The endocannabinoid system in brain reward processes

Abstract

Food, drugs and brain stimulation can serve as strong rewarding stimuli and are all believed to activate common brain circuits that evolved in mammals to favour fitness and survival. For decades, endogenous dopaminergic and opioid systems have been considered the most important systems in mediating brain reward processes. Recent evidence suggests that the endogenous cannabinoid (endocannabinoid) system also has an important role in signalling of rewarding events. First, CB₁ receptors are found in brain areas involved in reward processes, such as the dopaminergic mesolimbic system. Second, activation of CB₁ receptors by plant-derived, synthetic or endogenous CB₁ receptor agonists stimulates dopaminergic neurotransmission, produces rewarding effects and increases rewarding effects of abused drugs and food. Third, pharmacological or genetic blockade of CB₁ receptors prevents activation of dopaminergic neurotransmission by several addictive drugs and reduces rewarding effects of food and these drugs. Fourth, brain levels of the endocannabinoids anandamide and 2-arachidonoylglycerol are altered by activation of reward processes. However, the intrinsic activity of the endocannabinoid system does not appear to play a facilitatory role in brain stimulation reward and some evidence suggests it may even oppose it. The influence of the endocannabinoid system on brain reward processes may depend on the degree of activation of the different brain areas involved and might represent a mechanism for fine-tuning dopaminergic activity. Although involvement of the various components of the endocannabinoid system may differ depending on the type of rewarding event investigated, this system appears to play a major role in modulating reward processes.

Endogenous cannabinoid system

The endogenous cannabinoid (endocannabinoid) system is a signalling system composed of cannabinoid (CB) receptors, endogenous ligands for these receptors and proteins involved in the formation and deactivation of these endogenous ligands (Freund et al., 2003; Piomelli, 2003; Di Marzo et al., 2004; Fride, 2005). Although the endocannabinoid system is thought to serve multiple functions in the brain and in peripheral tissues (Freund et al., 2003; Piomelli, 2003; Di Marzo et al., 2004; Fride, 2005), in this review we will focus on its involvement in motivational processes. In particular, we will review the effects of activation and inactivation of the endocannabinoid system on behaviours maintained by natural, electrical and drug rewards and on the activity of the mesolimbic dopamine system, which is the brain neurotransmitter system most clearly involved in reward processes (Koob, 2000; Di Chiara et al., 2004; Wise, 2004; Everitt and Robbins, 2005; Gardner, 2005). We refer the reader to recent excellent reviews on the interactions of the endocannabinoid system with other neurotransmitters involved in reward processes such as endoopioids (Manzanares et al., 1999; Fattore et al., 2005; Gardner, 2005; Vigano et al., 2005; Solinas et al., 2007c), on the involvement of the endocannabinoid system in neuroplastic changes that may underlie drug addiction (Maldonado et al., 2006) and on the role of the endocannabinoid system in relapse (De Vries and Schoffelmeer, 2005).

CB receptors

For decades it was believed that the effects of the main active ingredient in cannabis, delta-9-tetrahydrocannabinol (THC), were due to alterations of cellular membrane structure. However, in the late 1980s, due to the availability of new synthetic CB receptor agonists, it was first suggested that specific CB receptors exist (Devane et al., 1988; Howlett et al., 1990). Soon after, the first CB receptor was sequenced and cloned (Matsuda et al., 1990). This receptor, named CB₁, is highly expressed in the brain (Herkenham et al., 1990, 1991) and mediates most, if not all, of the psychoactive/central effects of cannabis. A short time later, a second CB receptor, named CB₂, was discovered (Munro et al., 1993). Until recently it was thought that CB₂ receptors were present only in the periphery and did not mediate any central effect of CBs, but recent findings suggest that CB₂ receptors are present at low levels in some areas of the brain (Van Sickle et al., 2005; Gong et al., 2006; Onaivi et al., 2006). On the basis of studies showing certain behavioural and pharmacological effects of CB ligands that could not be explained exclusively by CB₁ and CB₂ receptors, it has also been hypothesized that additional non-CB₁ and non-CB₂ receptors might exist (Jarai et al., 1999; Zimmer et al., 1999; Breivogel et al., 2001; Hajos et al., 2001; Freund and Hajos, 2003; Fride et al., 2003; Begg et al., 2005; Ryberg et al., 2007). The potential involvement of CB₂ and non-CB₁ and non-CB₂ receptors in central effects of CBs needs further investigation and is not discussed in the present review.

CB₁ receptors are the most abundant G-protein-coupled receptors found in the brain (Howlett et al., 2002). They are metabotropic receptors coupled to Gi/o proteins, whose activation results in inhibition of adenylyl cyclase activity and in a consequent decrease in cytosolic cAMP content, closure of Ca²⁺ channels, opening of K⁺ channels and stimulation of kinases that phosphorylate tyrosine, serine and threonine residues in proteins (Howlett et al., 1998; McAllister and Glass, 2002). CB₁ receptors are localized preferentially at the presynaptic level and, thus, it is believed that they inhibit the release of glutamate, GABA and other neurotransmitters (Schlicker and Kathmann, 2001; Wilson and Nicoll, 2002).

The localization of CB₁ receptors in the brain is consistent with the known central effects of CBs, with highest concentrations in areas involved in memory (for example, hippocampus), motor coordination (for example, the cerebellum) and emotionality (for example, prefrontal cortex) (Herkenham et al., 1990; Tsou et al., 1998a). In the dopaminergic mesolimbic system, the best known circuit involved in motivational processes (Koob, 2000; Di Chiara et al., 2004; Wise, 2004; Everitt and Robbins, 2005), average to high concentrations of CB₁ receptors are found in the terminal region, the striatum, whereas low concentrations of CB₁ receptors are found in the origin, the ventral tegmental area (VTA) (Herkenham et al., 1990; Tsou et al., 1998a). These relatively low concentrations in the VTA do not necessarily indicate that CBs do not have important actions in this area. Several lines of evidence indicate that CB₁ receptor agonists have strong modulating effects on VTA neuron activity (Szabo et al., 2002; Cheer et al., 2003; Riegel and Lupica, 2004; Melis et al., 2004a, 2004b) and that CBs can produce rewarding effects when directly injected into this structure (Zangen et al., 2006).

