Progress in Opioid Reward Research: From a Canonical Two-Neuron Hypothesis to Two Neural Circuits

Abstract

Opioid abuse and related overdose deaths continue to rise in the United States contributing to the national opioid crisis in the USA. The neural mechanisms underlying opioid abuse and addiction are still not fully understood. This review discusses recent progress in basic research dissecting receptor mechanisms and circuitries underlying opioid reward and addiction. We first review the canonical GABA-dopamine hypothesis that was upheld for half a century, followed by major findings challenging this hypothesis. We then focus on recent progress in research evaluating the role of the mesolimbic and nigrostriatal dopamine circuitries in opioid reward and relapse. Based on recent findings that activation of dopamine neurons in the ventral tegmental area (VTA) and substantia nigra pars compacta (SNc) is equally rewarding and that GABA neurons in the rostromedial tegmental nucleus (RMTg) and the substantia nigra pars reticula (SNr) are rich in mu opioid receptors and directly synapse onto midbrain DA neurons, we proposed that the RTMg→VTA→ventrostriatal and SNr→SNc→dorsostriatal pathways may act as the two major neural substrates underlying opioid reward and abuse. Lastly, we discuss possible integrations of these two systems during initial opioid use, development of opioid abuse and maintenance of compulsive opioid seeking.

Opioids have become a fundamental part of modern medicine, because of their powerful analgesic properties (Clark, 2002; Kanjhan, 1995; Rosenblum et al., 2008). However, most of opioid-based medications (e.g., fentanyl, oxycodone) are among the most addictive substances in human history (Gruber et al., 2007; Volkow et al., 2019), often leading to opioid abuse and overdose. Skyrocketing numbers of opioid abuse have greatly contributed to the national opioid epidemic (White House, 2017), taking almost 450,000 American lives since 1999 (CDC, 2020; Colon-Berezin et al., 2019; Gladden et al., 2016; Jannetto et al., 2019; O’Donnell et al., 2017; Rudd et al., 2016). This demands for urgent scientific endeavors to better understand main cellular mechanisms and circuits involved in opiate addiction. The elucidation of underlying receptor mechanisms and circuits might provide fresh insights into pharmacotherapeutic targets and improve current treatments for opioid use disorder (OUD). Significant progress has been made in the last decade in opioid research, particularly in the development of various opioid receptor-transgenic (mutant or reporter) mice and opioid-based ligands (e.g., biased-opioid receptor ligands) to selectively target specific intracellular signaling pathways. In this review, we will focus on recent progress in research regarding the role of the midbrain circuitries in opioid reward. Specifically, we will review preclinical work mapping mu opioid receptor (MOR) expression in different phenotypes of midbrain neurons and extensively review the role of midbrain GABA neurons in opioid reward-driven behaviors. We will also summarize evidence suggesting that the integration of the mesolimbic and nigrostriatal pathways plays a critical role in opioid reward and opioid-motivated behaviors.

The mesolimbic DA pathway and opioid reward

Local GABA-DA hypothesis within the VTA

The neuronal substrates underlying opioid reward have been studied in great detail, but the precise circuitry is yet to be uncovered. Opioid receptors are G-protein coupled receptors that have been classified as mu, kappa, delta and nociception (Darcq and Kieffer, 2018; Kosterlitz and Paterson, 1980; Pasternak and Wood, 1986). Among them, stimulation of MORs produces the strongest analgesic and rewarding effects (Gruber et al., 2007).

In point of fact, abuse liability of opioids has initially been thought to derive from drug rewarding effects that are mediated primarily by dopamine (DA) neurons in the ventral tegmental area (VTA). Early electrophysiological and microdialysis studies indicate that stimulation of MORs directly inhibits VTA GABA neurons, leading to rapid excitation of neighboring VTA DA neurons (i.e., disinhibition) (Gysling and Wang, 1983; Johnson and North, 1992; Margolis et al., 2014) and release of DA in the nucleus accumbens (NAc) (Devine et al., 1993; Spanagel et al., 1992; Yoshida et al., 1993). For this reason, the rewarding effects of opioids were believed to be mediated by inhibition of VTA GABA interneurons and subsequent disinhibition of neighboring DA neurons (Gysling and Wang, 1983; Johnson and North, 1992; Margolis et al., 2014). This hypothesis was supported by a series of behavioral studies. For example, rats and mice can learn to self-administer MOR agonists directly into the VTA (Bozarth and Wise, 1981; David and Cazala, 1994; Devine and Wise, 1994; Welzl et al., 1989) and develop a conditioned place preference (CPP) for the environment associated with intra-VTA infusions of MOR agonists (Bals-Kubik et al., 1993; Mamoon et al., 1995; Phillips and LePiane, 1980; Zangen et al., 2002). Furthermore, intra-VTA infusions of naloxone, a MOR antagonist, block morphine CPP (Olmstead and Franklin, 1997). Pretreatment with α-flupenthixol (a non-selective DA receptor antagonist) can block morphine CPP induced by intra-VTA morphine injections in rats with a prior history of morphine exposure, but not in naive rats (Nader and van der Kooy, 1997). In addition to the VTA, morphine can also be self-administered directly into the NAc by rats and mice (Amalric and Koob, 1985; David and Cazala, 2000; Olds, 1982) and intra-NAc microinjections of methylnaloxonium (a peripherally limited opioid antagonist) block heroin-induced hyperlocomotion in rats (Amalric and Koob, 1985), suggesting an important role of the mesolimbic DA system in opioid reward and addition-relation behavior. Furthermore, opioid-induced molecular and cellular plasticity in VTA GABAergic synapses onto DA neurons have been correlated with learning and drug memory, and therefore, it has been suggested that this type of neuroplasticity might promote compulsive opioid-taking and opioid-seeking behavior (Langlois and Nugent, 2017; Mazai-Robison and Nestler, 2012).

