A Brain on Cannabinoids: The Role of Dopamine Release in Reward Seeking and Addiction

Abstract

Cannabis sativa, like all known drugs of abuse, leads to increased dopamine activation within the mesolimbic pathway. Consequent dopamine release within terminal regions of the striatum is a powerful mediator of reward and reinforcement and patterned dopamine release is critical for associative learning processes that are fundamentally involved in addiction. The endocannabinoid system modulates dopamine release at multiple sites, and the receptors, endogenous ligands, and synthetic and metabolic enzymes of the endocannabinoid system may provide key targets for pharmacotherapies to treat disorders of motivation including addiction. Disrupting endocannabinoid signaling decreases drug-induced increases in dopamine release as well those dopamine events evoked by conditioned stimuli during reward seeking. Advances in recording techniques for dopamine are allowing unprecedented examinations of these two interacting systems and elucidating the mechanisms of endocannabinoid modulation of dopamine release in reward and addiction.

All known drugs of abuse, including Δ⁹-tetrahydrocannabinol, the primary psychoactive component of Cannabis sativa, induce increases in dopamine concentrations at terminal regions of the mesolimbic pathway (Di Chiara and Imperato 1988; Pierce and Kumaresan 2006). This pathway consists of a group of dopamine neurons originating in area A10, the ventral tegmental area (VTA), and projecting to limbic regions primarily the nucleus accumbens (NAc) (Hillarp et al. 1966). Elevations in dopamine within the NAc are associated with the reinforcing properties of rewards, including food (Roitman et al. 2004), drugs of abuse (Phillips et al. 2003), and direct electrical stimulation of the medial forebrain bundle (Cheer et al. 2007). In addition, this key circuit is involved not only in the primary reinforcing properties of drugs and rewards, but also in associative learning processes. Cues associated with drugs can be powerful reinforcers themselves and repeated pairings of contextual cues imbue these stimuli with secondary reinforcer properties (Bindra 1968; Berridge and Robinson 1998; Flagel et al. 2011). Cues are capable of precipitating relapse by inducing craving and reinstating drug seeking and taking (Phillips et al. 2003; Nicola 2010). Transient dopamine events encode these reward-predictive cues as well as rewards themselves (Phillips et al. 2003; Owesson-White et al. 2009). In addition, the negative affective state that occurs during drug withdrawal is associated with a decrease in mesolimbic dopamine function that might lead to compulsive drug seeking (Weiss et al. 2001; Koob 2009; Wenzel et al. 2015). This article reviews cannabinoid interactions with the dopamine system and how, through CB1 receptors, cannabinoids and endocannabinoids (ECs) modulate responses to drugs of abuse and reward predictive cues. The EC system is a critical gatekeeper of the dopamine system during associative learning about both natural and drug rewards. A greater understanding of the mechanisms of EC modulation of dopamine release may provide new targets for treatment of disorders of motivation including addiction (Table 1).

Table 1.

Terminology and definitions used in text

Terminology	Definition
Cannabinoids	Pharmacologically defined as a class of chemical compounds—comprising phytocannabinoids, chemically synthesized cannabinoids, and endocannabinoids—that bind to the cannabinoid CB1/CB2 receptor.
Cre recombinase	An enzyme derived from P1 bacteriophage; commonly used in neuroscience to selectively express, excise, or manipulate portions of DNA between loxP sites. Cre-mediated recombination can be used as a powerful tool to selectively express receptor genes, proteins including fluorescent sensors and optogenetic tools.
GCaMP	A family of genetically encoded calcium indicators (GECIs). GCaMP is a circularly permutated form of green fluorescent protein (GFP) with a calmodulin-binding element. GCaMP fluoresces when calcium is elevated and is a sensor of neural activity.
GRAB-DA/dLight	Dopamine sensors GRAB-DA (Sun et al. 2018) and dLight (Patriarchi et al. 2018); these are modified from dopamine receptors and fluoresce when dopamine binds. They are reporters of dopamine release in terminal regions and can be targeted to specific populations using genetic tools delivered typically using adeno-associated viruses (AAVs).
Fast-scan cyclic voltammetry (FSCV)	An electrochemical detection method for dopamine and other molecules. A carbon fiber microelectrode is implanted into a brain region and voltage is applied. Characteristic oxidation and reduction of transmitters surrounding the electrode produces peaks in current flow, which are measured and analyzed.
Primary reinforcer	An event that increases the probability of a behavioral response. In the context of drug addiction, an injection of heroin or smoking a pipe might function as a primary reinforcer.
Secondary reinforcer	Also referred to as a “conditional cue,” a stimulus that acquires reinforcing properties through Pavlovian associations. In the context of drug addiction, a syringe or a pipe might function as a secondary reinforcer.
Optogenetics	Light-sensitive channels (opsins) are expressed in target neurons; these can excite (channelrhodopsins) or inhibit (halorhodopsins) activity when blue or yellow light is applied, respectively, by opening ion channels.

