Neural Substrates and Circuits of Drug Addiction

Abstract

Drug addiction is a chronic relapsing disorder, and a significant amount of research has been devoted to understand the factors that contribute to the development, loss of control, and persistence of compulsive addictive behaviors. In this review, we provide an overview of various theories of addiction to drugs of abuse and the neurobiology involved in elements of the addiction cycle. Specific focus is devoted to the role of the mesolimbic pathway in acute drug reinforcement and occasional drug use, the role of the mesocortical pathway and associated areas (e.g., the dorsal striatum) in escalation/dependence, and the contribution of these pathways and associated circuits to conditioned responses, drug craving, and loss of behavioral control that may underlie drug relapse. By enhancing the understanding of the neurobiological factors that mediate drug addiction, continued preclinical and clinical research will aid in the development of novel therapeutic interventions that can serve as effective long-term treatment strategies for drug-dependent individuals.

Drug addiction is a chronic relapsing disorder characterized by compulsive drug-seeking and drug-taking behaviors, loss of control over drug intake, and the emergence of negative affect (e.g., dysphoria, anxiety) when access to the drug is withheld (Koob and Le Moal 1997; Koob and Volkow 2010).⁵ According to the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition (American Psychiatric Association 2013), substance use disorders consist of symptoms primarily related to the inability to reduce or control drug use. The most recent National Survey on Drug Use and Health by the Substance Abuse and Mental Health Services Administration (SAMHSA) (2019) estimates that 53 million Americans 12 years of age or older, or 19.4% of the population, reported past year illicit drug use; 73 million Americans (26.7%) reported past year tobacco product use; and 179 million Americans (65.5%) reported past year alcohol use. Within these populations, SAMHSA estimates that 20.2 million Americans (7.4%) had a substance or alcohol abuse or dependence disorder. These estimates include 2.6 million Americans (1%) classified as dependent on or abusing both alcohol and illicit drugs, 5.4 million (2%) dependent or abusing illicit drugs but not alcohol, and 12.1 million (4.4%) dependent on or abusing alcohol but not illicit drugs.

Although the overall trend of substance abuse patterns has been relatively stable, the number of emergency department visits involving attention-deficit/hyperactivity disorder (ADHD) stimulant medications increased by 133% between 2005 and 2010 (Substance Abuse and Mental Health Services Administration 2013), and the number of suspected opioid overdoses increased by 30%–70% in some regions of the United States (Centers for Disease Control and Prevention 2018). Notably, cannabis remains the most prevalently used illicit drug with 40.4 million Americans reporting past year use (16.2%). Furthermore, recent legalization of recreational and medical cannabis has resulted in a small but significant increase in lifetime, past year, and past month cannabis use in Americans 12 years of age and older (Substance Abuse and Mental Health Services Administration 2019).

There is a high probability that an individual will relapse to drug taking following months or years of abstinence (Dackis and O'Brien 2001; Wagner and Anthony 2002). Given the detrimental societal and economic impact of substance abuse and addiction, significant research has been dedicated to understanding the neurobiological mechanisms that mediate the development and persistence of substance use disorders. In this review, we provide a brief overview of the theory, stages, neurocircuitry, and neuropharmacological mechanisms that likely underlie stages of the addiction cycle (e.g., acute reinforcement/drug use, escalation/dependence, withdrawal/relapse).

THEORIES AND STAGES OF ADDICTION

Early theories suggested that addictive behaviors develop in response to the acute rewarding effects of drugs of abuse, with dependence occurring as a function of a recurrent drive for reward (Wise 1980). Although positive reinforcement is initially involved in the development of a substance use disorder, long-term drug abuse often results in the emergence of aversive psychological and physiological effects if the drug is withheld, resulting in continued use as a means to avoid the aversive consequences of drug withdrawal (i.e., negative reinforcement) (Wikler 1973; Camí and Farré 2003). Thus, addictive behaviors reflect a gradual shift from positive reinforcement (impulsivity) to negative reinforcement (compulsivity) (Koob 2004). These conditioning theories cannot fully explain some aspects of drug dependence, such as the resumption of drug-seeking and drug-taking behaviors following a prolonged period of abstinence (i.e., relapse) after the dissipation of overt withdrawal symptoms. Thus, it has been theorized that prolonged drug use leads to a series of neuroadaptations that contribute to the enduring nature of the addictive state.

Robinson and Berridge (Robinson and Berridge 1993, 2008; Berridge and Robinson 1995) postulated in their “incentive sensitization” theory of drug addiction that chronic exposure to drugs of abuse results in alterations in a number of neural systems, including areas normally involved in motivation for natural reinforcers. These neural adaptations increase the sensitivity to drug-associated stimuli (Clark and Overton 1998) and lead to a shift from drug “liking” to “wanting,” with ensuing compulsive patterns of drug-seeking behavior. Alternatively, Koob and Le Moal (1997, 2001) hypothesized that chronic drug use shifts an individual's hedonic set point and elicits allostasis, a state of dysregulation characterized by enhanced sensitivity to drug-associated stimuli and drug opposite neural adaptations in brain reward systems. In allostasis, the ability to maintain homeostasis through adaptive change becomes disrupted, leading to a loss of control over drug intake and compulsive use. Other theories have centered on specific drug-induced neuroadaptations that may underlie maladaptive associative learning (Di Chiara 1999; Hyman and Malenka 2001), loss of prefrontal cortical control over behavior and decision-making (Jentsch and Taylor 1999; Franklin et al. 2002; Goldstein and Volkow 2002), aberrant stimulus–response learning resulting in drug habit formation (Wise 2002; Everitt and Robbins 2005; Volkow et al. 2006), and dysfunction in memory systems leading to pathologically salient and intrusive drug memories (Taylor et al. 2009; Sorg 2012; Everitt et al. 2018). Notably, none of these theories fully account for every element of the addiction cycle.