However, there is also evidence refuting the DA disinhibition hypothesis. For example, DA-deficient mice (i.e. unable to synthesize DA) display reduced locomotor activity in response to morphine and decreased sensitivity to the analgesic effects of morphine. However, these mice can develop robust CPP for morphine only when given morphine in conjunction with either caffeine or 1-dihydroxyphenylalanine (a DA precursor that restores DA throughout the brain) (Hnasko et al., 2005), suggesting that DA is critical for morphine-induced locomotion, but less important for morphine-induced reward. In agreement with these findings are studies reporting that chemical lesions of DA terminals in the NAc with 6-OHDA fail to alter heroin self-administration (Gerrits et al., 1994; Pettit et al., 1984). Likewise, the non-selective DA antagonists alpha-flupenthixol and haloperidol, at high doses, have been shown to significantly decrease cocaine, but not heroin, self-administration (Ettenberg et al., 1982; van Ree and Ramsey, 1987), suggesting that cocaine and heroin rewards are mediated by different mechanisms. This is further supported by the finding that pharmacological blockade or down-regulation of NAc DA D1 receptors attenuates cocaine, but not heroin, self-administration (Gerrits et al., 1994; Pisanu et al., 2015). Thus, the question whether the mesolimbic DA mechanisms alone are responsible for opioid reward is still passionately debated (Badiani et al., 2011; Blum et al., 2015; Nutt et al., 2015).

Opioids activate a subpopulation of VTA DA neurons

To further address the role of VTA GABA and DA neurons in opioid reward, a recent study using c-Fos immunohistochemistry and fiber photometry showed that a single injection of heroin can activate DA neurons in the medial part of the VTA and increase DA release in the medial shell of the NAc (Corre et al., 2018), suggesting that not all DA neurons in the VTA are activated by opioids and other unidentified neural substrates may also be involved in opioid action. Importantly, optogenetic activation of VTA DA neurons is rewarding as assessed by optical intracranial self-stimulation (oICSS), while activation of VTA GABA neurons is aversive, as assessed by optical real-time preference (Corre et al., 2018; Tan et al., 2012; van Zessen et al., 2012). These findings suggest that opioid-induced inhibition of VTA GABA neurons via MORs may cause disinhibition of VTA DA neurons, producing rewarding effects. Strikingly, chemogenetic inhibition of VTA DA neurons inhibits heroin self-administration (Corre et al., 2018), which provide supporting evidence for the role of VTA DA neurons in opioid reward. More recently, we used an optogenetic approach to directly manipulate neuronal activity of VTA DA neurons. We found that DAT-cre mice transfected with inhibitory halorhodophin (NpHR) in VTA DA neurons readily learn to press the lever for intravenous heroin. However, response-contingent optogenetic inhibition of VTA DA neurons significantly reduces heroin self-administration (Galaj et al., 2020). With optogenetic inhibition of VTA DA neurons, animals display a typical extinction-like pattern of responding-initial increases in drug intake followed by cessation of responding. These findings suggest that VTA DA neurons play an important role in maintenance of heroin self-administration. The important questions are: 1) whether the mesolimbic DA system is the only system mediating opioid reward and addiction, and 2) whether opioid-induced inhibition of VTA GABA neurons is the major mechanism that leads to activation of the mesolimbic DA system.

VTA GABA neurons play a limited role in opioid reward

To answer the above questions, it is critical to determine GABAergic inputs into the VTA. As stated above, VTA DA neurons receive inhibitory inputs from local GABA interneurons. However, autoradiography assays indicate that MOR binding level is very low in the VTA (Méndez et al., 2003; Tempel and Zukin, 1987) (Figure 1-A, B), suggesting that VTA GABA neurons may not be a major target of opioids. In line with these reports is our recent study in which we used a highly sensitive RNAscope ISH assay to examine GABA neuron distributions in the midbrain and the cellular distributions of MOR mRNA in the midbrain. We found that GABA neurons are mainly distributed in the SNr where over 90% of neurons are GABAergic (Figure 1C), while in the VTA, a majority of neurons (~70%) are dopaminergic (Galaj et al., 2020) (Figure 1D). RNAscope ISH assays did not detect MOR gene expression in midbrain DA neurons (Figure 1E). Within the VTA, less than 30% of GABA neurons express MORs (Figure 2), suggesting that the role of MORs in VTA GABA neurons in opioid reward might be limited.

Figure 1:

MOR and its gene (Oprm1) expression in the striatum and midbrain. A-B: [³H]-DAMGO binding results in autoradiography assays, indicating that high densities of MOR binding sites in the rat striatum and midbrain, particularly in the nucleus accumbens (NAc) and substantia nigra (SN) (Tempel and Zukin, 1987). C-D: RNAscope ISH results in rats, illustrating that GABA neurons (labeled by GAD1) are mainly distributed in the substantia nigra pars reticulata (SNr) and less in the ventral tegmental area (VTA), while DA neurons (labeled by DAT) are mainly expressed in the VTA and substantia nigra pars compacta (SNc). E: MOR RNAscope results, illustrating that the MOR gene – Oprm1, is not expressed in DA neurons in either the VTA or SNc. Abbreviations: FrPaSS, frontoparietal cortex, somatosensory area; cc, corpus callosum; CP, caudate-putamen; NAc, nucleus accumbens; SuG, superficial gray layer of the superior colliculus; CG, central gray; SNr, substantia nigra pars reticulata; VTA, ventral tegmental area; RN: red nucleus; Oprm1: opioid receptor mu 1 that encodes mu opioid receptor; GAD1: glutamate decarboxylase 1; DAT: dopamine transporter

Figure 2:

Oprm1 mRNA expression in GABA neurons in 6 different brain regions. A: Oprm1 mRNA is expressed in a subpopulation (~30%) of GABA neurons (labeled by GAD1) in the VTA and SNc; B: Oprm1 mRNA is expressed in ~50% of GABA neurons in the SNr and >70% of GABA neurons in the RMTg (i.e., the tail of the VTA); C: In the NAc, Oprm1 mRNA is mainly expressed in >80% of GABAergic D1-MSNs (labeled by Drd1), but not in D2-MSNs (labeled by Drd2).