INTRODUCTION TO DOPAMINE

Before delving into the interaction between cannabinoids and the dopamine system, it is important first to develop a general understanding of the patterns of dopamine signaling and the common methods used to monitor dopamine transmission in vivo. Dopamine neurons have a distinct functional fingerprint with two modes of activity, tonic and phasic firing (Roeper 2013). Tonic activity is driven by membrane currents, and consists of spontaneous single spikes, which are slow and pacemaker-like (1–5 Hz) (Grace and Onn 1989; Grace 1991; Dreyer et al. 2010). This provides a dopamine tone within terminal regions and acts on high-affinity D2 receptors (Keefe et al. 1993; Dreyer et al. 2010). Tonic dopamine levels within the NAc are detectable using techniques, such as in vivo microdialysis, that allow for neurochemical collection on a timescale of minutes. In contrast, when animals are presented with motivationally salient stimuli, such as conditioned cues that predict drug availability, midbrain dopamine neurons fire in high-frequency bursts (≥20 Hz), thereby producing transient increases in NAc dopamine concentration that are sufficiently high to occupy low-affinity dopamine D1 receptors (Richfield et al. 1989; Grace 1991; Phillips et al. 2003; Dreyer et al. 2010). These phasic dopamine events are detectable in vivo at the level of the dopamine neuron using single-unit electrophysiological recording techniques or at the neurochemical level within terminal fields of the mesolimbic dopamine system using fast-scan cyclic voltammetry (FSCV), an electrochemical technique that allows for the detection of dopamine on the millisecond timescale.

New advances in genetically encoded sensors have also emerged to record dopamine activity during behavior. These sensors include genetically encoded calcium indicators (GECIs), such as GCaMPs, which fluoresce when calcium is elevated during action potential events. GCaMP is a modified green fluorescent protein (GFP) construct, which produces measurable fluorescence when calcium binds to a calmodulin-binding site, and acts as a proxy for neuronal activity around the site of an implanted optical fiber (Chen et al. 2013). In addition, there are now several dopamine sensors (i.e., dLight, GRAB-DA [Patriarchi et al. 2018; Sun et al. 2018]), which consist of modified G-protein-coupled receptors (GPCRs) that are adapted to fluoresce when dopamine itself binds to them and these can be used in terminal regions to report phasic dopamine events. These encoded sensors are combined with light-based recording methods such as fiber photometry to measure fluctuations in fluorescence during behavior (Gunaydin et al. 2014; Lerner et al. 2015). The spatial, chemical, and temporal resolutions of these methods are not dissimilar to FSCV or dialysis. However, they offer a key advantage in that when using cre recombinase and other genetic approaches, expression of the sensor can be targeted to specific subpopulations of neurons in a transmitter and/or pathway-specific manner. These techniques afford a circuit specificity that previous methods were unable to discern and will be increasingly used to dissect the dopamine system.

A BRIEF INTRODUCTION TO THE ENDOCANNABINOID SYSTEM

The EC system consists of several endogenous GPCRs, lipid molecules that are ligands of these receptors (ECs), as well as several synthetic and metabolic enzymes involved in their production, release, and breakdown. There are two cannabinoid receptors, cannabinoid receptor 1 (CB1), found predominantly in the central nervous system (CNS) and the most abundant GPCR (Devane et al. 1988; Mechoulam and Parker 2013), and cannabinoid receptor 2 (CB2), more common in the periphery but also expressed in some neurons and glia (Munro et al. 1993; Atwood and Mackie 2010). Both CB1 and CB2 receptors are coupled to G_i/_o proteins, which promote inhibition of adenylyl cyclase as well as several voltage-dependent calcium channels, leading to inhibition of transmission within neurons expressing these receptors (Howlett et al. 2002). The two main endogenous ligands of the cannabinoid receptors are 2-arachidonoylglycerol (2-AG) and N-arachidonoylethanolamine ([AEA], or more commonly anandamide) (Devane et al. 1992; Mechoulam et al. 1995). The ECs 2-AG and anandamide have different synthetic and metabolic pathways (Howlett and Mukhopadhyay 2000). Namely, 2-AG is synthesized predominantly from 2-arachidonoyl-containing phospholipids (such as diacylglycerol lipase [DAGL]), whereas anandamide is produced from N-arachidonoyl phosphatidyl ethanol (NAPE) (Cravatt et al. 1996; Lu and Mackie 2016). They are also broken down by different means, 2-AG is metabolized by monoacylglycerol lipase (MAGL) mainly, and anandamide by the enzyme fatty acid amidohydrolase (FAAH) (Cravatt et al. 1996; Seifert et al. 2007; Lu and Mackie 2016).