In general, drug addiction can be considered to consist of three stages: acute reinforcement/drug use, escalation/dependence, and withdrawal/incubation/relapse. Although drug use does not necessarily result in drug dependence (e.g., social smoking/drinking), the acute appetitive effects of some drugs of abuse have some predictive validity for the transition to later stages of the addiction cycle. Importantly, the stages of drug addiction interact with one another and culminate in persistent drug addiction–related behaviors (Koob and Volkow 2010, 2016). Although significant advances in human neuroimaging and related techniques have allowed researchers to further our understanding of the neural substrates that underlie addiction, this review will primarily focus on data derived from various animal models to provide a broad overview of the neurobiology thought to underlie development of the different stages of addictive behaviors.

Drugs of abuse are generally classified into categories, including narcotics (e.g., opiates), cannabinoids (e.g., marijuana), depressants (e.g., ethanol), psychostimulants (e.g., nicotine, amphetamines, and cocaine), hallucinogens (e.g., lysergic acid diethylamide and ecstasy), and inhalants (e.g., toluene and nitrous oxide). Although all drugs of abuse have the potential to produce feelings of euphoria and relieve negative emotional states (Nesse and Berridge 1997), their behavioral and neuropharmacological properties are highly diverse. However, in terms of the mechanism for their acute reinforcing effects, most researchers have focused their attention on the mesocorticolimbic dopaminergic (DA) pathway (Wise 1980). This system, which mediates the rewarding effects of both natural reinforcers (e.g., food, drink, and sex) and drugs of abuse, consists of DA cell bodies that project from the ventral tegmental area (VTA) to various limbic (i.e., the mesolimbic pathway, including the nucleus accumbens [NAc], ventral pallidum, amygdala, hippocampus, and the bed nucleus of the stria terminalis [BNST]) and cortical (i.e., the mesocortical pathway, including the prefrontal cortex [PFC], orbitofrontal cortex [OFC], and the anterior cingulate) structures. Various drugs of abuse interact with this circuit at different levels (Camí and Farré 2003) through complex interactions with other neurotransmitter/neuromodulator systems, including opioid peptides, γ-aminobutyric acid (GABA), glutamate, and endocannabinoids. All of these play a significant role in the reinforcing effects of drugs of abuse (especially nonpsychostimulants) (Volkow et al. 2019). The mesolimbic and mesocortical circuits operate in parallel, with somewhat different roles in the addiction process (Camí and Farré 2003). Within the mesolimbic pathway, the NAc (Di Chiara 2002) and ventral pallidum are involved in the primary reinforcing effects of drugs of abuse (Volkow et al. 2003), whereas the amygdala (i.e., basolateral amygdala [BLA] [See 2005]) and hippocampus (both the dorsal [Fuchs et al. 2005; Meyers et al. 2006; Atkins et al. 2008] and ventral [Rogers and See 2007; Lasseter et al. 2010b] subregions) play key roles in discrete and/or contextual drug-associative learning. Within the mesocortical pathway, the anterior cingulate, prelimbic cortex, and OFC are involved in the regulation of emotional responses, cognitive control, and executive function (Volkow et al. 1993). Long-term drug use results in cellular adaptations in the PFC-NAc glutamatergic pathway, leading to persistent addictive behaviors, including the devaluation of natural rewards, diminished cognitive control, and hyperresponsiveness to drug-associated stimuli (Koob and Le Moal 2001; Kalivas and Volkow 2005; Carelli and West 2014). With respect to the stages of addiction outlined above, these data collectively suggest that the mesolimbic pathway (especially the VTA and NAc) is involved in acute drug reinforcement and occasional drug use and the mesocortical pathway and associated areas (e.g., the dorsal striatum) are involved in escalation/dependence, whereas both pathways and associated areas are involved in mediating conditioned responses, drug craving, and loss of behavioral control thought to underlie withdrawal and relapse.

NEUROCIRCUITRY INVOLVED IN ACUTE REINFORCEMENT/DRUG USE

In general, addiction theories suggest that acute reinforcement produced by drugs of abuse involves direct or indirect enhancement of DA neurotransmission in the NAc (e.g., Di Chiara and Imperato 1988; Koob and Bloom 1988). Interestingly, this system is similarly involved in the motivation and drive for natural reinforcers (Wise 2002); however, drugs of abuse produce more robust increases in accumbal DA that, unlike those produced by natural reinforcers, do not undergo adaptive change (i.e., tolerance) with repeated administration (Di Chiara 1999). These findings suggest that initiation of the addiction cycle involves “hijacking” of the reward system by drugs of abuse (Robbins and Everitt 1996) that is persistent and leads to the associative learning and the development of conditioned responses to drug-associated stimuli (Berke and Hyman 2000) that further promote the repeated use of the addictive substance.

Much of our neurobiological understanding of the acute rewarding effects of drugs of abuse can be traced to the discovery of a fundamental brain reward system by Olds and Milner (1954). These researchers showed that animals will perform instrumental responses (e.g., lever presses) to initiate the delivery of mild electrical current into a number of brain areas, the most sensitive of which being the medial forebrain bundle projecting from the VTA to the basal forebrain (Olds and Milner 1954). Research using the intracranial self-stimulation model has demonstrated that acute administration of drugs of abuse reduces brain stimulation reward thresholds (i.e., increased reward), and drugs with greater abuse potential produce greater reductions in reward thresholds (Kornetsky et al. 1979; Kornetsky and Bain 1990).