This premise is supported by our recent behavioral findings that optogenetic activation of VTA GABA neurons failed to alter heroin self-administration, but significantly inhibited cocaine self-administration in vGAT-cre mice (Galaj et al., 2020). Optogenetic activation of VTA GABA neurons produced aversive effects, as assessed by real-time place preference. These data suggest that VTA GABA neurons may play differential roles in cocaine and opioid reward. Furthermore, microinjections of the MOR antagonist naloxonazine into the VTA produced a significant increase in heroin self-administration maintained by a low (0.0125 mg/kg/infusion), but not by high, doses of heroin (Galaj et al., 2020). As animals increased heroin self-administration rates while maintaining an evenly distributed pattern of self-administration over a 3-h session, we interpret this increase in drug intake as compensatory responses to a reduction in drug reward (Galaj et al., 2020). These findings directly challenge the canonical VTA GABA-DA hypothesis and suggest that other pathways might be involved in opioid reward.

RMTg GABA neurons project to VTA DA neurons

In contrast, recent studies indicate that VTA DA neurons also receive intense GABA inputs from a number of other brain regions, including the NAc, ventral pallidum (VP) and the tail of the VTA (also called rostromedial tegmental nucleus, RMTg) (Bolam and Smith, 1990; Hjelmstad et al., 2013; Kalivas, 1993; Matsui et al., 2014; Tepper and Lee, 2007; Watabe-Uchida et al., 2012). The RMTg has unique characteristics that distinguish this region from the VTA. The RMTg represents relatively pure GABA cell population (Jhou et al., 2009a; Kaufling et al., 2009; Olson and Nestler, 2007; Perrotti et al., 2005) with high levels of MOR immunoreactivity and mRNA (Galaj et al., 2020; Jalabert et al., 2011; Jhou et al., 2009a, 2012) and high levels of the neuropeptide nociception (Jhou et al., 2012). RMTg GABA neurons send dense projections to the VTA and substantia nigra zona compacta (SNc) (Jhou et al., 2009b; Kaufling et al., 2010), where they form synapses with approximately 80% of DA neurons (Balcita-Pedicino et al., 2011).

Matsui and colleagues examined inhibitory GABA postsynaptic currents (IPSCs) evoked by selective optical stimulation of GABA projections from the VTA, NAc, or RMTg and found that inhibition of IPSCs induced by MOR agonists was pathway-dependent (Matsui et al., 2014). Specifically, morphine induced ~50% inhibition of IPSCs evoked by optical stimulation of the RMTg→VTA GABA pathway, ~20% inhibition with stimulation of the NAc→VTA pathway, and minimal (~10%) inhibition with stimulation of local VTA GABA neurons (Matsui et al., 2014). These findings suggest that opioid-induced disinhibition of DA neurons is mediated mainly by inhibition of RMTg GABA neurons that project to the VTA, but less by inhibition of VTA GABA interneurons (Jhou et al., 2012; Martin et al., 2002; Matsui et al., 2014; Steidl et al., 2017; Vaccarino et al., 1985) (Figure 3).

Figure 3:

Diagram showing the mesolimbic and nigrostriatal DA pathways that underlie opioid reward and addiction. Both pathways receive GABA inputs from the RMTg. The mesolimbic DA circuit (RMTg→VTA→NAc) originates with DA neurons in VTA and its neurons project to the NAc. VTA DA neurons receive GABAergic inputs from local VTA GABA neurons and other brain regions including the NAc, ventral pallidum (VP), and RMTg. The nigrostriatal DA circuit (SNr→SNc→dST) originates with DA neurons in the substantia nigra pars compacta (SNc) and projects to the dorsal striatum (dST). SNc DA neurons receive dense GABAergic inputs from multiple brain regions including the SNr and RMTg. SNr GABA neurons receive dense GABAergic input from striatal D1-MSNs. GABA neurons of the ventral striatum (NAc) project to the VTA directly or indirectly via the VP. Mu opioid receptors (MOR) are highly expressed in GABA neurons, particularly in the RMTg and SNr, and D1-MSNs in the NAc and dST. Opioids by binding to MORs inhibit GABA neuron activity and GABA release, which subsequently disinhibits DA neurons in the VTA and SNc. Early studies suggest that the local GABA-induced DA disinhibition within the VTA might underlie opioid reward and addiction. This long-held dogma has been changed by recent findings that opioid induce disinhibition of midbrain DA neurons by inhibiting GABA inputs from the RMTg, SNr, modestly from the NAc and VP.

Additional evidence suggesting the importance of the RMTg→VTA GABA pathway in opioid-driven behaviors comes from receptor ligand autoradiography and transgenic mCherrry MOR knock-in reporter mouse studies, indicating that MORs are highly expressed in the NAc, VP, RMTg and SNr (Erbs et al., 2015; Kitchen et al., 1997; Mansour et al., 1987). More recently, we used a highly sensitive RNAscope in situ hybridization technique to map MOR gene, Oprm1, distributions in different phenotypes of neurons in the midbrain. We found that >70% of GABA neurons in the RMTg express high density of Oprm1 mRNA, which is in a strike contrast to that in the VTA where <30% of GABA neurons express Oprm1 mRNA and in the SNr where ~50% of GABA neurons express Oprm1 mRNA (Galaj et al., 2020). This anatomical distribution of MORs suggests that MORs on RMTg GABA neurons might play a more important role in opioid reward and addiction than those on VTA GABA neurons.

