Among the most insidious characteristics of drug addiction is the recurring desire to take drugs even after many years of abstinence. Equally sinister is the compromised ability of addicts to suppress drug seeking in response to that desire even when confronted with seriously adverse consequences, such as incarceration. The enduring vulnerability to relapse is a primary feature of the addiction disorder and has been identified as a point where pharmacotherapeutic intervention may be most effectively employed +(1). In order to fashion rational pharmacotherapies it is necessary to understand the neurobiological underpinnings of craving, relapse, choice, and control, and the last decade has seen significant advances toward achieving this goal. However, as the pursuit for the neural basis of addiction advances, it is clear that the search intimately involves understanding the neurobiological basis of motivation and choice for biological rewards, such as food and sex, as well as more cognitively and experientially based rewards, such as friendship, family, and social status. Moreover, the fact that vulnerability to relapse in addicts can persist after years of abstinence implies that addiction is caused by long-lasting changes in brain function as a result of pharmacological insult (repeated drug use), genetic disposition, and environmental associations made with drug use (learning). Therefore, comprehending the basis of addiction also requires understanding physiological mechanisms of enduring neuroplasticity. Accordingly, this review begins with an overview of the neural circuitry and mechanisms of neuroplasticity that underlie the generation of adaptive behavioral responses to motivationally relevant events.

After outlining the neurobiology of motivated behavior, we will describe the pathological dysregulation of cellular and circuitry functions produced by addiction. As early as the 1970s it became clear that the acute administration of most drugs of abuse increases dopamine transmission in the basal ganglia +(2), which is essential for these drugs to reinforce behavior, and thereby promote addiction +(3, +4). Therefore, dopamine projections to the basal ganglia and cortex are important in facilitating the encoding of learned associations necessary for the development of addiction +(5). In contrast, once a person is addicted the uncontrollable urge to obtain drugs and relapse arises from a pathological form of the plasticity in excitatory transmission +(6–+9). Alterations in excitatory transmission occur physiologically when learned associations with motivationally relevant events are formed. Similarly, addicts’ extreme difficulty in resisting the desire to use drugs of abuse is encoded by changes in excitatory synapses, and it will be proposed that the molecular underpinnings of dysregulated excitatory transmission may be fruitful pharmacotherapeutic targets in ameliorating addiction.

Seeking food and companionship and avoiding physical and psychological discomfort are examples of motivated adaptive behaviors. Motivated behavior classically implies both an activation of the organism by environmental or interoceptive stimuli and a directed behavioral output +(10). Thus, the neurobiological search for the antecedents of motivated behavior involves defining the neural substrates that 1) attach sufficient importance (salience) to an integrated stimulus that behavior is "activated" and 2) "direct" this state of activation toward a specific behavioral response. While we have made substantial progress toward identifying the neural circuits and cellular foundations responsible for activating behavior, we have been only marginally effective at understanding the substrates that cause one behavior to be favored over another behavior (direction of behavior).

Neurobiology has focused on three brain regions in the activation of behavior: the amygdala, prefrontal cortex, and nucleus accumbens. The amygdala emerged from studies showing involvement in fear-motivated behaviors +(11), while the nucleus accumbens was identified from a connection with reward-motivated behaviors +(12). The prefrontal cortex is less involved in establishing whether a stimulus is positive or negative (valence); rather, it regulates the overall motivational salience and determines the intensity of behavioral responding +(13, +14). More recent studies have blurred the linkage between positive and negative emotional valence in the amygdala and nucleus accumbens, and they have revealed a neuronal circuit consisting of glutamatergic interconnections among the amygdala, nucleus accumbens, and prefrontal cortex and dopaminergic afferents to all three regions +(15, +16). +Figure 1 illustrates this circuit and includes three additional components. The accumbens has dense projections carrying γ-aminobutyric acid (GABA) and neuropeptides to the ventral pallidum that are critical for the expression of motivated behaviors +(17). Another GABA/neuropeptide component of the circuit is the extended amygdala, which is a cluster of interconnected nuclei, including the central amygdala nucleus, bed nucleus of the stria terminalis, and shell of the nucleus accumbens, that is in part a conduit for environmental and interoceptive stressors +(18). It is important to note that while the shell of the accumbens possesses some functional and anatomical characteristics of the extended amygdala (especially in terms of the neurocircuitry of addiction), it is also anatomically distinct from the other nuclei in terms of some aspects of connectivity and histochemistry +(19). Finally, there is a series subcircuit consisting of GABA-ergic projections from the ventral pallidum to the mediodorsal thalamus and a reciprocal glutamatergic projection between the thalamus and prefrontal cortex that mediates reintegration of information exiting the circuit back into the prefrontal cortex +(20).