It should be noted that anandamide, along with a variety of other lipids, can also activate transient receptor potential vanilloid type 1 (TRPV1) vanilloid receptors (Zygmunt et al., 1999; Di Marzo et al., 2001a). However, the role of these receptor channels in the behavioural and neurochemical effects of anandamide in brain reward processes remains largely undefined (Piomelli, 2003; Solinas et al., 2006a, 2007b).

Endogenous ligands for CB receptors

In the early 1990s, anandamide (Devane et al., 1992; Di Marzo et al., 1994) and 2-arachidonoylglycerol (2-AG) (Mechoulam et al., 1995; Sugiura et al., 1995; Stella et al., 1997) were discovered and characterized as the first endogenous ligands for CB receptors. Subsequently, other possible endocannabinoids have been proposed, such as noladin ether (Hanus et al., 2001), virodhamine (Porter et al., 2002) and arachidonoyldopamine (Porter et al., 2002), but their natural occurrence and their roles are still unclear.

Anandamide and 2-AG have different structures, different biosynthesis and degradation pathways and, in addition, appear to be formed under different conditions and to be differently affected by several manipulations, including pharmacological stimulation, as reviewed elsewhere (Freund et al., 2003; Piomelli, 2003; Di Marzo et al., 2004; Fride, 2005). In addition, a recent paper has shown that anandamide inhibits the metabolism and the effects of 2-AG levels in the stiratum (Maccarrone et al., 2008). Thus, it has been proposed that anandamide and 2-AG might play different roles in physiological and pathophysiological conditions (Piomelli, 2003).

A peculiarity of the endocannabinoids, which makes them an interesting target for the discovery of new drugs, is that they are not present in vesicular stores but instead, are formed ‘on demand' and undergo rapid metabolic deactivation, so that drugs that target this system would act predominantly when and where altered levels of endocannabinoids are present (Piomelli, 2003; Piomelli et al., 2006).

Synthesis and degradation of endocannabinoids and endocannabinoid transport

Endocannabinoid signalling is terminated by a two-step process consisting of (1) transport inside the cell and (2) metabolic degradation by specific enzymes. Although this has not been established, it is believed that anandamide and 2-AG might use the same intracellular transport as the first step of degradation (Freund et al., 2003; Piomelli, 2003; Di Marzo et al., 2004; Fride, 2005). In contrast, it has been clearly demonstrated that their metabolic degradation diverges (Freund et al., 2003; Piomelli, 2003; Di Marzo et al., 2004; Fride, 2005).

Transport

Because of their lipophilic nature, anandamide and 2-AG can diffuse passively through lipid membranes. However, it appears that diffusion is accelerated by a rapid and selective carrier system that would use a facilitated diffusion mechanism (Beltramo et al., 1997; Hillard et al., 1997). This carrier might work bi-directionally and could also facilitate the release of endocannabinoids (Hillard et al., 1997). Endocannabinoid transport is the target of pharmacological tools such as AM404, VDM11, UCM707 and AM1172 that inhibit it and produce elevations in endocannabinoid levels, CB-like effects, and enhanced and prolonged effects of exogenously administered endocannabinoids (Piomelli et al., 1999; De Petrocellis et al., 2000; Lopez-Rodriguez et al., 2003; Fegley et al., 2004; Glaser et al., 2005). However, the protein(s) responsible for this transport has not yet been isolated or characterized.

Anandamide degradation

Once taken up into the cell, anandamide is broken down into arachidonic acid and ethanolamine by fatty acid amide hydrolase (FAAH), which has been molecularly cloned and characterized (Desarnaud et al., 1995; Hillard et al., 1995; Cravatt et al., 1996; Bracey et al., 2002). FAAH is widely distributed in the CNS and its distribution partially overlaps with that of CB₁ receptors, with FAAH being present mainly at the postsynaptic level and CB₁ receptors at the presynaptic level (Tsou et al., 1998b; Egertova et al., 2003). The availability of FAAH-deficient mice (Cravatt et al., 2001) and of selective FAAH inhibitors (Kathuria et al., 2003) has allowed characterization of the critical role played by this enzyme in the degradation of anandamide (Cravatt et al., 2001; Kathuria et al., 2003; Fegley et al., 2005; Solinas et al., 2006a, 2007b).

2-AG degradation

Once taken up into the cell, 2-AG is broken down into arachidonic acid and glycerol, mainly by monoacylglycerol lipase (MGL), which has been cloned and characterized (Dinh et al., 2002a, 2002b). MGL is widely distributed in the CNS and its distribution partially parallels that of CB₁ receptors (Dinh et al., 2002a). However, both MGL and CB₁ receptors appear to be situated at the presynaptic level (Dinh et al., 2002a). This localization, together with the fact that FAAH and MGL only partially overlap, suggests that anandamide and 2-AG play different roles in the modulation of neurotransmission.

Role of CB receptors in brain reward processes

CB₁ receptors appear to play an important role in brain reward processes. One long-standing line of evidence for the role for CB₁ receptors in brain reward processes is that CB₁ receptor agonists, such as the active ingredient in cannabis, THC, have rewarding effects in humans and animals (Maldonado and Rodriguez de Fonseca, 2002; Tanda and Goldberg, 2003; Gardner, 2005; Justinova et al., 2005a; Solinas et al., 2007c). The reinforcing effects of THC have been extensively reviewed elsewhere (Tanda and Goldberg, 2003; Justinova et al., 2005a). Here, we focus on recent evidence for a modulatory role of endocannabinoids on the rewarding effects of drugs of abuse, food and electric brain stimulation.

Effects of CB₁ receptor agonists on reward processes

CB₁ receptor agonists, such as THC, WIN 55,212-2 and HU-210, can facilitate the rewarding effects of drugs. For example, administration of THC or WIN 55,212-2 increases the reinforcing effects of heroin (Solinas et al., 2005), nicotine (Valjent et al., 2002) and alcohol (Gallate et al., 1999; Colombo et al., 2002). Concerning psychostimulants, one study in rats has shown that administration of WIN 55,212-2 decreased self-administration of cocaine under a fixed-ratio (FR1) schedule (Fattore et al., 1999). However, as a decrease in the number of drug injections self-administered under a FR1 schedule can be interpreted either as a decrease or an increase in reinforcing efficacy (Arnold and Roberts, 1997), definitive conclusions cannot be drawn from these experiments.

