The NAc is a key structure of the ventral striatum that is traditionally divided into central core and surrounding shell regions (NAcore and NAshell, respectively) based on heterogeneity in structure, function, and connectivity (Scofield et al., 2016; Záborszky et al., 1985; Zahm, 1999). Most neurons within the NAc (~90–95%) are GABAergic medium spiny neurons (MSNs) that can be differentiated by expression of dopamine D1- or D2-receptor expression (Gerfen & Surmeier, 2011). Remaining neurons include interneurons expressing parvalbumin, calretinin, and choline acetyltransferase, as well as those that co-express somatostatin, neuropeptide Y, and neuronal nitric oxide synthase (Burke et al., 2017; Tepper et al., 2010). Finally, the NAc contains glia that regulate extracellular glutamate and influence synaptic plasticity (Jarvis et al., 2020; Kalivas, 2009; Kruyer & Kalivas, 2021; McGrath & Briand, 2019).

Connectivity differs between the NAcore and NAshell. The NAcore integrates inputs from cortical and allocortical structures involved in reward processing, goal-directed behavior to earn reward, and reinstatement (Scofield et al., 2016). The majority of NAcore inputs arise from glutamatergic afferents including the prefrontal cortex (PFC), the insular cortex (IC), the basolateral amygdala (BLA), the hippocampus (HIPP) and midline thalamus (Brog et al., 1993; Sesack et al., 1989), whereas, the ventral tegmental area (VTA), ventral pallidum (VP), and to a minor extent PFC, provide GABAergic input (Lee et al., 2014; Scofield et al., 2016). Finally, VTA and substantia nigra (SN) provide dopaminergic inputs to NAcore (Brog et al., 1993). Reciprocal projections from the NAcore innervate the SN and VP, and to a lesser extent, lateral VTA (Bernal-Gamboa et al., 2017; Groenewegen et al., 1999; Heimer et al., 1991; Tripathi et al., 2010).

The NAshell integrates inputs from a variety of brain regions central to reward learning and motivation (Castro & Bruchas, 2019). Glutamatergic afferents of the NAshell include cortical (anterior insular cortex, dorsal peduncular cortex, infralimbic cortex, orbitofrontal cortex, and prelimbic cortex) and allocortical (BLA and ventral HIPP) regions, as well as a input from the paraventricular thalamus, with quantitatively smaller glutamatergic input from the VTA (Beckstead et al., 1979; Brog et al., 1993; Mcdonald, 1991; Sesack et al., 1989). The NAshell also receives dopaminergic input from the VTA, and GABAergic inputs from the central nucleus of the amygdala (CeA), lateral hypothalamus (LH), VP, and VTA (Brog et al., 1993; Groenewegen et al., 1999). NAshell D2-MSNs project primarily to the VP, but also innervate the VTA and LH, while D1-MSNs project to the VP, VTA and LH (Gibson et al., 2018; O’Connor et al., 2015).

Several functions within the NAc have direct relevance for understanding SUDs. The rewarding and reinforcing effects of different abused substances (Di Chiara & Imperato, 1988; Volkow et al., 2011c) and natural rewards (Volkow et al., 2011b; Wise, 2006) have long been associated with increased NAc dopamine release. However, individuals with SUDs often show paradoxical blunting of NAc dopamine in response to abused substances relative to nonaddicted individuals (Volkow et al., 1997, 2014) but increased NAc dopamine responses to drug-paired stimuli which correlates with drug craving (Volkow et al., 2014). This finding is supported by a study in rats where NAc lesions reduced drug SA mediated by conditioned stimuli, but had little effect on maintenance of SA without cues (Ito et al., 2004). This increased propensity for drug-paired stimuli to drive behavior has been termed incentive salience and is thought to contribute to the prepotency of drug seeking behavior (Koob & Volkow, 2016). NAc sensitivity to drug stimuli is of particular interest to SUD research due to the prominent role of drug-associated stimuli in occasioning drug craving and relapse.