In agreement with electrophysiological and neuroanatomical evidence are studies demonstrating that the RMTg→VTA GABA pathway directly influences VTA DA neuronal activity. Inhibition of the RMTg was reported to increase the activity of VTA DA neurons (Jalabert et al., 2011) but stimulation of the RMTg decreases such activity (Hong et al., 2011; Lecca et al., 2012; Matsui and Williams, 2011). Optogenetic and electrophysiological studies also demonstrated that morphine excites VTA DA neurons by targeting MOR receptors localized on the soma of RMTg GABA neurons as well as their projection terminals in the VTA (Jalabert et al., 2011; Lecca et al., 2012, 2012; Matsui and Williams, 2011), further challenging the canonical VTA GABA-DA disinhibition hypothesis (Figure 3).

The role of the RMTg→VTA pathway in opioid reward is also supported by other behavioral findings. For example, intra-RMTg infusions of MOR agonist support self-administration and CPP (Jhou et al., 2012). Intra-RMTg infusions of morphine also increase open-field locomotion (Steidl et al., 2017), but reduce intravenous heroin self-administration rates (Steidl et al., 2015), an effect that is most likely due to enhanced opioid reward. Likewise, a follow-up study with DREADDS demonstrated that chemogenetic inhibition of RMTg neurons facilitates morphine-induced locomotion, whereas chemogenetic stimulation of these neurons inhibits such behavior (Wasserman et al., 2016).

It is important to note that the RMTg→VTA pathway has been also implicated in the development of opioid tolerance and withdrawal (Matsui et al. 2014; Kaufling and Aston-Jones, 2015). A number of studies indicate that naloxone-precipitated withdrawal from morphine induces c-Fos expression in the RMTg (an indicative of neuronal activity) (Sánchez-Catalán et al., 2017) and increases GABA release in the VTA (Kaufling and Aston-Jones, 2015). During opioid withdrawal VTA DA neurons are activated by disinhibition of GABA projections from the RMTg (de Guglielmo et al., 2015; Kaufling and Aston-Jones, 2015; Lecca et al., 2012). In addition, phosphorylated CREB activity in RMTg GABA neurons is known to be elevated during morphine-withdrawal and is positively correlated with the severity of withdrawal symptoms (Bobzean et al., 2019). A recent study has shown that the RMTg→VTA pathway undergoes neuroadaptations after chronic opioid exposure (Kaufling and Aston-Jones, 2015). Specifically, optogenetic stimulation of RMTg neurons can directly inhibit VTA DA neurons in drug naive and withdrawn rats. However, optogenetic inhibition of RMTg GABA neurons appear to increase VTA DA activity only in naive, but not in withdrawn rats, suggesting that the RMTg→VTA GABA neurons lose their capacity to disinhibit VTA DA neurons after chronic morphine exposure (Kaufling and Aston-Jones, 2015).

The NAc D1-MSN→VTA pathway

In addition to GABA afferents from the RMTg, VTA DA neurons also receive GABA inputs from medium-spiny neurons (MSNs) in the NAc (Beier et al., 2015; Watabe-Uchida et al., 2012; Yang et al., 2018) (Figure 3). The importance of the NAc in opioid-driven behaviors has been studied in great detail. Kainic acid lesions of the NAc disrupt intravenous heroin self-administratin (Zito et al., 1985) and electrolytic lesions to the NAc block the development of morphine CPP (Kelsey et al., 1989). Rats can learn to self-administer opioids directly into the NAc (Olds, 1982). Intra-NAc opioid agonists enhance brain stimulation reward (van Wolfswinkel and van Ree, 1985), while intra-NAc infusions of MOR antagonists cause compensatory increases in opioid self-administration due to reduced opioid reward (Corrigall and Vaccarino, 1988; Vaccarino et al., 1985). These findings indicate an important role of the NAc in opioid reward.

The striatal MSNs are segregated into D1 receptor (D1-MSNs) and D2 receptor (D2-MSNs). The D1-MSNs directly project to the VTA, while D2-MSNs indirectly project to the VTA through VP GABA neurons. A recent study using conditional MOR-KO mice (D1x5/6-Cre × Oprm1-flox) demonstrated that selective deletion of MORs from both D1- and D2-MSNs caused significant increases in heroin self-administration and decreases in locomotor response to heroin (Charbogne et al., 2017). The animals’ ability to develop opioid CPP was not disrupted but their responding for heroin under a progressive ratio (PR) schedule of reinforcement and their drug seeking during reinstatement was enhanced, suggesting that deletion of MORs in both D1- and D2-MSNs significantly alters motivation for heroin rather than heroin reward. Furthermore, D1- and D2-MSNs in the NAc core and shell show differential AMPA receptor neuroadaptations following morphine exposure and during withdrawal from morphine, suggesting that both types of the neurons play a different role in opioid-driven behaviors (Madayag et al., 2019, Graziane et al., 2016).

To dissect the role of MORs in D1- vs. D2-MSNs in opioid reward and addiction, Cui and colleagues used conditional BAC-mediated transgenic rescue strategy to reexpress MOR expression in D1-MSNs in full MOR-KO mice. Complete deletion of MORs in full MOR-KO mice abolished opioid CPP and physical dependence, while ectopic rescue of MOR expression in striatal D1-MSNs successfully restored opioid-induced CPP, locomotor sensitization, and partially restored opioid self-administration (Cui et al., 2014), suggesting that D1-MSNs are critical players in opioid reward. In contrast to this finding, the Ferguson lab has recently reported that chemogenetic manipulations of D1-MSNs failed to alter motivation to self-administer heroin, but inactivation of the direct MSN pathway reduced cue-induced reinstatement of heroin seeking in high-risk rats (i.e., vulnerable to addiction) (O’Neal et al., 2020). Similarly, selective deletion of MORs from D1-neurons reduced opioid (morphine or oxycodone)-induced hyperlocomotion but did not affect self-administration of remifentanil or oxycodone (Severino et al., 2020). These findings suggest that the direct D1-MSN pathway may play a more important role in heroin seeking and hyperactivity, rather than in opioid reward.