In addition to anecdotal evidence that cannabis increases appetite, especially for sweet food, in recreational cannabis smokers (Abel, 1975), several preclinical studies have shown that CB₁ receptor agonists facilitate food reward, in particular, the hedonic response to sweet food that Berridge and Robinson (1998) call ‘liking'. For example, THC increases the intake of food (Williams et al., 1998; Williams and Kirkham, 2002b) and increases the consumption of sweet solutions (Gallate et al., 1999) in rats. In addition, low doses of THC increase hedonic reactions to sucrose and decrease aversive reactions to bitter quinine solutions (Jarrett et al., 2005, 2007) and THC increases the palatability of sucrose (Higgs et al., 2003) in rats. Also, we recently found that the motivation to respond for food, as measured by break points in responding for food under a progressive-ratio schedule (Arnold and Roberts, 1997), is increased by administration of THC (Solinas and Goldberg, 2005). Interestingly, enhancement of the motivation to respond for food by THC is dependent on actual food consumption (Solinas and Goldberg, 2005), suggesting that both appetitive and consummatory aspects of food reward may involve the endocannabinoid system. Taken together, these findings provide a rationale for the clinical use of CB₁ receptor agonists such as Marinol in anorexic cancer and HIV patients (Croxford and Miller, 2003).

There are contradictory reports on the effects of CBs on brain stimulation reward. Brain stimulation reward or intracranial self-stimulation is an operant procedure where animals have to press a lever to receive a small electrical current in restricted areas of the brain (Wise, 2002; Kornetsky, 2004). Brain stimulation reward is arguably the most robust form of reinforcement and is believed to derive from the ability of electrical currents to activate, probably indirectly, the dopaminergic mesolimbic system (Wise, 2002; Kornetsky, 2004). In support of this hypothesis, drugs that activate the dopaminergic system and increase dopamine levels in the nucleus accumbens (NAcc) (namely drugs of abuse) also facilitate brain stimulation reward (increase thresholds for self-stimulation), whereas drugs that block dopamine receptors (but also opioid antagonists, see Schaefer, 1988) reduce thresholds for self-stimulation (Wise, 2002; Kornetsky, 2004). Concerning, CB₁ receptor agonists, Gardner and colleagues found that THC facilitates brain stimulation reward (Gardner et al., 1988; Lepore et al., 1996), whereas other investigators found no effects of the synthetic agonists CP55,940 or AMG-3 (Arnold et al., 2001; Antoniou et al., 2005) and some investigators have found a reduction of brain stimulation reward with CB₁ receptor agonists (Kucharski et al., 1983; Vlachou et al., 2005, 2007). These discrepancies, which are likely due to procedural differences, remain to be resolved.

A caveat of all experiments with directly acting CB₁ receptor agonists is that, for several reasons, these drugs do not provide a realistic picture of the physiological role of the endocannabinoids. First, anandamide is a partial agonist (Mackie et al., 1993), whereas synthetic CB₁ receptor agonists (not THC) are often full agonists and have higher affinities for CB₁ receptors (Childers, 2006). Second, anandamide and 2-AG have very short half-lives (Deutsch and Chin, 1993; Dinh et al., 2002a), whereas THC and synthetic CB₁ receptor agonists have relatively long half-lives. Finally, systemic injections of these compounds result in the activation of all brain areas containing CB₁ receptors, whereas physiological activation of endocannabinoid synthesis and release is likely to be region, neuron or even synapse specific (Piomelli, 2003; Di Marzo et al., 2004). The availability of mice genetically deprived of CB₁ receptors in a tissue-specific manner (Marsicano et al., 2003; Monory et al., 2006, 2007) may help address this possibility.

Effects of CB₁ receptor antagonists on reward processes

Assessment of the roles the endocannabinoid system plays in brain reward processes was greatly facilitated by the discovery of selective CB₁ receptor antagonists/inverse agonists such as rimonabant (SR141716A) (Rinaldi-Carmona et al., 1994) and AM251 (Gatley et al., 1996). CB₁ receptor antagonists decrease the rewarding effects of a wide variety of abused drugs under certain conditions. For example, the rewarding effects of opioids are generally decreased in both intravenous self-administration (Navarro et al., 2001; Caille and Parsons, 2003; De Vries et al., 2003; Solinas et al., 2003) and conditioned place preference procedures (Chaperon et al., 1998; Mas-Nieto et al., 2001; Navarro et al., 2001; Singh et al., 2004). There have also been reports that rimonabant and AM251 reduce the rewarding effects of methamphetamine (Vinklerova et al., 2002), alcohol (Arnone et al., 1997; Colombo et al., 1998; Gallate and McGregor, 1999; Freedland et al., 2001; Cippitelli et al., 2005; Gessa et al., 2005; Economidou et al., 2006; Lallemand and De Witte, 2006) and nicotine (Cohen et al., 2002; Le Foll and Goldberg, 2004; Forget et al., 2005).

The rewarding effects of cocaine are reduced in a very specific way by CB₁ receptor antagonists. CB₁ receptor antagonists do not generally alter self-administration of cocaine under low fixed-ratio schedules (Tanda et al., 2000; De Vries et al., 2001; Filip et al., 2006) or conditioned place preference procedures (Chaperon et al., 1998), but AM251 has been found to significantly reduce self-administration of cocaine under progressive-ratio schedules (Xi et al., 2007) and rimonabant prevents relapse to cocaine-induced and cue-induced cocaine-seeking behaviour (De Vries et al., 2001). This suggests that the appetitive and conditioned effects of cocaine, but not its direct reinforcing effects, depend on CB₁ receptor activation.