Projections from the ventral tegmental area release dopamine throughout the circuit in response to a motivationally relevant event +(21, +22). The release of dopamine signals the circuit to initiate adaptive behavioral responses to the motivational event, and in doing so it facilitates cellular changes that establish learned associations with the event +(5). In this way the organism can more effectively emit an adaptive behavioral response should the event reoccur. However, in contrast to repeated drug administration, as a motivational event becomes familiar by repeated exposure, dopamine release is no longer induced by that particular event +(23). In this case, although the behavioral response remains goal directed, it is well learned and further dopamine-induced neuroplastic changes are not necessary. However, it is important to note that conditioned stimuli predicting the event continue to trigger release of dopamine +(23–+25). Therefore, in most natural situations where learned associations accompany a repeatedly encountered motivational event, dopamine will likely be released as part of the overall experience. In sum, dopamine can be seen as serving two functions in the circuit: 1) to alert the organism to the appearance of novel salient stimuli, and thereby promote neuroplasticity (learning), and 2) to alert the organism to the pending appearance of a familiar motivationally relevant event, on the basis of learned associations made with environmental stimuli predicting the event +(23, +26).

The amygdala is especially critical in establishing learned associations between motivationally relevant events and otherwise neutral stimuli that become predictors of the event +(27). Interactions between the basolateral and central amygdala nucleus involve autonomic and endocrine associations by means of projections from the central nucleus to the brainstem, the hypothalamus, and dopamine neurons in the ventral tegmental area +(28, +29). In contrast, the glutamatergic projections from the basolateral amygdala to the prefrontal cortex and accumbens are required for learned associations to influence more complex behavioral responses +(16, +20). The functional integration between the amygdala and prefrontal cortex has been demonstrated in many neuroimaging studies in which healthy subjects were exposed to stimuli associated with motivationally relevant events, ranging from food and sex to social cooperation +(30–+32).

The anterior cingulate and ventral orbital cortices in the prefrontal cortex are recruited by motivationally relevant events, as well as stimuli that predict such events, and contribute to whether a behavioral response will be emitted and the relative intensity of that response +(14, +33). Consistent with involvement of dopaminergic afferents, the activation of the prefrontal cortex by rewarding stimuli is strongly influenced by the predictability of the reward +(34, +35).

The accumbens contains two functionally distinct subcompartments, termed the shell and core +(36). The shell is strongly interconnected with the hypothalamus and ventral tegmental area and is correspondingly important in regulating ingestive behaviors +(21, +36). Reciprocal dopamine innervation from the ventral tegmental area to the accumbens shell is important in modulating motivational salience and contributes to establishing learned associations between motivational events and concurrent environmental perceptions +(37, +38). In contrast, the core compartment is anatomically associated with the anterior cingulate and orbitofrontal cortex and appears to be a primary site mediating the expression of learned behaviors in response to stimuli predicting motivationally relevant events +(36, +39). Moreover, the obligatory involvement of the accumbens core in expressing adaptive behavior depends not on dopaminergic afferents but, rather, on glutamatergic afferents from the prefrontal cortex +(40). Although not an obligatory event, dopamine is released into the core in response to stimuli predicting a rewarding event and likely modulates the expression of adaptive behaviors +(41, +42).