NAc D2-MSNs are implicated in refraining from drug seeking (Roberts-Wolfe et al., 2018, 2019), and D2 receptor availability is reduced in individuals with SUDs (Volkow et al., 2002). Further, D2 receptor downregulation in individuals with SUDs is associated with decreased activity in cortical areas associated with decision making (Volkow et al., 2011a), and increases or decreases in D2 expression in animals results in reduced or increased sensitivity to abused substances, respectively (Bello et al., 2011; Thanos et al., 2001, 2008). These findings indicate a potential role for NAc in regulating compulsive seeking of abused substances, another hallmark symptom of SUDs.

Whether the nature of drug seeking is primarily goal-directed or habitual in individuals with SUDs is an area of growing contention. Drug seeking tends to prevail over other non-drug-related behaviors in individuals with SUDs, despite the finding that addicted individuals often describe a reduced desire to consume drugs (Berridge & Robinson, 2016). Further, drug seeking shows relative insensitivity to aversive consequences, devaluation, and reward unavailability in animal models, leading many to conclude that drug seeking becomes habitual in nature, such that drug-related stimuli drive a habitual response pattern that terminates in procuring and using drugs (Everitt & Robbins, 2016). However, a recent study indicates that drug seeking in humans tends to be highly flexible, potentially detracting from habit-based accounts of addiction (Hogarth, 2020). Regardless of the outcome of this debate, the NAc is likely to play a prominent role in the maintenance of persistent drug seeking. The NAc is involved both in habitual responding for drugs (Belin et al., 2013; Everitt et al., 2008) and in complex tasks related to outcome ambiguity or uncertainty (e.g. Mascia et al., 2019), and is thought to aid in response selection based on relative salience of inputs (Floresco, 2015). Thus, the NAc is likely involved in persistent drug-seeking behaviors regardless of whether they are described as habitual or goal-directed.

Finally, while the NAc has been shown to regulate behavior motivated by both drug and natural rewards (Kelley, 2004; Robinson & Berridge, 2000), drug-specific adaptations that occur within the NAc over the course of drug SA, extinction/abstinence, and relapse are not seen in identical sucrose SA experiments. Many studies have characterized drug-specific adaptations in pre- and post-synaptic compartments, extracellular matrix, and astroglia (i.e. the tetrapartite synapse) that lead to changes in glutamate signaling and in turn regulate drug-seeking behaviors (for detailed reviews see, Bobadilla et al., 2017b; Kruyer et al., 2020; Neuhofer & Kalivas, 2018). Taken together, existing data strongly suggest that the NAc is a critical brain nucleus for regulating the behavioral symptoms of SUDs.

Drug-induced adaptations in the NAcore are necessary for the expression of seeking behaviors related to SUDs (Scofield et al., 2016). Because the NAcore is also involved in behaviors motivated by natural reward (Berridge, 2009; Salamone et al., 2003), investigations directly comparing adaptations and the effects of manipulations on drug versus natural reward in NAcore are especially beneficial for understanding how drugs, but not natural rewards lead to pathological behaviors in SUDs. Alasmari et al. (2018) found adaptations in glial glutamate regulation in the NAcore induced by ethanol and nicotine, but not sucrose. They also demonstrated that nicotine + sucrose, nicotine + ethanol, and ethanol alone decreased expression of glutamate transporter (GLT-1) and cystine glutamate antiporter (xCT) and increased expression of metabotropic glutamate receptor (mGluR1) in the NAcore relative to sucrose only or water controls. These results are consistent with the important role of the NAcore glutamate in relapse (Kalivas, 2009), as well as the role of glia in regulating glutamate in drug, but not sucrose seeking (Kruyer et al., 2020). Martin et al. (2006) found that cocaine but not food SA inhibited long-term depression in both the NAcore and NAshell after only 1 day of forced abstinence, but selectively in the NAcore after 21 days of abstinence, suggesting that adaptations specifically in the NAcore may contribute to incubation of drug craving. Moreover, Cameron and Carelli (2012) used in vivo electrophysiological recordings to demonstrate that neurons in the NAcore exhibited mostly reward-specific phasic activity in a multiple schedule of cocaine and sucrose reinforcement. Re-exposure to the multiple schedule after 30d of abstinence revealed that most neurons remained selectively activated, but that the percentage coding for cocaine-motivated behaviors increased, and the percentage coding for sucrose-motivated behavior decreased. Further, while the relative increase in the number of neurons coding for cocaine and decrease in neurons coding for sucrose occurred in both the NAcore and NAshell, increased cocaine coding was more dramatic in the NAcore while decreased sucrose coding was more dramatic in the NAshell. A recent study by Bobadilla et al. (2020) used a c-Fos-TRAP (Targeted Recombination in Active Populations) strategy to investigate drug-specific NAcore neuronal ensembles alongside a within-subject behavioral model in which mice were trained to self-administer cocaine and sucrose across alternating sessions. This study showed that cued reinstatement of cocaine or sucrose seeking formed NAcore ensembles that were mostly distinct, overlapping by only ~30%, and that ensembles primarily consisted of D1-expressing neurons during cued reinstatement and D2-expressing neurons during extinction. Further, while cocaine and sucrose seeking ensembles were similar in size during cued reinstatement when each reward was administered independently, mice exposed to both rewards exhibited a larger cocaine-seeking than sucrose-seeking ensemble, and the magnitude of cued reinstatement was greater for cocaine than sucrose when both cues were simultaneously available. They also noted that the size of reward-specific ensembles was significantly correlated with the magnitude of reinstated behavior. These results suggest that D1- and D2-expressing NAcore neurons encode seeking and refraining from seeking, respectively, and that mostly non-overlapping neuronal ensembles in the NAcore code for cocaine versus sucrose seeking. Finally, these data indicate that differences in recruitment to reward-specific ensembles may explain preferential relapse effects for drug compared to natural reward. Together, the studies by Alasmari et al. (2018), Bobadilla et al. (2020), Cameron and Carelli (2012), and Martin et al. (2006) suggest that drugs of abuse can lead to adaptations in the NAcore that are not shared with natural rewards, that these adaptations correspond with behavior, and that adaptations within NAc subregions may differ for drug versus natural rewards.