The effects of rimonabant on opioid reward may be mediated primarily in the NAcc, as blockade of CB₁ receptors in this area reduces heroin self-administration (Caille and Parsons, 2006). On the other hand, the modulation of ethanol reward by the CB system appears to take place both in the NAcc (Caille et al., 2007) and in the prefrontal cortex (Hansson et al., 2007). The brain sites where CBs act to alter the rewarding effects of nicotine and psychostimulants are not known at present.

Drugs of abuse share the ability to elevate extracellular levels of dopamine in the shell of the NAcc, as measured by in vivo microdialysis, and this effect is believed to play an important role in their reinforcing effects (Koob, 2000; Di Chiara et al., 2004; Wise, 2004; Everitt and Robbins, 2005). CB₁ receptor antagonists have been shown to block the elevations of accumbal dopamine levels induced by administration of nicotine or ethanol (Cohen et al., 2002), but not by administration of heroin (Tanda et al., 1997; Caille and Parsons, 2003), morphine or cocaine (Caille and Parsons, 2006). Transient surges of dopamine in the NAcc, as measured by cyclic voltammetry, are also produced by drugs of abuse (Cheer et al., 2007) and are believed to be involved in drug seeking (Phillips et al., 2003). Interestingly, the transient increases in dopamine produced by administration of nicotine, ethanol and cocaine in the shell of the NAcc of freely moving rats are all blocked by CB₁ receptor antagonists (Cheer et al., 2007).

Consistent with a role for endocannabinoids in the rewarding effects of food and in the regulation of appetite and food intake (Di Marzo and Matias, 2005), blocking endocannabinoid tone with CB₁ receptor antagonists reduces intake of food and sweet solutions (Arnone et al., 1997; Colombo et al., 1998; Simiand et al., 1998; Freedland et al., 2001; McLaughlin et al., 2003; Thornton-Jones et al., 2005, 2006, 2007). Also, injection of rimonabant within 24 h of birth completely prevents milk intake and causes almost 100% mortality in mouse pups (Fride et al., 2001, 2003). The motivational effects of food measured by a progressive-ratio schedule of food reinforcement (Solinas and Goldberg, 2005) and the appetitive aspects of food reward (Thornton-Jones et al., 2005) are significantly reduced by rimonabant in rats, indicating that some aspects of food intake regulation involve reward and motivational processes. In addition, AM251 decreases hedonic reactions to sucrose and increases aversive reactions to quinine (Jarrett et al., 2007). Consistent with these preclinical findings, rimonabant has been found to be effective in the clinical treatment of obesity (Padwal and Majumdar, 2007), although the clinical efficacy of this agent appears to be primarily due to its ability to alter peripheral lipid metabolism, rather than to reduce food intake (Cota et al., 2003a, 2003b).

As in the case of CB₁ receptor agonists, the effects of CB₁ receptor antagonists on brain stimulation reward are somewhat controversial. Rimonabant has been shown to increase the threshold for brain stimulation reward in some studies (Deroche-Gamonet et al., 2001; De Vry et al., 2004) or to produce no change in brain stimulation reward thresholds in other studies (Arnold et al., 2001; Vlachou et al., 2003, 2005, 2007; Antoniou et al., 2005).

It should be noted that both rimonabant and AM251 act as inverse agonists at CB₁ receptors and, therefore, can produce effects that are opposite to those of CB₁ receptor agonists even in the absence of endogenous ligands (Landsman et al., 1997; Pan et al., 1998). Thus, it is possible that certain effects of these agents might be due to their inverse agonist properties. The development of new antagonists devoid of inverse agonist properties (Pertwee, 2005; Bergman et al., 2008) will undoubtedly help to provide a more definitive description of the effects of CB₁ receptor blockade on brain reward processes.

Effects of genetic ablation of CB₁ receptors on brain reward processes

Mice genetically engineered to lack CB₁ receptors (Ledent et al., 1999; Zimmer et al., 1999; Marsicano et al., 2002) do not show dramatic changes in body weight, food consumption or fertility (Ledent et al., 1999; Zimmer et al., 1999; Marsicano et al., 2002), suggesting that CB₁ receptors modulate rather than mediate basic reward functions or that other systems can compensate for their absence.

By using CB₁-null mice, the role of CB₁ receptors in the rewarding effects of drugs of abuse has been confirmed. For example, in these mutant mice morphine is not self-administered (Ledent et al., 1999; Cossu et al., 2001), does not induce conditioned place preferences (Martin et al., 2000) and does not elevate dopamine levels in the NAcc (Mascia et al., 1999). Also, the rewarding effects of alcohol are reduced in CB₁-null mice, as demonstrated by data showing that CB₁-null mice do not develop conditioned place preference with this drug (Houchi et al., 2005; Thanos et al., 2005) and that they do not prefer it over water in a two-bottle free-choice paradigm (Wang et al., 2003; Naassila et al., 2004; Thanos et al., 2005; Lallemand and De Witte, 2006). However, another study (Racz et al., 2003) reported that CB1-null mice show slight and short-lasting decreases in preference for ethanol. Development of conditioned place preferences with cocaine is unaltered in CB₁-null mice (Martin et al., 2000; Houchi et al., 2005). On the other hand, development of cocaine self-administration behaviour under fixed-ratio schedules of reinforcement in CB₁-null mice was reported to be unaltered when mice were restrained (Cossu et al., 2001), but was reduced in freely moving mice (Soria et al., 2005). In addition, cocaine self-administration was significantly reduced under a progressive-ratio schedule of self-administration in CB₁-null mice (Soria et al., 2005). These contrasting results highlight the fact that, when working with genetically modified mice such as CB1-null mice, not only methodological differences but also differences in the genetic background may result in very different and sometimes contrasting behavioural outputs (Bucan and Abel, 2002; Bailey et al., 2006).

It is interesting to note that CB₁-null mice show normal elevations in dopamine levels in the NAcc following administration of cocaine (Soria et al., 2005), but no elevations following administration of morphine (Mascia et al., 1999) or ethanol (Hungund et al., 2003), compared with wild-type controls. Finally, CB₁-null mice do not develop conditioned place preferences to nicotine (Castane et al., 2002), but they self-administer the drug like wild-type controls (Cossu et al., 2001). It remains to be seen whether increasing the effort needed to obtain nicotine, as with cocaine (Soria et al., 2005), could reveal a role of CB₁ receptors in some aspects of nicotine reinforcement as suggested by the results with CB₁ receptor antagonists (Cohen et al., 2005).

