been shown to alter dendritic spine morphology; however, the mechanism did not involve 2-AG synthesis changes but altered protein-protein interactions between DAGLα and ankyrin-G (Yoon, Myczek, & Penzes, 2021).

DAGLα-mediated 2-AG synthesis is essential for regulation of synaptic activity by ECS (Tanimura et al., 2010; Yoshino et al., 2011). However, the effects of DAGLα inhibition and genetic deletion also provide evidence of a broader role for 2-AG/CB1 receptor signaling in brain. For example, postsynaptic DAGLα activity is required for the growth of inhibitory boutons at dendrites receiving strong excitation in the CA1 region (Hu et al., 2019). Astrocytic DAGLα expression is involved in the regulation of adult hippocampal neurogenesis, through the synthesis of astroglial 2-AG and activation of neuronal CB1 receptor (Schuele et al., 2022).

Regulation of AEA synthesis.

Several pathways have been described to contribute to the synthesis of AEA depending on the physiologic conditions and tissue specificity (reviewed in (Maccarrone, 2017)). The most well-studied pathway involves formation of N-acyl phosphatidylethanolamide (NAPE) by N-acetyltransferase (NAT), followed by hydrolysis of the NAPE by N-acyl phosphatidylethanolamide-specific phospholipase D (NAPE-PLD). This results in the generation of a family of N-acylethanolamines, including N-palmitoylethanolamine (PEA) and N-oleoylethanolamine (OEA) (Rahman, Tsuboi, Uyama, & Ueda, 2014). Because arachidonic acid is not found at high amounts at the N-acyl position in NAPE, AEA is a relatively minor product of this pathway (Schmid, Reddy, Natarajan, & Schmid, 1983). While this has led to questions regarding the importance of this pathway in regulating AEA concentrations in brain, the studies using NAPE-PLD inhibitors described below suggest it is an important pathway for the synthesis of signaling AEA.

Mice with genetic ablation of NAPE-PLD have been studied to explore the role of this pathway of AEA synthesis in the CNS. However, although loss of this enzyme reduces brain concentrations of AEA and other NAEs, the concentrations of many other lipids are altered as well (Leishman, Mackie, Luquet, & Bradshaw, 2016). Indeed, results of other studies suggest that elevated NAPE rather than reduced NAE concentrations may have greater implications for cell function (Palese, Pontis, Realini, & Piomelli, 2019, 2021). Thus, these mice may not be useful for the study of AEA.

On the other hand, a brain-penetrant NAPE-PLD inhibitor, LEI-401, has been developed that has the potential to greatly enhance understanding of the specific roles of AEA in ECS (Mock et al., 2021; Mock et al., 2020). Indeed, treatment of rodents with LEI-401, which significantly reduced brain N-acylethanolamine contents, including AEA, and was found to have effects mimicking CB1 receptor antagonism, including impaired extinction of aversive memories and excessive activation of the hypothalamus-pituitary-adrenal (HPA) axis (Mock et al., 2020).

2-AG Catabolism.

The hydrolysis of 2-AG into its component parts (arachidonic acid and glycerol) is primarily mediated in the CNS by monoacylglycerol lipase (MAGL). MAGL is found in presynaptic terminals of neurons (Gulyas et al., 2004), microglia (Kouchi, 2015), astrocytes (Viader et al., 2015) and oligodendrocytes (Moreno-Luna et al., 2021). Other serine hydrolases that also contribute to 2-AG hydrolysis include α,β hydrolase-domain containing (ABHD) 6 and ABHD 12 (Blankman, Simon, & Cravatt, 2007; Marrs et al., 2010; Marrs et al., 2011). Pharmacological inhibition of ABHD6 has anti-epileptic effects in a spontaneous seizure mouse model (Naydenov et al., 2014), and improves coordination and memory performance in a traumatic brain injury model (Tchantchou & Zhang, 2013). Additionally, anti-inflammatory effects have been documented after treatments with the ABHD6 inhibitor WWL70 in autoimmune pathologies (Alhouayek, Masquelier, Cani, Lambert, & Muccioli, 2013; Bottemanne, Paquot, Ameraoui, Alhouayek, & Muccioli, 2019; Wen, Ribeiro, Tanaka, & Zhang, 2015).

Irreversible MAGL inhibitors including JZL184 (Long et al., 2009) and MJN110 (Feja et al., 2020), and reversible inhibitors based upon a diphenylsulfide-benzoylpiperidine scaffold (Bononi et al., 2021; Granchi et al., 2021; Ikeda et al., 2021) have been developed. Considerable evidence suggests that MAGL inhibition can have significant therapeutic benefit for the treatment of pain, inflammation, and neurodegenerative diseases (reviewed in (Deng & Li, 2020)). Recent studies have demonstrated beneficial effects of MAGL inhibitors in alleviating cancer-induced bone pain (Thompson et al., 2020), HIV-associated neuropathic pain (Aly & Masocha, 2021) and migraine-like pain (Greco, Demartini, Francavilla, Zanaboni, & Tassorelli, 2021). MAGL inhibition shows promise for the treatment of neurodegenerative diseases such as Alzheimer’s disease (Hashem, Hu, Zhang, Gao, & Chen, 2021), Huntington’s disease (Covey et al., 2018), and amyotrophic lateral sclerosis (Pasquarelli et al., 2017). Inhibition of MAGL could present challenges, though, since prolonged, complete inhibition of MAGL activity is associated with down-regulation of the CB1 receptor (Schlosburg et al., 2010). However, use of recently developed reversible inhibitors could mitigate this problem.

Degradation of AEA.

