In animals, the intensity of somatic withdrawal is quantified by scoring the presence or severity of several physical signs for 10–30 min immediately following precipitated withdrawal, or every 6–9 h for several days following spontaneous withdrawal (Maldonado et al., 1996). To facilitate the quantification of the withdrawal syndrome, Gellert and Holtzman (1978) developed a weighted scale consisting of graded symptoms including percentage of weight loss, number of escape jumps, number of wet dog shakes, number of abdominal constrictions, and checked signs including diarrhea, facial fasciculations/teeth chattering, swallowing, salivation, chromodacryorrhea, ptosis, abnormal posture, erection/ejaculation/genital grooming, and irritability. In assessing the ability of pharmacological cannabinoid manipulations to alleviate the intensity of somatic withdrawal, most studies use a variation of this scale to determine whether individual signs or a global rating of withdrawal has significantly decreased.

A review of the literature reveals that activation of the cannabinoid system acutely prior to withdrawal, or chronic inhibition of the system during the development of dependence, is most effective in reducing withdrawal severity (see Table 1 for specific agonists and antagonists and original references). In the majority of cases, acute treatment with a cannabinoid agonist (in particular agonists of the CB₁ receptor) immediately prior to the induction of withdrawal in highly dependent animals (whether precipitated or spontaneous) is able to attenuate several aspects of the syndrome. Notably, a reduction in the incidence of escape jumps, paw tremors, weight loss, and diarrhea were most commonly reported in rodents (see Table 1 for original references). Exceptions to the reductions in withdrawal occurred only when the agonist (Δ⁹THC) was administered chronically during the development of morphine dependence (Cichewicz and Welch, 2003; Gerak and France, 2016), when rats were made acutely dependent on morphine prior to withdrawal (Hine et al., 1975b), or when the cannabinoid tested, cannabidiol, was not a CB₁ receptor agonist (Hine et al., 1975d,e; Chesher and Jackson, 1985).

Somewhat contradictory, CB₁ receptor knock-out (KO) mice show reduced somatic opioid withdrawal and cannabinoid antagonists (rimonabant and AM251) interfere with somatic opioid withdrawal (see Table 1). However, contrary to the cannabinoid agonists, antagonists were most effective in reducing withdrawal when delivered chronically (not acutely) during the development of opiate dependence. Indeed, when given chronically prior to precipitating withdrawal with naloxone, rimonabant and AM251 reliably reduced withdrawal severity (Rubino et al., 2000; Mas-Nieto et al., 2001; Trang et al., 2006). In contrast, when delivered acutely, antagonists were either without effect (Navarro et al., 1998; Mas-Nieto et al., 2001; Trang et al., 2006), or their ability to attenuate withdrawal was markedly reduced (Trang et al., 2006). In a case of spontaneous withdrawal (Navarro et al., 1998), acute rimonabant treatment was even found to precipitate withdrawal in morphine-dependent rats, however, this finding was not replicated when tested in mice (Lichtman et al., 2001). Ultimately, these findings, along with the results from the agonist studies, suggest that the endocannabinoid system plays an important role in the development of somatic opioid dependence and activation of the system during withdrawal, or chronic blockade during opioid dependence, can mitigate some of its adverse effects.

Unfortunately, CB₁ receptor (CB₁R) agonists and inverse agonists/antagonists are known to produce undesirable side effects (e.g., psychoactivity, depression) which limit their therapeutic potential (Moreira et al., 2009). Therefore, it is fortunate that non-psychoactive treatments (including Fatty Acid Amide Hydrolase [FAAH] and monoacylglycerol lipase [MAGL] inhibitors) which act to enhance endogenous cannabinoid tone have also been effective in alleviating somatic withdrawal (see Table 1 for specific agents and original references). While inhibition of the MAGL enzyme (which elevates endogenous 2-arachidonoyl glycerol [2-AG]) produced the most robust effects (Ramesh et al., 2011; Gamage et al., 2015), inhibitors or KO mice of the catabolic FAAH enzyme (which elevates endogenous anandamide [AEA]) reduced a subset of withdrawal symptoms in most cases (Ramesh et al., 2011; Shahidi and Hasanein, 2011), whereas the AEA transport inhibitor, AM404, was without effect (Del Arco et al., 2002). However, unlike FAAH inhibitors, MAGL inhibitors have been found to produce cannabimimetic side effects (e.g., hypomotility, hyperreflexia; Long et al., 2009) and can lead to the development of dependence and tolerance with repeated administration (Schlosburg et al., 2010). In light of this, Ramesh et al. (2013) and Gamage et al. (2015) tested the combinations of low doses of FAAH and MAGL inhibitors, or dual FAAH/MAGL inhibitors, for their effectiveness in reducing withdrawal maximally without additional side effects. Indeed, this combination of catabolic enzyme inhibitors proved to be highly effective in reducing withdrawal (including jumping, paw flutters, head shakes, diarrhea, and weight loss) but was absent of adverse effects. Consequently, when considering pharmacological interventions that may aid in the treatment of somatic aspects of opiate withdrawal, dual FAAH/MAGL inhibition (at low doses) is most promising.