Consistent with the findings by Cameron and Carelli (2012), other data indicate that substances of abuse induce adaptations in the NAcore during cued reinstatement (Koya et al., 2006; Madsen et al., 2012) contextual reinstatement (Edwards et al., 2011) and CPP (Mattson & Morrell, 2005; Zombeck et al., 2008) that are not recapitulated by natural reward. Likewise, several studies show that perturbations of NAcore activity preferentially affect responding motivated by drug but not natural reward. McFarland and Kalivas (2001) found that GABA agonism within the NAcore prevented primed reinstatement of cocaine but not sucrose seeking. Spencer et al. (2014) showed that reducing excitatory transmission in NAc by inhibiting voltage-gated calcium channels attenuated primed but not cued reinstatement of cocaine seeking, and did not affect primed sucrose seeking. Sinclair et al. (2012) found that blocking NAcore glutamate receptor mGluR5 prevented cued reinstatement of alcohol but not sucrose seeking. Similarly, Peters and Kalivas (2006) showed that while mGluR2/3 antagonism reduced primed reinstatement of both cocaine and sucrose seeking, cocaine primed reinstatement was more substantially reduced than sucrose reinstatement. Czachowski (2005) showed that NAcore administration of serotonin_1B agonist decreased ongoing alcohol seeking, but had little effect on consumption of freely-available alcohol. Conversely, a serotonin_1A agonist decreased alcohol consumption, but not ongoing seeking. Importantly, neither agonist had an effect on sucrose consumption or seeking. Consistent with the findings of Alasmari et al. (2018) discussed above, Scofield et al. (2015) found that chemogenetic activation of NAcore glia inhibited cued reinstatement of cocaine but not sucrose seeking. Building on this data, Kruyer et al. (2019) discovered that glia were retracted from NAcore synapses following extinction of heroin but not sucrose seeking, and that cued reinstatement of heroin, but not sucrose seeking produced a transient reassociation of glia with synapses. Further, blocking glial reassociation with synapses reduced cued heroin seeking. Taken together, these results strongly indicate that the necessity of NAcore is drug selective, and support the hypothesis that the NAcore is an essential nucleus for regulating the behavioral symptomology of SUDs.