Further evidence for the role of DA and the mesolimbic pathway in acute drug reinforcement has been provided from drug self-administration studies, in which animal subjects perform instrumental responses to initiate drug delivery. Using this model in combination with in vivo microdialysis, studies have shown increases in extracellular DA in the NAc during psychostimulant (Hurd et al. 1989; Pettit and Justice 1989; Weiss et al. 1992; Di Ciano et al. 1995; Meil et al. 1995; Pontieri et al. 1995, 1996; Wise et al. 1995b) and opiate (Hemby et al. 1995; Pontieri et al. 1995; Wise et al. 1995a) self-administration. Similar effects occur during the self-administration of a variety of other drugs of abuse, including nicotine (Pontieri et al. 1996; Lecca et al. 2006b), cannabinoids (Fadda et al. 2006; Lecca et al. 2006a), and ethanol (Weiss et al. 1993; Melendez et al. 2002; van Erp and Miczek 2007). Causal evidence for the involvement of DA comes from studies indicating that systemic administration of DA synthesis inhibitors (Pickens et al. 1968; Wilson and Schuster 1974) or DA receptor antagonists (Yokel and Wise 1975; Woolverton 1986; Corrigall and Coen 1991; Rassnick et al. 1992; Richardson et al. 1994) reduces the self-administration of a variety of drugs of abuse, including psychostimulants, opiates, nicotine, and ethanol. Moreover, excitotoxic lesions of the NAc inhibit cocaine, heroin (Zito et al. 1985), and morphine self-administration (Dworkin et al. 1988b), with similar effects shown using 6-hydroxydopamine (6-OHDA) to selectively lesion DA terminals in the NAc (Roberts et al. 1977, 1980; Lyness et al. 1979; Corrigall et al. 1992) or DA cell bodies in the VTA (Roberts and Koob 1982). Finally, a number of studies have demonstrated that animals will reliably self-administer intracranial microinjections of various drugs of abuse (Myers 1974; Bozarth 1987), including psychostimulants, opiates, and ethanol, into the NAc (Monaco et al. 1981; Olds 1982; Hoebel et al. 1983; Chevrette et al. 2002; Rodd-Henricks et al. 2002) or VTA (van Ree and de Wied 1980; Bozarth and Wise 1981, 1982; Gatto et al. 1994; Rodd et al. 2004a,b, 2005). Moreover, these effects appear anatomically selective and dependent on drug reinforcement in that switching the response contingency or replacing the drug with artificial cerebrospinal fluid eliminates or inhibits these behaviors.

Although abundant evidence indicates that the mesolimbic DA system plays a role in the rewarding and reinforcing effects of psychostimulants, there has been some debate regarding its necessity for the acute rewarding and reinforcing effects of drugs of abuse (Koob 1992; Nestler 2005). For example, DA-deficient mice exhibit psychostimulant- and opiate-conditioned place preference (Hnasko et al. 2005, 2007). Furthermore, 6-OHDA lesions of the NAc reliably impair the self-administration of psychostimulants but fail to alter the self-administration of opiates (Ettenberg et al. 1982; Pettit et al. 1984; Smith et al. 1985; Dworkin et al. 1988a) or ethanol (Lyness and Smith 1992; Rassnick et al. 1993). Similarly, DA antagonist administration in the NAc fails to significantly alter opioid self-administration (Ettenberg et al. 1982; Gerrits et al. 1994). These findings suggest that the reinforcing effects of all major drugs of abuse involve DA-independent mechanisms to some extent. For instance, the endogenous opioid and cannabinoid systems likely mediate some aspects of drug reinforcement. In support of this, systemic or intracranial μ opioid receptor antagonism attenuates heroin (Bozarth and Wise 1983; Corrigall and Vaccarino 1988) and ethanol (Altshuler et al. 1980; Samson and Doyle 1985) self-administration. Furthermore, genetic knockout of various opioid receptors and peptides disrupts cocaine-, nicotine-, and THC-conditioned place preference and cannabinoid self-administration (Charbogne et al. 2014). Similarly, genetic knockout or antagonism of CB1 receptors (Caille and Parsons 2003; Colombo et al. 2005) can reduce opiate (Ledent et al. 1999; Cossu et al. 2001), ethanol (Hungund et al. 2003; Thanos et al. 2005), and cocaine self-administration (Soria et al. 2005; Xi et al. 2008). Moreover, CB1 receptor antagonism, but not genetic deletion (Cossu et al. 2001), can impair nicotine self-administration (Schindler et al. 2016).

The acute reinforcing and rewarding effects of all drugs of abuse, and thus the initiation of the addiction cycle, critically involve the mesolimbic DA system (Johnson and North 1992; Weiss et al. 1993; Koob and Le Moal 2006) as well as DA-independent mechanisms (Van Ree et al. 1999). Furthermore, sex differences (Calipari et al. 2017) and individual differences (Pitchers et al. 2017) in the acute and conditioned rewarding effects of drugs of abuse are related to differences in sensitivity to the effects of drugs of abuse on VTA DA function. However, DA may play a different role in regulating the reinforcing (i.e., “liking”) and persistent motivational effects of drugs of abuse (i.e., “wanting”) in chronic addictive states, following the development of drug-induced and learning-related pathophysiology in these systems (Berridge 2007; Berridge and Robinson, 2016).

