Imaging and Genetic Tools for the Investigation of the Endocannabinoid System in the CNS

1.1. Cannabinoid Receptors

Endocannabinoids can bind a wide variety of receptors, including the core cannabinoid receptor type 1 and 2 (CB1R and CB2R) as well as others such as G-protein-coupled receptors (GPR55, GPR18, GPR119), transient receptor potential vanilloid type 1 (TRPV1), and peroxisome proliferator-activated receptors (PPARα and PPARγ) (Table 1).

CB1R and CB2R are considered to be the main target for endogenous, plant-derived, and synthetic cannabinoids and are thus the best characterised receptors of the ECS [9]. As G-protein-coupled receptors (GPCRs), their activation triggers different effectors such as inwardly rectifying potassium channels (GIRKs), arrestins, adenylyl cyclase, mitogen-activated protein kinases, and some voltage-dependent calcium channels [7]. As a result, CBR activation can regulate a wide variety of cellular functions such as transcription, synaptic plasticity, and general cell physiology [10].

CB1R in humans is a 472-amino-acid protein, encoded by the CNR1 gene. It is highly expressed on axon segments and terminals of interneurons and glutamatergic, cholinergic, glycinergic, and serotonergic neurons [11]. Furthermore, CB1R distribution varies between brain regions such as the cortex, hippocampus, basal ganglia, and cerebellum. Recently, CB1R has been shown to be associated with subcellular organelles such as mitochondria [12]. CB2R is encoded by CNR2 in humans, with a full length of 360 amino acids. It is far less abundant than CB1R in the brain and is mainly found in microglia and vascular elements of the CNS, but substantial increases in CB2R expression can be induced with injury or inflammation [7,13,14].

The other receptors are not as well characterised as CB1R and CB2R and are rather considered as part of the extended ECS, or endocannabinoidome. However, CB1R/CB2R alone fail to explain the totality of eCB functions [15] and therefore the influence of other related receptors deserves further study. TRPV1 is a TRP channel activated by the eCB anandamide (N-arachidonoylethanolamine; AEA), among other endogenous ligands. It is regarded as an ionotropic CBR due to increasing evidence for its interaction with cannabinoids. This receptor often colocalizes with CB1R, suggesting a simultaneous activation by AEA. Interestingly, these two receptors mediate diametrically opposite effects on neuronal activity: TRPV1 tends to increase activity by regulating calcineurin and ATM serine–threonine kinase and promoting cation influx, while CB1R frequently inhibits synaptic activity. This contrast may contribute to the inconsistent role of AEA in anxiety and fear studies. Other receptors from the TRP channel family have been reported to bind cannabinoids. For example, AEA can bind to TRPM8, while the main psychoactive constituent of Cannabis sativa, Δ⁹-tetrahydrocannabinol (THC), can exert physiological actions through TRPV2, TRPV3, and TRPV4 binding. However, current knowledge on their mechanisms and roles in pathophysiology is scarce, necessitating further studies to understand the importance of TRP channels in the ECS [16,17].

The nuclear PPARα, PPARγ, and PPARδ receptors are also reported to be cannabinoid targets, with potentially a significant influence on gene transcription [18]. However, despite recent evidence implicating PPARs’ involvement in cannabinoid-induced neuroprotection and analgesia [8], the role of PPARs as true CBRs is still disputed.

Other GPCRs such as GPR55, GPR18, and GPR119, are also reportedly regulated by endocannabinoids but their interactions require more extensive study [19,20]. In particular, GPR55 can bind cannabinoids and some researchers designate this receptor as a putative CB3R [21]. However, GPR55 also interacts with non-cannabinoid ligands, shares a low amino acid homology with CB1R/CB2R, and interacts with different signalling pathways involving Rho-associated protein kinase (ROCK); thus, its designation as a true CBR is still disputed. GPR18 presents some sequence similarities to GPR55 in critical positions and binds a wide range of eCBs mediating therapeutic effects. Similarly, however, the GPR18 pharmacology and mechanism of action have yet to be clarified [22]. Therefore, this review will focus on CB1R and CB2R.