To further address the role of D1-MSNs in opioid action, we and others have recently examined the cellular distributions of MORs in the striatum. Double fluorescent localization of MORs and GFP in the GENSAT Drd1-GFP and Drd2-GFP BAC mice shows MOR expression in genetically labeled striatal D1-MSNs, not in D2-MSNs (Cui et al., 2014). These findings are in line with fluorescence-activated cell sorting (FACS) assays indicating that the Oprm1 is expressed in Pdyn-GFP (D1-MSNs), not in Drd2-GFP mice (Cui et al., 2014). More recently, we used highly sensitive and specific RNAscope ISH assays to examine Oprm1 expression in the NAc. We confirmed the previous findings that MORs are expressed mainly in D1-MSNs, not in D2-MSNs, of the NAc (Figure 2E). These findings support the above behavioral findings that that striatal D1-MSNs may play an important role in opioid action.

The cellular mechanisms through which the direct NAc→VTA pathway modulate VTA neuron activity are yet to be fully understood. Studies combining optogenetic and electorphysiological approaches demonstrated that NAc D1-MSNs target non-DA neurons directly, which subsequently regulate VTA DA neuronal activity through disinhibition (Bocklisch et al., 2013; Chuhma et al., 2011; Xia et al., 2011; Yang et al., 2018). In the VTA, prolonged optogenetic stimulation of D1-MSN terminals leads to decreased activity of VTA GABA neurons and increase in the firing rate of VTA DA neurons during initial stimulation (Soares-Cunha et al., 2019). In contrast, it was reported that some D1-MSNs in the medial part of the NAc shell directly project to VTA DA neurons and functionally inhibit VTA DA neuron activity (Edwards et al., 2017; Yang et al., 2018). Using a multidisciplinary approach combining retrograde tracing, electrophysiology, optogenetics and behavioral assays to determine the architecture and function of the NAc→VTA pathway, the Yang lab reported that D1-MSNs in the medial part of the NAc shell form synapses directly on DA neurons in the lateral and medial parts of the VTA (Yang et al., 2018). Optogenetic stimulation of the medial NAc→VTA GABA pathway inhibits medial VTA DA neurons via GABA_A receptors and lateral VTA DA neurons via GABA_B receptors and produced a general state of behavioral suppression, but failed to produce rewarding effects, as assessed by real-time place preference and optical intracranial stimulation (oICSS) (Yang et al., 2018). In contrast, the NAc→VTA pathway originating in the lateral part of the NAc synapses directly onto VTA GABA neurons via GABAa receptors (Yang et al., 2018). Optogenetic stimulation of the lateral NAc→VTA pathway in D1-cre mice has been shown to produce rewarding effects. These findings suggest that the NAc→VTA pathway originating from distinct NAc subnuclei induces distinct behavioral phenotypes. Thus, the impact of D1-MSNs in the direct NAc→VTA pathway on VTA DA neurons appears to depend on the balance between direct inhibition and indirect disinhibition of VTA DA neurons (Figure 3).

The NAc D2-MSN→VP pathway

As stated above, D2-MSNs may indirectly modulate VTA DA neuron activity via VP GABA neurons. Given that MORs are not detected in D2-MSNs in the NAc (Cui et al., 2014) (Figure 2), the role of D2-MSNs in opioid action could be minimal. However, experimental manipulations of D2-MSNs may also modulate opioid addiction-related behaviors. Using dual viral-mediated gene transfer of DREADDs, O’Neal et al show that transient activation of D2-MSNs is capable of suppressing cue-induced reinstatement of heroin-seeking, while inactivation of D2-MSNs exacerbates cue-induced reinstatement of heroin-seeking in high- but not low-risk rats (O’Neal et al., 2020). However, similar manipulations of D2-MSNs did not alter heroin self-administration (O’Neal et al., 2020), suggesting that the manipulation of NAc D2-MSNs in the indirect pathway may alter heroin seeking but not opioid reward.

A recently published paper by Walwyn group showed that selective deletion of MORs from different neuron phenotypes produces differential effects on opioid reward. Selective deletion of MORs from D2-neurons or CHAT neurons affected neither opioid (remifentanil or oxycodone) self-administration nor opioid-induced hyperlocomotion (Severino et al., 2020). Interestingly, deletion of MORs from adenosine 2a (A2a)-expressing neurons (including D2-MSNs) resulted in enhanced opioid-induced locomotion (Severino et al., 2020). During the extinction, mice lacking MORs in A2a-expressing neurons or CHAT neurons showed higher drug seeking than control (Severino et al., 2020). These findings suggest that other A2a-expression neurons outside the striatum may play a role in opioid reward-related behaviors.

The VP→VTA pathway

There is evidence that the VP GABA neurons receive inputs from D2-MSNs and project to the VTA (Kalivas et al., 1993). Electrophysiological studies demonstrated that the VP→VTA projections synapse onto DA and non-DA neurons. Optogenetic stimulation of VP GABA terminals in the VTA elicits GABA inhibitory postsynaptic currents (IPSCs) in both VTA DA and non-DA neurons, and these IPSCs can be inhibited by DAMGO (a MOR agonist) (Hjelmstad et al., 2013). Approximately a half of VP GABA neurons can be inhibited by local application of DAMGO and these neurons are involved in the development of opioid sensitization (Mickiewicz et al., 2009). Rats exposed to intra-VP injections of morphine develop morphine CPP but not when pretreated with naloxone (a MOR antagonist) (Zarrindast et al., 2007), which suggest that the VP, rich in MORs, may also play an important role in opioid reward. Likewise, ibotenic acid lesions of the VP disrupted responding for heroin under fixed-ratio (FR5) and PR schedules of reinforcement (Hubner and Koob, 1990). In addition, the VP→VTA pathway has been implicated in natural and drug reward to drug and alcohol (Prasad et al., 2020). Chemogenetic inhibition of the VP→VTA pathway reduces alcohol (Prasad et al., 2020) and cocaine seeking (Mahler et al., 2014). Although the VP→VTA connection is well defined, there is no direct evidence that this pathway is involved in opioid self-administration and relapse. Therefore, it requires future study.