The primary mechanism for catabolism of AEA and other NAEs in the CNS is hydrolysis into arachidonic acid and ethanolamine by fatty acid amidohydrolase (FAAH) (Cravatt et al., 1996). FAAH is an intrinsic membrane protein present primarily on intracellular membranes, and not on the plasma membrane (Hillard, Wilkison, Edgemond, & Campbell, 1995). Although several studies suggest that FAAH activity can be enhanced by PKA-mediated phosphorylation (Gray et al., 2015; Rossi et al., 2007), a precise mechanism for this has not been uncovered.

Interestingly, there is a relatively prevalent single nucleotide polymorphism in the coding regions for the FAAH gene in humans that results in instability of the FAAH protein (Chiang, Gerber, Sipe, & Cravatt, 2004) and increased circulating AEA concentrations (Spagnolo et al., 2016). A transgenic mouse model of this polymorphism exhibits elevated brain AEA concentrations and reduced anxiety (Dincheva et al., 2015).

Although FAAH acts as a homodimer, recent studies demonstrate that inhibition of one homomer is sufficient to completely inhibit enzymatic activity, consistent with allosteric interactions between the dimers (Dainese et al., 2020).

FAAH inhibitors are the most mature of the therapeutics that act through changes in eCB synthesis and degradation and have been studied in several human clinical trials. PF-04457845 was found to reduce cannabis withdrawal symptoms and cannabis use at 4 weeks post treatment in cannabis dependent men (D’Souza et al., 2019). Another inhibitor, JNJ-42165279, was studied in individuals with social anxiety and found to be well-tolerated and to produce moderately positive results (M. E. Schmidt et al., 2021). PF-04457845 administration to healthy humans attenuated stress-induced sympathetic activation and negative affect (Mayo et al., 2020). These studies suggest that FAAH inhibition may have usefulness in some psychiatric conditions, particularly when anxiety is a core symptom.

On the other hand, although inhibition of FAAH has demonstrated beneficial effects in pain in preclinical models (Woodhams, Chapman, Finn, Hohmann, & Neugebauer, 2017), clinical trials have not found efficacy for FAAH inhibitors in neuropathic pain (Bradford et al., 2017), pain associated with osteoarthritis (Huggins, Smart, Langman, Taylor, & Young, 2012) or bladder pain in cystitis (Houbiers et al., 2021).

Preclinical studies with FAAH inhibitors and transgenic models with genetic inactivation of the enzyme have highlighted a potential therapeutic role for FAAH inhibition in the treatment of brain injury and neurodegeneration. Recent studies have shown that inhibition of FAAH reduces hippocampal ischemia-reperfusion injury (Abdel Mageed, Ammar, Nassar, Moawad, & Kamel, 2021) and prevents quinolinic acid-induced excitotoxic damage (Chavira-Ramos et al., 2021). Elevation of AEA has demonstrated neuroprotective properties in epilepsy, where synaptic function and behavioral alterations are rescued by FAAH inhibition (Colangeli, Morena, Pittman, Hill, & Teskey, 2020). Additionally, FAAH inhibition reduces motor impairments (Celorrio et al., 2016) and behavioral and biochemical deficits (Escamilla-Ramirez et al., 2017) in Parkinson’s Disease models. Similarly, FAAH inhibition reduces neuroinflammation (Chiurchiu et al., 2021) and genetic deletion of FAAH ameliorates β-amyloidopathy and rescues synaptic function (Ruiz-Perez et al., 2021) in Alzheimer’s Disease models.

Given that FAAH inhibitors used at reasonable doses in humans are well tolerated and without serious adverse effects, it is likely that positive results from clinical trials, particularly focusing on neuroprotection and reducing anxiety, will be forthcoming.

AEA versus 2-AG signaling.

The CNS CB1 receptor is activated by two endogenous ligands, an unusual feature of this signaling system. While both AEA and 2-AG function as agonists of the receptor, differences in the cellular distribution of their primary synthetic and catabolic enzymes support the hypothesis that they are not redundant activators of the receptor. As discussed in the preceding sections, DAGLα is enriched in dendrites that are in proximity to CB1 receptor-expressing axons, while MAGL is enriched in axon terminals that co-express the CB1 receptor. This molecular arrangement, demonstrated clearly in the hippocampus (Nyilas et al., 2008), is consistent with 2-AG functioning as the primary mediator of retrograde signaling by the ECS. On the other hand, the synthetic enzyme for AEA (NAPE-PLD), is concentrated in presynaptic terminals, in close association with axonal calcium stores (Nyilas et al., 2008) while FAAH is primarily expressed postsynaptically (Egertova, Cravatt, & Elphick, 2003). This arrangement of synthetic and catabolic enzymes suggests that AEA is an anterograde signaling molecule, and could have as it primary target dendritic or somatic pools of CB1 receptor (Ong & Mackie, 1999).

Endocannabinoid binding and transport proteins.

Given that both 2-AG and AEA are both highly hydrophobic molecules, it has been hypothesized that they require transporters to diffuse in the cytosol and across the synapse (Kaczocha & Haj-Dahmane, 2021). In support of this hypothesis, several binding proteins for the eCBs have been identified. As their name implies, fatty acid binding proteins (FABPs) are a family of proteins that primarily bind fatty acids (H. Xu, Diolintzi, & Storch, 2019). FABPs are found in tissues and the circulation and play important roles in metabolism and metabolic diseases. AEA and 2-AG have been found to bind to several of the FABPs, including FABP5 which is present in the adult brain (Kaczocha, Vivieca, Sun, Glaser, & Deutsch, 2012). Subsequent studies using the FABP5 inhibitor SBFI-26 (Hsu et al., 2017) have provided compelling evidence that FABP5 is involved in ECS in some brain regions. For example, inhibition of FABP5 reduces both tonic and phasic CB1 receptor regulation of glutamate release in the dorsal raphe nucleus (Haj-Dahmane et al., 2018). On the other hand, inhibition of FABP5 in a murine model of high intraocular pressure reduced pressure in a CB1-receptor dependent manner (Miller et al., 2020). Thus, it is possible that FABP5 could play roles in both the release and inactivation of the eCBs.

