Sufficiency of Mesolimbic Dopamine Neuron Stimulation for the Progression to Addiction (2015)

Introduction

Addiction is a disease that evolves in several steps (Everitt et al., 2008, George et al., 2014). The diagnosis is made when recreational use becomes compulsive, persisting despite negative consequences. While a leading addiction hypothesis posits that drugs of abuse cause the disease because they excessively increase the concentration of dopamine (DA) in the brain, it is unclear whether triggering this system is sufficient to drive the transitions from recreational use to addiction (Di Chiara and Bassareo, 2007, Volkow and Morales, 2015). The supporting evidence for the DA hypothesis for drug reinforcement has accumulated over several decades and relies on the initial effect of drugs. For example, addictive drugs reduce the threshold for intracranial self-stimulation (ICSS) of the medial forebrain bundle, a fiber tract containing, among others, ascending DA projection from the midbrain (Stein, 1964, Crow, 1970, Kornetsky et al., 1979). Pharmacology and lesion studies then identified the mesocorticolimbic DA system as the origin of this circuit (Wise and Bozarth, 1982). In the late 1980s, a direct measure of the extracellular DA concentration with microdialysis confirmed that addictive drugs shared the property of evoking a DA surge in the NAc (Di Chiara and Imperato, 1988). This led to the proposal of a mechanistic classification of addictive drugs (Lüscher and Ungless, 2006).

Much less is known of how these initial effects of drug use facilitate the transition to addiction. DA-independent mechanisms have been considered because addictive drugs have other pharmacological targets. For instance, cocaine, in addition to inhibiting the DA transporter (DAT), also binds to SERT (serotonin transporter) and NET (norepinephrine transporter) to decrease serotonin and norepinephrine reuptake, respectively, thus increasing the concentration of all major monoamines (Han and Gu, 2006, Tassin, 2008). Similar concerns may apply to other psychostimulants. Moreover, there is a claim that opiates are, at least in the initial phase, DA independent (Badiani et al., 2011, Ting-A-Kee and van der Kooy, 2012). The DA hypothesis has also been challenged based on genetic mouse models, where, after interference with the DA system, some forms of drug-adaptive behavior were still apparent. For example, DAT knockout mice self-administer cocaine (Rocha et al., 1998), and abolishing DA synthesis either pharmacologically (Pettit et al., 1984) or genetically (Hnasko et al., 2007) failed to prevent drug self-administration or conditioned place preference. While better characterization of these transgenic mice and generation of double monoamine transporters knockouts have resolved some of these issues (Rocha, 2003, Thomsen et al., 2009), the sufficiency of DA to trigger cardinal features of addiction is unknown. To circumvent issues of non-specificity, we have therefore decided to allow mice to self-stimulate VTA DA neurons using an optogenetic approach.

Recent studies have shown that activation of DA neurons in the midbrain can induce place preference (Tsai et al., 2009) or reinforce instrumental behavior (Adamantidis et al., 2011, Witten et al., 2011, Kim et al., 2012, Rossi et al., 2013, McDevitt et al., 2014, Ilango et al., 2014). While this selective activation of DA pathways confirms intracranial self-stimulation (ICSS) studies carried out more than 30 years ago in delineating the reward system (Fouriezos et al., 1978), they fall short demonstrating the induction of late-stage adaptive behavior that defines addiction, nor did they identify the underlying neuronal adaptations. Here we used optogenetic manipulation not only to allow for direct testing of the sufficiency criterion for phasic DA signaling in initiating reinforcement, but also to test for the transition to addiction.