Contrary to the drug selectivity of the NAcore, studies examining the NAshell produced mixed results. Some data indicate drug selective adaptations (Crombag et al., 2005, 2008; Mattson & Morrell, 2005) and others indicate similar adaptations between drug and natural reward (Madsen et al., 2012; Zombeck et al., 2008). Manipulation studies also find mixed results, with some studies demonstrating a necessity for the NAshell in regulating behaviors motivated by drug but not natural reward. Pascoli et al. (2014) found that neurons projecting from infralimbic cortex to NAshell showed cocaine-evoked plasticity (i.e. reduced AMPA/NMDA ratio and increased rectification mediated by GluA2 lacking AMPA receptors). Optogenetic reversal of plasticity in these neurons prior to testing eliminated cue-induced reinstatement of cocaine but not sucrose seeking. Liechti et al. (2007) found downregulation of mGluR2/3 in NAshell following nicotine but not sucrose SA, and mGluR2/3 antagonism reduced ongoing nicotine but not sucrose seeking. McFarland et al. (2004) found that GABA agonism in NAshell reduced stress-primed cocaine seeking but not food-primed food seeking. However, other data indicate that the NAshell is necessary for both drug- and natural-reward-motivated behaviors. Liechti et al. (2007) reported that mGluR2/3 antagonism in the NAshell decreased cued reinstatement of both nicotine and sucrose seeking. Similarly, Guercio et al. (2015) found that deep brain stimulation of the NAshell attenuated cued reinstatement of both cocaine and sucrose seeking. Thus, the role of the NAshell in governing drug versus natural-reward motivated behaviors is not entirely clear. One potential explanation for these mixed results could be that only particular NAshell subcircuits, cell types, or neurotransmitter systems show drug selectivity. This possibility is supported by the drug selectivity of IL-NAshell (Pascoli et al., 2014) but not BLA-NAshell (Millan et al., 2017; reviewed below) projections in regulating cocaine seeking. Another potential explanation is that differences in methodology (e.g. deep brain stimulation vs. reversible inactivation, differences in behavioral paradigms) are responsible for discrepant findings.

The prefrontal cortex is critical for decision making and executive control, and regulates reward seeking in part via glutamatergic projections to the NAc (Kalivas et al., 2005). Involvement of the PFC in SUDs is well established, where deficits associated with attribution of salience, impulsivity, motivational arousal, and self-control are accompanied by increased PFC activation by drugs or drug cues, blunted PFC activation by natural reward, and poor performance on PFC-mediated cognitive tasks (Goldstein & Volkow, 2011). Further, individuals with an extended recreational cocaine-use history but not a diagnosable SUD show increased PFC grey matter volume while individuals with cocaine use disorder show reduced PFC grey matter volume relative to non-drug-using controls (Ersche et al., 2013a). These results have led some to suggest that variations in PFC functionality may confer vulnerability or resilience to developing SUD following exposure to abused substances (Ersche et al., 2013b, 2020). Recent animal studies provide some support for this hypothesis. PFC-dependent task performance and PFC orexin receptor expression prior to beginning methamphetamine SA correlated with future methamphetamine preference (Tavakkolifard et al., 2020). Similarly, De Laat et al. (2018) found that pre-SA performance on a cognitive task and levels of PFC glutamate and glycine were associated with future rates of cocaine intake. Together these studies indicate a potential role for PFC in conferring vulnerability or resilience to the development of SUDs.

The PFC has also been described as a critical component of the final common pathway for relapse of drug seeking (Kalivas & Volkow, 2005). A serial pathway from dorsomedial PFC, to NAcore and VP is necessary for drug seeking, as perturbations of this circuit reduce seeking across different types of triggering events and drug classes (e.g. Cordie & McFadden, 2019; Doncheck et al., 2020; Hernandez et al., 2020; Ma et al., 2014; Palombo et al., 2017; Struik et al., 2019), highlighting the importance of PFC for the expression of relapse to drug seeking. Overall, these studies strongly suggest that PFC mediates pathological behaviors associated with SUDs.