In vitro data suggest that sterol carrier protein 2 (SCP-2) is also an intracellular transport protein for the eCBs. SCP-2 binds to both AEA and 2-AG with K_i values in the nanomolar range (Hillard et al., 2017) and overexpression of SCP-2 enhances the cellular accumulation of AEA (Liedhegner, Vogt, Sem, Cunningham, & Hillard, 2014). SCP-2 is present in brain and is enriched in synaptosomal preparations (Myers-Payne et al., 1996), observations that support its potential for regulating the ECS system at the synapse. Studies using SCP-2 knock out mice fed a high fat diet suggest that SCP-2 plays a role in regulating eCB concentrations in the brain; however, the effect may be secondary to changes in arachidonic acid concentrations in the circulation (Martin et al., 2019).

Early studies of the characteristics of cellular accumulation of labeled AEA hypothesized the existence of a plasma membrane protein transporter and a facilitated diffusion process (Hillard & Jarrahian, 2003). The opposing hypothesis posits that the eCBs easily diffuse through membranes, and small molecule inhibitors of cell accumulation likely inhibit intracellular processes and perturb the eCB concentration gradient, thereby affecting accumulation (Glaser, Kaczocha, & Deutsch, 2005). However, the concept of a protein that contributes to AEA cellular accumulation is strongly supported by recent data using a novel compound WOBE437 (Chicca et al., 2017). The compound mimics many effects of cannabinoid agonists, including analgesic, anxiolytic and anti-inflammatory effects. Recent reports also demonstrate that WOBE437 is effective in a mouse model of multiple sclerosis (Reynoso-Moreno et al., 2021) and reduces intraocular pressure (Miller et al., 2020). Importantly, a photoaffinity probe based upon the WOBE437 structure was synthesized that competes with WOBE437, providing a potential avenue for identification of the elusive eCB membrane transport protein.

Another potential player in the cellular homeostasis of AEA is the plasma membrane channel pannexin 1 (Bialecki et al., 2020). Pannexin 1 is present on post-synaptic terminals of CA1 hippocampal neurons and data from this study indicate that AEA can pass through the channel. Furthermore, Src-mediated phosphorylation of pannexin 1 reduces AEA conductance, resulting in an increase in AEA-mediated activation of TRPV1 receptors.

Integration of ECS Within the CNS

The data we discussed in the section above demonstrates that the ECS system does far more than subserve retrograde synaptic plasticity. In this section, we will expand on this theme and discuss in greater detail how ECS contributes to communication between astrocytes and neurons and emerging findings that suggest similar contribution to microglial-neuronal communication.

ECS in Astrocyte/Neuronal Communication.

Morphological evidence that astroglial processes are in close proximity to the presynaptic axon and postsynaptic dendrite led to the concept of the tripartite synapse (Serrat et al., 2021). Subsequent data strongly support a functional role for astrocytes in synaptic activity and demonstrate bidirectional communication between neurons and astrocytes at many synapses (Serrat et al., 2021). In fact, more than 50% of all synapses in the rat hippocampus are tripartite (Ventura & Harris, 1999) and a single human astrocyte can contact up to two million neuronal synapses (Oberheim et al., 2009).

Astrocytes express the CB1 receptor, which responds to neuronal-derived eCBS and participates in the regulation of synaptic activity by astrocytes. In particular, eCBs released by neurons activate astrocytic CB1 receptor to increase intracellular calcium, triggering glial glutamate release and activation of NMDA receptors in hippocampal pyramidal neurons (M. Navarrete & Araque, 2008, 2010). Like retrograde neuronal-neuronal eCB signaling, astrocytic regulation of synaptic activity via ECS is localized and synapse-specific (Lines, Covelo, Gomez, Liu, & Araque, 2017). Recent data suggest that astrocyte CB1 receptor activation can also trigger ATP (Covelo & Araque, 2018) and adenosine (Hablitz, Gunesch, Cravetchi, Moldavan, & Allen, 2020) release, also as a result of calcium mobilization. Interestingly, the temporal relationship between glutamate and ATP release from the same astrocyte is dependent upon neuronal firing patterns, which allows for the astrocytic contribution to synaptic activity to be well-integrated with neuronal activity; this relationship is mediated by ECS (Covelo & Araque, 2018).

Since a single astrocyte can contact many synapses, they can transmit information from one synapse to many others in a process known as lateral synaptic regulation. Recent data indicate that activation of astrocyte CB1 receptor exerts effects on neurotransmission far from the site of eCB release through this process (Covelo, Eraso-Pichot, Fernandez-Moncada, Serrat, & Marsicano, 2021). ECS-activated lateral synaptic regulation has been demonstrated in the amygdala (Martin-Fernandez et al., 2017). The outflow nucleus of the amygdala circuit is the central amygdala (CeA); activity in the CeA is regulated by excitatory input from the basolateral amygdala (BLA) and inhibitory input from the lateral subdivision of the central amygdala (CeL) (Ehrlich et al., 2009; Gilpin, Herman, & Roberto, 2015). Astroglial CB1 receptor activation in the CeM reciprocally depressed excitatory input from the BLA and enhanced inhibitory input from the CeL (Martin-Fernandez et al., 2017). Consistent with these findings, CB1 receptor activation in CeM astrocytes in vivo reduced CeM neuronal firing rate and reduced fear expression (Martin-Fernandez et al., 2017).