A key feature of EC signaling is that these molecules are not stored in vesicles as classical neurotransmitters. Instead, 2-AG and anandamide are synthesized and released de novo in times of sustained neuronal activity (Freund et al. 2003). Increased activity of dopamine—or other—neurons, increases intracellular calcium, which activates enzymes (such as DAGL) leading to EC synthesis (Wilson and Nicoll 2002; Melis et al. 2004; Alger and Kim 2011). Once synthesized, ECs traverse the plasma membrane into the extrasynaptic space, in which they retrogradely activate cannabinoid CB1 receptors on presynaptic terminals (Wilson and Nicoll 2001). This unique mode of action makes ECs uniquely placed modulators of neuronal activity as a negative feedback mechanism to filter and select afferent inputs to given neurons (Alger 2002). Cannabis sativa contains several molecules that interact with the EC system, most notably Δ⁹-tetrahydrocannabinol, which predominantly interacts as a partial agonist at CB1 receptors. Dopamine neurons within reward-related pathways contain the molecular machinery required for EC synthesis and action, and this is increasingly implicated in normal reward processing and learning as well as in disorders of motivation such as addiction.

CANNABINOIDS INCREASE TONIC DOPAMINE LEVELS

A long-held misconception was that cannabinoids, such as Δ⁹-tetrahydrocannabinol, fail to increase dopamine concentrations in the same manner as other drugs of abuse (Wickelgren 1997). Currently, however, the existence of an overwhelming body of neurochemical evidence (Ng Cheong Ton et al. 1988; Chen et al. 1990, 1991, 1993; Tanda et al. 1997; Malone and Taylor 1999; Cheer et al. 2004) unequivocally shows that cannabinoids do, indeed, increase dopamine concentrations in the NAc. Importantly, these cannabinoid-induced increases in dopamine are most prominently observed in the shell region of the NAc and occur in a cannabinoid CB1 receptor–dependent manner. As shown in Figure 1A, the cannabinoid CB1 receptor agonist WIN55,212-2 increased dopamine concentrations within the shell, an effect that was blocked by the coadministration of the cannabinoid CB1 receptor antagonist rimonabant (Cheer et al. 2004). Cannabinoid-induced increases in NAc dopamine concentration are thought to arise from an increase in the mean firing rate of dopamine neurons within the VTA. In accordance with this theory, using single-unit recording techniques, multiple reports indicate that cannabinoids increase the firing rate of VTA dopamine neurons (French 1997; French et al. 1997; Gessa et al. 1998; Cheer et al. 2000a, 2003; Wu and French 2000). These parallel increases in cannabinoid-induced neural activity are shown in Figure 1B; specifically, WIN55,212-2 dose-dependently increased the frequency of dopamine neural firing (Gessa et al. 1998). Importantly, cannabinoid-induced increases in dopamine neural activity were abolished following administration of rimonabant, which shows that cannabinoids increase dopamine neural activity through a CB1 receptor–dependent mechanism (Gessa et al. 1998; Cheer et al. 2004).

Figure 1.

Cannabinoid effects on tonic and phasic dopamine release. Cannabinoids increase tonic and phasic release. (A) The cannabinoid receptor 1 (CB1) agonist WIN55,212-2 (0.3 mg/kg intravenous [i.v.], blue circles) increased tonic dopamine concentrations in the shell region of the nucleus accumbens (NAc) in comparison to vehicle (magenta diamonds). The cannabinoid antagonist rimonabant (1 mg/kg subcutaneous [s.c.], black triangles) prevented the WIN-induced increase in accumbal dopamine concentration (constructed from data in Tanda et al. 1997). (B) WIN55,212-2 dose-dependently increased the neuronal activity of an antidromically identified ventral tegmental area dopamine neuron. Rimonabant reversed the WIN-induced increase in dopamine neuronal activity (constructed from data in Gessa et al. 1998). (C) WIN55,212-2 (0.125 mg/kg i.v.) increased the frequency of phasic dopamine events detected in the shell of the NAc. To assess cannabinoid-induced changes in dopamine uptake, dopamine release was evoked by electrical stimulation (0.4 sec duration, 60 Hz, ±120 µA, black bars) applied to the medial forebrain bundle. WIN failed to increase the width of electrically evoked dopamine events, suggesting that cannabinoids do not increase dopamine by decreasing uptake (constructed from data in Cheer et al. 2004). (D) Δ⁹-Tetrahydrocannabinol (THC) (blue-filled bars) and WIN55,212-2 (lined bars) increased phasic dopamine neural activity. Both cannabinoids increased the number of bursting events and the numbers of impulses occurring per burst. *P < 0.05 (constructed from data in Gessa et al. 1998).