Other studies have identified the role of heterogeneous regions of the prefrontal cortex in drug seeking and relapse. While a complete characterization of the role of these subregions in SUDs is beyond the scope of this review, a brief overview is provided here to aid in the discussion of drug selective effects below. PFC subregions with a notable role in SUDs and in which studies investigating drug vs natural reward exist include medial PFC prelimbic (PL) and infralimbic (IL) cortices, and the lateral prefrontal insular cortex (IC). In general, existing data implicate PL in relapse to drug seeking (Moorman et al., 2015). IL is involved in suppression of drug seeking (Peters et al., 2008), and experimentally manipulating IL activity affects relapse of drug seeking (Augur et al., 2016; Ma et al., 2014; Muller Ewald & LaLumiere, 2018). Further, Warren et al. (2016) showed that distinct ensembles of IL neurons control food SA and extinction, and the same lab later showed that distinct ensembles regulating cocaine SA and extinction are mostly composed of IL-NAcore and IL-NAshell projection neurons, respectively. Finally, IC is thought to be involved in interoception and drug craving (Paulus & Stewart, 2014) and is strongly implicated in nicotine craving in smokers (Droutman et al., 2015; Kenny, 2011). Overall, these data suggest that heterogeneous cortical subregions are important for regulating drug-related behaviors in individuals with SUDs.

Drug selective adaptations have been noted in medial prefrontal cortex regions (mPFC; PL, IL). Koya et al. (2006) found increased immediate early gene (IEG) expression in mPFC following cued reinstatement for heroin but not sucrose relative to extinction controls. Wedzony et al. (2003) also noted increased Fos expression in mPFC following protracted abstinence from alcohol but not sucrose SA, specifically near the border of PL and IL (i.e. ventromedial PFC). In addition, Crombag et al. (2005) found increased spine density in the mPFC following amphetamine SA relative to sucrose self-administering and untreated controls. Using an innovative in vivo single-cell calcium imaging approach (discussed below), Siciliano et al. (2019) characterized activity patterns in neurons projecting from the mPFC to the dorsal periaqueductal grey that predicted susceptibility to compulsive alcohol drinking. These observations led the authors to optogenetically manipulate the circuit and conclude that compulsive alcohol seeking is likely driven by reduced aversion signaling. Importantly, they showed that manipulating this circuit had selective effects on alcohol but not water seeking. Because drug selective adaptations occur in several mPFC subregions, and manipulations of neurons in the mPFC affect behaviors motivated by drug but not natural reward, we will next explore the whether mPFC subregions are selectively necessary for behaviors motivated by drug but not natural rewards.

The amygdala (AMY) is critical for memory-processing, emotional responses, decision making, and drug seeking via connections to the NAc and PFC (Peters et al., 2009). The amygdala has been implicated in addiction, with particular roles in negative affect during withdrawal and preoccupation with abused substances (Koob & Volkow, 2016). Indeed, simultaneous downregulation of PFC and increased activity in AMY likely confer vulnerability and contribute to negative affect and relapse (Ruisoto & Contador, 2019). Like NAc and PFC, activity in AMY is increased when individuals with SUDs are exposed to cues related to abused substances (Jasinska et al., 2014) and resting state functional connectivity is increased in chronic heroin users relative to non-using controls (Ma et al., 2009). Finally, a review of neuroimaging studies by Mihov and Hurlemann (2012) revealed consistent increases in AMY activity during abstinence from nicotine, greater reactivity to nicotine-paired cues than neutral cues, reduced AMY activation by harm signals, and increased AMY activity during relapse prevention therapy in smokers. These findings led the authors to suggest nicotine cue-reactivity and decreased sensitivity to harm-signals as potential vulnerability biomarkers for smoking relapse. Taken together, the studies above suggest that AMY plays a central role in the pathological behaviors associated with SUDs, with particular involvement in stress and emotional processing. Importantly, these effects are mediated by different subregions within AMY, with components of the extended amygdala (i.e. central nucleus of the amygdala [CeA], bed nucleus of the stria terminalis [BNST]) controlling negative affect and stress reactivity (Centanni et al., 2019), and the basolateral amygdala mediating cue reactivity (See et al., 2003) and reward valuation (Wassum & Izquierdo, 2015). Thus, we will explore the necessity of these subregions for regulating behaviors motivated by drug and natural reward next.