1.2. Endocannabinoids: Endogenous CBR Ligands

Endocannabinoids (eCBs) are the endogenous lipid signalling molecules that exert their physiological effects by interacting with CBRs. They are often contrasted and compared to plant-derived (phyto-) cannabinoids such as THC, which exerts most of its effects through CB1R agonism. 2-Arachidonoylglycerol (2-AG) and AEA are the first-described and best characterised eCBs, and thought to mediate the majority of ECS function [28,29].

AEA was discovered first in 1992, followed by 2-AG in 1995, while researchers were searching for the endogenous ligands of CBRs. Both compounds are key autocrine and paracrine second messengers derived from phospholipid metabolism [19,30].

While almost all neurotransmitters are pre-synthesized and stored in vesicles until their release, eCBs seem to be unique in that they are produced “on demand” at the synapse to modulate synaptic activity. This is mainly derived from the observation that the first steps in eCB biosynthesis are mediated by enzymes activated during membrane depolarization upon neurotransmitter binding [29]. However, the “on demand” release model remains controversial due to the lack of clarity regarding the molecular mechanisms involved and virtual inability to experimentally distinguish between on-demand synthesis and canonical vesicular release.

All eCBs are enzymatically synthesized from abundant membrane phospholipids, predominantly conjugated with polyunsaturated fatty acids (PUFAs). Specifically, 2-AG is majorly derived from membrane phosphatidylinositols (PIs), substituted with arachidonic acid, while AEA is synthesized from N-arachidonoyl phosphatidylethanolamine (PE) [3,31]. Between these two major eCBs, 2-AG is much more abundant and has a significantly higher agonist efficacy for both CB1R and CB2R than AEA [13]. Even though 2-AG and AEA are considered as the major eCBs, many of their congeners from the N-acylethanolamine (NAE) and monoacylglycerol (MAG) signalling lipid classes are also thought to interact with the ECS (Table 1; Figure 1) [32]. These molecules include 2-O-arachidonoylethanolamine (virodhamine or O-AEA), N-docosahexaenoylethanolamine (DHEA), N-eicosapentaenoylethanolamine (EPEA), N-palmitoylethanolamine (PEA), N-oleoylethanolamine (OEA), N-stearoylethanolamine (SEA), docosatetraenoylethanolamide (DEA), N-linoleoylethanolamine (LEA), dihomo-γ-linolenoyl-ethanolamide (DGLEA), arachidonyl glycerol ether (noladin ether or 2-AGE), and 2-oleoylglycerol (2-OG). The discovery of two endogenous CB1R agonists, N-arachidonoyldopamine (NADA) and 2-arachidonoyl lysophosphatidylinositol (2-ALPI), that are not NAEs or MAGs implies that many other bioactive lipids with a similar structure may also act as eCBs. These compounds are still poorly studied in comparison to 2-AG and AEA in the context of the ECS and CNS; thus, this review will focus on 2-AG and AEA.

2.1. The Most Used Tools in ECS Imaging

Many initial studies on ECS proteins employed radioisotope-labelled ligands (radioligands) to investigate their tissue distribution and infer their pharmacological properties (Table 2). Radioligands are coupled with non-invasive techniques such as PET to obtain 3D organ images or ex vivo autoradiography or radioligand binding assays. Therefore, they present relevant alternatives to immunocytochemistry (ICC) or immunohistochemistry (IHC), especially in the absence of validated antibodies [50,51].

ICC and IHC are still some of the most widely used techniques to study the cellular and subcellular ECS organization [1,52]. Utilizing conventional fluorescence light microscopy, these tools facilitate visualization of specific target structures with high sensitivity and adequate resolution. The potential limitations of immunofluorescence include limited resolution, low tissue penetration, and potential interference by endogenous molecules that impede affinity and selectivity. Fluorescence resonance energy transfer (FRET) greatly improves the resolution of optical microscopy, creating a midway between ICC and expensive electron microscopes [52]. This technique exploits an energy transfer between two closely positioned fluorophores and is mainly used to determine the proximity of two or more biomolecules to each other. Thus, one of its main limitations is a high dependence on the target distance while it also lacks appropriate labelling for intracellular proteins.

The next tier in optical power comes with the electron microscope. Transmission electron microscopy (TEM) has been the most widely exploited tool in this category for ECS investigations. TEM has a very high resolution but, compared to optical microscopy, is more expensive, time-consuming, and involves laborious sample preparation with high risk of contamination [52].