While our understanding of brain mechanisms responsible for activating goal-directed behavior is considerable, the mechanisms by which the circuit in +Figure 1 determines or "chooses" the direction of behavior are less clear. Choice is initiated in part by means of the prefrontal cortex, as some studies have shown that activation of the prefrontal cortex precedes behavior +(33, +43, +44). Glutamatergic efferents from the prefrontal cortex stimulate behavioral output by accessing accumbens-thalamocortical circuitry +(45, +46). It has long been proposed that distinct neuronal ensembles within the accumbens encode the relationship between discrete stimuli and behavioral responses +(47). However, only recently was it demonstrated that different subsets of neurons in the accumbens respond differentially to cues associated with distinct motivationally relevant stimuli, such as water versus cocaine +(48). How these ensembles are formed and organized is unclear. However, the intensity and quality of behavioral output are strongly influenced by both dopaminergic and glutamatergic input to the accumbens, and activity at these synapses produces morphological changes in the dendrites of accumbens spiny cells +(49). Changes in dendritic spine density occur in cellular and in vivo models of learning, and they correlate roughly with excitatory synaptic contacts +(50). In addition to morphological changes, in vitro models of neuroplasticity reveal intracellular changes that can augment or diminish excitatory transmission +(51, +52). Recent studies demonstrate that addiction is associated with neuroplasticity in these cellular mechanisms of synaptic organization, and we will discuss them in detail.

Repeated use of addictive drugs induces reorganization in the circuit shown in +Figure 1 to establish behaviors characteristic of addiction, and +Figure 2 illustrates connections in the circuit that are particularly critical for craving and drug seeking. In preclinical studies the most widely used animal model is training rodents to self-administer a drug +(18). The behavior is extinguished, making it possible to reinstate drug seeking by exposing the animals to stimuli similar to those that elicit craving in human addicts, such as a cue previously associated with drug delivery, a mild stressor, or a single dose of the drug +(18). In great measure the circuit in +Figure 2 was assembled by inactivating specific nuclei in animals tested for reinstatement of drug seeking, as well as using cerebral blood flow or blood-oxygenation-level-dependent responses in addicts elicited by the presentation of cues previously associated with drug use +(15). Three general principles emerge from the circuit in +Figure 2.

First, this glutamatergic projection is a final common pathway for initiation of drug seeking. Inactivation of the prefrontal cortex in rats prevents the reinstatement of drug seeking by all three modalities of reinstating stimuli +(29, +53+–55). Further supporting an obligatory role of this projection in drug seeking, AMPA glutamate receptor antagonists in the nucleus accumbens prevent drug- and cue-induced reinstatement +(40, +56, +57). Moreover, increased release of glutamate in the nucleus accumbens occurs following drug- and stress-primed reinstatement, and treatments that prevent the release of glutamate also prevent drug seeking +(29, +58). Within the accumbens the region most strongly associated with drug seeking is the core subcompartment, consistent with the role of this region in emitting learned behavioral responses (see the preceding). Neuroimaging studies support a strong linkage between the prefrontal cortex and drug seeking. The magnitude of change in metabolic activity in both the orbitofrontal and anterior cingulate cortices statistically correlates with the intensity of the self-reported cue-induced craving +(33, +35, +59, +60). It is interesting that the craving-associated increase in prefrontal activity is on a drug-free background of reduced activity +(13, +15, +61, +62). Moreover, activation of the anterior cingulate and orbitofrontal cortex in addicts is inhibited in experimental situations of decision making +(63, +64) and in response to biologically relevant rewards, such as sexually evocative cues +(30). Together, these data indicate that dysregulation in the anterior cingulate and orbitofrontal cortex is critically involved in the prepotent motivation by stimuli predicting drug availability relative to stimuli associated with biological rewards, as well as the difficulty experienced by addicts in cognitive control over drug seeking. Indeed, just as hyperactivity of the anterior cingulate and orbitofrontal cortex contributes to compulsive behaviors in patients with obsessive-compulsive disorders, the relative hypermetabolism in response to drug-related stimuli could trigger compulsive drug intake +(65).