Recent studies have focused on the role of mitochondrial CB1 receptors in astrocytes on the regulation of astrocytic calcium concentrations. As in other cell types, mitochondria have a role in the regulation of calcium in astrocytes, particularly regulation of microdomain calcium concentrations in astroglial process that are in contact with synapses (Agarwal et al., 2017). CB1 receptor expressing mitochondria in astrocytes are localized close to synapses, which indicates that mitochondrial CB1 receptors play a role in the regulation of calcium in astrocytic processes and thereby contribute to synaptic regulation (Gutierrez-Rodriguez et al., 2018). In support of this notion, mitochondrial CB1 receptor activation was recently shown to be required for lateral synaptic potentiation between two hippocampal CA1 pyramidal neurons that were more than 60 μm apart (Serrat et al., 2021). Further studies suggest that the mitochondrial CB1 receptor mediates mitochondrial uptake of calcium from the endoplasmic reticulum through the mitochondrial calcium uniporter (Serrat et al., 2021).

Several recent studies have expanded the list of roles that the astrocytic CB1 receptor plays in the effects of the cannabinoids on the CNS. Studies using astrocyte-selective rescue of CB1 receptor expression in otherwise CB1 receptor null mice suggest that perceived aversive effects of CB1 receptor agonists are mediated by astrocytic CB1 receptors (Cong et al., 2021). Astrocyte CB1 receptor in the ventral horn of the spinal cord is the site of action for intrathecal WIN 55,212–2 to reduce tremors in a mouse model of essential tremor (Carlsen et al., 2021). Electroacupuncture-induced neuroprotection involves mobilization of eCBs and activation of astrocyte CB1 receptor in the vicinity; the subsequent increase in glutamate was found to be neuroprotective (T. T. Yang et al., 2021). Activation of CB1 receptor-induced release of astrocytic glutamate in the nucleus accumbens has been shown to normalize glutamate homeostasis after chronic cocaine administration and reduce relapse to cocaine seeking (L. Y. Zhang et al., 2021). The astrocytic CB1 receptor is required for spike-timing dependent long-term depression in the rodent somatosensory cortex (Manninen, Saudargiene, & Linne, 2020) and regulates neuronal activity and clock timing in the suprachiasmatic nucleus (Hablitz et al., 2020). These data suggest that the astrocytic CB1 receptor, primarily through an ability to regulate gliotransmitter release, plays an important role in the regulation of brain and spinal cord function under physiological and pathophysiological conditions.

Astroglia also express several of the enzymes involved in the synthesis and degradation of the eCBs. Astrocyte-specific MAGL knock out mice have demonstrated several functions for MAGL in astrocytes that differ from the role of neuronal MAGL. Since MAGL hydrolyzes 2-AG to arachidonic acid, its activity regulates both ECS and concentrations of arachidonate metabolites, several of which are neurotoxic (Yagami, Koma, & Yamamoto, 2016). Pharmacological inhibition and genetic deletion of MAGL both result in a significant reduction of the proinflammatory prostaglandin, PGE2, in brain and impair inflammatory responses to LPS and MPTP treatment in mice (Nomura et al., 2011). Recent studies indicate that astrocyte MAGL is mainly responsible for generating the arachidonic acid that supports prostaglandin synthesis (Viader et al., 2015) particularly in response to inflammatory challenge (Grabner et al., 2016). Inhibition of MAGL protects against huntingtin-induced neuronal damage of striatal neurons in vivo; this requires astrocyte MAGL, as astrocyte-specific deletion of MAGL abrogated the beneficial effects of MAGL inhibition (Ruiz-Calvo et al., 2019).

Together with neuronal MAGL, astrocyte MAGL also plays a role in regulating the spread of synaptically evoked suppression of excitation (Chen et al., 2016) and terminating DSE at parallel fiber to Purkinje cell (PC) synapses and DSI at stellate cell to PC synapses in the mouse cerebellum (X. Liu et al., 2016).

There is also evidence that astrocytes express DAGL and can synthesize 2-AG (Hegyi et al., 2018; Schuele et al., 2021). Transient heterosynaptic depression is a fast form of hippocampal synaptic plasticity that requires astrocytes for its elaboration; recent studies suggest that this form of plasticity requires astrocytic synthesis and release of 2-AG, which targets neuronal CB1 receptor (Smith, Bekar, & Nedergaard, 2020). Thus, similar to glutamate which is both a neuro- and gliotransmitter, these data suggest that 2-AG should also be considered both. Interestingly, CB1 receptor activation of astrocytes isolated from spinal cord induces calcium-dependent release of 2-AG (Hegyi et al., 2018), suggesting that astrocytes could function as a feedforward system for the ECS system by amplifying 2-AG spread when CB1 receptors are activated.

While astrocyte-specific DAGLα knock out mice suggest that astrocyte 2-AG synthesis only contributes a small amount to steady state 2-AG concentrations, loss of this synthetic pathway results in increased depressive-like behavioral responses and changes in maternal behavior (Schuele et al., 2021).

Transcriptomic analyses of astrocytes from mice with experimental autoimmune encephalomyelitis (EAE), a model of multiple sclerosis, demonstrate significant changes in the expression of genes for both 2-AG and AEA synthesis and degradation (Moreno-Garcia et al., 2020). Interestingly, expression of FAAH mRNA is upregulated early in the disease course, prior to emergence of symptoms, while expression of both DAGL isoforms and MAGL are significantly reduced during acute disease and remain depressed in the recovery phase (Moreno-Garcia et al., 2020).

Thus, there is compelling evidence that astroglia are an important cell of the CNS ECS system. Furthermore, ECS plays a prominent role in the regulation of neuronal communication by astrocytes, and evolving data suggest that 2-AG should be considered both a neurotransmitter and gliotransmitter.

ECS in Microglial-Neuronal Communication.