The nigrostriatal DA pathway and opioid reward

The SNc→dST pathway

In addition to the VTA, midbrain DA neurons are densely distributed in a neighboring region, the substantia nigra pars compacta (SNc), and they project to the dorsal striatum (dST) giving the origin to the nigrostriatal pathway (Hattori et al., 1991; Ungerstedt, 1976) (Figure 3). Early studies with brain stimulation reward and intracranial injections emphasized the importance of the SNc in reward. Rats or mice would readily work for electrical or optogenetic ICSS that involves direct stimulation of SNc neurons (Corbett and Wise, 1980; Galaj et al., 2020; Ilango et al., 2014; Rossi et al., 2013b, 2013a; Wise, 1981) and would show increased rates of cocaine self-administration when local SN D1 receptors are blocked by a local infusions of a D1R antagonist (Quinlan et al., 2004). It is important to note that optogenetic stimulation of SNc DA neurons produces real time place preference and oICSS in TH-cre or DAT-cre mice to the similar degree as optogenetic stimulation of VTA DA neurons (Galaj et al., 2020; Ilango et al., 2014), suggesting that midbrain DA neurons both in the VTA and SNc play an equally important role in drug reward. In contrast, inhibition of SNc DA neurons can produce aversion (Ilango et al., 2014), which is also seen with inhibition of VTA GABA neurons (Galaj et al., 2020). These findings nicely complement clinical research showing that the striatal DA are involved in craving for cocaine (Volkow et al., 2006; Wong et al., 2006) and the striatal neurons respond to alcohol (Brumback et al., 2015; Vollstädt-Klein et al., 2010) and nicotine cues (Yalachkov et al., 2009).

The SNr→SNc pathway

The SNr is the home to GABAergic neurons that form reciprocal synaptic connections with DA neurons in the SNc (Deniau et al., 1978; Mailly et al., 2003; Sanderson et al., 1986; Savasta et al., 1986; Tepper and Lee, 2007) (Figure 3). Early studies showed that SNr neurons that receive a major inhibitory input from the dST participate in the regulation of SNc DA activity and striatal DA content (Chan, 1985; Kamata et al., 1986; Reid et al., 1990; Waszczak et al., 1980; You et al., 1994). Systemic administration of morphine was shown to cause naloxone reversible increases in the firing rate of SNc DA neurons and decreases in the activity of neighboring SNr GABA neurons (Hommer and Pert, 1983; Melis et al., 2000; Walker et al., 1987), suggesting possible GABA-mediated disinhibition of SNc DA neurons from the SNr. Interestingly, acute morphine fails to affect ICSS in rats implanted with electrode in the SNc. However, repeated daily administration of morphine lead to increases in the ICSS threshold (Nazzaro et al., 1981). Autoradiography binding study revealed that chronic treatment with morphine leads to a significant reduction of [³H]-SCH-23390 and [³H]-spiperone binding to D1 and D2Rs, respectively, in the rat SNc (Spampinato et al., 1988). Thus. one mechanism responsible for the reduced rewarding effects of ICSS in morphine-treated rats might be due to morphine-induced down-regulation of nigral D1 and D2Rs.

Local administration of the D1R agonist SKF 38393 into the dorsal striatum was shown to increase extracellular dynorphin B and GABA levels in the ipsilateral SNr region, followed by prolonged decreases in the striatal DA level (You et al., 1994). Ibotenic acid lesions or intracranial injections of GABA or dynorphin A directly into the SNr were shown to inhibit DA levels in the striatum (Reid et al., 1990), whereas intra-SNr application of morphine to excite SNc DA neurons and increase the striatal DA content (Matthews and German, 1984; You et al., 1996). Likewise, intravenous administration of the GABA agonist musimol leads to inhibition of SNr GABA neurons and simultaneously to excitation of SNc DA neurons (Waszczak et al., 1980). Direct infusion of GABA, morphine or enkephalin into the SNr alter circling behavior that can be antagonized by ipsilateral lesions or pharmacological blockade of SNc DA neurons (Kaakkola, 1980; Martin et al., 1978). Thus, it appears that the SNr is an important hub that controls the activity of neighboring SNc DA neurons and striatal DA [for a comprehensive review see (Tepper and Lee, 2007)].

In our recent report, we found that approximately 50% of SNr GABA neurons express MORs (Galaj et al., 2020) (Figure 2), suggesting the importance of SNr MORs in opioid reward and opioid-driven behaviors. A number of studies showed that systemic or intravenous administration of morphine inhibits spontaneous activity of SNr neurons, presumably GABA neurons and this effect is often accompanied by excitation of SNc DA neurons that can be reversed by naloxone (a MOR antagonist) (Finnerty and Chan, 1979; Hommer and Pert, 1983). Thus, it appears that one of the mechanisms by which opioids produce rewarding effects involves direct MOR-induced inhibition of SNr GABA neurons, which in turns leads to disinhibition of SNc DA neurons and an increase in striatal DA neurotransmission (Figure 3).