More recent tools have enabled huge improvements in imaging resolution, giving access to information with nanoscale distribution and unprecedented detail. A notable example from recent ECS studies is Stochastic Optical Localization Microscopy (STORM). Specifically, this technique has high lateral, axial, and spatial resolution and can be used for 3D analyses, but requires specialized fluorophores and produces enormous datasets, which impedes rapid workflows [1,52].

Mass spectrometry imaging is the main technique used to image eCBs specifically and has high sensitivity, mass resolution, and accuracy, but has no spatiotemporal resolution and limited quantitative power. Recently, bioengineered sensors have overcome these limitations, providing high selectivity and sensitivity and favourable kinetics, but their potential is yet to be fully exploited.

2.2. CB1R Imaging

Radioactive ligands have been a primary and extensively used tool, especially for receptor imaging. Therefore, there are several radioactively tagged agonist and antagonist ligands available for imaging with PET, autoradiography, or radioligand binding. For in vivo or ex vivo CB1R visualization, studies have mainly used PET imaging with [¹⁸F]- or [¹¹C]-labelled ligands [53] (Table 3). For example, Burns et al. were able to investigate the selective inverse agonist [¹⁸F]MK-9470 both with PET and autoradiography, confirming its potential as an imaging tool for CB1R. Similarly, DPR₂₁₁ was exploited for PET scanning studies by Chang et al., showing significant enrichment in the brain [54,55]. Another interesting [¹⁸F]-labelled radioligand is FMPEP-d₂ that has been used to assess CB1R distribution in Alzheimer’s disease mouse models, among other neurodegenerative diseases. [¹¹C]-labelled radioligands offer the advantage of a faster synthesis, but they have a shorter half-life, which may limit quantification of radioligand kinetics. The most promising radioligands in this category are [¹¹C]-OMAR and [¹¹C]COCl₂ [56,57].

Predictably, conventional fluorescence light microscopy has been extensively used for CB1R imaging, i.e., for elucidating its distribution from the organ to subcellular level with IHC. In the CNS, CB1R has been found in a wide range of regions (e.g., spinal cord, neocortex, olfactory bulb, striatum, amygdala, hippocampus, prefrontal cortex, cerebellum, and hypothalamus) and in different cell types ranging from glutamatergic neurons to astrocytes, microglia, and oligodendrocytes. For example, Tsou et al. generated a highly specific polyclonal antibody targeting the N-terminal CB1R to demonstrate its enriched distribution in the rat forebrain [58]. Egertová and Elphick analysed the expression of CB1R utilising C-terminal antibodies in rat brains, revealing the receptor’s subcellular locations [59]. Katona et al. assessed CB1R expression in the amygdala and elucidated its role in GABAergic synaptic transmission [60]. Navarrete et al. exploited rabbit polyclonal CB1R antibodies to demonstrate eCB-mediated neuron–astrocyte communication [61]. At the subcellular level, CB1R expression was shown at both the postsynaptic and presynaptic membranes as well as on mitochondrial-associated membranes and neuronal soma [24]. Marinelli and colleagues were able to confirm eCB-mediated regulation of interneurons and pyramidal cells using a similar strategy [62]. It should be noted that antibodies targeting the N- and C-terminal of CB1R have shown different effectivity for different applications. While an in-depth analysis thereof is outside the scope of this review, we refer the reader to the studies cited above for more detail.

Other types of probes have also been used for optical imaging. These are divided into covalent probes (i.e., photoactivable, electrophilic, and bifunctional) and fluorescent probes. Covalent probes for CB1R have mainly been used in affinity assays with relatively low application in imaging studies. Fluorescence probes, however, allow real-time assessment with a high spatiotemporal precision and have been used in imaging studies. Their downside is that a link between the pharmacophore and the fluorescent tag is required as the probe affects various physicochemical and photophysical properties of the final conjugate [53]. An interesting application of fluorescent probes was found by Martin-Couce et al. where a biotinylated 2-AGE analogue complex was successfully used to visualize CB1R in neurons [47].

FRET provided additional resolution on interactions between CBRs and other receptors. Many studies focused on the potential heteromerization of CB1R with orexin receptor 1 (OX-1R), considering their shared role in ERK1 and ERK2 activation and thus eCB signalling regulation. For example, Imperatore et al. demonstrated the formation of the OX-1R/CB1R heteromer [48]. Hojo et al. investigated another interaction between CB1R and the μ-opioid receptor (μOR), demonstrating the formation of a μOR/CB1R complex that transmits a signal though a common G protein [63].