CANNABINOIDS INCREASE PHASIC DOPAMINE EVENTS

The majority of neurochemical studies have measured changes in tonic dopamine levels using in vivo microdialysis. To assess whether cannabinoids increase “phasic” dopamine release events, Cheer et al. (2004) measured NAc dopamine concentrations in the behaving rat using FSCV. As shown in Figure 1C, WIN55,212-2 increased the frequency of phasic dopamine events detected in the NAc shell (Cheer et al. 2004). Rather than resulting from the regular pacemaker firing that characterizes tonic dopamine signaling (e.g., Fig. 1B), these transient increases in accumbal dopamine release are thought to arise from high-frequency bursts of dopamine neural activity (Gonon 1988; Sombers et al. 2009). As would be predicted, therefore, Figure 1D shows that the cannabinoids Δ⁹-tetrahydrocannabinol and WIN55,212-2 both increased the frequency of bursts in addition to the number of impulses occurring during each burst of dopaminergic neural activity (Gessa et al. 1998).

DISRUPTING ENDOCANNABINOID SIGNALING DECREASES CUE-EVOKED DOPAMINE RELEASE

Pavlovian associations formed between drugs of abuse and environmental cues are fundamental mediators in addiction. Drug seeking can be increased by cues and these have a powerful influence on relapse. The effects of CB1 antagonism on both rewards and their predictive cues have been studied in multiple paradigms involving animals responding for drugs of abuse, natural rewards such as food, and intracranial self-stimulation (Justinova et al. 2008). During intracranial self-stimulation, rats press a lever to receive brief electrical activation of the mesolimbic pathway—which is reinforcing. Cheer and colleagues (Oleson et al. 2012) used a light cue to predict the availability of brain stimulation and they observed consistent terminal elevations of dopamine in the NAc to both reward and the conditioned cue. Systemic CB1 receptor antagonism reduced the impact of cues on reward-related behaviors for both intracranial stimulation of the mesolimbic pathway and in response to food rewards (Oleson et al. 2012). Furthermore, this was the result mainly of the effects of 2-AG rather than anandamide, as increasing 2-AG but not anandamide has the opposite effect of CB1 antagonism in increasing reward-seeking (behavioral) and dopamine responses to both reward delivery and reward-predictive cues (neural). The VTA EC system is critical in the regulation of dopamine signaling mediating reward-directed behavior. Not only can ECs mediate primary reward responses but they also modulate responses to secondary, conditioned predictors of reward.

2-ARACHIDONOYLGLYCEROL AND ANANDAMIDE DIFFERENTIALLY REGULATE DOPAMINE SIGNALING DURING CUE-MOTIVATED BEHAVIOR

The cannabinoids anandamide and 2-AG may have opposing effects on cue-related responding and dopamine release. Multiple studies using pharmacological tools that specifically increase anandamide levels report that these compounds reduce the potency of cues to motivate drug-seeking behavior (Scherma et al. 2008; Forget et al. 2009; Gamaleddin et al. 2011). VDM11 is a drug that selectively increases anandamide by inhibiting transport mechanisms (Van Der Stelt et al. 2006). Increasing anandamide with VDM11 has a similar effect to CB1 blockade, with decreased cue-evoked dopamine release and reward seeking (Oleson et al. 2012). Figure 2A shows the effects of VDM11 on accumbal dopamine concentrations in an animal responding for brain stimulation reward in a cued intracranial self-stimulation task. Anandamide decreases dopamine signaling during reward seeking. In contrast, 2-AG may act to enhance the effects of cue-evoked dopamine. Figure 2B shows a representative surface plot of FSCV dopamine peaks in the same task with enhanced 2-AG (using the MAGL inhibitor—JZL184). These data suggest that 2-AG is the primary EC involved in modulating cue-evoked dopamine release during reward-seeking behavior. Cue-evoked dopamine transients progressively increase in magnitude and occur with a shorter latency across the session with JZL184, and Figure 2C shows this increase compared with vehicle. It is possible that anandamide, which is a partial agonist for CB1 receptors, functions as a competitive antagonist in the presence of 2-AG; this conclusion is consistent with the observation that it is the primary EC involved in mediating synaptic plasticity in multiple brain regions (Melis et al. 2004; Tanimura et al. 2010).

Figure 2.

Anandamide decreases cue-evoked dopamine concentrations and reward seeking, while 2-arachidonoylglycerol (2-AG) increases cue responses. VDM11, a drug that selectively increases the endocannabinoid anandamide, dose-dependently decreased dopamine events occurring in response to a reward-predictive cue. As the cue-evoked dopamine concentration decreased, the behavioral response for brain stimulation reward shifted away from the conditioned cue. (A) (Left) Dopamine concentration measured by fast-scan cyclic voltammetry in response to cue presentation with VDM11. (Right) Representative dopamine traces for vehicle (dark blue) and VMD11 560 µg (light blue). (b) baseline, (v) vehicle (300–560 µg/kg intravenous [i.v.]). (B) Enhancement of 2-AG with JZL184 had opposing effects, as seen in a representative surface plot showing changes in dopamine concentration (z-axis) occurring across trials (y-axis), while responding is maintained by brain stimulation reward in a cued intracranial self-stimulation task. JZL184 at 10 mg/kg, i.v., which selectively increases 2-AG), increased dopamine responses to the cue and reduced latency to respond. (Black line) baseline, (red line) vehicle, (orange line) JZL184. (C) Increasing 2-AG levels produces robust dopamine increases to cues. (Left) Cue-evoked dopamine responses under baseline, vehicle, and with JZL184 (10 mg/kg i.v., orange bar). (Right) Representative dopamine traces for vehicle (red) and JZL184 (orange) show that augmenting 2-AG produces increased cue-evoked dopamine responses. *P < 0.05. (Constructed from data in Oleson et al. 2012.)