NEUROCIRCUITRY INVOLVED IN DRUG DEPENDENCE

In addition to mediating the acute drug reinforcement, DA activity in the mesocorticolimbic system may promote continued, persistent drug-seeking behaviors by enhancing the formation of associations between actions, the rewarding and motivational effects of drugs of abuse, and discrete and contextual environmental stimuli involved in the drug-taking process (Everitt et al. 2001; Fuchs et al. 2005). With repeated drug taking (i.e., as the state of drug dependence develops), drug-associated stimuli eventually begin to dominate the behavioral output, independent of goal-directed action (Everitt and Robbins 2005). Interactions between Pavlovian and instrumental learning processes (Everitt et al. 2001) can lead to a shift from “stimulus–reward” to “stimulus–response” learning that underlies the habitual drug-seeking characteristic of chronic drug dependence. As such, this loss of controlled drug-seeking and drug-taking behavior, which is the hallmark of clinical drug addiction/dependence, has become a focus of interest for preclinical researchers (Wolffgramm and Heyne 1995; Deroche-Gamonet et al. 2004; Vanderschuren and Everitt 2004).

The concept of drug addiction as a maladaptive habit is supported by multiple lines of evidence. Studies suggest that habit formation is associated with the progressive recruitment of the dorsal striatum (i.e., caudate-putamen), likely through spiraling connections between the ventral striatum (i.e., NAc), midbrain, and dorsal striatum (Haber et al. 2000; Ikemoto 2007). In contrast to the ventral striatum, the dorsal striatum (especially the lateral regions) appears to have little, if any, involvement in the acute reinforcing effects of drugs of abuse, but it becomes engaged in the course of chronic drug taking (Everitt et al. 2008; Hodebourg et al. 2019). For example, 6-hydroxy-dopamine (Roberts 1992; Gao et al. 2013) or excitotoxic (Gabriele and See 2011) lesions of the dorsal striatum or functional disconnection of the dorsal and ventral striatal subregions (Belin and Everitt 2008) do not have any effect on cocaine or morphine self-administration under fixed ratio schedules of reinforcement that maintain goal-directed drug seeking. However, both the dorsal and ventral striatum support cocaine and morphine self-administration under progressive ratio schedules of reinforcement (Suto et al. 2011). Regarding the possible mechanisms of increased dorsal striatal engagement, the dorsal and lateral regions of the striatum exhibit progressively greater activation, as assessed by 2-deoxyglucose autoradiography in rhesus monkeys, following 100, but not 5, days of cocaine self-administration (Porrino et al. 2004). These changes are paralleled by progressive ventral-to-dorsal changes in DA transporter (Letchworth et al. 2001) and receptor (Nader et al. 2002) expression.

Engagement of the dorsal striatum is particularly important for cue-induced responses established in the course of repeated drug self-administration. In support of this, following chronic cocaine self-administration under a second-order schedule of reinforcement, response-contingent presentation of a cocaine-paired stimulus increases extracellular DA levels in the dorsal striatum, but not in the NAc (Ito et al. 2000, 2002). Functional disconnection of the dorsal and ventral striatum or dopamine receptor antagonism in the dorsal striatum also attenuates cocaine and heroin seeking under second-order schedules of reinforcement that critically depend on the conditioned reinforcing effects of drug-paired stimuli (Di Ciano and Everitt 2004b; Vanderschuren et al. 2005; Belin and Everitt 2008; Hodebourg et al. 2019). Similarly, the functional integrity of the dorsolateral striatum (Fuchs et al. 2006a; See et al. 2007; Gao et al. 2013) is necessary for context-induced cocaine-seeking behavior following abstinence, when habitual responding is a significant factor in drug-seeking behavior, whereas the recruitment of the NAc appears to vary across drugs of abuse (See et al. 2007; Gao et al. 2013). Finally, inactivation of the dorsolateral striatum reduces cocaine seeking only after the drug-taking response is devalued to isolate habitual goal-directed behavior (Zapata et al. 2010). Combined with clinical data indicating that activation of the dorsal, but not ventral, striatum positively correlates with cue-induced craving in cocaine-dependent patients (Volkow et al. 2006; Wong et al. 2006), these findings imply that the dorsal striatum gains control over drug seeking during the development of impulsive and compulsive behaviors characteristic of chronic drug addiction and dependence. However, recent data also suggests that the development of some aspects of addiction-like behavior do not inevitably rely upon stimulus–response seeking habits (Singer et al. 2018).

In addition to the dorsal striatum, impairments in prefrontal cortical areas mediate aspects of compulsive drug-seeking behavior, in particular the loss of behavioral control and inhibition. Although habitual behaviors primarily involve cortico-striato-thalamic circuits (Jog et al. 1999; Canales 2005), these behaviors rely on higher level cortical processing of new information to determine whether the behavior needs to be modified according to its adaptive function. If the behavior becomes maladaptive, information of this outcome from the PFC should lead to a modification or cessation of the behavior. However, despite a number of negative consequences, drug taking continues in drug-dependent individuals. Interestingly, a number of studies have shown decreased gray matter density and reduced baseline blood glucose metabolism in frontal cortical areas, including the anterior cingulate and OFC, of chronic drug users (London et al. 1999; Volkow and Fowler 2000; Franklin et al. 2002; Matochik et al. 2003; Thompson et al. 2004; Ersche et al. 2011). Furthermore, chronic cocaine abusers exhibit deficits in inhibitory control and decision-making processes (Bolla et al. 2003; Hester and Garavan 2004). In animal models, lesions of the medial PFC (mPFC), encompassing the prelimbic and infralimbic (IL) cortex, result in enhanced acquisition of cocaine self-administration and increased responding for cocaine on a second-order schedule of reinforcement (Weissenborn et al. 1997). These effects do not appear to be caused by alteration in reinforcement sensitivity (Burns et al. 1993), but rather a loss in behavior/inhibition processes (Dalley et al. 2004). In addition, there is an abundance of preclinical data to suggest that OFC dysfunction is involved in addiction-related behavioral changes (Schoenbaum et al. 2006; Everitt et al. 2007; Olausson et al. 2007; Lucantonio et al. 2012). Excitotoxic lesions of the OFC result in compulsive drug-seeking behavior that is resistant to extinction in rats (Fuchs et al. 2004b), and similar decision-making problems and perseveration are observed in patients with OFC damage (Rogers et al. 1999). Preexisting deficits in prefrontal cortical function (Volkow and Fowler 2000; Kaufman et al. 2003; Hester and Garavan 2004) may lead to poor decision capacities contributing to the development of the addiction. On the other hand, experimental studies in rodents and nonhuman primates suggest that frontal cortical dysfunction is likely a consequence of chronic drug history (Jentsch and Taylor 1999; Schoenbaum et al. 2006). In animals withdrawn from chronic psychostimulant exposure, profound alterations have been observed in dendritic morphology (e.g., dendritic length and spine density) in the OFC and PFC (Kolb et al. 2004; Crombag et al. 2005) and in expression of the transcription factor ΔFosB (Winstanley et al. 2007). Chronic cocaine exposure also results in impaired decision-making, including cognitive inflexibility and impulsivity (Roesch et al. 2007; Simon et al. 2007). In fact, cocaine-induced deficits in reversal learning and extradimensional shifting are similar to those seen following damage to the OFC (Calu et al. 2007; Izquierdo et al. 2010; Porter et al. 2011) and medial PFC (Parsegian et al. 2011), respectively. Thus, functional impairments in frontal cortical areas following prolonged drug use may exacerbate addiction-related behaviors by diminishing adaptive decision-making and behavioral control. However, these brain regions also exhibit hyperactivity upon exposure to drug-predictive conditioned stimuli (George et al. 2001; Franklin et al. 2007; Ray et al. 2010) and likely contribute to drug seeking and relapse in drug-dependent individuals by enhancing the salience or motivational significance of these stimuli.