CB₁-null mice have also provided evidence for the involvement of the endocannabinoid system in food reward. For example, CB₁-null mice eat less than their wild-type control littermates after food restriction (Di Marzo et al., 2001b). Moreover, CB₁-null mice respond less for sucrose in a two-lever paradigm, have lower break points under progressive-ratio schedules of sucrose delivery (Sanchis-Segura et al., 2004) and show less preference for sucrose over water in a two-bottle free-choice procedure (Poncelet et al., 2003; Sanchis-Segura et al., 2004). Genetic ablation of CB₁ receptors results in small reduction in body weight, reduction in adiposity and resistance to diet-induced obesity (Ravinet Trillou et al., 2004). However, as in the case of CB₁ receptor antagonists, these effects appeared to be related more to increased metabolic energy consumption than to differences in rewarding effects of food or hypophagia (Cota et al., 2003a, 2003b; Ravinet Trillou et al., 2004). Interestingly, in the study by Fride et al. (2003), administration of rimonabant in mice pups lacking CB₁ receptors still induced a decrease in milk intake and survival rate, suggesting that some of the effects of CBs on food intake may be mediated by still uncharacterized CB receptors.

To date, no study has investigated the effects of CB₁ receptor deletion on brain stimulation reward. On the other hand, it should be noted that CB₁-null mice show increased anhedonia after chronic mild stress (Martin et al., 2002), a measure of reduced activity of the reward system and a model of depression (Willner, 2005), further supporting a role for the endocannabinoid system in brain reward functions.

Role of anandamide in brain reward processes

Effects of systemically administered anandamide on reward processes

We have recently shown that anandamide and the metabolically stable anandamide analogue, methanandamide, can serve as powerful reinforcers of self-administration behaviour in non-human primates when injected intravenously (Justinova et al., 2005b) and they both increase extracellular levels of dopamine in the shell of the NAcc when injected intravenously, a neurochemical effect common to many rewarding stimuli (Solinas et al., 2006a). In addition, anandamide and methanandamide produce THC-like discriminative effects in two-lever choice drug discrimination procedures (Wiley et al., 1997; Solinas et al., 2006b, 2007b). Importantly, these effects of anandamide appear to be mediated by CB₁, but not TRPV1, receptors and are potentiated by pharmacological inhibition of FAAH activity, but not by administration of the anandamide transport inhibitor AM404 (Solinas et al., 2006a, 2007b).

In addition to its direct reinforcing effects, systemically administered anandamide can potentiate the effects of food reward. Systemic administration of anandamide was found to increase food intake (Hao et al., 2000) and induce overeating in sated rats (Williams and Kirkham, 1999, 2002a). Also, hypothalamic (Jamshidi and Taylor, 2001) or intra-shell injections of anandamide into the NAcc increase food intake (Mahler et al., 2007). Similarly to THC, anandamide appears to increase food intake, at least in part, by increasing the palatability of food, as measured by changes in patterns of sucrose drinking (Higgs et al., 2003). The effects of systemically administered anandamide on food palatability, as measured by hedonic reactions to food, appear to be mediated by the more dorsal part of the shell of the NAcc (Mahler et al., 2007).

We are not aware of any study investigating the effects of systemically administered anandamide on brain stimulation reward.

Effects of abused drugs, food or electrical brain stimulation on brain anandamide levels

Although measurements of the effects of systemic or intracranial injections of anandamide provide useful information on the functions of the endocannabinoid system, to provide support for a role of neurally released anandamide in brain reward processes it is also important to measure anandamide released in the brain by different abused drugs, by food and by electrical brain stimulation. Release of neurotransmitters such as dopamine or glutamate can be readily measured by microdialysis techniques, but only a few studies have employed microdialysis techniques to measure extracellular brain levels of endocannabinoids (Giuffrida et al., 1999; Caille et al., 2007). Thus, most information on the modification of endocannabinoid levels comes from measurements of tissue levels in different brain areas. Measuring tissue level of anandamide has the limitation that only one time point can be established at a time, limiting information on the pattern of endocannabinoid release.

Using tissue levels, it has been demonstrated that chronic administration of several drugs of abuse leads to region-specific increases in anandamide levels. For example, when administered chronically, THC decreases levels of anandamide in the striatum (Di Marzo et al., 2000), ethanol decreases levels of anandamide in the midbrain but not in the striatum (Gonzalez et al., 2002), nicotine decreases levels of anandamide in the striatum but not in the midbrain (Gonzalez et al., 2002), and cocaine and morphine do not alter anandamide levels, either in the striatum or midbrain (Gonzalez et al., 2002; Vigano et al., 2003). However, it is difficult to determine from these studies whether measured levels of endocannabinoids reflected the consequences of chronically administered drug or withdrawal.

Vigano et al. (2004) compared the effects of chronic versus acute administration of morphine on endocannabinoid levels in the brain. They found that acute administration of morphine increased anandamide levels in the striatum, whereas chronic treatment with the drug failed to do so. In addition, they found that chronic treatment with morphine did not alter the ability of a challenge dose of morphine to increase anandamide levels in the striatum; that is, repeated administration of morphine did not induce sensitization or tolerance to this effect (Vigano et al., 2004). Thus, chronic administration of drug followed by withdrawal, chronic administration of drug followed by an acute drug challenge and acute administration of drug can lead to very different changes in brain anandamide levels. Such profiles of release may indicate that anandamide is released in response to relevant changes in homoeostasis but not when an adaptive response has already occurred. This suggests that it may be preferable to assess the role of anandamide in drug reward processes by measuring changes in anandamide levels directly produced by administration of the drug rather than changes in anandamide levels produced after chronic drug exposure, when alterations may be opposite to those after acute administration and may reflect biochemical alterations associated with withdrawal. This is supported by a recent study in which Caille et al. (2007) employed microdialysis techniques to monitor extracellular endocannabinoid levels in the brain during active self-administration of drugs and found that anandamide levels were elevated during heroin self-administration, consistent with findings by Vigano et al. (2004). Interestingly, extracellular levels of anandamide were not altered during cocaine or ethanol self-administration (Caille et al., 2007). Differences in the molecular target and mechanism of actions of drugs of abuse could account for the different effects on accumbal levels of endocannabinoids.