As reviewed in the section above, many studies reveal that ECS mediates synaptic function through astrocytic signaling. Recent studies suggest that ECS also contributes to microglial influences on synaptic function. Communication between neurons and microglia is bidirectional, with microglia expressing receptors for many neurotransmitters and neurons responding to molecules released from microglia (Eyo & Wu, 2013). Microglia contain multiple types of neurotransmitter receptors including glutamatergic, GABAergic, noradrenergic, and cannabinoidergic receptors, among others. Microglia are highly secretory and release various neuroactive chemicals, such as BDNF, as well as free radicals, chemokines, and cytokines, depending upon the state of microglial activation and the signaling cascade activated by the neurotransmitter. The subsequent neuronal effects of these molecules can be neuromodulatory, neurotoxic, or neuroprotective. Some of the neuroactive chemicals released by microglia critically participate in the regulatory processes of synaptic transmission. This includes modulation of homeostatic synaptic plasticity, the development of neuronal networks (Gabrielli et al., 2015), response to injury (Kettenmann, Kirchhoff, & Verkhratsky, 2013) and inflammation (Parkhurst et al., 2013).

One mechanism by which microglia regulate synaptic activity is through the release of extracellular vesicles (EVs) (Simons & Raposo, 2009). EVs are a method to convey hydrophobic signaling molecules from the place of manufacture to the target destination within the nervous system. There is evidence that the eCBs and CB receptor are present in EVs released by activated microglia that presumably target neurons in the vicinity (Bianco et al., 2009; Gabrielli et al., 2015). There is evidence that microglial secretion of AEA-containing EVs and subsequent activation of CB1 receptor signaling in neurons is integral to the positioning and circuit connectivity of CB1R-expressing interneurons during brain development (Gabrielli et al., 2015). EVs that carry AEA on their surface have been shown to bind to the CB1 receptor of cultured GABAergic neurons, resulting in suppression of inhibitory postsynaptic current (mIPSC) frequency, consistent with a decrease in the number of spontaneously active synapses (Gabrielli et al., 2015).

Microglia have many important roles in the CNS, including important roles in synaptic stripping or loss during development and aging (Augusto-Oliveira et al., 2022). Microglial CB2 receptor, through regulation of microglial activation state as discussed above, could indirectly alter synapse numbers through effects on synapse stripping. While a role for the ECS system in regulating this process has not been demonstrated directly, CB2 receptor knock out mice exhibit reduced synapse number and decreased spine density (Garcia-Gutierrez et al., 2013; Y. Li & Kim, 2016) while chronic treatment with CB2 receptor agonist increases spine density in the hippocampus (Kim & Li, 2015). Interestingly, a recent study found that neuronally-derived interleukin-33 (IL-33) engages microglial IL1RL1 which increases phagocytosis of extracellular matrix proteins by microglia (P. T. Nguyen et al., 2020). As a result, extracellular space becomes “emptied” and available for remodeling of dendritic spines. A recent study suggests that neuronal CB1 receptors suppress the release of IL-33 while microglial CB2 receptors suppress expression of IL1RL1 (Du et al., 2020). Thus, elevation of 2-AG could affect this process by coordinating changes in both cell types.

Understanding of the regulation of microglia by the CB2 receptor is rapidly expanding, as is understanding of the bidirectional role of the ECS system in regulation of microglial-neuronal communication (Komorowska-Muller & Schmole, 2020). Given the extremely important role of microglia in regulating neuroinflammation, synapses, neurodevelopment and aging, this is an area of research that requires considerably more attention.

ECS and Oligodendrocytes.

Oligodendrocytes (OLs) are glial cells that form myelin sheath in the CNS (Sherman & Brophy, 2005). OLs differentiate from oligodendrocyte precursor cells (OPCs); OPCs arise from neural stem cells during embryonic development (R. J. M. Franklin & Ffrench-Constant, 2017). OLs are essential for proper signal propagation and also contribute to axonal integrity and maintenance through exosomal vesicle-mediated transfer of proteins, lipids and other cargo from OLs to neuronal axons (Fruhbeis et al., 2020). When a demyelinating injury occurs, OPCs undergo proliferation and migration to the site of OL loss, followed by differentiation and recovery of myelin (Spaas et al., 2021).

OLs and OPCs express both CB1 and CB2 receptor mRNA as well as transcripts for the synthetic and degradative enzymes for the eCBs (Molina-Holgado et al., 2022). Recent studies have begun to elucidate the functional role of the ECS in these cell types. For example, treatment of newborn rat pups (PND6–9) with THC resulted in enhanced OL maturation, by promoting cell cycle exit and differentiation of OPCs in the corpus callosum (Huerga-Gomez et al., 2021). THC treatment also promoted myelination and increased expression levels of myelin-associated proteins in the sub-cortical white matter; effects that were blocked by both CB1 and CB2 receptor antagonists (Huerga-Gomez et al., 2021). Activation of ECS can also promote remyelination; in particular, a low dose (0.5 mg/kg) of the mixed CB1/CB2 receptor agonist, WIN-55,212–2, increased myelinated axons and increased OPCs in cuprizone-treated mice (Tomas-Roig et al., 2020). However, chronic treatment with a slightly higher dose of WIN-55,212–2 (1 mg/kg) itself caused demyelination, perhaps as a result of CB1 receptor downregulation (Tomas-Roig et al., 2020). These data, in combination with the large body of evidence that the ECS plays an important role in neuronal differentiation and development (Galve-Roperh et al., 2013), suggest that further studies of the ECS in OL’s and their precursors are warranted.

ECS and neurovascular coupling.

The neurovascular unit (NVU), formed by cerebral endothelial cells interacting with pericytes, astrocytes and neurons, forms the interface between the peripheral circulatory system and the CNS (Kho, Glass, & Graham, 2017). Given that the CNS depends almost completely on oxygen for fuel, the NVU plays a vital role in coupling cerebral blood flow to local metabolic needs. Multiple studies have provided evidence that ECS components are expressed in the cell types involved in the NVU. CB1 receptor expression has been observed in perivascular astrocytes, endothelial cells and pericytes (Benyo, Ruisanchez, Leszl-Ishiguro, Sandor, & Pacher, 2016). Similarly, CB2 mRNA expression has been detected in endothelial cells and perivascular microglia (Kho et al., 2017). In addition to neurons, microglia and astrocytes, endothelial cells and blood-derived cells such as platelets, macrophages or lymphocytes are also sources of eCBs at the NVU (Benyo et al., 2016). Not surprisingly, the role of ECS in regulation of the NVU is complex and likely dependent upon circumstances.