Electron microscopy, particularly TEM, has been exploited in different investigations to provide a very high-resolution description of CB1R [64,65,66]. Nyíri et al. confirmed previous results on CB1R distribution, showing significant receptor density on the preterminal and terminal axonal segments in GABAergic and glutamatergic cells, as well as astrocytes [67]. Moreover, new tools such as immunogold imaging coupled with TEM provide even further opportunities. As previously stated, Benard et al. utilised this combination to probe mtCB1R and studied their energetic dynamics, demonstrating a potential link between specific neuronal functions and their energetic metabolism [12]. Immunogold labelling was also used by Katona and Freund to identify CB1R on other intracellular compartments such as Golgi and endoplasmic reticula [68].

STORM is another advance in optical imaging that is considered to be one of the most promising tools to image the ECS. For example, Dudok et al. employed STORM to study the synaptic ECS architecture and CB1R distribution with very high resolution, confirming its power to resolve cellular and subcellular molecule distribution at nanoscales [69]. Other interesting applications of this technique were shown by Zoldi and Katona, who were able to couple STORM imaging with behavioural pharmacology experiments, and Zhou et al., who investigated the receptor tyrosine kinase (RTK) transactivation mediated by CB1R [70,71].

2.3. CB2R

Radioligands can also be used to image CB2R; note the abundance of PET studies assessing its role in neuroinflammation (Table 4). To date, no CB2R radioligand has been approved due to poor specificity in the CNS, but some show significant potential [73]. Here, we review the main radioligand categories and cite the most promising candidates. CB2R radioligands are usually divided into five main subcategories based on their chemical class: oxoquinolines, oxadioles, thiophene and thiazole derivatives, pyridines, and indole derivates. Oxoquinoline compounds have significant potential for in vitro PET studies, but strong adverse effects impede their in vivo translatability. The most promising compound in this category is [¹¹C] RS-020, as a hydrophilic derivative of [¹⁸F]RS-126. This radioligand showed significant stability in the brain of Huntington’s disease (HD) mouse models and human amyotrophic lateral sclerosis (ALS) spinal cord tissues [74]. Oxadiazole, pyridine, and indole derivatives show promising preliminary results in vitro but fail in vivo. In the class of thiophene and thiazole derivatives, the most promising radiotracer is [¹¹C]A-836339 that has been exploited in Alzheimer’s disease (AD) mouse models to imagine neuroinflammation [75].

As for CB1R, multiple IHC and ICC techniques have been used to image CB2R coupled with light microscopy. The very first distribution studies with this tool detected CB2R expression in the spleen only; thus, it was classified as an exclusively peripheral protein. However, recent gene expression and immunocytochemistry studies have shown CB2R expression in the brain; thus, the exact distribution of CB2R remains a controversial topic. Wu and Wang used rabbit polyclonal antibodies to image the spatiotemporal changes of the receptor in the rat hippocampus, but an interesting specificity assessment by Zhang et al. demonstrated that available CB2R antibodies have poor specificity, bringing these and similar results into question [76,77].

Beyond canonical IHC and ICC, many different agonist or antagonist fluorophores have been used to study CB2R and have been gradually optimized to increase their efficacy. The fluorescent properties of biarylpyrazole derivatives have been exploited for this purpose. Bai et al. developed NIR-mbc94, which emits in the near-infrared (NIR) spectrum and can bind endogenous CB2R with high affinity in mouse-derived microglial and neuronal primary cultures [78]. An updated version was created by Wu et al., known as ZW760-mbc94, with significantly increased fluorescence intensity [79]. Other oxyquinoline and indole fluorophore derivates that show significant affinity and selectivity have mainly been used to assess inflammatory cells. Interestingly, derivatives of the THC molecular core have also been exploited for imaging studies, exemplified by Singh et al. who identified such a ligand and used it to successfully visualize CB2R in living cells with confocal microscopy [73,80].

FRET has also been used to assess whether CB2R can form heterodimers to understand its effects on downstream pathways. To that end, Mensching et al. investigated cAMP responses at the single cell level through CB2R [81].