We recently provided behavioral evidence supporting this hypothesis. We demonstrated that optogenetic inhibition of SNr GABA neurons or activation of neighboring SNc DA neurons produces rewarding effects, as assessed by real-time place preference and oICSS in vGAT-cre and DAT-cre mice, respectively (Galaj et al., 2020). The NpHR-induced inhibition of SNr GABA neurons most likely involves disinhibition of SNc DA neurons, leading to an increase in the extracellular striatal DA level. We also demonstrated that SNr GABA neurons play a critical role in heroin-self administration and relapse. In vGAT-cre mice transfected with inhibitory NpHR in SNr GABA neurons, optogenetic inhibition of SNr GABA neurons caused compensatory increases in heroin intake and reductions in heroin-primed reinstatement of drug seeking (Galaj et al., 2020). Similar effects were observed with a pharmacological approach such that intra-SNr naloxone (a MOR antagonist) or naloxonazine (an irreversible MOR antagonist), which produced a compensatory increase in heroin self-administration and heroin intake (Galaj et al., 2020). However, we found no evidence for the involvement of SNr GABA neurons in cue-induced relapse as naloxonazine failed to alter cue-induced reinstatement of drug seeking, suggesting that the SNr plays a role in the acute effects of opioids rather than conditioned cues. In addition, systemic administration of heroin (1 mg/kg; i.p.) significantly increased locomotor activity that was reduced by optical activation of SNr GABA neurons in vGAT-re mice, suggesting that these neurons play a role in heroin-driven behaviors (Galaj et al., 2020). These new findings are in the agreement with a previous report that intra-SNr infusions of morphine produces morphine CPP (Baumeister et al., 1993). Furthermore, studies with blood oxygen level dependent (BOLD) imaging and c-fos immunohistochemistry showed enhanced neuronal activation in the SNr region in response to systemic administration of oxycodone (Moore et al., 2016) or escalating doses of morphine (Bai et al., 2012), suggesting that the SNr plays a role in opioid action.

It is worth noting that SNr GABA neurons have been also implicated in morphine analgesic effects (Baumeister, 1991, 1991; Baumeister et al., 1987) and withdrawal from alcohol and opioids (Baumeister et al., 1992, 1989; Chen et al., 2011). Intense wet-dog shakes, irritability to touch, teeth chattering, diarrhea and locomotion were observed in morphine-dependent rats after intra-SNr infusions of naloxone (a MOR antagonist) (Baumeister et al., 1989). Similar withdrawal symptoms were observed in morphine-dependent animals after intra-SNr infusions of the MOR antagonist CTOP (Baumeister et al., 1992). In addition, intra-SNr infusions of DAMGO, a MOR agonist, produced antinociceptive effects that were then reversed by systemic administration of beta-funaltrexamine (Baumeister et al., 1992). Morphine, its metabolite morphine-6-glucuronide (M6G) or opioid peptides (e.g, enkephalin and dynorphin), when injected directly into the SNr, also produced analgesia (Baumeister et al., 1993, 1987), as assessed in the tail flick and or hot plate assays, further suggesting that SNr MORs play a critical role in analgesic effects of opioids. Taken together, our new findings combined with other previous reports suggest that MORs in the SNr GABA neurons play a critically important role in the number of opioid driven behaviors.

The dST →SNr pathway

One of the major GABA afferents to the SNr comes from the dorsal striatum (dST) (Figure 3). The dST, which receives DA inputs from the SNc and sends GABA projections to the SNr, has been implicated in opioid-related behaviors. Lesions to the dST were shown to decrease PR responding for intravenous morphine or cocaine (Suto et al., 2011) and lower the threshold of reward (for morphine doses that are typically not rewarding in rats) to the point that the rewarding effects of morphine could be now detected (Glick et al., 1975). In addition to reductions in morphine self-administration, rats with dST lesions showed a reduction in naloxone-precipitated withdrawal from opioids (Glick et al., 1975), suggesting that the dST plays a role in opioid abuse and dependence. As mentioned above, transgenic D1 ×-MOR mice lacking MORs in MSN afferents showed significant alterations in heroin-driven behaviors: self-administration, reinstatement, locomotor response to heroin, but not in opioid-induced analgesia, CPP or withdrawal (Charbogne et al., 2017). Some of these alterations can be reversed by the ectopic rescue of MORs in the striatal direct pathway (Cui et al., 2014), which suggest that striatal neurons play a critical role in behaviors driven by opioids.

Recent studies showed that compulsive cocaine-, alcohol-, and heroin-seeking progressively recruits anterior dST (Everitt and Robbins, 2005; Hodebourg et al., 2019; Giuliano et al., 2019). Bilateral microinjections of flupenthixol (a non-selective DA antagonist) into the anterior dST can disrupt drug seeking but not initial drug taking (Everitt and Robbins, 2005; Giuliano et al., 2019; Hodebourg et al., 2019). Similarly, dorsolateral but not dorsomedial microinjections of SCH23390 (a D1R antagonist) attenuated context-induced reinstatement of heroin seeking in rats (Bossert et al., 2009), suggesting that heroin-relapse involves the striatal DA. In addition, the dST appears to be also involved in compulsive oxycodone seeking. Rats given long access to oxycodone self-administration exhibited incubation of oxycodone seeking after 4 weeks of abstinence. Compared to rats with short-access, rats with long access also showed greater mRNA expression of fibroblast growth factors (FGFs) and immediate early genes (c-fos and jun-B) in the dST in response to drug cues or context (Blackwood et al., 2019b). Rats with long access to oxycodone also showed decreases in protein levels of striatal MOR and delta receptors, as compared to saline and short access rats. In contrast, long access rats show an increase mRNA expression of striatal MORs (Blackwood et al., 2019a), suggesting that the dST undergoes neuroadaptation after long exposure to oxycodone. These findings suggest that the dST plays a critical role in opioid-driven behaviors. Given that SNr is one of the major outputs for dST neurons, the above studies also implicate the involvement of the dST→SNr pathway in opioid-related behaviors.