Early studies demonstrated through a variety of methods that CB1 receptor activation causes vasodilation of the cerebral circulation (Wagner, Jarai, Batkai, & Kunos, 2001) through effects on both vascular smooth muscle cells (Gebremedhin, Lange, Campbell, Hillard, & Harder, 1999) and brain capillary endothelial cells (Chen et al., 2000). In addition, AEA causes vasodilation through effects on capsaicin-sensitive sensory nerves and TRPV1 receptors in arteries (Zygmunt et al., 1999), a mechanism that may contribute to cerebral edema (Cernak et al., 2004). Similarly, treatment of human brain endothelial cells with either 2-AG or AEA increases intracellular calcium concentrations, effects that are inhibited by CB1 and CB2 receptor antagonists, and by inhibition of TRPV1 (Golech et al., 2004).

Few studies have directly examined the role of the ECS in mediating blood flow recruitment in response to increased neuronal activity. Electrical stimulation was used to activate neurons in the nucleus accumbens of awake rats and local changes in oxygen and pH changes were used as indices of changes in cerebral blood flow (Cheer, Wassum, & Wightman, 2006). Systemic administration of the CB1 receptor agonist WIN-55212–2 reduced these measures, as did the AEA clearance inhibitor, AM404. Systemic administration of a CB1 receptor antagonist did not have any effect alone in this study, but was shown to increase the hyperemic response to whisker stimulation in another study (Patel, Gerrits, Muthian, Greene, & Hillard, 2002). The design of these studies did not allow for conclusions regarding the site or mechanism of action of the cannabinoid ligands.

A study employing laser Doppler flowmetry to examine local changes in blood flow together with direct application of drugs to the surface of the sensory cortex suggests that the primary effect of CB1 activity is to alter excitatory neurotransmission (Ho et al., 2010). In this paradigm, WIN 55212–2 potentiated while CB1 receptor antagonists inhibited the hyperemic response in the sensory cortex to whisker stimulation (Ho et al., 2010). Other evidence in this study points to neuronal CB1 receptors, specifically those on GABAergic terminals, as the primary site of action of the cannabinoid ligands to alter blood flow recruitment. Thus, under physiological conditions, ECS regulation of the balance of excitatory and inhibitory neurotransmission affects local metabolic drive and thereby indirectly regulates blood flow to the area (Benyo et al., 2016). This conclusion is supported by other data showing no effect of CB1 receptor antagonism on stimulation-induced astrocytic calcium transients (Schipke, Haas, & Kettenmann, 2008).

On the other hand, there is evidence that ECS plays a direct role in neurovascular coupling in pathologic conditions. For example, CB1 receptor activation, which increases endothelial calcium under physiological conditions (Golech et al., 2004), counteracts the increase evoked by endothelin-1 following ischemia (McCarron et al., 2006). Similarly, CB1 receptor blockade exhibited a potent effect on autoregulation of cerebral blood flow following hypoxia and hypercapnia, but was without effect under control conditions (Iring et al., 2013). One hypothesis that follows from these observations is that brain ischemia results in mobilization of ECS, specifically, increases eCBs. Several studies support this hypotheses; for example occlusion of the middle cerebral artery results in a significant increase in brain AEA concentrations (Muthian, Rademacher, Roelke, Gross, & Hillard, 2004) while traumatic brain injury increases 2-AG contents (Mechoulam & Shohami, 2007). Brain ischemia also increases CB1 receptor mRNA in astrocytes and microglia, and reduces expression in neurons, suggesting that non-neuronal signaling is enhanced during pathological circumstances (W. Schmidt, Schafer, Striggow, Frohlich, & Striggow, 2012). A recent study exploring the beneficial effects of electroacupuncture (EA) on stroke outcomes found that EA increased brain eCB contents and that eCB activation of penumbral astrocyte CB1 receptors provided protection against neuronal loss in a middle cerebral artery occlusion model (C. Yang et al., 2021).

Evolving data suggests a role for CB2 receptors in the regulation of cerebral blood flow (Kho et al., 2017). CB2 receptor activation attenuates the early brain injury after subarachnoid hemorrhage (Fujii, Sherchan, Krafft, et al., 2014; Fujii, Sherchan, Soejima, et al., 2014) and cerebral ischemia/reperfusion (M. Zhang et al., 2009). CB2 receptor agonists reduce edema formation and increase cerebral blood flow in a model of traumatic brain injury (Braun et al., 2018). The mechanism or site of action of CB2 receptor agonists is not completely clear, but likely involves reduction of neuronal inflammation (Amenta, Jallo, Tuma, & Elliott, 2012; Braun et al., 2018; Elliott, Tuma, Amenta, Barbe, & Jallo, 2011) secondary to reduced leukocyte recruitment and microglial activation (Amenta et al., 2012; Amenta, Jallo, Tuma, Hooper, & Elliott, 2014; M. Zhang et al., 2009). Further support for a peripheral mechanism of action of CB2 receptor activation, recent data in humans with ischemic stroke demonstrate CB2 receptor up-regulation in monocytes, particularly those expressing CD16 (Greco, Demartini, Zanaboni, et al., 2021).

ECS and Dopaminergic Circuits.