A striking observation of the later stages of the disease is that even with the most addictive drugs, only a fraction of users becomes addicted (Warner et al., 1995, O’Brien, 1997). Human addicts will continue drug consumption despite negative consequences (see American Society for Addiction Medicine’s “Definition of Addiction,” DSM5, American Psychiatric Association, 2013), typically related to social and psychological defeats that are often delayed in time. Similarly, in rodents roughly one out of five animals that acquire self-administration of cocaine are eventually classified as addicted (Deroche-Gamonet et al., 2004, Kasanetz et al., 2010; but see George et al., 2014). Perseverance of drug intake despite negative consequences can also be modeled in rodents by introducing a simple aversive stimulus to the consumption schedule. While the human disease is more complex, associating punishment with consumption is a straightforward model of a core component of addiction.

Here, we used a mild foot shock to evaluate its consequence on self-administration of cocaine, sucrose, and optogenetic self-stimulation. We further investigate whether DA neuron self-stimulation can induce two addictive-related behaviors—cue-associated reward seeking and compulsivity associated with consumption despite negative consequences—and characterize the neural plasticity associated with these behaviors.

Results

To control DA neuron activity, we injected a Cre-inducible adeno-associated virus (AAV) with a double-floxed inverted open reading frame (DIO) containing ChR2 fused to enhanced yellow fluorescent protein (eYFP) (Atasoy et al., 2008, Brown et al., 2010) into the VTA of DAT-Cre mice. In addition, an optic fiber was placed to target the VTA (ChR2, see Experimental Procedures). Specificity of the ChR2 expression was confirmed by the co-localization of eYFP with Tyrosine Hydroxylase (TH), an enzyme required for DA synthesis (Figure 1A). 

First, to establish the laser stimulation protocol, mice were placed in an operant box where they could press an active lever, which triggered a number of laser stimulations that was varied (1, 2, 8, 32, 60, or 120 bursts) every two sessions. To emulate phasic firing pattern (Hyland et al., 2002, Mameli-Engvall et al., 2006, Zhang et al., 2009) typically induced by natural reward (Schultz, 1998), we used burst stimulation. One burst consisted of five laser pulses of 4 ms, at 20 Hz, and was repeated twice per second. We found that mice adapted their lever-pressing behavior as a function of bursts per laser stimulation, thus controlling the total number of bursts received (Figure 1B). This behavior was reminiscent of self-administration of addictive drugs, when the dose per infusion was varied (Piazza et al., 2000). For the subsequent experiments, we chose to administer 30 bursts per lever press, yielding a half-maximal number of bursts (Figure 1B). To mimic the delay in DA increase typically observed when drugs are administered intravenously (Aragona et al., 2008), we delayed the laser stimulation by 5 s and added a flashing cue light for 10 s (Figure 1C).

During 12 consecutive days, mice were allowed to self-stimulate a maximum of 80 times in 2 hr. Mice quickly increased the rate of laser stimulation, reaching 80 laser stimulations (LS) before the end of the first hour of a session (Figures 1D and 1E). The distinction between the active and inactive lever was rapidly acquired and the number of active lever presses increased accordingly with increasing fixed ratio (FR1, 2, 3) schedules (Figures 1F and 1G). In control experiments using DAT-Cre− mice or mice that expressed ChR2 in γ-aminobutyric acid (GABA) neurons (GAD-Cre+ mice, to target the inhibitory neurons of the VTA), rates of self-stimulation were low and continuously decreased across sessions. This also applied to two Cre+ animals where post hoc validation showed that the VTA was not infected with ChR2-eYFP (not shown). Moreover, no discrimination between the active and inactive lever was detected (Figures S1A and S1B).

We observed that DAT-Cre+ mice pressed more often on the active lever than required for the laser stimulation. In fact such “futile” active lever presses accounted for more than 30% of all active lever presses (Figure S2A) and occurred—as sessions progressed—mostly between cue and laser stimulation onset (Figures S2B and S2C). This singular behavior developed during acquisition and may reflect impulsive responses.

Taken together, burst activity in VTA DA neurons strongly reinforces lever responding.