In animals, affective opioid withdrawal can be measured using a number of motivational paradigms including the conditioned place aversion (CPA), intracranial self-stimulation, and operant responding for food (Maldonado et al., 1996). In evaluating the role of the endocannabinoid system in affective opioid withdrawal, the CPA paradigm has been most commonly employed. This paradigm typically involves pairing naloxone-precipitated morphine withdrawal (in acutely or chronically dependent animals) with a specific environmental context, such that, upon re-exposure to this context in a drug-free state, animals will preferentially avoid the withdrawal paired context versus a context that was previously paired with a placebo saline injection (Sanchis-Segura and Spanagel, 2006). This avoidance of the withdrawal context is used as a measure of the intensity of the aversive affect experienced during opioid withdrawal, and can be used to assess the potential of pharmacological treatments to reduce affective withdrawal. Indeed, pharmacological treatments that are currently used in the treatment of opioid withdrawal (e.g., buprenorphine) are effective in reducing the establishment of avoidance behavior in this paradigm (Stinus et al., 2005). Also employed is the operant responding for food paradigm (Navarro et al., 2001). In this paradigm, a pharmacological treatment is deemed effective in reducing affective withdrawal if it is able to suppress a withdrawal-induced reduction in operant responding for food (Maldonado et al., 1996).

A review of the literature reveals a more complicated role for the endocannabinoid system in its ability to alleviate affective opioid withdrawal than was observed for somatic withdrawal (see Table 1); granted the studies conducted thus far are fairly limited. Indeed, as with somatic withdrawal, both activation and inhibition of the cannabinoid system has been found effective in reducing affective opioid withdrawal, but with greater inconsistencies. Contrary to somatic withdrawal, the cannabinoid agonist, Δ⁹THC, and the FAAH enzyme inhibitors, PF3845 and URB-597, tested were unable to modify the establishment of a naloxone-precipitated CPA (Wills et al., 2014; Gamage et al., 2015). However, while activation of the cannabinoid system via exogenous cannabinoid administration or endogenous elevation of AEA proved to have little efficacy, inhibition of MAGL activity and concomitant elevation of 2-AG showed mixed (Wills et al., 2014, 2016; Gamage et al., 2015) but more promising results. Given that these discrepant findings were obtained by different laboratories using different procedures, compounds (JZL-184 vs. MJN110), and species (mice vs. rats), more research into the potential of MAGL inhibition in reducing affective opioid withdrawal will be required in order to elucidate its role. Furthermore, while the only dual FAAH/MAGL inhibitor (SA-57) investigated was unable to modify a naloxone-precipitated CPA (Gamage et al., 2015), additional research into the ability of such compounds to reduce affective withdrawal should continue owing to their effectiveness in reducing somatic withdrawal with minor adverse effects (noted in previous section).