The second principle in the circuitry of drug seeking is that different modes of stimuli inducing drug seeking involve distinct components of the circuit. In contrast to the obligatory role of the glutamate projection from the prefrontal cortex to the accumbens core in drug seeking, distinct nuclei in +Figure 2 regulate reinstatement in response to selective stimuli. For example, cue-primed drug seeking requires involvement of the basolateral amygdala +(66–+68), while stress- and drug-primed drug seeking do not +(29, +55, +69). Also, stress-induced drug seeking selectively engages nuclei in the extended amygdala +(29, +70). Consistent with a role for the amygdala in recognition of cue association with drug use and not in determining cue salience or the intensity of the behavioral responding, neuroimaging studies reveal that the amygdala is not consistently correlated with the reported intensity of craving +(33, +59, +71).

The last principle is that all three modalities of drug-seeking stimuli require dopamine transmission. Since drug seeking is inhibited by inactivation of the ventral tegmental area regardless of the stimulus modality employed, the mesocorticolimbic dopamine projection is obligatory for reinstatement +(29, +55, +72). However, while the rewarding effects accompanying the acute administration of most drugs of abuse depend on increased dopamine release in the accumbens +(2, +3), the reinstatement of drug seeking requires dopamine release in the prefrontal cortex and amygdala +(29, +53, +55, +73), not in the nucleus accumbens core +(40, +55, +57). Dopamine release in the prefrontal cortex is antecedent to activation of the projection from the prefrontal cortex to the accumbens core since preventing cortical dopamine release prevents glutamate release in the nucleus accumbens by a stress or drug prime +(29, +58). Also, reinstatement elicited by dopamine release in the prefrontal cortex is blocked by glutamate antagonists in the accumbens core +(56). Data from imaging studies support the idea that once a person is addicted, dopamine release into the accumbens is not critical for craving. By using positron emission tomography and the D_2/3 dopamine receptor ligand [¹¹C]raclopride in combination with a dopamine reuptake inhibitor such as methylphenidate, it is possible to estimate dopamine release. These studies have corroborated in humans that increases in dopamine in the striatum are associated with the reinforcing effects of stimulants (as evidenced by self-reported "high") (+Figure 3). However, in relation to comparison subjects, cocaine-addicted subjects showed less dopamine release in parallel with fewer self-reports of a methylphenidate-induced "high" +(74). In contrast, an intense methylphenidate-induced cocaine craving in cocaine abusers but not in comparison subjects indicates that addiction is associated not with either enhanced drug-induced dopamine release in the striatum or an augmented pleasurable response to the drug but, rather, with enhanced motivation to procure the drug. As predicted by the animal studies, methylphenidate-induced craving in cocaine-addicted subjects is not associated with dopamine increases in the striatum but with increased activity in the orbitofrontal cortex (+Figure 3). In addition, further contributing to the hypofunction of striatal dopamine is low D₂ receptor availability in the striatum of addicted subjects +(65). Given the role we have described for dopamine transmission in the accumbens, to signal the salience of novel motivational events, the decrease in dopamine release and reception combined with the reduced activation of the prefrontal cortex in response to biologically relevant stimuli (see the preceding) may explain the reduced sensitivity of addicted subjects to "natural" reinforcers.

The portrait of addiction drawn by the studies we have discussed indicates that dopamine release in the accumbens is required for the drug high and for the initiation of addiction but that repeated use of a drug causes gradual recruitment of the prefrontal cortex and its glutamatergic efferents to the accumbens.

The switch from dopamine- to glutamate-based behavior reveals that the development of addiction occurs in a chronological sequence during which different parts of the circuit are preeminent. Similarly, cellular adaptations occur in a chronological sequence. Three temporally sequenced stages of addiction are illustrated in +Figure 4: 1) acute drug effects, 2) transition from recreational use to patterns of use characteristic of addiction, and 3) end-stage addiction, which is characterized by an overwhelming desire to obtain the drug, a diminished ability to control drug seeking, and reduced pleasure from biological rewards.