The hippocampus (HIPP) is heavily implicated in learning and memory and sends glutamatergic projections to NAshell (ventral HIPP), and to a lesser extent to NAcore (dorsal HIPP; Britt et al., 2012; Groenewegen et al., 1987; Kelley & Domesick, 1982). HIPP is also involved in SUDs, where increased activity is associated with the formation of salient drug-stimulus associations and decreased activity is associated with drug withdrawal, potentially contributing to relapse (Kutlu & Gould, 2016). In addition, several classes of abused substances can affect neurogenesis in HIPP, which is thought to contribute to inflexible decision making, negative affect, and relapse in SUDs (Canales, 2012). Indeed, blocking neurogenesis in HIPP increased cocaine SA and cued reinstatement in mice (Deroche-Gamonet et al., 2019). A recent review emphasized the role of HIPP in stress-, context-, and cue-induced relapse, and suggested that HIPP may regulate drug SA and relapse via inputs to PFC (Goode & Maren, 2019). In addition, ventral HIPP projections to NAshell regulate cocaine seeking after abstinence (Pascoli et al., 2014). Overall, these findings indicate that HIPP plays an important role in the behavioral symptomology of SUDs.

The ventral tegmental area (VTA) is the major source of mesocorticolimbic dopamine, is critical for reward processing, motivational salience, and learning, and sends inputs to the NAc (Russo & Nestler, 2013). The role of VTA in drug- (Lüscher & Malenka, 2011) and natural-reward-motivated behavior (Morales & Margolis, 2017) is well established, and abused substances modulate both excitatory and inhibitory effects within VTA (Oliva & Wanat, 2016). In humans, Gu et al. (2011) showed that VTA resting state functional connectivity was reduced in cocaine users relative to healthy controls, and notably VTA connectivity to NAc was reduced. While these results indicate a role for VTA in SUDs, it is important to assess drug selectivity of VTA due to its known involvement in mediating behavior motivated by natural reward.

The thalamus mediates cognitive function as well as goal-directed and motor behaviors via projections to the PFC and striatum which also establish a role for the thalamus in reward circuitry (Huang et al., 2018). A recent review by Huang et al. (2018) highlights several adaptations in the thalamus of individuals with SUDs. First, grey matter is reduced and this reduction is correlated with drug craving, length of substance use, time in abstinence, and relapse. Differences in thalamic activity were observed with a variety of methods, and differences in functional connectivity were noted between the thalamus and other regions discussed here, including AMY, NAc and PFC. Thalamic activity was reduced in response inhibition tasks, and this decrease was correlated with SUD severity. Finally, thalamic activity was increased in response to drug-paired stimuli in individuals with SUDs relative to healthy controls.

The hypothalamus also plays an important role in regulating behaviors characteristic of SUDs. Orexin/hypocretin is a neuropeptide originating the lateral hypothalamus (LH) and orexin has been shown to play an important role in highly-motivated responding (including for drugs) and in negative affect, stress, and anxiety (Hopf, 2020). Interestingly, orexin neurons appear to be activated preferentially by cocaine but not highly palatable food in one study (Matzeu & Martin-Fardon, 2018), but other studies find mixed evidence for activation of orexin neurons by natural reward (for discussion see, Hopf, 2020). Finally, Orisni et al. (2018) reported that increased functional connectivity during early abstinence from cocaine vs. sucrose seeking in rats varied with respect to subregions of the thalamus and hypothalamus. Thus, regions within the diencephalon appear to play a role in characteristic behaviors of SUDs. We explore drug selectivity in some of these regions of the diencephalon next.

The ventral pallidum (VP) has reciprocal connections with NAc and VTA and sends outputs directly to the thalamus (Smith et al., 2009). VP is implicated in both hedonic “liking” and incentive motivational “wanting” of reward (Smith et al., 2009), as well as drug seeking, stimulus discrimination, working memory, and relapse (Root et al., 2015). Thus, VP is an important nucleus in regulating behaviors motivated by both drug and natural reward. Heinsbroek et al. (2020) reported that specific cell types in VP are involved in refraining from drug seeking and cued reinstatement. Specifically, in vivo single-cell calcium imaging of VP glutamate neurons revealed the highest activity during extinction, and chemogenetic stimulation of these neurons attenuated cued reinstatement of cocaine seeking. VP GABA and enkephalin neurons were most active during cued reinstatement, and chemogenetic stimulation of these neurons reinstated cocaine seeking. Similarly, Creed et al. (2016) showed that VP synapses from NAc D1-neurons were potentiated while synapses from NAc D2-neurons were depressed following cocaine exposure. Further, they found that reversing potentiation of NAc D1-neuron synapses in VP reduced cocaine sensitization, and reversal of depression in NAc D2-neuron synapses in VP increased motivation and decreased negative affective responses to natural reward (i.e. orofacial responses to sucrose). Together, these studies establish that VP is critically involved in mediating behavioral symptoms of SUDs.