There is only limited information on the effects of food reward on brain levels of anandamide. In one study, anandamide levels in the NAcc, but not in the hypothalamus, increased after 24-h food deprivation but not during active feeding (Kirkham et al., 2002). This suggests that appetitive aspects of food reward involve release of anandamide, but consummatory aspects of reward do not. However, it is also possible that, as in the case of dopamine (Di Chiara et al., 2004), natural rewards produce much lower increases in anandamide release than pharmacological rewards and that increases in anandamide occur with food reward but are lower than the limit of detection with current techniques.

Administration of addictive drugs or exposure to food increases dopamine levels in the NAcc (Koob, 2000; Di Chiara et al., 2004; Wise, 2004; Everitt and Robbins, 2005) and activation of dopamine D₂ receptors in this region increases extracellular levels of anandamide (Giuffrida et al., 1999). This suggests that increases in anandamide levels in the NAcc after administration of abused drugs or after eating may be secondary to increases in dopamine levels. However, Caille et al. (2007) have recently shown that drugs such as cocaine produce large increases in dopamine levels but do not alter anandamide levels in the NAcc.

Finally, to the best of our knowledge, the release of CBs during electrical brain self-stimulation remains to be investigated. However, indirect pharmacological evidence indicates that, at least in the VTA, 2-AG and not anandamide is released after application of brain stimulation that would sustain self-administration in rats (Melis et al., 2004a; Pillolla et al., 2007).

Effects of inhibiting anandamide degradation by FAAH on reward processes

As anandamide is quickly degraded, studying its normal roles can be facilitated by blocking the mechanisms of degradation and, thus, magnifying and prolonging its actions. Importantly, increases in anandamide levels would presumably be obtained only in those brain regions where anandamide is released. As previously described, the intracellular step of anandamide degradation is mediated mainly by FAAH activity (Freund et al., 2003; Piomelli, 2003; Di Marzo et al., 2004; Fride, 2005). Thus, the specific contribution of anandamide to reward functions can be more readily investigated when FAAH is inhibited by drugs such as URB597 (Kathuria et al., 2003). It should be noted that URB597 has no rewarding effects by itself in conditioned place preference paradigms, does not produce THC-like discriminative effects (Gobbi et al., 2005) and does not alter dopamine levels in the shell of NAcc (Solinas et al., 2006a, 2007b) in rats.

The effects of FAAH inhibition on drug reward have been the focus of several recent papers. One study found that the directly acting CB receptor agonists such as THC and WIN 55,212-2 increased the reinforcing effects of heroin, but FAAH inhibition by URB597 had no significant effect, suggesting that heroin-induced release of anandamide is not necessary for opiate reward (Solinas et al., 2005). Another study found that administration of URB597 increased ethanol preference and intake in a two-bottle free-choice procedure (Blednov et al., 2007). In a recent study, we found that nicotine can produce THC-like discriminative effects when given in combination with URB597, indicating that nicotine does release anandamide and that, under conditions of FAAH inhibition, can actually produce THC-like behavioural effects (Solinas et al., 2007a). However, it has to be established whether the reinforcing effects of nicotine itself, rather than its ability to produce THC-like discriminative effects, are altered by stimulation of the endocannabinoid system, as another study found that the discriminative effects of nicotine are not altered by administration of URB597 or combinations of URB597 plus anandamide (Zaniewska et al., 2006). Finally, one study has found that the effects of ethanol on the mesolimbic dopaminergic system are reduced by URB597 in mutant FAAH-null mice (Perra et al., 2005); however, the behavioural relevance of these neurochemical alterations remains to be established as mutant FAAH-null mice show a higher preference for ethanol (Basavarajappa and Hungund, 2005; Blednov et al., 2007).

The effects of URB597 on food reward have not yet been systematically studied, but one study investigated the effects of administration of URB597 on brain stimulation reward. FAAH inhibition increased the threshold for intracranial self-stimulation, indicating a decrease in the reinforcing effects of brain stimulation (Vlachou et al., 2006). However, these effects were obtained at doses of URB597 3–10 times higher than those required to produce significant inhibition of FAAH (Kathuria et al., 2003) and dramatic potentiation of the effects of exogenously administered anandamide (Solinas et al., 2006a, 2007b) and, thus, the interpretation of these results is difficult and may be related to loss of selectivity at high doses.

Effects of genetic deletion of FAAH on reward processes

In parallel with results obtained with URB597, the availability of FAAH-null mice (Cravatt et al., 2001) has further expanded possibilities of assessing anandamide's role in brain reward processes. As with CB₁-null mice, FAAH-null mice do not show dramatic changes in body weight, food consumption or fertility (Cravatt et al., 2001), suggesting, again, that the endocannabinoid system plays only a modulatory role in basic reward functions.

There are only a few studies on the effects of genetic ablation of FAAH on reward processes. Among drugs of abuse, information is available only regarding ethanol. In one study, it was found that ethanol consumption and preference increase in female, but not male, FAAH-null mice (Basavarajappa and Hungund, 2005), whereas in another study it was found that ethanol consumption and preference increase in both female and male FAAH-null mice (Blednov et al., 2007). Interestingly, conditioned place preferences were similarly obtained in males, but not in females, when ethanol was injected intraperitoneally (Blednov et al., 2007), indicating that the consequences of FAAH deletion on ethanol's rewarding effects may be complex. To date, no study has investigated the effects of FAAH deletion on food or brain stimulation reward. The finding that FAAH-null mice show a slightly antidepressant-like phenotype (Naidu et al., 2007) is consistent with a role for anandamide in mood regulation (Kathuria et al., 2003; Gobbi et al., 2005).