Although the deleterious reputation of cannabinoid antagonists/inverse agonists may limit their therapeutic potential (Bergman et al., 2008), blockade of the endocannabinoid system may be effective for reducing affective opioid withdrawal. Indeed, acute CB₁R antagonism (with both AM251, and the neutral CB₁ antagonists, AM4113, and AM6527) was effective in preventing the establishment of a naloxone-precipitated CPA (Wills et al., 2014, 2016). This is in contrast to the ability of acute CB₁ antagonism to reliably reduce somatic withdrawal, but could be attributed to the fact that the CPA paradigm may provide a more sensitive measure of opioid withdrawal (Azar et al., 2003). Of special consideration is that both antagonists (which have been found to possess inverse agonist activity) and neutral antagonists (void of intrinsic activity; Sink et al., 2008) were effective in preventing the establishment of the CPA. This is particularly important since these ‘neutral’ antagonists have also been found to lack the adverse qualities attributed to typical cannabinoid antagonists (Bergman et al., 2008). While the only evaluations of chronic inhibition of the endocannabinoid system (CB₁ KO mice) yielded mixed results (Ledent et al., 1999; Martin et al., 2000), chronic antagonism of the endocannabinoid system during the development of dependence using ‘neutral’ antagonists should be investigated given that acute treatments have been reported to precipitate both spontaneous affective and somatic withdrawal (Navarro et al., 1998, 2001).

The neuroanatomical substrates most sensitive to the appearance of somatic opioid withdrawal include the periaqueductal gray (PAG) and locus coeruleus (LC; see Maldonado et al., 1996, for a review). Indeed, microinjection of methylnaloxonium, a hydrophilic opiate antagonist, into these regions was most sensitive in eliciting the opiate withdrawal syndrome, with active symptoms such as jumping being particularly prominent (Maldonado et al., 1992). In revisiting the withdrawal symptoms most commonly attenuated by cannabinoid modulation (e.g., jumping and paw tremors), it becomes apparent that these neuroanatomical substrates may represent regions through which cannabinoids interact with the opioid system to attenuate somatic withdrawal.

Much evidence indicates the PAG as a locus for cannabinoid–opioid interactions. Indeed, anatomical immunolabeling has co-localized the CB₁R and μ-opioid receptor (MOR) within this region. Furthermore, 8% of immunoreactive MOR PAG neurons received immunoreactive CB₁R appositions, indicating a role for presynaptic cannabinoid modulation (Wilson-Poe et al., 2012). In addition, sub-chronic CB₁R agonist treatment with THC or AM356 was reported to increase proenkephalin mRNA levels in the PAG (Manzanares et al., 1998). A reciprocal effect following repeated morphine treatment is also suggested given the ability of the CB₁R antagonist AM251 to increase the frequency of spontaneous miniature inhibitory postsynaptic currents (IPSCs) in morphine, but not saline treated animals, suggesting an elevation of endocannabinoid tone (Wilson-Poe et al., 2014). Chronic (up to 7 days) CB₁R agonist treatment has also been reported to up-regulate MOR density (Viganò et al., 2005), with tolerance developing with prolonged (14 day) treatment (Corchero et al., 2004). Finally, chronic THC pre-opiate exposure or acute rimonabant treatment has been found to induce or attenuate Fos immunoreactivity in the PAG from acute heroin or morphine administration, respectively (Singh et al., 2004, 2005).

While the investigation of cannabinoid–opioid interaction in the LC is less extensive than in the PAG, a recent anatomical study confirmed the co-existence of MOR and CB₁R immunoreactivity in somatodendritic compartments of catecholaminergic neurons. Additionally, as in the PAG, immunoreactive CB₁R axon terminals formed synaptic contacts with MOR dendrites (Scavone et al., 2010). Ultimately, these findings suggest a role for cannabinoid–opioid interactions in the PAG and LC via common signal transduction mechanisms, or potential receptor heterodimerization given their close proximity and reports of this occurrence (Schoffelmeer et al., 2006). Additionally, as discussed above, interactions in endogenous neurotransmitter release are also evident in the PAG. These mechanisms of interaction correspond to the general ability of CB₁R agonism or chronic CB₁R antagonism to reduce symptoms of somatic opioid withdrawal.

The brain structures most sensitive to the motivating or affective component of opioid withdrawal include the nucleus accumbens (NAc) and the amygdala (see Maldonado et al., 1996, for a review). Indeed, microinjection of hydrophilic opiate antagonists into these regions was most sensitive in producing the establishment of a CPA (Stinus et al., 1990), and suppressing operant responding in morphine dependent rats (Koob et al., 1989). An evaluation of cannabinoid–opioid interactions within these regions also reveals these sites to be crucial in the ability of the cannabinoid system to modulate affective opioid withdrawal.