NEUROBIOLOGY UNDERLYING RELAPSE

Relapse to drug use following prolonged periods of abstinence constitutes one of the most significant problems for the long-term treatment of drug-dependent individuals (Dackis and O'Brien 2001; Wagner and Anthony 2002). A number of factors contribute to craving and relapse, including exposure to environmental stimuli previously paired with drug use (i.e., conditioned drug cues), negative mood states or stress, and exposure to small amounts of the drug. For example, abstinent cocaine users report increases in drug craving when exposed to cocaine-associated stimuli (Childress et al. 1993), following stressful life events (Sinha et al. 1999), and in response to noncontingent cocaine infusions (Jaffe et al. 1989). These triggers have also been employed in animal models of drug relapse, including the extinction-reinstatement model (de Wit and Stewart 1981; Kalivas and McFarland 2003; Fuchs et al. 2018) and voluntary abstinence models (Venniro et al. 2016, 2019). In the extinction/reinstatement model, animals that previously self-administered a drug (usually paired with a discrete stimulus, such as a light and/or tone) undergo extinction sessions, during which responding on the previously drug-paired lever no longer results in drug reinforcement. Response-contingent presentation of drug-paired conditioned stimuli or exposure to environmental stressors or drug administration can robustly reinstate the extinguished drug seeking (i.e., induce relapse) as indexed by an increase in responding on the previously drug-paired operandum (Erb et al. 2001; Shaham et al. 2003; Fuchs et al. 2018). In versions of the model that aim to assess reinstatement upon exposure to a drug-predictive environmental context, animals are trained to self-administer a drug in a distinct environmental context, undergo extinction training in a different context, and are tested for reinstatement upon passive exposure to the drug-associated context (Fuchs et al. 2005, 2008, 2018). Finally, in voluntary abstinence models, responding is suppressed following self-administration training through punishment (i.e., foot shock [Panlilio et al. 2003; Cooper et al. 2007]) or access to an alternate, nondrug reinforcer (i.e., food or social reward [Venniro et al. 2019]). These animal models have been useful for extensive exploration of the neural circuitry underlying relapse-like behaviors (Meil and See 1997; Neisewander et al. 2000; McFarland and Kalivas 2001; McFarland et al. 2004).

Research using reinstatement models has revealed an important role for a number of mesocorticolimbic brain regions, including the BLA, PFC, OFC, NAc core, and the hippocampus in drug seeking. Reinstatement of cocaine seeking in rats exposed to discrete cocaine-associated cues positively correlates with increased Fos expression in the BLA (Neisewander et al. 2000; Kufahl et al. 2009), consistent with the result of neuroimaging studies demonstrating increases in metabolic activity in the amygdala in abstinent cocaine users exposed to drug-associated cues or drug-related imagery (Grant et al. 1996; Childress et al. 1999; Kilts et al. 2001, 2004; Bonson et al. 2002), Moreover, permanent lesions or reversible inactivation of the BLA has a number of effects on discrete cue-induced drug seeking, including decreases in responding for stimuli associated with cocaine reinforcement (Whitelaw et al. 1996; Meil and See 1997; Grimm and See 2000) and prevention of the acquisition (Kruzich and See 2001), consolidation (Fuchs et al. 2006b; Gabriele and See 2010), and expression (Kruzich and See 2001; Fuchs and See 2002; McLaughlin and See 2003; Rogers et al. 2008) of cocaine- or heroin-seeking behavior. The BLA mediates cue-induced drug seeking through interactions with a number of other brain regions. Specifically, monosynaptic inputs from the BLA to the prelimbic cortex and NAc core (Stefanik and Kalivas 2013) are necessary for cue-induced cocaine seeking, whereas BLA inputs to the OFC are involved in cue-induced reward expectations (Lichtenberg et al. 2017). The BLA plays a similar role in contextual drug seeking (Fuchs et al. 2005, 2009) through interactions with other brain regions implicated in cue-mediated drug seeking, including the PFC (Fuchs et al. 2007; Lasseter et al. 2011) and the hippocampus (Fuchs et al. 2005; Wells et al. 2011).