The acute rewarding effects of drugs involve supraphysiological dopamine release throughout the motive circuit, which induces changes in cell signaling. The prototype signaling cascade in this regard is D₁ dopamine receptor stimulation resulting in the activation of cAMP-dependent protein kinase (PKA), PKA-induced phosphorylation of the transcriptional regulator cAMP response element binding protein (CREB), and the induction of immediate early gene products, such as cFos +(75, +76). The induction of Fos and other immediate early genes promotes short-term neuroplastic changes in response to the acute drug injection that persist for a few hours or days after drug administration +(77). Thus, these molecular consequences of acute drug administration are widely distributed in the motive circuit and initiate cellular events leading up to addiction, but they do not mediate the enduring behavioral consequences of addiction.

The transition from recreational drug use to addiction is associated with changes in neuron function that accumulate with repeated administration and diminish over days or weeks after discontinuation of drug use. The most well-studied molecular adaptation in this category is D₁-receptor-mediated stimulation of proteins with long half-lives, such as DFosB +(78). ΔFosB is a transcriptional regulator that modulates the synthesis of certain AMPA glutamate receptor subunits and cell-signaling enzymes +(79, +80). Recently, the pattern of gene expression induced by long-term induction of ΔFosB was shown to overlap considerably with the pattern of changes induced in the accumbens by chronic cocaine administration, strongly implicating ΔFosB in mediating the transition to addiction +(81). In addition to the effects of ΔFosB, elevation of the GluR1 glutamate receptor subunit in the ventral tegmental area for a few days after discontinuation of cocaine administration may contribute to the development of addiction +(82). Also, there are alterations in the content and function of various proteins directly involved in dopamine transmission that endure for a few days after drug administration stops; these proteins include tyrosine hydroxylase, dopamine transporters, RGS9-2, and D₂ autoreceptors +(83, +84). However, these changes in dopamine transmission appear to be compensatory and may not directly mediate the transition to addiction.

Vulnerability to relapse in end-stage addiction endures for years and results from equally enduring cellular changes. It is interesting that, like the locomotor sensitization and drug-seeking behaviors associated with addiction +(85, +86), changes in protein content and/or function in this category often become greater with increasing periods of withdrawal +(87–+89). This temporal characteristic is consistent with the possibility that the more temporary changes in protein expression that mediate the transition to addiction (see the preceding) may induce changes in protein expression that convert vulnerability to relapse from temporary and reversible into permanent features of addiction.

The involvement of the glutamate projection from the prefrontal cortex to the accumbens core as a final common pathway for initiating drug seeking poses molecular changes in the projection as potential mediators of the uncontrollable desire to take drugs that characterizes addiction. Moreover, the search for cellular changes in the projection from the prefrontal cortex to the accumbens that are involved in the recruitment of this pathway by repeated drug use may reveal pharmacological targets for ameliorating craving and relapse.

Withdrawal from repeated administration of psychostimulants or opioids results in dysmorphisms in the dendritic tree of pyramidal cells in the prefrontal cortex +(49, +90). The enduring drug-induced morphological plasticity indicates long-lasting alterations in neurotransmission +(91). Accordingly, there is reduced cell signaling through transmitter receptors coupled to G_i (a G protein subtype) +(92, +93) that is mediated in part by an elevation in the G protein binding protein AGS3 +(87). It is important that an elevated AGS3 level increases the activity of prefrontal glutamatergic projections to the nucleus accumbens and that normalizing the level of AGS3 in cocaine-addicted rats prevents the reinstatement of drug seeking +(87). Recently, it was postulated that the increased excitability of pyramidal cells may result from AGS3 inhibition of D₂ receptor signaling and a corresponding relative increase in D₁ signaling +(94). Supporting this possibility are the findings that D₁ receptor blockade in the prefrontal cortex prevents the reinstatement of drug seeking +(53) and that withdrawal from repeated amphetamine administration renders pyramidal cells more excitable +(95).