One limitation of the current review is that the extant studies examined only comparisons between drug reward and non-pathological behaviors motivated by natural reward. A growing literature is directed at understanding the similarities and differences in neurobiological factors contributing to so-called behavioral addictions (e.g. gambling disorders, eating disorders, exercise addiction) and SUDs (e.g., Chamberlain et al., 2016; Hadad & Knackstedt, 2014; James & Tunney, 2017). While recent discussions have drawn parallels between the characteristic behaviors and neurobiological mechanisms of both pathologies (e.g., Fletcher & Kenny, 2018; Kuhn et al., 2019), studies employing protocols for directly comparing pathological natural and drug reward seeking do not yet exist.

An obvious limitation is that there exist relatively few studies directly comparing the necessity of a brain nucleus or circuit in drug and natural reward. Below we outline behavioral paradigms that will facilitate this comparison in future studies. The vast majority of the work we discuss compared reinstated seeking of a drug (most often cocaine) in one group versus sucrose in another group and employed relatively broad manipulation techniques to assess the necessity of the specified region for controlling behavior. Thus, future studies utilizing more finely-grained analyses are needed to refine the proposed drug selective circuitry we compile in Figure 1C. In the following sections, we discuss promising protocols and technologies that will allow investigators to draw more nuanced conclusions regarding the role of brain circuits in regulating drug versus natural reward seeking.

Polysubstance use is highly prevalent among people suffering from substance use disorders (Crummy et al., 2020; Liu et al., 2018). The rise of the opioid epidemic in the United States has indeed precipitated polydrug-induced overdoses, notably due to the use of opiates as drug-cutting ingredients (Meier et al., 2020; Nolan et al., 2019). Besides polysubstance abuse, humans have complex lives comprising many sources of non-drug rewards, such as food, water, social interaction, or sex. Like drugs, these rewards drive and influence behavior constantly. It is therefore important for future studies to determine how the simultaneous use of multiple drugs or competing rewards affects the brain. Mounting evidence indicates that chronic exposure to unmixed drugs of abuse induces drug selective effects that may occur only at particular stages of addiction (e.g., Cameron & Carelli, 2012; De Laat et al., 2018; Yager et al., 2019), in particular modalities of relapse (e.g., Spencer et al., 2014; Wunsch et al., 2017), or in particular subsets of subjects (e.g., Cain et al., 2008; Hernandez & Moorman, 2020). Thus, strategies that allow for drug:drug and drug:natural-reward comparisons in the same animal across experimental phases will shed light on how individual differences and drug-induced adaptations in neurobiology selectively regulate drug reward, drug seeking, and refraining from drug seeking. Making comparisons between drugs of different classes or drugs and natural rewards in animal models can be challenging due to differences in routes of administration (e.g. oral versus intravenous), onset of action (e.g. quick for cocaine, slow for alcohol), and direct effects different rewards on behavior (e.g. increased activity by stimulants and decreased activity by depressants). However, these difficulties can be minimized with careful arrangement of experimental paradigms and validation of interesting findings in more societally-relevant paradigms. Indeed, any interesting findings with the potential to further understanding of the neurobiological underpinnings of SUDs or contribute to the development of novel therapies should be validated using procedures high in face validity.

To increase the face validity of animal models of SUDs, preclinical models designed to investigate specific endophenotypes underlying SUDs have recently emerged, including drug-induced alterations in motivation for natural reward (e.g., Creed et al., 2016; Hart et al., 2018), relapse following analogues of human behavioral interventions using natural rewards (e.g., Nall et al., 2018; Venniro et al., 2019), and resistance to aversive effects of drugs (e.g., Marchant et al., 2019; Nall & Shahan, 2020). However, there is currently little work directly comparing drug versus natural reward using these preclinical models. Thus, while polyreward models are more complex than non-mixed reward paradigms, they more closely approximate key features of SUDs in humans. Future work using these behavioral models along with the finely-grained biological assays discussed below could be fruitful for determining how drugs of abuse, but not natural reward, lead to pathological behaviors indicative of SUDs.