There is substantial research supporting cannabinoid–opioid interactions within the NAc. As in the PAG and LC, immunolabeling has revealed evidence for CB₁R and MOR localization on the same neurons in the NAc shell and core, as well as expression on synaptically linked neurons (Pickel et al., 2004). Furthermore, CB₁R and MOR allosterically interact in this region causing non-additive and synergistic effects on glutamate and GABA release, respectively. These effects were reversed by their respective antagonists, but antagonist co-administration blocked these antagonistic effects, suggesting the potential for G-protein coupled heterodimeric receptor complexes (Schoffelmeer et al., 2006). Consistent with these findings, reports also suggest CB₁R and MOR synergy on cAMP/PKA signaling that is mediated through βγ dimers (Yao et al., 2006). Chronic opioid and cannabinoid administration has also been found to alter cannabinoid and opioid stimulated G-protein receptor coupling, respectively; though while cannabinoids enhance MOR binding (Viganò et al., 2005), opioids have produced more mixed effects on CB₁R binding (Viganò et al., 2003, 2005; Fattore et al., 2007). Reports of changes in receptor densities have also been noted, with chronic opioid administration increasing CB₁R density (Gonzalez et al., 2002), mRNA and protein levels (Ren et al., 2009), and cannabinoid agonist treatment increasing MOR density (Corchero et al., 2004; Fattore et al., 2007; Molaei et al., 2016). However, while cannabinoid agonists seem to consistently increase MOR density, the effects of CB₁R gene deletion have been less consistent in reducing expressivity (Urigüen et al., 2005; Lane et al., 2010). Cannabinoid–opioid interactions on neuronal activation and dopamine release have also been reported in the NAc. Indeed, cannabinoid agonists increase while antagonists decrease Fos expression induced by acute morphine (Singh et al., 2004, 2005), and CB₁R KO mice report decreased morphine-induced dopamine release (Mascia et al., 1999). Finally, cross-talk on endogenous neurotransmitter and peptide release has also been described, with increased AEA and decreased 2-AG content following chronic opioid administration, and increased Met-enkephalin immunoreactivity (Valverde et al., 2001), proenkephalin mRNA (Manzanares et al., 1998), and β-endorphin levels (Solinas et al., 2004) following acute or chronic cannabinoid (THC or CP-55,940) administration. These findings are consistent with the ability of cannabinoid agonists to decrease affective opioid withdrawal and provide evidence for interactions within the NAc at the neurotransmitter, receptor and signal transduction level.

A second region of interest attributed to mediating affective opioid withdrawal is the amygdala. Although there is less research investigating cannabinoid–opioid interactions within this region in comparison to the NAc, effects on receptor density and functionality have been noted. Specifically, acute and repeated cannabinoid agonist (THC or WIN 55,212-2) administration produced increases in MOR density and G-protein receptor coupling in the amygdala (Corchero et al., 2004; Fattore et al., 2007). Similar findings were also reported on the effects of chronic opioid administration on CB₁R density and receptor coupling when considering the amygdala as a whole (Fattore et al., 2007), however, decreases in CB₁R density have been described when analyzing the basolateral amygdala (BLA) individually (Gonzalez et al., 2002). In addition to receptor interactions, alterations in Fos immunoreactivity (Fos IR; a marker of neuronal activation) have also been noted. Indeed, acute treatment of the cannabinoid antagonist, rimonabant, attenuated acute morphine-induced Fos immunoreactivity in the amygdala as a whole (Singh et al., 2004), while investigations of the central nucleus of the amygdala (CeA) specifically revealed decreases in acute heroin-induced Fos IR following chronic THC pre-treatment (Singh et al., 2005), and additive effects on naloxone-induced Fos IR following acute THC exposure (Allen et al., 2003). Although the above studies do not make any direct comparisons between the BLA and CeA on cannabinoid–opioid interactions, a recent experiment suggests a functional double dissociation between these regions in the ability of the cannabinoid system to modulate affective opioid withdrawal. Indeed, cannabinoid agonism (via inhibition of MAGL activity with MJN110) within the BLA, while cannabinoid antagonism (AM251) within the CeA, is able to interfere with the establishment of a naloxone-precipitated CPA in acutely morphine dependent rats (Wills et al., 2016). Though the mechanisms mediating the dissociation between these regions were not investigated, this finding suggests potential differences in cannabinoid–opioid modulation within different sub-regions of the amygdala.