In the nucleus accumbens there are two categories of adaptation in glutamate transmission, those that promote presynaptic glutamate release and those altering postsynaptic responsiveness to released glutamate. Increased release of glutamate in response to a stimulus that induces drug seeking arises in part from adaptations that reduce inhibitory presynaptic regulation by metabotropic glutamate (mGluR2/3) inhibitory autoreceptors +(96, +97), and perhaps from alterations in the organization of vesicles in presynaptic terminals +(98–+100). There is a reduction in basal extracellular glutamate in the nucleus accumbens after withdrawal from cocaine that results from reduced glial cystine-glutamate exchange +(96). The cystine-glutamate exchanger is the rate-limiting step in glutathione synthesis +(101) and is responsible for the majority of extracellular glutamate outside the synaptic cleft +(102). Thus, the exchanger is the primary contributor to maintenance of tone on mGluR2/3 inhibitory autoreceptors and thereby inhibits the release of synaptic glutamate +(103). It is noteworthy that activation of cystine-glutamate exchange by procysteine drugs prevents the reinstatement of drug seeking by restoring extrasynaptic glutamate and stimulating inhibitory presynaptic mGluR2/3 +(96, +103).

Postsynaptic responses to glutamate in the accumbens of animals withdrawn from cocaine reveal enduring adaptations in postsynaptic receptor associated proteins (postsynaptic density) that can alter glutamate receptor intracellular signaling and trafficking to the membrane. This includes reductions in scaffolding proteins such as PSD-95 +(104) and Homer +(89, +105). Together, these changes in postsynaptic density proteins may account for the dendritic dysmorphisms produced in the accumbens by withdrawal from psychostimulants or morphine +(49, +90). The cellular ramifications of changes in postsynaptic density induced by cocaine withdrawal remain to be fully investigated. However, animals with constitutive Homer2 gene deletion show remarkable similarities to animals withdrawn from chronic cocaine, including reduced cystine-glutamate exchange, increased AGS3 levels, increased releasability of glutamate, as well as augmented behavioral responsiveness to cocaine +(106). Moreover, many of the effects of Homer2 gene deletion are normalized when Homer2 is restored to the accumbens by viral transfection. Similarly, the deletion of the PSD-95 gene augments the acute behavioral response to cocaine +(104).

The enduring changes in proteins regulating excitatory transmission in the projection from the prefrontal cortex to accumbens core point to a number of novel targets for treating addiction. However, it is important to remember that additional pharmacotherapies could target other components of the circuit shown in +Figure 2 that pertain selectively to stress- or cue-induced relapse. These include drugs that 1) decrease the motivational value of the drug, 2) increase the salience and motivational value of nondrug reinforcers, or 3) inhibit conditioned responses to stimuli predicting drug availability. Most advanced in this regard are drugs that restore inhibitory presynaptic regulation of excitatory transmission as a means to interfere with the enhanced salience of the drug or drug-related cues. +Figure 5 illustrates that this has been approached by three different mechanisms, all of which have been shown to reduce drug seeking in the reinstatement model of relapse. Procysteine drugs such as N-acetylcysteine have been used clinically to stimulate cystine-glutamate exchange and restore glutathione following acetaminophen overdose +(107). N-Acetylcysteine administration to rats withdrawn from cocaine self-administration restores glutamatergic tone on mGluR2/3 inhibitory presynaptic receptors and abolishes cocaine-induced reinstatement of drug seeking +(96). It is noteworthy that in animal models it is effective only in cocaine-trained subjects and is without effect on control animals or animals trained in a food-seeking paradigm +(96). Direct stimulation of mGluR2/3 with systemically active agonists also reduces the reinstatement of drug seeking +(108). Finally, a portion of decreased mGluR2/3 inhibitory tone on presynaptic glutamate release arises from the increased levels of AGS3 produced by withdrawal from repeated cocaine, and restoring AGS3 levels to normal in the prefrontal cortex blocks cocaine-induced reinstatement of drug seeking +(87). Unfortunately, the pharmacological techniques for manipulating AGS3 levels in vivo require intracellular transduction of active protein, and the currently available technologies to accomplish this (e.g., Tat fusion proteins, viral transfection, or oligonucleotide/iRNA infusions) are not available for clinical use.