DISRUPTING ENDOCANNABINOID SIGNALING DECREASES DRUG-INDUCED INCREASES IN PHASIC AND TONIC DOPAMINE SIGNALING

All drugs of abuse increase phasic dopamine events, which promotes drug seeking (Cheer et al. 2007; Aragona et al. 2008). Decreasing drug-induced phasic dopamine events by disrupting EC signaling might, therefore, prove to be a successful pharmacological approach for the treatment of addiction (de Vries and Schoffelmeer 2005; Solinas et al. 2008). To test whether disrupting EC signaling decreases drug-induced phasic dopamine events, Cheer et al. (2007) monitored drug-induced increases in phasic dopamine release events in the NAc shell using FSCV. All drugs assessed, including cocaine, nicotine, amphetamine, and ethanol, increased the frequency of phasic dopamine events. Remarkably, disrupting EC signaling by coadministering rimonabant (CB1 antagonist/inverse agonist) attenuated these drug-induced increases in phasic dopamine release. As shown in Figure 3C, cocaine-induced increases in phasic dopamine events (top) were diminished by rimonabant (middle) and not observed following vehicle administration alone (bottom). If disrupting EC signaling decreases drug-induced phasic dopamine events by preventing the disinhibition of dopamine neural activity within the VTA, tonic dopamine signaling should also be suppressed. In accordance with this theory, ethanol-induced increases in tonic accumbal dopamine concentrations are indeed blocked by rimonabant (Hungund et al. 2003). These findings are in agreement with the electrophysiology literature. For example, Pistis and colleagues (Perra et al. 2005) showed that ethanol-induced increases in dopamine neural activity are reduced following rimonabant treatment. There is increasing evidence that antagonism of CB1 receptors abolishes phasic dopamine responses to many drugs of abuse including cocaine, ethanol, nicotine, cannabinoids, and amphetamine (Cheer et al. 2007; Wang et al. 2015; Covey et al. 2016), and may underlie their propensity to suppress drug reinforcement (Lupica and Riegel 2005; Lazary et al. 2011). Together, these findings support that developing drugs designed to disrupt EC signaling might decrease drug-induced increases in phasic dopamine release, which is thought to promote drug seeking, in addition to tonic dopamine release, which is thought to mediate the primary rewarding and reinforcing effects of drugs of abuse. However, it is worth noting that rimonabant, which reached clinical trials and was licensed in several countries as an anti-obesity and smoking cessation drug in 2006, was later withdrawn because of substantial psychiatric side effects including increased anxiety and depression in humans and in animal models (Viveros et al. 2005; Moreira and Lutz 2008; Moreira and Crippa 2009). Targeting the CB1 receptor directly may be responsible for such effects. There is some evidence that rimonabant has an inverse agonist action at the CB1 receptor. Unlike neutral agonists that competitively inhibit ECs at the CB1 receptor, rimonabant has intrinsic activity of its own; for example, in vitro it inhibits GTPγS binding to membranes of cell preparations in which no ECs are present (Pertwee 2003; Xie et al. 2007). The extent to which these side effects are caused by rimonabant's inverse agonist actions is still unknown and poorly understood. However, recently, drugs targeting the enzymatic pathways in synthesis and breakdown of ECs have been developed in attempts to reduce some of these effects seen when directly interacting with CB1 receptors and may show promise in obesity and addiction treatments.

Figure 3.

Endocannabinoids are necessary for drug-induced increases in tonic and phasic dopamine release. (A) Endocannabinoids are required for ethanol-induced increases in tonic dopamine concentrations in the nucleus accumbens (NAc). When administered independently, ethanol (1.5 g/kg, green squares) increased dopamine concentrations (constructed from data in Hungund et al. 2003). (B) Disrupting endocannabinoid signaling with rimonabant (1 mg/kg intravenous [i.v.], bottom) reversed ethanol-induced (0.5 g/kg i.v., top) increases in the neural activity of the ventral tegmental area dopamine neurons (constructed from data in Perra et al. 2005). (C) Endocannabinoids are required for drug-induced phasic dopamine events. Cocaine (3 mg/kg, i.v., top) significantly increased transient increases in NAc dopamine concentration. Rimonabant coadministration (0.3 mg/kg, i.v., middle) significantly attenuated the cocaine-induced increases in phasic dopamine release. Vehicle alone failed to alter phasic dopamine (bottom) (constructed from data in Cheer et al. 2007).