The majority of the studies discussed in this review compared natural reward with drug reward for purposes of controlling for effects mediated by the behavioral protocol. However, it is becoming increasingly apparent that groups of neurons (i.e. neuronal ensembles) within the same brain nucleus of an animal may encode both drug and natural rewards, while other ensembles may selectively motivate responding for one or another reward (Bobadilla et al., 2020; Cameron & Carelli, 2012; Carelli et al., 2000; Cruz et al., 2013; DeNardo & Luo, 2017; Kane et al., 2020; Pfarr et al., 2018). While some studies indicate a relatively small overlap in drug and natural-reward ensembles (e.g., Bobadilla et al., 2020; Cameron & Carelli, 2012), others indicate larger overlap (e.g., Pfarr et al., 2018). Ensembles may also differentially code behaviors related to SA and refraining from seeking during extinction. Indeed, there is evidence that self-administration and extinction of both cocaine (Warren et al., 2019) and food seeking (Warren et al., 2016) depend, in part, on mostly independent ensembles of IL neurons. Further, cocaine SA and extinction ensembles are composed primarily of IL-NAcore and IL-NAshell projection neurons, respectively (Warren et al., 2019). As such, comparisons of individual neuron activity within-subject could provide particularly precise data regarding drug selectivity and differential control over drug-related behaviors. Further, a variety of approaches, such as TRAP (DeNardo et al., 2019) or Calcium Modulated Photoactivatable Ratiometric Integrator (CaMPARI; Moeyaert et al., 2018) strategies have been developed that allow these neuronal ensembles to be tagged and manipulated in vivo, providing rich data on how specific cellular ensembles control motivated behavior (for discussion see, Cruz et al., 2013; Whitaker & Hope, 2018).

Individual cell types can also differentially control behavior. For example, a growing body of evidence indicates differential roles for D1- and D2-MSNs in NAcore in regulating behavior, with specific D1-MSN circuits mostly promoting drug seeking and D2-MSN circuits mostly inhibiting seeking (e.g., Bock et al., 2013; Gibson et al., 2018; Heinsbroek et al., 2017; Kravitz et al., 2012; Lobo & Nestler, 2011; Roberts-Wolfe et al., 2018). The dual-reward study discussed above by Bobadilla and colleagues (2020) revealed that reward- seeking ensembles were mostly comprised of D1-MSNs within NAcore during seeking, and mostly D2-MSNs during extinction, showing that opposing control over behavior by NAcore D1- and D2-MSNs is also recapitulated in recruitment of those cell types to specific cellular ensembles. Recent evidence also indicates that individual cell types in VP differentially control behavior. As detailed above, Heinsbroek et al (2020) used a calcium imaging approach to demonstrate that activity in glutamatergic VP cells was high while activity in GABAergic or an enkephalin-expressing subpopulation of GABAergic VP cells was low during extinguished cocaine seeking. Chemogenetic stimulation of VP glutamate cells attenuated extinguished cocaine seeking, while stimulation of VP GABA and enkephalin cells reinstated cocaine seeking. Importantly, recent follow-up experiments showed that while activation of VP glutamate cells also inhibited sucrose seeking, activation of enkephalin neurons did not augment sucrose seeking and activating GABA neurons produced mixed results on sucrose seeking (Figure 1A). These data also highlight the power of the relatively new in vivo single-cell calcium imaging approach, that can provide information on cellular activity in real time across different phases of drug-seeking (see also, Siciliano & Tye, 2019). Further, calcium imaging allows for tracking cells over time, and clustering of individual cells into ensembles based on similar patterns of calcium activity. Such analyses can provide rich within-subject data regarding changes in ensemble activity across experimental phases and in response to both drug and natural reward. Together, the studies discussed in this section indicate the utility of studying drug selective and nonselective involvement of brain circuits in neuronal subpopulations within each nucleus or in subpopulations that have distinct axon terminal fields or unique activity patterns. Ideally, these more nuanced analyses would be utilized in conjunction with a within-subjects protocol (see above) to directly compare behaviors motivated by drug and natural rewards across experimental phases.