Role of 2-AG in brain reward processes

Although 2-AG was discovered only 3 years after anandamide, much less is known about the role of this endogenous compound in reward processes and most knowledge comes from indirect evidence. Hopefully, the development of new tools such as drugs selectively targeting enzymes involved in 2-AG formation and degradation or genetically modified mice will help to characterize the physiological and physiopathological roles of this endocannabinoid.

Effects of systemically administered 2-AG on reward processes

Only one study has reported behavioural effects related to reward processes for systemically administered 2-AG. Kirkham et al. (2002) found that 2-AG directly injected into the shell of the NAcc increased food intake.

Effects of abused drugs, food or electrical brain stimulation on 2-AG levels

The best evidence for a role of 2-AG in brain reward processes comes from studies that measured changes in brain levels of 2-AG produced by drugs, food or brain stimulation (with all the limitations of these measurements described for anandamide, section ‘Effects of inhibiting anandamide degradation by FAAH on reward processes').

Chronic administration of many abused drugs alters levels of 2-AG in brain areas involved in reward. For example, it has been shown that chronic administration of (1) THC decreases tissue levels of 2-AG in the striatum (Di Marzo et al., 2000), (2) ethanol decreases tissue levels of 2-AG in the midbrain but not in the nucleus accumbens (Gonzalez et al., 2002), (3) nicotine or (4) cocaine do not alter 2-AG levels either in the midbrain or the striatum (Gonzalez et al., 2002) and (5) morphine decreases 2-AG levels in the striatum without altering 2-AG levels in the striatum (Vigano et al., 2003). Moreover, Vigano et al. (2004) found that acute injection of morphine decreased tissue levels of 2-AG in the striatum more than chronic administration and that a challenge injection of morphine, after a 2-week period of withdrawal, in rats chronically treated with morphine, had effects similar to those of an acute injection. On the other hand, no changes in extracellular levels of 2-AG were found in the NAcc of rats self-administering heroin, but there was an increase in 2-AG levels in rats self-administering ethanol (Caille et al., 2007). The general picture that can be drawn from these data is that 2-AG is involved in some consequences of repeated administration of abused drugs but it may be more important for the rewarding effects of ethanol than for the rewarding effects of other abused drugs, such as heroin, cocaine or nicotine.

One study investigated the effects of manipulation of food intake and food deprivation on brain levels of 2-AG and found that tissue levels of 2-AG were increased by fasting but were decreased by feeding (Kirkham et al., 2002), indicating that 2-AG would be more important for appetitive aspects than consummatory aspects of food reward.

Finally, no study has investigated brain levels of 2-AG during electrical brain self-stimulation reward. However, one study has shown that electrical stimulation of the medial prefrontal cortex produces a decrease in presynaptic glutamate release that might be explained by release of 2-AG (Melis et al., 2004a). Future studies are needed to verify that these neurochemical results have behavioural relevance.

Endocannabinoid transport

As noted in section ‘Synthesis and degradation of endocannabinoids and endocannabinoid transport', endocannabinoid transport remains one of the less understood features of the endocannabinoid system. Although endocannabinoid transport was proposed long ago and drugs thought to inhibit the transport of endocannabinoids through the cell membrane have been available for a decade (Beltramo et al., 1997; Hillard et al., 1997), the molecular entity(ies) mediating this transport is still unknown and the nature of endocannabinoid transport is debated (Glaser et al., 2003, 2005; Moore et al., 2005). However, drugs such as AM404, VDM11, UCM707 and AM1172 (Piomelli et al., 1999; De Petrocellis et al., 2000; Lopez-Rodriguez et al., 2003; Fegley et al., 2004; Glaser et al., 2005), which are thought to block endocannabinoid uptake into the cell, can be important tools for investigating the role of the endocannabinoid system in brain reward processes.

Unlike the FAAH inhibitor URB597, the uptake inhibitor AM404 has some rewarding effects by itself using a conditioned place preference procedure, although its rewarding effects measured with this procedure are clearly smaller than those of the direct CB receptor agonist WIN 55,212-2 (Bortolato et al., 2006). Interestingly, AM404 induced conditioned place preferences in rats at a dose that did not significantly increase tissue levels of anandamide or 2-AG in the brain areas investigated (Bortolato et al., 2006). On the other hand, AM404 does not produce THC-like discriminative effects and does not alter dopamine levels in the shell of NAcc in rats (Solinas et al., 2007b). It is, thus, surprising that the effects of AM404 or other uptake inhibitors on drug, food or brain stimulation reward have seldom been studied. To our knowledge, only one study has investigated the effects of uptake inhibition on drug reward and in that study AM404 did not increase heroin self-administration under a progressive-ratio schedule and, instead, produced a small decrease (Solinas et al., 2005). In that study, CB₁ receptor agonists such as THC and WIN 55,212-2 potentiated heroin self-administration and URB597 had no effect on heroin self-administration. One interpretation of these results is that enhancement of endocannabinoid tone has either an inhibitory or a neutral effect on opioid reward, whereas broad activation of CB₁ receptors has a facilitatory effect on opioid reward.

The effects of blockade of anandamide transport on food intake and electrical brain stimulation reward have been investigated using the uptake inhibitors VDM-11 and OMDM-2. It was found that VDM-11 does not alter food intake in rats (Chambers et al., 2004) and that high doses (30 mg/kg i.p.) of OMDM-2 increase the threshold for electrical brain self-stimulation reward (Vlachou et al., 2006).

Obviously, more research is needed to interpret the effects of inhibition of endocannabinoid uptake on brain reward processes and to understand its role in the regulation of the activity of the endocannabinoid system. For example, we have recently found that some effects of anandamide, such as its ability to produce THC-like discriminative effects and to increase dopamine levels in the NAcc, are potentiated by URB597 but not by AM404 (Solinas et al., 2007b), suggesting that regional differences may exist in the relative roles of uptake inhibition and FAAH inhibition.

Summary

Involvement of the endocannabinoid system in drug reward

The importance of the endocannabinoid system in drug reward varies greatly depending on the drug studied, with a pronounced role for the endocannabinoid system in opioid and ethanol reward and more subtle roles in nicotine and psychostimulant reward.

