The Ventral Pallidum: Proposed Integrator of Positive and Negative Factors in Cocaine Abuse

Abstract

In this issue of Neuron, Creed et al. (2016) describe how cocaine produces divergent forms of plasticity at synapses between specific neurons in nucleus accumbens and ventral pallidum, and how these changes are associated with positive and negative reward behaviors.

Addiction involves the transition from recreational drug use to compulsive and maladaptive drug seeking that is maintained despite negative social, financial, and health consequences (Koob and Le Moal, 2008). Cessation of drug use is associated with the emergence of a withdrawal syndrome characterized by negative affect that promotes further drug use via negative reinforcement processes (Koob and Le Moal, 2008). A complex interplay between “positive” and “negative” symptoms contributes to the chronic relapsing nature of addiction. Despite significant effort, effective pharmacotherapies for psychostimulant addiction are still lacking. Such slow progress is likely due in part to the propensity of preclinical investigations to study individual positive or negative addiction-related behaviors. Development of novel treatments for addiction would be significantly enhanced by identification of brain systems that modulate the expression of complex endophenotypes that encompass both positive and negative symptoms.

Repeated drug use is thought to remodel synapses in brain circuits involved in reward and motivation. Such changes are perhaps best characterized in the nucleus accumbens (NAc), which undergoes characteristic forms of plasticity in part due to high levels of dopamine transmission in this region following drug use (Steketee and Kalivas, 2011). GABAergic medium spiny neurons (MSNs) account for ~95% of all NAc neurons and provide the primary projections from this region. These MSNs are typically divided into two subpopulations based on their expression of either D1 or D2 dopamine receptors, their releasable peptides and their axonal projection targets (Lobo and Nestler, 2011). The canonical understanding is that these subpopulations influence motivated behavior via two parallel pathways: the direct pathway, consisting of projections from D1-MSNs directly to basal ganglia output nuclei, and the indirect pathway, composed of D2-expressing MSNs that influence output nuclei indirectly via synapses in the ventral pallidum (VP) (Lobo and Nestler, 2011). Cocaine potentiates synaptic input onto the direct pathway, which contributes to the expression of “positive” addiction-related behaviors such as behavioral sensitization and cued drug seeking (Pascoli et al., 2015). On the other hand, alterations in D2-MSNs are associated with the expression of depression-like behavior (Francis et al., 2015). Because of their distinct projection patterns, it has been assumed that D1 -and D2-MSNs regulate these divergent states via separate circuits.

Recent work questions the anatomical segregation of D1- and D2-MSN outputs, with evidence that both D1- and D2-MSNs arising from NAc core innervate the dorsal VP (dVP) (Kupchik et al., 2015). This potentially places the VP in a strategic anatomical position to integrate D1- and D2-MSN activity and relay this information to key motivational circuits. In this issue of Neuron, Creed et al. (2016) suggest that the VP may be a site of convergence for drug-evoked synaptic changes that contribute to both positive and negative symptoms following cocaine experience.

In their paper, Creed et al. (2016) focus on the ventromedial subdivision of the VP (vmVP), as this subregion sends projections to the ventral tegmental area (VTA)—a pathway recently shown to be important for drug-seeking behavior (Mahler et al., 2014). Because this subdivision of VP is predominantly innervated by the shell component of the NAc (distinct from the NAc core to dVP projections in Kupchik et al., 2015), Creed et al. (2016) first confirmed that vmVP is innervated by both D1- and D2-MSNs. Retrograde tracer cholera toxin B (CTb) was injected into the vmVP of transgenic mice that expressed a reporter protein in either D1- or D2-MSNs. In D1-reporter mice, approximately 50% of NAc neurons that projected to vmVP arose from D1-MSNs. An almost identical pattern was observed in D2-reporter mice, indicating that both neuron subtypes project to vmVP in similar proportions. Because these experiments were carried out in separate animals, it was unclear whether D1- and D2-MSNs innervate the same VP neurons or whether their projections segregate onto separate VP target neurons. Creed et al. (2016) recorded responses of VP cells following local optogenetic stimulation of either D1- or D2-MSN terminals and found that almost all (92.8%) VP cells received innervation from D1-MSNs, whereas 75% of VP cells were innervated by D2-MSNs. These findings indicate that the D1- and D2-MSNs innervate overlapping populations of vmVP neurons. This is consistent with the finding from Kupchik et al. (2015), who showed that up to 50% of dVP neurons are innervated by both D1- and D2-MSN afferents.