ENDOCANNABINOIDS INCREASE DOPAMINE RELEASE BY DISINHIBITING DOPAMINE NEURON ACTIVITY

Dopamine cell bodies lack CB1 receptors (Herkenham et al. 1991; Julian et al. 2003), so it is unlikely that increases in dopamine seen following CB1 receptor activation are caused by direct effects. Instead it has been proposed by Lupica and colleagues that cannabinoids act to increase dopamine through indirect mechanisms, notably through disinhibition of dopamine neurons within the VTA by actions at GABA interneurons (Lupica and Riegel 2005). Under standard conditions, dopamine neurons within the VTA receive substantial inhibitory GABA inputs (Paladini and Tepper 2017). Disinhibition, through removal of this GABAergic input, removes this tonic inhibition and promotes dopamine release. Selective optogenetic inhibition of GABA inputs to VTA dopamine neurons evokes dopamine release in the NAc (Nieh et al. 2016). In addition, such manipulations drive appetitive behavior and support reward seeking (Stamatakis et al. 2013; Nieh et al. 2016).

Application of cannabinoids to VTA brain slices decreases GABAergic inhibitory postsynaptic currents in a GABA_A receptor-dependent manner (Szabo et al. 2002). Furthermore, pretreatment with GABA_A antagonists abolishes the ability of CB1 agonists to increase dopamine neural activity (Cheer et al. 2000a). ECs binding to presynaptic cannabinoid CB1 receptors is known to result in the suppression of GABA-mediated inhibition (Wilson and Nicoll 2001). Within the VTA, this suppression of inhibition results in net disinhibition of neuronal activity (Lupica and Riegel 2005) and subsequent increased dopamine release. Figure 4 shows a proposed disinhibition mechanism, phasic activation of dopamine neurons, and consequent increased intracellular Ca²⁺ activates DAGL leading to EC synthesis (Wilson and Nicoll 2002; Melis et al. 2004; Alger and Kim 2011). Synthesized ECs activate cannabinoid CB1 receptors on presynaptic terminals on GABA neurons (Wilson and Nicoll 2001). There is a growing body of evidence suggesting that 2-AG is the primary EC involved in mediating such forms of synaptic plasticity (Melis et al. 2004; Tanimura et al. 2010). Several drugs of abuse including cocaine (Wang et al. 2015; Tung et al. 2016) and nicotine (Buczynski et al. 2016) also promote 2-AG synthesis within the VTA, which may contribute to this disinhibitory mechanism.

Figure 4.

Disinhibition of ventral tegmental area (VTA) dopamine neurons by cannabinoid receptor 1 (CB1) receptor activation. Under typical conditions, dopamine neurons within the VTA are inhibited by GABA through activation of GABA_B receptors. When animals are presented with motivational salient stimuli (e.g., a drug-associated cue), dopamine neurons fire in high-frequency phasic bursts. (1) Consequently, intracellular calcium levels increase, which results in the activation of endocannabinoid-synthesizing enzymes (e.g., diacylglycerol lipase [DAGL]). (2) As a result, 2-arachidonoylglycerol (2-AG) is synthesized and released into the extrasynaptic space. (3) 2-AG retrogradely activates G_i/o-coupled cannabinoid CB1 receptors on GABAergic terminals. (4) GABA release is suppressed. This GABA suppression results in disinhibition of the dopamine neuron, which presumably promotes the occurrence of phasic dopamine events.

ENDOCANNABINOIDS MODULATE DOPAMINE RELEASE AT MULTIPLE SITES THROUGH INHIBITION OF GABA AND GLUTAMATE ACTIVITY

Mesolimbic dopamine neurons terminate in the NAc, in which release of dopamine subserves important functions in reward and associative learning about cues relating to the availability of both natural and drug rewards as discussed above. There are several modulatory influences on terminal dopamine release, which can occur independently of cell body firing (Cachope et al. 2012; Threlfell et al. 2012). Two key nodes worth discussing here are cortical glutamate afferents to the NAc and cholinergic interneurons (CINs) within the NAc. These are two possible points of interaction for ECs to modulate dopamine release at NAc terminals.

CIN activation produces robust dopamine release in the NAc independently of VTA firing (Cachope et al. 2012). Mateo et al. (2017) used optogenetics to selectively activate CINs within the NAc while recording with FSCV; they found in vivo and in vitro that agonism of CB1 receptors with WIN55,212-2 significantly decreased CIN-driven dopamine release and this was prevented by pretreatment with CB1 antagonist AM251 showing a CB1 dependence (Mateo et al. 2017). However, CINs themselves do not express CB1 receptors (Uchigashima et al. 2007; Hohmann and Herkenham 2000; Mateo et al. 2017). This finding was supported by data showing genetic deletion of CB1 receptors from this population of CINs (using CB1^flx/flx mice crossed with ChAT::cre mice) does not affect the WIN55,212-2 effect.