Presentation of stimuli associated with cocaine availability results in a significant elevation in extracellular DA overflow in the BLA (Weiss et al. 2000). Consistent with the functional importance of this neurochemical response, direct intra-BLA administration of DA receptor antagonists attenuate cocaine seeking under a second-order schedule of reinforcement (Di Ciano and Everitt 2004b). Similar treatments have also been shown to attenuate the acquisition (Berglind et al. 2006) and expression (See et al. 2001) of discrete cue-induced reinstatement. Moreover, DA D1 receptor antagonism can reverse elevations in Fos (Neisewander et al. 2000; Ciccocioppo et al. 2001) and Fos-related antigen (Franklin and Druhan 2000) expression noted when animals are exposed to contextual or discrete cocaine-associated conditioned stimuli (Ciccocioppo et al. 2001). Other neurotransmitters, including glutamate (See et al. 2001; Feltenstein and See 2007) and acetylcholine (See et al. 2003), have also been shown to be involved in BLA-mediated cue-induced associative learning and drug-seeking behavior. Taken together, these observations suggest that the BLA plays a prominent role in drug-cue associative learning and relapse, with different neurotransmitter systems likely mediating unique aspects of amygdalar processing of drug-paired stimuli.

The PFC is critically important for cue-induced motivation, similar to the amygdala. Functional neuroimaging studies have demonstrated that metabolic activity increases in the PFC (specifically the dorsolateral regions) when drug-dependent individuals are exposed to drug-associated cues (Grant et al. 1996), and this activation is positively correlated with craving (Maas et al. 1998; Bonson et al. 2002). In animal models, increases in reinstatement behavior and Fos protein or Arc mRNA expression have been noted in the dorsomedial PFC (dmPFC) when animals are exposed to discrete cocaine- (Zavala et al. 2008; Kufahl et al. 2009) or heroin-paired cues (Koya et al. 2006), as well as discriminative stimuli predicting cocaine or ethanol availability (Ciccocioppo et al. 2001; Dayas et al. 2007). DA D1 receptor antagonism (Ciccocioppo et al. 2001) or inactivation of the dmPFC using the sodium channel blockers, tetrodotoxin (McLaughlin and See 2003; Fuchs et al. 2005) or lidocaine (Di Pietro et al. 2006), or the GABA_A and GABA_B agonists, muscimol and baclofen (Fuchs et al. 2007; LaLumiere and Kalivas 2008; Rogers et al. 2008), inhibits the reinstatement of drug seeking when animals are exposed to discrete or contextual cocaine- or heroin-associated stimuli. Importantly, the dmPFC mediates cue-induced reinstatement through glutamatergic innervation of the NAc core, in that inhibition of heroin seeking following dmPFC inactivation prevents concomitant increases in extracellular glutamate levels within the NAc core (LaLumiere and Kalivas 2008). Confirming these results, selective optogenetic inhibition of prelimbic cortex projections to the NAc core disrupts reinstatement in response to cocaine and conditioned stimuli (Stefanik and Kalivas 2013). Studies examining the putative role of the IL cortex in cue-mediated drug seeking have been mixed (for review, see Lasseter et al. 2010a). Finally, the lateral, but not the medial, OFC is thought to play a critical role in evaluating the incentive value of stimuli (Gallagher et al. 1999), and inactivation studies show that it is critically involved in both discrete and contextual cue-induced cocaine seeking (Fuchs et al. 2004b; Lasseter et al. 2009). Although the functional connectivity of the OFC has been less well-studied, the OFC mediates cue-induced drug seeking at least in part through monosynaptic projections to the BLA (Arguello et al. 2017).

A substantial amount of data suggests that the NAc is significantly involved in cue-induced drug seeking. Exposure to discrete or contextual cocaine- (Neisewander et al. 2000; Kufahl et al. 2009) or ethanol-associated (Dayas et al. 2007) cues elicits Fos protein expression in the NAc. Numerous studies have demonstrated a functional role for the NAc core in that reversible inactivation of the NAc core attenuates cocaine seeking (Di Ciano and Everitt 2004a; Fuchs et al. 2004a; Di Ciano et al. 2008). With respect to specific neurotransmitter systems, studies have indicated that DA and glutamate in the NAc mediate these behaviors. For example, exposure to discriminative stimuli predicting cocaine availability results in robust elevation in DA in the NAc (Weiss et al. 2000). Similarly, fast scan cyclic voltammetry indicates cue-induced increases in phasic DA release specifically in the NAc core, as well as concomitant decreases in phasic DA release is in the NAc shell, after conditioning with cocaine (Aragona et al. 2009). Critical DA inputs to the NAc core originate from the VTA (Stefanik et al. 2013). Within the NAc core, DA D1 receptor antagonism decreases cue-induced heroin seeking (Bossert et al. 2007), and chemogenetic stimulation of D1 receptor-bearing medium spiny neurons (MSNs) potentiates cue-induced cocaine seeking (Heinsbroek et al. 2017). Consistent with the contribution of glutamate, direct infusion of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) or N-methyl-D-aspartic acid (NMDA) agonists into the NAc results in reinstatement of cocaine seeking (Cornish et al. 1999), whereas AMPA/kainate and/or NMDA receptor antagonism blocks cocaine seeking following exposure to drug-associated cues under a second-order schedule (Di Ciano and Everitt 2001; Bäckström and Hyytiä 2007). These data suggest that cue-associated reinstatement relies on DA and glutamate interactions in the NAc core, with likely contributions from other neurotransmitter systems (e.g., acetylcholine [Zhou et al. 2007]). Interestingly, DA in the NAc shell does not play a critical role in drug seeking (Bossert et al. 2007; Aragona et al. 2009), but the NAc shell is a component of the relapse circuitry. Consistent with this, synaptic inputs from the ventral subiculum to the NAc shell are necessary for cue-induced alcohol seeking (Marchant et al. 2016), and the stimulation of NAc shell projections to the lateral hypothalamus enhances the motivation to seek cocaine (Larson et al. 2015).