Electron microscopy has also been employed to study CB2R. Onaivi et al. utilised immunoelectron microscopy to elucidate the CB2R subcellular location in rodent brains. Esteban et al. used a similar technique in AD mouse models, detecting the receptors on microglial cells [82,83]. Considering the controversy regarding CB2R expression in the brain, microscopy tools have mainly been exploited to study the receptor’s structure. Xing et al. described the 3D structure of CB2R-G_i and compared it to CB1R-G_i, showing overall similar interaction profiles between the two receptors. These structural results are critical to understanding CB2R function and, if complemented with further structural analyses and computational docking simulations, could help in synthetizing drugs to selectively target the receptor in CNS disorders [84]. To the best of our knowledge, super resolution techniques such as STORM have yet to be applied to CB2R [52].

2.4. Imaging the Endocannabinoids

Spatiotemporal distribution of eCBs remains elusive due to a lack of tools capable of directly visualizing them. This is due to the lipid nature of eCBs, the plethora of structural analogues in tissues, and the short half-life of functional eCBs [29,41]. To date, even the simplest immunofluorescence approaches are impeded due to a lack of commercial lipid-targeting antibodies.

Traditional metabolomics/lipidomics techniques such as liquid chromatography followed with mass spectrometry (MS) have been adapted to identify and quantify eCBs with high sensitivity and specificity but since these techniques often involve analyte extraction from the biological matrix, they offer very limited spatiotemporal resolution [85,86,87]. Nielsen et al. were able to exploit MS imaging (MSI) with matrix-assisted laser desorption/ionization (MALDI-MSI). This technique measures ions grid-wise across a 2D tissue matrix to detect high levels of brain MAG during the early phase of ischemia. However, they could not distinguish between 1-AG or 2-AG, and operated on the assumption that their analysis reflected the natural abundance of 2-AG over 1-AG in brain tissue [88] (Table 5).

Other methods have been used to indirectly assess spatiotemporal eCB dynamics such as electrophysiology or microdialysis coupled with pharmacological or genetic manipulations, but both present important limitations preventing precise measurement of the ECS in vivo [89]. Due to these critical limitations, even key assumptions regarding eCB signalling such as its duration (in the order of seconds) and spread (in the order of micrometres) have yet to be unambiguously confirmed [1].

Fortunately, new tools are continuously in development. High-specificity sensors developed through protein engineering have enabled, at least partially, bridging of these methodological gaps. Recently, Dong et al. developed such bioengineered fluorescent sensors based on GPCRs and circular-permutated fluorescent proteins (cpEGFP) to detect extracellular levels of dopamine, norepinephrine, acetylcholine, and serotonin [89]. Replicating this strategy, they created a new GPCR-activation-based eCB sensor known as GRAB_eCB2.0 that is based on the human CB1R and cpEGFP. This sensor has high selectivity, sensitivity, and favourable kinetics for detecting AEA and 2-AG in both healthy and diseased neuronal cultures, brain slices, and specific brain structures such as the amygdala and the hippocampus. Furthermore, GRAB_eCB2.0 can visualize eCBs’ dynamics during physiological and pathological processes in vitro and in vivo, opening a door to overcome current limitations in investigating eCB spatiotemporal dynamics.

3.1. C. elegans as a Model

The nematode C. elegans is an important invertebrate model for a wide range of biological research applications due to its practical advantages over more complex species. Its small size and cost-effective maintenance make it practical for storage in limited laboratory spaces [107]. They produce multiple offspring in a 3-day reproduction cycle, while thousands of commercially available mutant strains facilitate genetic and pharmacological studies. Of particular importance to CNS research, C. elegans has a fully characterised nervous system which consists of 302 neurons with a completely sequenced genome [108].

3.2. Zebrafish as a Model for ECS Studies

Danio rerio, commonly known as zebrafish, are small tropical fish native to Southeast Asia. Despite the limitations of using primitive fish as a model for complex human diseases, recent genetic advances underscore the numerous advantages that have established zebrafish as an increasingly important model. Some of these advantages include its rapid development and fecundity, well-characterised genome and orthologs to human genes, ease of genome editing, scalability for high-throughput pharmacological screens, as well as cost- and space-efficiency [121,122].

The authors apologize to all colleagues whose work was not discussed in this review.