It is clear that activity of CB₁ receptors is important for opioids to produce maximal reinforcing effects (Fattore et al., 2005; Vigano et al., 2005). Although opioids release anandamide in the NAcc (Caille et al., 2007), the importance of this release remains unclear given that compounds that increase brain concentrations of anandamide and prolong anandamide's actions do not alter heroin self-administration (Solinas et al., 2005). Thus, interactions between CB₁ receptors and mu-opioid receptors (Berrendero et al., 2003; Kathmann et al., 2006) may be more important than endocannabinoid release in the modulation of opioid effects.

It is also clear that activity of CB₁ receptors is important for ethanol to produce maximal reinforcing effects (Mechoulam and Parker, 2003). Although it has been reported that ethanol self-administration increases 2-AG but not anandamide levels in the NAcc (Caille et al., 2007), FAAH-null mice show higher preferences for ethanol but no higher ethanol-induced conditioned place preferences (Blednov et al., 2007). Thus, although release of endocannabinoids appears to participate in ethanol reward, the relative importance of 2-AG and anandamide needs to be established.

CB₁ receptors may be critically involved in the rewarding effects of nicotine (Castane et al., 2002; Viveros et al., 2005). The role of release of endocannabinoids is less clear, however, as we have reported that URB597 unmasks some THC-like discriminative effects of nicotine (Solinas et al., 2007a), whereas another study found that URB597 did not potentiate the discriminative effects of nicotine itself (Zaniewska et al., 2006). Importantly, unpublished results from company-sponsored clinical trials suggest that blockade of CB₁ receptors may be effective in promoting smoking cessation and abstinence (http://en.sanofi-aventis.com/press/ppc_1960.asp). Finally, the few studies available have not supported a role of CB₁ receptors in the primary reinforcing effects of cocaine. In addition, in vivo microdialysis studies have found that neither anandamide nor 2-AG is formed in the shell of the NAcc during active self-administration of cocaine (Caille et al., 2007). However, CB₁ receptors appear to be required for the incentive motivational effects of cocaine, as measured by self-administration under progressive-ratio schedules (Soria et al., 2005) and by reinstatement of extinguished cocaine self-administration (De Vries et al., 2001).

Involvement of the endocannabinoid system in food reward

Food reward also clearly depends on CB₁ receptors and almost every aspect of food reward is affected by activation or inactivation of CB₁ receptors (Di Marzo and Matias, 2005). Again, data from clinical trials appear to support the preclinical findings that CB₁ blockade is effective in promoting weight loss, although the peripheral and hormonal effects of rimonabant may be more important that the central effects on reward (Padwal and Majumdar, 2007). However, the involvement of released endocannabinoids may be limited to appetitive aspects of food reward, as concentrations of both anandamide and 2-AG increase in the NAcc during fasting but not during consumption of food. In addition, inhibitors of anandamide uptake do not increase food intake (Chambers et al., 2004), further indicating that endogenous anandamide, at least, may be insufficient to drive food intake.

Involvement of the CB system in electrical brain stimulation reward

Brain stimulation reward remains the most obscure aspect of CB-related reward. In contrast to other neurotransmitter systems involved in brain stimulation reward functions, such as dopaminergic and opioid systems (Kornetsky, 2004), the effects of endocannabinoid system alterations in different studies have been contradictory and, in fact, activation of the endocannabinoid system appears to reduce the rewarding effects of electrical brain stimulation. Further studies are needed, but it is worth noting that in some instances dopaminergic systems and CB systems appear to have opposite antagonistic effects. For example, dopamine D₂ receptor activation in the dorsal striatum releases anandamide, which might act to modulate or counterbalance the effects of dopamine (Giuffrida et al., 1999). Also, glutamate release in the VTA activates dopaminergic neurons and, at the same time, leads to the release of 2-AG that in turn reduces glutamate release (Melis et al., 2004a, 2004b). On the other hand, CB agonists induce release of dopamine in the shell of the NAcc (Tanda et al., 1997; Solinas et al., 2006a) and have rewarding effects when administered locally both in the NAcc and the VTA (Zangen et al., 2006). Thus, it is possible that, depending on the balance between endocannabinoids and dopamine and the intensity of stimulation of the region, the systems facilitate or oppose each other. This could be a mechanism for fine-tuning of dopaminergic activity. As electrical brain stimulation is a very strong excitatory stimulus, it is possible that the endocannabinoid system acts to counteract and oppose such stimulation.

Future directions

The endocannabinoid field has been in active expansion since the early 1990s and promises to continue expanding at a rapid pace. The intellectual and financial investments in the field are providing new tools that will allow scientists working in the field to understand better the physiological and pathophysiological roles of the endocannabinoids. Among the innovations that will improve our knowledge are (1) improvement of tools already present, such as more selective FAAH and anandamide transport blockers, compounds that do not pass the blood–brain barrier and conditional and region-specific CB₁- and FAAH-null mice; (2) the molecular identification of the components comprising the endocannabinoid transport system; (3) availability of potent and selective MGL blockers that are active in vivo and of MGL-null mice and (4) availability of pharmacological and genetic tools to regulate the synthesis of endocannabinoids.

Conclusions

The involvement of the endocannabinoid system in brain reward processes can be inferred from at least four lines of evidence: (1) CB₁ receptors are highly expressed in areas of the brain involved in reward functions such as the mesolimbic dopaminergic system; (2) CB₁ receptor agonists such as THC produce reinforcing effects; (3) pharmacological or genetic activation of CB₁ receptors facilitates, whereas pharmacological or genetic inactivation of CB₁ receptors antagonizes, several types of brain reward functions and (4) levels of endocannabinoids in brain areas such as the mesolimbic dopamine system are altered following manipulations that involve reward processes. The involvement of various components of the endocannabinoid system may differ from one type of rewarding stimulus to another. Notwithstanding these differences and the need for further investigations, the data available strongly suggest that the endocannabinoid system plays an important role in brain reward processes.

Abbreviations

2-AG

2-arachidonoylglycerol

CB

cannabinoid

FAAH

fatty acid amide hydrolase

NAcc

nucleus accumbens

THC

delta-9-tetrahydrocannabinol

TRPV1

transient receptor potential vanilloid type 1

VTA

ventral tegmental area