To understand how cocaine may affect signaling from D1- and D2-MSNs to VP neurons, Creed et al. (2016) first characterized these synapses under drug-free conditions. Activity-dependent plasticity of VP neurons was assessed following high-frequency optogenetic stimulation (HFS) of D1- or D2-MSN terminals. Recordings were carried out only in VP neurons that project to VTA as identified by retrograde labeling. D1-MSN-VP synapses showed activitydependent long-term potentiation (LTP), whereas identical patterns of stimulation produced long-term depression (LTD) at D2-MSN-VP synapses. These data raised a critical question: how can the same stimulation protocol produce divergent forms of plasticity at D1- versus D2-MSN-VP synapses? Creed et al. (2016) show that signaling through D1Rs is necessary for LTP expression at D1-MSN-VP synapses, consistent with previous findings from this group that ambient signaling at the D1R is permissive for plasticity in other reward regions (Creed et al., 2015). With respect to LTD at D2-MSN-VP synapses, Creed et al. (2016) note that these neurons differ from D1-MSNs in that they corelease enkephalin which acts at presynaptic δ-opioid receptors (DORs). They hypothesize that DORs might act as autoreceptors when opioids are released following HFS, decreasing the probability of transmitter release from the presynaptic terminal. Consistent with this idea, application of a DOR agonist induced LTD at D2-MSN-VP synapses, whereas a DOR antagonist blocked HFS-induced LTD.

If D1- and D2-MSN projections to VP are involved in addiction behaviors, then plasticity at these synapses should be susceptible to drug exposure. To test this, Creed et al. (2016) subjected mice to five daily injections of cocaine followed by a 10 day withdrawal period and then assessed activity-dependent plasticity as before. Remarkably, this treatment prevented the induction of either LTP or LTD, perhaps via occlusion at these synapses. Consistent with this, cocaine withdrawal treatment was associated with increased release probability at D1-MSN-VP synapses, but decreased release probability at D2-MSN-VP synapses. In other words, cocaine exposure potentiated the output of D1-MSNs onto VP neurons and simultaneously decreased the output of D2-MSNs.

Could these divergent changes in synaptic transmission at D1- and D2-MSN-VP synapses underlie the elevated positive and negative symptoms that characterize addiction? First, it was important to demonstrate that the cocaine regimen utilized for the plasticity experiments induced behaviors that reflect these divergent symptoms. Behavioral sensitization, or the augmentation of motor-stimulant responses with repeated drug exposure, is considered by some investigators to be a “positive symptom” as it is associated with increased drug craving and relapse propensity (Steketee and Kalivas, 2011). The authors report that cocaine-exposed mice showed behavioral sensitization with repeated drug exposure that persisted at 10 days post-withdrawal. With respect to negative symptomology, both clinical and preclinical evidence indicates that chronic drug use is associated with a reduction in the reinforcing efficacy of natural reward (Garavan et al., 2000). Therefore, Creed et al. (2016) assessed animals on tests of motivation and hedonic impact of a natural reward. Following 10 day withdrawal, cocaine-exposed mice showed significantly reduced break points for sucrose, a commonly used measure of motivation for a natural reward. In a separate task, where mice had free access to sucrose, cocaine-withdrawn mice exhibited significantly fewer stereotyped hedonic tongue protrusions, reflecting reduced “liking” of the sucrose solution. The simultaneous expression of positive and negative reward symptoms provided an opportunity for investigating the possibility that these behaviors are differentially regulated by changes in D1- and D2-MSN transmission in VP.