One candidate mechanism for CB1 agonist-induced reduction in CIN-modulated dopamine release is that cortical glutamatergic afferents onto CINs, which possess CB1 receptors, underlie this effect. In fact, cortical activation of prefrontal cortical glutamate afferents stimulates CINs and this enhances dopamine release via nicotinic ACh receptors (nAChRs) (Kosillo et al. 2016). Selective deletion of CB1 receptors in cortical neurons can be achieved with a cre-recombinase-driven approach. In ChAT::cre/CB1^flox/flox mice, an adeno-associated virus (AAV) encoding cre under a CaMKIIα promoter is transduced into the medial prefrontal cortex (mPFC) and this selectively deletes CB1 receptors in cortical projection neurons to the NAc (Chiarlone et al. 2014; Mateo et al. 2017). Pharmacological agonism of CB1 receptors no longer decreases CIN-evoked dopamine release in animals without CB1 receptors in cortical glutamate terminals. In these experiments, optical CIN stimulation is modulated by cortical glutamate afferents, which in turn can be influenced by the agonism of the CB1 receptors that these cortical afferents express.

Not only does pharmacological agonism of CB1 receptors decrease CIN-driven dopamine release, but this effect is also recapitulated by raising tissue levels of endogenous cannabinoids such as 2-AG, suggesting a role for the EC system in curtailing CIN-driven release at the terminals (Mateo et al. 2017). Figure 5 summarizes these proposed mechanisms for terminal modulation of dopamine release by CB1 receptors. ECs, via their actions at CB1 receptors, are powerful modulators and exert significant modulatory influences on striatal dopamine release and motivated behavior, acting to indirectly disinhibit dopamine neurons in the midbrain as well as via complex interactions at glutamatergic cortical terminals in the NAc (Cheer et al. 2000b, 2004; Lupica et al. 2004; Riegel and Lupica 2004; Mateo et al. 2017). Despite several candidate mechanisms, it is still unclear exactly how ECs are activated and preferentially modulate dopamine release within terminal regions. Further research is needed to dissect these mechanisms. In addition, it remains to be determined how different cell body and terminal EC actions are integrated in the behaving animal.

Figure 5.

Proposed mechanisms for terminal control of dopamine (DA) release by cannabinoid 1 (CB1) receptors. Both cholinergic and glutamatergic influences on DA release at terminals within the nucleus accumbens. (1) Activation of cholinergic interneurons (CINs) by electrical or optogenetic stimulation, produces release of acetylcholine (ACh). (2) ACh drives glutamate release through actions at nicotinic ACh receptors (nAChRs) on cortical glutamatergic afferents. (3) Glutamate is released, and activates AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) receptors on DA varicosities and potentiates DA release. (4) Binding of 2-arachidonoylglycerol (2-AG) to CB1 receptors inhibits activity of cortical glutamate terminals inhibiting glutamate release and reducing DA activity. (mAChR) Muscarinic acetylcholine receptor, (Glu) glutamate, (PFC) prefrontal cortex.

CB1 RECEPTORS ON GABA AND GLUTAMATE NEURONS HAVE DISSOCIABLE EFFECTS ON COCAINE SELF-ADMINISTRATION

The presence of CB1 receptors on both glutamate and GABA neurons and the different consequences for dopamine modulation and the ability to selectively ablate these using modern genetic viral strategies has led to new evidence that CB1 modulation of addictive behaviors is nuanced and complex. In a cocaine self-administration paradigm—combined with microdialysis detection of dopamine fluctuations—Martín-García et al. (2016) selectively knocked out CB1 receptors in two mouse lines, from either forebrain GABA neurons or from cortical glutamate neurons. They found that these two different populations of neurons expressing CB1 receptors were involved in different aspects of cocaine reinforcement. Namely, GABA CB1 knockout increased both the primary reinforcing properties of cocaine and the dopamine responses to cocaine in the NAc. Conversely, glutamate CB1 knockout facilitated cocaine cue-associated learning and responses to the secondary reinforcer, the light cue. CB1 receptors on forebrain GABA neurons preferentially modulate the sensitivity to primary reinforcing properties of drugs of abuse, whereas those present on cortical glutamate neurons are implicated in the associative learning processes involved in drug taking.

DOPAMINE AND ENDOCANNABINOIDS IN NEGATIVE REINFORCEMENT AND AVERSION

Withdrawal, a key feature of addiction, is a negative emotional state that drives persistent relapsing drug seeking (Childress et al. 1988; Koob et al. 1998), and cannabinoids, like other drugs of abuse, can produce some withdrawal symptoms (Jones et al. 1976; Budney et al. 1999; Haney et al. 1999). The “cannabis-withdrawal syndrome” manifests as anxiety, decreased appetite, restlessness, sleep difficulties, chills, depressed mood, stomach pain, shakiness, and sweating (Budney and Hughes 2006). These symptoms contribute to cannabis dependence through negative reinforcement processes. Withdrawal occurs in association with a decrease in mesolimbic dopamine function (Weiss et al. 2001; Wenzel et al. 2015). During withdrawal from drugs, including cocaine, ethanol, morphine, and amphetamine, experimental animals show decreased tonic dopamine concentrations in the NAc as assessed by in vivo microdialysis (Rossetti et al. 1992; Weiss et al. 1992, 1996). In the case of cannabinoids, precipitated withdrawal induced by a rimonabant challenge results in immediately observable withdrawal symptoms (e.g., wet dog shakes) (Aceto et al. 1995; Tsou et al. 1995). This is accompanied by decreased VTA neural activity as assessed using single-unit recording (Diana et al. 1999). Decreased dopamine neural activity occurs along the same time course as decreased tonic dopamine concentrations observed in the NAc with microdialysis (Tanda et al. 1999).