Similar to cue-induced drug seeking, a considerable amount of research suggests that reinstatement following exposure to the previously self-administered drug alone is mediated by the PFC, NAc core, and VTA. Inactivation of these brain regions attenuates cocaine- (Grimm and See 2000; McFarland and Kalivas 2001; Capriles et al. 2003) and heroin-primed reinstatement (Rogers et al. 2008). DA and glutamate play pivotal roles in drug-primed relapse-like behaviors. For example, microdialysis studies have shown increases in accumbal DA and glutamate overflow following a priming injection of cocaine concomitant with reinstatement (McFarland et al. 2003). Microinfusions of DA or glutamate, as well as AMPA agonists, into the NAc produce robust elevations in drug seeking (Cornish et al. 1999; McFarland and Kalivas 2001). Conversely, antagonism of DA D1 receptors in the NAc shell (Anderson et al. 2003) or AMPA/kainic receptors in the NAc core (Cornish and Kalivas 2000) attenuates cocaine-primed reinstatement. Interestingly, DA or NMDA antagonism in the NAc core were without effect (Cornish and Kalivas 2000; Anderson et al. 2003), collectively suggesting divergent roles for DA and glutamate (via AMPA receptors) in the NAc shell and core, respectively, in mediating cocaine-primed reinstatement.

With respect to prefrontal cortical areas, dmPFC inactivation impairs drug-primed heroin, cocaine, methamphetamine, and nicotine seeking (McFarland and Kalivas 2001; Capriles et al. 2003; Zavala et al. 2003; Hiranita et al. 2006; Rogers et al. 2008). Moreover, direct infusions of DA, cocaine, or amphetamine into the dmPFC are sufficient to elicit cocaine seeking, whereas DA D1 and D2 receptor antagonists disrupt cocaine-primed reinstatement and conditioned place preference (McFarland and Kalivas 2001; Park et al. 2002; Sanchez et al. 2003; Sun and Rebec 2005, but see Capriles et al. 2003). The prelimbic cortex mediates these behaviors at least in part through glutamatergic inputs into the NAc core, because enhancement in glutamate release and drug seeking following cocaine- or heroin-priming are both reversed by dmPFC inactivation (McFarland et al. 2003; LaLumiere and Kalivas 2008). Furthermore, optogenetic inhibition of prelimbic cortical projections to the NAc core impairs drug-primed reinstatement (Stefanik et al. 2013).

Growing evidence indicates that neural activity in the vmPFC (i.e., IL cortex) and NAc shell can inhibit drug seeking. For instance, GABA agonist–induced inactivation or optogenetic inhibition of the IL alone is sufficient to reinstate extinguished cocaine seeking (Peters et al. 2008; Müller Ewald et al. 2019), whereas stimulation of the IL using an AMPA receptor–positive allosteric modulator inhibits cue-induced reinstatement of heroin seeking (Chen et al. 2016). These effects are likely mediated through the NAc shell, as chemogenetic stimulation of monosynaptic IL inputs into the NAc shell also impairs cocaine seeking (Augur et al. 2016). However, some evidence indicates that the IL can facilitate drug seeking under certain conditions. In particular, earlier research has shown that GABA agonist–induced IL inhibition attenuates cocaine seeking after forced abstinence, and GABA antagonist–induced IL stimulation reduces cocaine seeking (Koya et al. 2008). A potential resolution of these inconsistencies in the literature comes from a recent study that demonstrated the coexistence of IL neural ensembles that increase cocaine seeking with those that inhibit this behavior (Warren et al. 2019). Future research will need to investigate whether similar ensembles can bidirectionally control reinstatement of heroin seeking. However, these findings hint at the possibility that the NAc shell plays a complex modulatory role in drug seeking.

In addition to cue and drug exposure, stress plays a prominent role in the initiation and maintenance of substance use disorders (Higgins and Marlatt 1975; Russell and Mehrabian 1975; Koob and Le Moal 2001). In animal models of stress-induced drug relapse (Fuchs et al. 2018), drug seeking is elicited by a variety of stressors, including exposure to physical stressors (e.g., foot shock) (Erb et al. 1996; McFarland et al. 2004; Buffalari and See 2009), pharmacological stressors (e.g., yohimbine) (Lee et al. 2004; Shepard et al. 2004; Lê et al. 2005; Buffalari et al. 2012) and, most recently, social stressors (i.e., social defeat stress) (Blouin et al. 2019). Examination of the neurocircuitry involved in stress-induced reinstatement using these methods has suggested that overlapping, yet distinct, neural systems mediate these behaviors relative to other forms of reinstatement (Stewart 2000; Shaham et al. 2003). Inactivation of the PFC, NAc, or VTA decreases stress-induced reinstatement, with a unique contribution noted for the BNST and central nucleus of the amygdala (CeA) in mediating these behaviors in response to physical stressors (Capriles et al. 2003; McFarland et al. 2004). In contrast, a recent report indicates that social stress–induced potentiation of drug-primed methamphetamine seeking relies on the BLA (Blouin et al. 2019).