The changes in postsynaptic proteins induced by cocaine withdrawal also point to potential pharmacotherapeutic interventions. Notably, blockade of AMPA glutamate receptors in the nucleus accumbens prevents cocaine- or cue-primed reinstatement of drug seeking in animal models +(40, +56, +57). In addition, while tests in the reinstatement model have not yet been conducted, restoration of the scaffolding proteins Homer and PSD-95 may also inhibit drug seeking +(104, +106). Unfortunately, akin to AGS3, reagents to manipulate scaffolding proteins are not yet available for use in humans. A final promising discovery is that mice with a deletion of the mGluR5 gene show reduced responsiveness to cocaine +(109), and systemic administration of an mGluR5 antagonist reduces cocaine and nicotine self-administration +(110).

The cardinal behavioral feature of drug addiction is continued vulnerability to relapse after years of drug abstinence. Vulnerability arises from an intense desire for the drug and reduced capacity to control that desire. Addiction can be viewed as a pathology in how importance is attached to stimuli that predict drug availability and how the brain regulates (chooses) behavioral output in response to those stimuli. Thus, end-stage addiction is characterized by the excessive motivational importance of drug seeking. The glutamatergic projection from the prefrontal cortex to the accumbens is a final common pathway for eliciting drug seeking. This anatomical locus of pathology is consistent with behavioral dysregulation in addiction, since the prefrontal-accumbens projection is credited with providing the properties of motivational salience and direction to normal goal-directed behavior. Recently, some of the cellular mediators of the pathology in the prefrontal-accumbens glutamate projection have been identified. These include 1) alterations in G protein signaling in the prefrontal cortex that increase the excitability of neurons projecting to the accumbens, 2) augmented presynaptic glutamate release in the accumbens due to reduced inhibitory presynaptic regulation and increased releasability of synaptic vesicles, and 3) alterations in postsynaptic proteins that result in rigid dendritic morphology and signaling. These findings combined with functional imaging studies in addicts reveal a situation whereby prefrontal regulation of behavior is reduced in basal conditions, thereby contributing to the reduced salience of nondrug motivational stimuli and reduced decision-making ability. However, when stimuli predicting drug availability are presented, there is profound activation of the prefrontal cortex and glutamatergic drive to the nucleus accumbens. Combined with the cellular neuroadaptations in the accumbens that render excitatory synapses relatively immune to regulation, the increased prefrontal drive contributes overpowering motivational salience to drug-associated stimuli and correspondingly impels craving and drug seeking. The combined functional and cellular evidence predicts that pharmacotherapeutic agents regulating prefrontal glutamatergic drive to the nucleus accumbens have the potential to ameliorate both the excessive motivational importance given to stimuli that predict drug availability and the reduced capacity of addicts to inhibit drug intake.

Presented at the 157th annual meeting of the American Psychiatric Association, New York, May 1–6, 2004. Received July 7, 2004; revision received Aug. 26, 2004; accepted Sept. 9, 2004. From the Department of Neurosciences, Medical University of South Carolina; and the National Institute on Drug Abuse, Bethesda, Md. Address correspondence and reprint requests to Dr. Kalivas, Department of Neurosciences, Medical University of South Carolina, 173 Ashley Ave., BSB-403, Charleston, SC 29425; kalivasp@musc.edu (e-mail). Supported in part by National Institute on Drug Abuse grants DA-12513, DA-05369, and DA-03906.