In recent years, there has been considerable debate around how NAc dopamine encodes aversive stimuli with some conflicting data from multiple methods (Ikemoto and Panksepp 1999; Joseph et al. 2003; Wenzel et al. 2015). Microdialysis studies in rats have shown increased dopamine in response to aversive stimuli including foot and tail shock (Abercrombie et al. 1989; Young et al. 1992). Electrophysiological studies, however, have reported mixed findings, with both inhibitory responses (Mirenowicz and Schultz 1994; Ungless et al. 2004; Cohen et al. 2012) and excitatory responses reported in different paradigms (Horvitz 2000; Brischoux et al. 2009; Matsumoto and Hikosaka 2009), although others have shown decreased NAc dopamine transients using FSCV recordings in the NAc (Roitman et al. 2008; Budygin et al. 2012; McCutcheon et al. 2012).

McCutcheon et al. (2012) assessed dopamine transients within the NAc shell; they investigated responses to sucrose under different learning history conditions. In half of the rats, sucrose was paired with lithium chloride (conditioned taste aversion). For those animals in which sucrose had been paired with an aversive consequence, dopamine transients were reduced compared with those rats receiving unadulterated sucrose (McCutcheon et al. 2012). Dopamine responses to the same stimulus were opposite in animals with an aversive learning history. Moreover, Cheer and colleagues have shown that phasic dopamine increases accompany cues that predict aversive foot shock (Wenzel et al. 2018). These transients play a role in driving active avoidance, an adaptive response to aversive events. Furthermore, ECs are required for such dopamine transients and the avoidance behavior that they facilitate (Wenzel et al. 2018). These dopamine events (and their modulation by cannabinoids) were only present during learning and not once was the avoidance behavior fully established. Dopamine and EC mechanisms thus appear crucial specifically in the learning of active responses to aversive stimuli rather than the stimuli itself.

Some of the discrepancies in the studies discussed have been attributed to methodological differences, for example, the intensity of aversive stimuli (an air puff vs. an electric shock). There may also be important species differences (between monkeys often used in electrophysiological experiments and rodents in microdialysis or FSCV), as well as the paradigms in which they are tested (head fixed animals vs. freely moving). In electrophysiological studies, there may be some misidentification of dopamine neurons (Margolis et al. 2006; Ungless and Grace 2012). Finally, in microdialysis, samples often include both the onset and offset of an aversive stimulus; they then do not differentiate between the aversive stimulus itself and its termination or relief (however, see Young 2004). Regional differences in the VTA or NAc may also account for some discrepant results. For example, Brischoux et al. (2009) performed electrophysiological recordings in the VTA during noxious foot shock and found two distinct, anatomically segregated responses. Dorsal VTA neurons were inhibited, whereas ventral cells increased their firing rate. A further distinction can be made based on the NAc projection targets of these VTA neurons. Lammel et al. (2011) found reward stimuli preferentially activated VTA neurons projecting to the NAc medial shell, whereas those projecting to the NAc lateral shell respond to both rewarding and aversive stimuli. The heterogeneity of VTA dopamine neurons and their NAc targets, both anatomically and functionally, is an ongoing and active area of research.

CONCLUDING REMARKS

Cannabinoids, such as Δ⁹-tetrahydrocannabinol, affect the mesolimbic dopamine system similarly to other common drugs of abuse resulting in increased terminal release within the NAc. In addition, the EC system, comprising CB1 and CB2 receptors, 2-AG, anandamide, and their various synthetic and metabolic enzymes, is a key modulator of both drug- and cue-related dopamine release. Little is known concerning the relative contributions of specific ECs or their exact signaling mechanisms, and we are just beginning to dissect these multiple influences. We are, however, now aware of compelling evidence showing that the ECs modulate the mesolimbic dopamine system and of their potential impact in disorders of motivation. Future studies must be conducted to dissect the precise roles of ECs in this modulation and how they influence dopamine transmission in animal models. New methodological advances, particularly in sensor-based techniques, which allow genetic targeting of sensor expression, will allow the dissection of these mechanisms in detail never before possible. A greater understanding of the basic functioning of the EC system and its interaction with other key transmitters, particularly dopamine, will help elucidate mechanisms of disorders of motivation, especially addiction.