Dopaminergic neurons form a dense and complex neuronal network originating in the ventral midbrain (VTA and substantia nigra) and projecting widely to limbic and cortical regions (Reynolds & Flores, 2021). Dopaminergic neurotransmission is involved in a wide variety of behaviors, including reward and motor and cognitive regulation. The ECS system modulates the activity of dopaminergic neurons in the midbrain nuclei and regulates DA release in target nuclei (reviewed in (Covey, Mateo, Sulzer, Cheer, & Lovinger, 2017)). Dopaminergic neurons do not express CB1 receptor (Julian et al., 2003), so regulation of dopaminergic neuronal activity is largely mediated by CB1 receptor regulation of GABA and glutamate release onto dendrites and axons. Overall, the effect of CB1 receptor activation is enhanced DA release.

There is also evidence of direct interactions of the ECS system and dopaminergic systems at the receptor level. D2 receptor can interact with CB1 receptor, leading to the formation of heterodimers (Kearn, Blake-Palmer, Daniel, Mackie, & Glass, 2005). CB1/D2 heterodimers, widely studied in the striatum of rats and primates, have specific pharmacological properties. Stimulation of either receptor separately results in the activation of Gαi, while co-activation switches the signaling to Gαs (Bagher, Laprairie, Kelly, & Denovan-Wright, 2016; Bagher, Laprairie, Toguri, Kelly, & Denovan-Wright, 2017). Each member of these heterodimers can modulate the activity of the other receptor (Blume et al., 2013) and prolonged activation can alter formation of the heterodimers (Przybyla & Watts, 2010). Recent evidence has demonstrated D2-CB1 receptor interactions could contribute to pathophysiology. For example, perinatal exposure to THC upregulates mRNA expression of CB1 and D2 receptors and reduces DNA methylation of the Drd2 gene in adult rats (Di Bartolomeo et al., 2021). Interestingly, THC exposure also potentiates the formation and increases activation of D1-D2 receptor heteromers in the striatum (Hasbi et al., 2020). These studies suggested a role for D1-D2 heterodimers in the symptoms of cannabis use-related disorders

It has long been recognized that ECS modulates the reward circuit and many of these effects are mediated by enhanced dopaminergic signaling (Covey et al., 2017; Everett, Gomez, Hamilton, & Oleson, 2021). A recent study demonstrated that mGluR5-mediated increases in 2-AG synthesis in D1 receptor expressing neurons in the nucleus accumbens resulted in the formation of LTD of glutamatergic transmission at prefrontal cortico-accumbal synapses (Bilbao et al., 2020).

In contrast to CB1 receptor potentiation of dopaminergic signaling, there is evidence that CB2 receptor signaling has opposite effects on the activity of dopaminergic neurons, inhibiting neuronal firing and reducing neurotransmitter release (H. Y. Zhang et al., 2014; H. Y. Zhang et al., 2017). CB2 receptor agonists reduce the behavioral effects of both acute and chronic cocaine treatment (Xi et al., 2011; H. Y. Zhang et al., 2021). Similarly, recent data find that CBD, which can act as a CB2 receptor PAM, reduces the rewarding effects of cocaine through a mechanism partially blocked by CB2 receptor antagonism (Galaj, Bi, Yang, & Xi, 2020). The CB2 receptor agonist, Xie2–64, also reduces extracellular dopamine levels in the nucleus accumbens and reduces the rewarding effect of cocaine (Jordan et al., 2020). Recent data indicate that prolonged activation of CB2 receptor upregulates D2 receptor expression through a mechanism requiring GRK5, β-arrestin 2, and ERK1/2 protein signaling (J. M. Franklin, Broseguini de Souza, & Carrasco, 2021).

CB1 receptor-dopamine interactions are also relevant in the regulation of movement through effects on dorsal striatal circuits (reviewed in (Garcia, Palomo-Garo, Gomez-Galvez, & Fernandez-Ruiz, 2016)). Recent studies have found that the CB2 receptor agonist, AM1241, can protect dopaminergic neurons in a rodent model of Parkinson’s Disease via regulation of the Xist/miR-133b-3p/Pitx3 axis (He et al., 2020). Activation of the CB2 receptor resulted in a reduction of striatal neuroinflammation and vulnerability of dopamine neurons in a 6-OHDA model (Rentsch, Stayte, Egan, Clark, & Vissel, 2020; Yu et al., 2021).

ECS system and Neurotrophism.

Neural progenitor cells (NPCs) can generate functional neurons in the adult brain via a process called neurogenesis which occurs mainly in the sub-ventricular zone (SVZ) and sub-granular zone within the dentate gyrus (DG) (Rodrigues et al., 2017). In adult neurogenesis, neural stem cells (NSCs) or NPCs must proliferate, differentiate, survive, mature, and integrate into existing circuitry (C. Zhao, Deng, & Gage, 2008). Studies suggest important roles for the ECS system in neurogenesis. For example, in NSCs from early postnatal rats, DG cell proliferation requires CB1 receptor and CB2 receptor co-activation while SVZ cell proliferation is increased with CB1 receptor activation and lost with co-activation (Rodrigues et al., 2017). Genetic deletion of DAGLα from NSC and NPCs in the hippocampal DG results in strongly impaired neurogenesis, producing a 50% decrease in proliferation of cells (Schuele et al., 2022), suggesting that 2-AG-mediated signaling is required for proper neurogenesis. On the other hand, ABHD4, a serine hydrolase that can contribute to non-NAPE-PLD-mediated AEA synthesis (Simon & Cravatt, 2006), is necessary and sufficient for anoikis of radial glia progenitor cells that have inappropriately detached from the ventricular wall in response to alcohol exposure (Laszlo et al., 2020). Although the precise mechanism of this effect is not yet understood, these findings further emphasize the importance of the enzymatic cascades that regulate the eCBs and related lipids at all stages of life.