Several neurotransmitter systems have been implicated in stress-induced reinstatement (Shaham et al. 2000a). DA neurotransmission in the prelimbic cortex (Capriles et al. 2003; Sanchez et al. 2003) and OFC (Capriles et al. 2003) is critical for this phenomenon, because DA receptor antagonist administration into these brain regions blocks stress-induced drug seeking. Corticotrophin-releasing factor (CRF), corticosterone (CORT), norepinephrine (NE), endocannabinoids, and endogenous opioids are also critically involved in stress-induced reinstatement. In line with this, intracerebroventricular (ICV) administration of CRF alone is sufficient to induce reinstatement of heroin (Shaham et al. 1997) or cocaine seeking (Buffalari et al. 2012). Systemic or intra-NAc administration of CORT potentiates drug-primed reinstatement of cocaine seeking (Graf et al. 2013; McReynolds et al. 2017), and systemic or intra-BLA glucocorticoid receptor antagonism inhibits stress-induced and drug context–induced reinstatement of cocaine seeking, respectively (Polter et al. 2014; Stringfield et al. 2016). Similarly, ICV or intra-BNST administration of CRF antagonists (Shaham et al. 1997; Erb et al. 1998; Erb and Stewart 1999; Lê et al. 2000) attenuates foot shock–induced reinstatement. Attenuation in drug seeking has also been noted following systemic, intra-BDNF, or intra-CeA pretreatment with compounds that attenuate noradrenergic activity (Erb et al. 2000; Shaham et al. 2000b; Highfield et al. 2001; Leri et al. 2002). Moreover, systemic cannabinoid receptor type 1 or κ opioid receptor antagonism inhibits the stress-induced potentiation of cocaine-primed cocaine seeking (McReynolds et al. 2016) and stress-induced cocaine reinstatement (Polter et al. 2014), respectively. Thus, well established mediators of stress responses within the extended amygdala circuit critically contribute to stress-induced drug craving and relapse.

In summary, it appears that three distinct, yet overlapping, neurocircuits mediate relapse-like behaviors following exposure to drug-associated cues, drug, or stressful events (Kalivas and McFarland 2003; Shaham et al. 2003). Although there are a number of unique aspects involved in each type of reinstatement, these data collectively suggest that projections from the VTA (all forms of reinstatement), limbic regions including the BLA and dorsal hippocampus (DH) (cue reinstatement), CeA, BNST, and NAc shell (stress reinstatement) converge on motor pathways involving the dmPFC and NAc core. The dmPFC and NAc core represent a “final common pathway” for relapse behavior. It should be noted, however, that recent discoveries have implicated other brain regions and neurotransmitter systems in drug-seeking behaviors. As mentioned previously, a considerable amount of research has implicated the dorsal (Fuchs et al. 2005, 2007; Meyers et al. 2006; Atkins et al. 2008; Ramirez et al. 2009; Xie et al. 2010; Wells et al. 2011) and ventral (Rogers and See 2007; Lasseter et al. 2010b) hippocampus in cue-induced relapse, including discrete and contextual forms of associative learning. Additionally, the mesopontine rostromedial tegmental nucleus (RMTg; also known as the tail of the VTA) (Jhou et al. 2009; Kaufling et al. 2009; Lavezzi and Zahm 2011) has been implicated in reward-related behaviors. The RMTg receives projections from a number of regions, including the PFC, ventral pallidum, BNST, and lateral habenula (Kaufling et al. 2009). GABAergic neurons of the RMTg project to the VTA and have a modulatory role on VTA-mediated drug reward and relapse (Jalabert et al. 2011; Lecca et al. 2011) as well as negative affective states related to drug withdrawal (Kaufling and Aston-Jones 2015; Glover et al. 2019). Another region that has received interest for its potential role in addiction, particularly nicotine dependence, is the pedunculopontine tegmental nucleus (PPTg) (Maskos 2008; Mark et al. 2011). Providing cholinergic innervation to the VTA, it has been suggested that the posterior PPTg is involved in drug reinforcement (Alderson et al. 2006), with lesion and/or selective nicotinic receptor antagonism studies demonstrating an attenuation of nicotine (Lança et al. 2000; Corrigall et al. 2001) and cocaine self-administration behaviors (Corrigall et al. 2002). The hypothalamic neuropeptide orexin (also known as hypocretin) plays a significant role in mediating drug addiction and relapse (James et al. 2017), including cocaine- (Boutrel et al. 2005; Smith et al. 2009, 2010; Zhou et al. 2012; James et al. 2019), heroin- (Smith and Aston-Jones 2012; Schmeichel et al. 2015), ethanol- (Lawrence et al. 2006; Richards et al. 2008; Shoblock et al. 2011; Moorman 2018), and nicotine-seeking behaviors (Plaza-Zabala et al. 2010). Finally, the neuropeptide oxytocin has also received some interest recently for its potential as a therapeutic for drug addiction (Cox et al. 2017; Leong et al. 2018; Zanos et al. 2018).

CONCLUSIONS

Drug addiction is a chronic disease that encompasses a large number of social, economic, and medical issues that can persist even years after abstinence (Meyer 1996). A considerable amount of research has been devoted to understanding the behavioral and neurobiological mechanisms that mediate the transition from casual drug use to the loss of control and persistence of drug-seeking behaviors that characterize drug addiction and dependence. Using various animal models of addiction, researchers have been able to determine aspects of the fundamental neurobiology involved in drug seeking across the entire addiction cycle, including the acute reinforcing effects of drugs, the neuroadaptations that occur during the transition to drug dependence, and finally the relatively permanent alterations in these systems that leave an individual susceptible to relapse. Although the role of neurotransmitters/neuromodulators and neural systems can vary across drugs of abuse and stages of the addiction cycle, evidence shows that the mesocorticolimbic pathway, including the VTA, NAc, amygdala, and prefrontal cortices, via DA and glutamate pathways, play a significant role in addiction. With a better understanding of the neurobiological factors that underlie drug addiction, continued clinical and preclinical research will greatly facilitate the development of novel therapeutic interventions that may result in better, more effective treatment strategies for drug-dependent individuals.