Integration of the mesolimbic and nigrostriatal systems

Although the mesolimbic system is deemed for its critical role in reward and motivation and the nigrostriatal system in reward and motor processes, both systems are integrated at key brain regions implicated in opioid reward, motivation and goal-directed behaviors. First and for most, both systems are integrated at the midbrain level through dendro-dendritic DA connections. DA neurons in the VTA and SNc are known for somatodendritic release of DA (Ford et al., 2010) and dendro-dendritic connections (Bayer and Pickel, 1990; Wilson et al., 1977). Immunocytochemistry assays indicate that unlike many neurotransmitters and neuromodulators that are synthesized in the soma and later transported to axon terminals, tyrosine hydroxylase (TH) can be seen in somatodendritic compartments and terminal fields (Witkovsky et al., 2009). In addition, regulation of TH activity by phosphorylation occurs in all parts of the cell, including soma, dendrites and axons, indicating that DA is synthesized for local somatodendritic or axonal release (Salvatore and Pruett, 2012). The release of dendritic DA has been also verified by fast-scan voltammetry (Chen and Rice, 2001; Cragg et al., 1997; Ford et al., 2010; Patel et al., 2009), microdialysis (Klitenick et al., 1992), and electrophysiology (Beckstead et al., 2007, 2004; Courtney et al., 2012; Ford, 2014; Ford et al., 2010). Dendritic DA release elicits inhibitory postsynaptic current in DA neurons that can be blocked by D2R antagonists (Beckstead et al., 2004) and can stimulate D1 and D2 receptors GABA terminals (Cameron and Williams, 1993; Koga and Momiyama, 2000). Interestingly, intra-VTA application of morphine reduces VTA GABA concentration and increases DA release (Klitenick et al., 1992). Therefore, it is conceivable that the integration of the mesolimbic and nigrostiratal systems occurs in the midbrain through dendrodentritic DA connections as they themselves play a role in opioid-driven behaviors.

The second potential region in which such integration might occur is the RMTg. The RMTg that sends major GABA projections to the VTA and SNc (Jalabert et al., 2011; Jhou et al., 2009b, 2009a) (Figure 3), whose DA neurons have been implicated in reward and addiction, as described above. RMTg GABA neurons are strongly inhibited by MOR agonists (e.g., morphine) and this is followed by disinhibition of midbrain DA neurons (Jalabert et al., 2011; Lecca et al., 2012, 2011; Matsui and Williams, 2011), leading to arousal and rewarding effects. Thus, the RMTg→VTA and RMTg→SNc projections might be the integral part of both systems (Figure 3), controlling the midbrain DA activity and greatly contributing to the acute effects of opioids.

The third potential region in which the mesolimbic and nigrostriatal system integrate to mediate opioid-motivated behaviors is the striatum. It is well established that the ventral striatum (NAc) is involved in motivated, goal-directed behaviors, including voluntary drug use, while the dST in habit formation and compulsive behaviors (Everitt and Robbins, 2016, 2005). During the initial voluntary drug use, the NAc is strongly implicated, as it receives dopaminergic projections from the VTA that controls the rewarding effects of drugs. As addiction progresses, the dST with its major dopaminergic innervations from the SNc takes over compulsive drug seeking that can persist for a prolonged time (Blackwood et al., 2019b; Bossert et al., 2009; Everitt and Robbins, 2016, 2013, 2005; Giuliano et al., 2019; Hodebourg et al., 2019; Suto et al., 2011; Zhang et al., 2018, 2017, 2014). A number of studies observed significant neuroplasticity in the NAc following exposure to drugs of abuse (Luscher et al., 2000; Wolf, 2016). Similar and other abnormalities were also observed in the NAc during late stages of addiction and compulsive drug seeking, including oxycodone taking and seeking (Zhang et al., 2018, 2017, 2014). Thus, the striatum that has historically been divided into the ventral part, implicated in rewarding effects of drugs, and the dorsal part, implicated in compulsive drug taking, might be another hub through which the mesolimbic and nigrostriatal systems integrate and together contribute to the development and maintenance of opioid addiction.

Conclusions

While the exact circuitry underlying opioid reward and relapse remains to be fully understood, the rewarding effects of heroin that lead to compulsive drug taking and seeking appear to depend on both mesolimbic and nigrostriatal systems. VTA DA neurons that send strong projections to the NAc play a critical role in opioid reward and relapse. The significant elements of the circuitry are strong GABA projections from the RMTg, NAc, VP and to some extend local interneurons that control neuronal activity of DA neurons. These projections with high density of MORs are readily inhibited by MOR agonists, which in turn leads to disinhibition of VTA DA neurons. Excitation of VTA DA neurons ultimately elicits arousal and strong rewarding effects. A more lateral part of the midbrain circuitry involves the SNc and SNr, whose neurons have been implicated in reward as well. SNr GABA neurons appear to be an important element of the SNr→SNc→dST circuitry that controls opioid-driven behaviors. Thus, cooperativity of the parallel systems that originate in the midbrain and send strong projections to the striatum contributes to the drug use initiation, development of addiction and maintenance of compulsive drug seeking.

Highlights:

• Opioid reward has long-time been believed to be mediated by inhibition of VTA GABA interneurons, which subsequently disinhibits VTA DA neurons

• This canonical two-neuron hypothesis has been upheld for over a half century, but challenged by recent research

• MORs are highly expressed in GABA neurons in the RMTg and SNr, but less in the VTA

• Activation of dopamine neurons in both the VTA and SNc is equally rewarding

• The RTMg→VTA→ventrostriatal and SNr→SNc→dorsostriatal pathways may act as the major neural substrates underlying opioid reward