AM1241, a CB2 receptor agonist, induced neurogenesis in vitro and in GFAP/Gp120 transgenic mice (Avraham et al., 2014). HIV1 envelope protein Gp120 is associated with HIV-associated dementia and is known to decrease hippocampal NPC proliferation. AM1241 increased survival of human and murine NPCs and rescued neurogenesis in vivo as evidenced by an increase in neuroblasts and neuronal cells (Avraham et al., 2014). Due to the concurrent decrease in astrogliosis and gliogenesis, the protective effects of AM1241 were thought to be mediated by a decrease in neuroinflammation (Avraham et al., 2014). In addition to a role in proliferation, there is evidence that the CB2 receptor plays a role in survival of NSCs (Oddi, Scipioni, & Maccarrone, 2020). The CB2 receptor agonist, MDA7, restored expression of the NSC marker, Sox2, in the DG in a double transgenic APP/PS1 mice model of Alzheimer’s Disease, suggesting CB2 receptor activation rescued impaired neurogenesis (Oddi et al., 2020).

Brain-derived neurotrophic factor (BDNF) has been shown to induce release of eCBs at dendrites through its binding to tropomyosin receptor kinase B (trkB) (Lemtiri-Chlieh & Levine, 2010). BDNF binding to trkB increases activation of the phospholipase Cγ (PLCγ) signaling pathway (Pan, Zhao, & McNamara, 2019; Y. Wu et al., 2020; L. Zhao & Levine, 2014; Zhong et al., 2015), leading to the DAGL-mediated production of 2-AG (Y. Wu et al., 2020). 2-AG then acts as a retrograde messenger, binding presynaptic CB1 receptor. This leads to inhibition of adenylyl cyclase (AC) and protein kinase A (PKA) and suppresses presynaptic calcium influx, ultimately decreasing the probability of presynaptic neurotransmitter release (Y. Wu et al., 2020) (Figure 4). These effects have been shown in both layer 2/3 (L. Zhao & Levine, 2014) and layer 5 (Yeh, Selvam, & Levine, 2017) pyramidal neurons in the prefrontal cortex, in the hippocampus (Pan et al., 2019), and in calyx-type neurons in the brainstem (Yeh et al., 2017).

Figure 4. CB1R activation leads to an increase in BDNF.

CB1 receptor activation leads to an increase in PI3K-Akt-mTORC1 signaling, which ultimately leads to an increase in BDNF production.

This signaling has important effects on electrophysiological processes. In VTA dopamine neurons, increased BDNF-trkB signaling facilitates DSI and long-term depression of inhibition (iLTD) (Zhong et al., 2015). At glutamatergic corticostriatal synapses, trkB activation is required for the expression and scaling of both eCB-LTD and eCB-LTP (Gangarossa et al., 2020). Theta burst stimulation in layer 2/3 of the mouse somatosensory cortex results in a form of eCB-mediated iLTD, which induces endogenous BDNF-trkB signaling rather than mGluR activation (Zhong et al., 2015). In the rat barrel cortex, which processes sensory information from the whiskers, eCBs are also crucial for the induction of BDNF-dependent LTP at excitatory layer 5 pyramidal neuron synapses (Maglio, Noriega-Prieto, Maraver, & Fernandez de Sevilla, 2018).

There is also evidence that CB1 receptor activation leads to increased BDNF release (D’Souza, Pittman, Perry, & Simen, 2009) via the PI3K/Akt/mTORC1 pathway (Blazquez et al., 2015; D. P. Wang et al., 2021), which may have important implications for eliciting the neuroprotective effects of BDNF (Figure 5). Increased expression of CB1 receptor in the dorsolateral striatum in a mouse model of Huntington’s disease resulted in increased expression of BDNF and rescued neuropathologic deficits (Blazquez et al., 2015). In a mouse model of oxygen-glucose deprivation, indirect CB1 receptor activation through FAAH inhibition led to enhanced BDNF signaling and enhanced the neuroprotective effects of VEGF-A (D. P. Wang et al., 2021). In a mouse model of ALS, inhibition of MAGL increased expression of BDNF in the spinal cord and in primary neurons, resulting in a neuroprotective effect (Pasquarelli et al., 2017).

Figure 5. BDNF recruits synaptic endocannabinoid signaling.

BDNF binding to the TrkB receptor results in the DAGL-mediated production of 2-AG from the post-synaptic neuron. 2-AG then travels in a retrograde manner, binding the CB1R on the presynaptic terminal. CB1R activation leads to decreased presynaptic neurotransmitter release through inhibition of voltage-gated calcium channels and inhibition of adenylyl cyclase with the resulting inactivation of PKA. Abbreviations: 2-AG, 2-arachidonylglycerol; ATP, adenosine triphosphate; BDNF, brain-derived neurotrophic factor; cAMP, cyclic adenosine monophosphate; CB1, cannabinoid receptor type 1; DAG, diacylglycerol; DAGL, diacylglycerol lipase; PKA, protein kinase A; PLCγ1, phospholipase Cγ; trkB, tropomyosin receptor kinase B.

Additionally, there is evidence that during exercise, CB1 receptor activation triggers the production of BDNF, resulting in enhanced cognition (Ferreira-Vieira, Bastos, Pereira, Moreira, & Massensini, 2014). This has also been suggested through a study in human cyclists that found that increased plasma AEA following exercise might be one of the factors involved in the exercise-induced increase in peripheral BDNF (Heyman et al., 2012).

Summary.

Collectively, the many scientists intrigued by the ECS system have made incredible progress in understanding of an enormously important modulatory system in the CNS. Preclinical models strongly suggest that pharmacological manipulation of ECS holds great promise to treat a some of the most devasting diseases of the 21^st century, including mental disorders, neurodegeneration and metabolic disorders. While the ubiquitous expression of the ECS system underlies this great promise, it also presents the problem of how to selectively alter signaling. However, more selective approaches, including indirect agonists, NAMs and PAMs, are on the horizon. Like other fields of neuroscience, discovery and refinement of knowledge regarding the ECS system has been and will continue to be driven by advances provided by medicinal chemistry, new animal models, high resolution imaging, structural biology, omics and countless other developments. It is hard to imagine what the next 40 years will bring!