To address this question, Creed et al. (2016) examined whether normalizing transmission at both D1- and D2-MSN-VP synapses could reverse the behavioral effects of cocaine withdrawal. In a series of experiments first carried out in brain slices of cocaine-naive mice, the authors showed that they could selectively decrease transmission at D1-MSN-VP synapses by providing intermittent, low-frequency (1 Hz) optogenetic stimulation at this synapse. At D2-MSN-VP synapses, medium frequency (10 Hz) stimulation resulted in increased transmission of inhibitory output. These stimulation protocols were then applied in vivo in cocaine-withdrawn animals. Remarkably, application of these stimulation protocols in cocaine-withdrawn rats completely normalized transmission at both D1- and D2-MSN outputs in the VP.

Can reversing the effects of cocaine at these synapses modify the expression of addiction behaviors? Creed et al. (2016) show that low-frequency stimulation that normalized transmission at D1-MSN-VP synapses, but not at D2-MSN-VP synapses, abolished the sensitized locomotor response to cocaine on day 10 of withdrawal. Conversely, medium frequency stimulation that potentiated throughput at D2-MSN-VP synapses, but not at D1-MSN-VP synapses, increased break points on the sucrose progressive ratio task and increased the amount of hedonic tongue protrusions in the sucrose free access task. Taken together, these data indicate that changes in transmission at D1 and D2-MSNs converge in VP to integrate both positive and negative changes in reward processing during cocaine withdrawal.

The importance of accumbal projections to VP in the control of motivated behavior has been known for over 30 years. However, increasingly sophisticated genetic techniques, such as those utilized here, are providing new insight into the intricacies of this circuit. Indeed, the elegant series of experiments described by Creed et al. (2016) provides a synaptic-level understanding of cocaine-induced changes in VP function and how these may contribute to the complex cluster of symptoms that characterize addiction. However, these findings also raise a new set of important questions. For example, how is it that D1- and D2-MSNs innervate the same VP neurons but mediate two distinct reward-related behaviors, presumably via distinct VP outputs? It is possible that VP neurons exclusively innervated by D1-MSNs are especially involved in the expression of behavioral sensitization, whereas VP neurons that receive input only from D2-MSNs drive negative affective states. This may seem unlikely given that both MSN subtypes seem to innervate many of the same VP neurons, but there is evidence of small and overlapping neuronal ensembles in other brain regions playing distinct roles in the expression of reward behaviors (Cruz et al., 2013). Another possibility is that D1- and D2-MSNs might innervate different compartments on the VP neuron, or modulate VP neuronal signaling so that distinct patterns of impulse activity, or selectively altered responsiveness to specific inputs, yield different addiction-associated behaviors. Clearly, future studies are required to address these issues.

It is also interesting to consider that D2 is a Gi-coupled receptor, meaning that the increased dopamine availability after acute cocaine likely inhibits D2-MSNs. How, then, do these neurons come to release enkephalins at levels that are sufficient to act at presynaptic delta-opioid receptors, resulting in LTD at these synapses? Such neuropeptide release would normally follow high-frequency neural activity rather than neuronal inhibition. Finally, it will be important for future studies to explore the involvement of D1- and D2-MSN outputs in VP using more comprehensive behavioral assessments of positive and negative addiction symptoms. Creed et al. (2016) provide evidence that plasticity occurs at D1- and D2-MSN-VP synapses in animals trained to self-administer cocaine; however, it remains to be shown that reversal of these processes affects behaviors such as drug seeking. Also, do these changes in synaptic plasticity underlie addiction behaviors associated with long-term drug abuse, such as escalation of cocaine intake, demand, and punished (compulsive) drug taking (Bentzley et al., 2014)? Similarly, how do the changes in motivation for sucrose shown here relate to behavior on more direct measures of negative affective state such as assessments of depression- and anxiety-like behavior? Regardless, the findings of Creed et al. (2016) provide novel and exciting insights into the role of VP in cocaine withdrawal and highlight sophisticated synaptic mechanisms that may be attractive targets for therapies designed to normalize behavior following cocaine.