Occlusion of VTA DA Neuron Self-Stimulation by Cocaine

To test whether VTA DA neuron self-stimulation hinges on the same brain circuits that are targeted by addictive drugs to reinforce behavior, we injected cocaine intraperitoneally (i.p.) immediately prior to self-stimulation sessions (free access to laser for 45 min, Figure 2A). At baseline, well-trained animals pressed about 400 times to obtain 85 LS in 45 min under the FR3 schedule. After cocaine injection, the performance decreased significantly in a dose-dependent fashion to about 30 LS for 100 lever presses with the highest dose (Figure 2B). This occlusion was most pronounced during the first 30 min of the session, reflecting the pharmacokinetics of the drug (Figure 2C). This experiment indicates that reinforcement by optogenetic self-stimulation and reinforcement by cocaine share underlying neural circuits.

To further compare optogenetic self-stimulation to addictive drugs, we next asked whether mice would relapse to self-stimulation of VTA DA neurons following several weeks of withdrawal. Since cue-associated drug seeking is an established model of relapse (Epstein et al., 2006, Soria et al., 2008, Bossert et al., 2013), we placed mice back into the operant chamber 30 days after the last self-stimulation session, where active lever pressing now triggered the cue light without laser stimulation (Figure 3A). Robust cue-associated seeking behavior, demonstrated by a high rate of active lever presses, was only apparent in mice with expression of eYFP-ChR2 in VTA DA neurons (DAT-Cre+ but not DAT-Cre− mice, Figure 3B).

Previous studies have shown the causal link between cue-associated relapse and synaptic plasticity evoked by cocaine in a subtype of NAc neurons expressing the DA D1R (Pascoli, Terrier et al., 2014). Therefore, to evaluate this synaptic plasticity, we generated DAT-Cre mice crossed with Drd1a-tdTomato mice to identify the medium-sized spiny neurons (MSNs) subtype in the NAc. Instead of the seeking test, slices of the NAc were prepared where D1R-MSNs were red, contrasting with green fibers from VTA DA neurons infected with flox-ChR2-eYFP (Figure 3C). Whole-cell patch-clamp recordings ex vivo revealed a rectifying current voltage relationship for AMPAR-evoked postsynaptic currents (AMPAR-EPSCs) and an increased AMPAR/NMDAR ratio (Figures 3D and 3E), in the D1R-MSNs but not in the D2R-MSNs. Similar findings previously obtained after withdrawal from cocaine self-administration were shown to indicate the combined insertion of GluA2 lacking and GluA2 containing AMPARs, at separate inputs onto D1R-MSNs (Pascoli, Terrier et al., 2014).

Self-Stimulation despite Punishment

Substance use despite negative consequences is anoher crucial defining feature of addiction (see DSM5 definition, American Psychiatric Association, 2013). Rat models have been established (Deroche-Gamonet et al., 2004, Pelloux et al., 2007, Pelloux et al., 2015, Chen et al., 2013) where an electric shock introduced in the cocaine self-administration schedule suppresses cocaine consumption in some animals. Following 12 days of initial exposure (acquisition), mice were allowed to have three additional sessions at FR3 but with a reduced session cut-off (60 min or 40 rewards maximum). These three sessions served as a baseline for the subsequent four sessions, where every third laser stimulation was paired with a foot shock (500 ms; 0.2 mA) predicted by a novel cue (Figure 4A). The intensity and duration of the foot shock were adjusted to completely suppress lever pressing for sucrose reward (see also data below). The punishment schedule led to two opposite behavioral responses (Figure 4B). Some mice rapidly stopped responding when the punishment was introduced (called “sensitive”), whereas others continued responding to obtain the maximum number of laser stimulations and can be considered as “resistant” to punishment. The two clusters of animals fully emerged at the end of the four punishment sessions (Figure 4C). “Resistant mice” maintained the number of laser stimulations (less than 20% reduction) while “sensitive mice” decreased self-stimulation by more than 80%. With these criteria, only one animal (gray dots) could not be assigned. This observation demonstrates that forced burst activity evoked by self-stimulation of VTA DA neurons is sufficient to induce perseverance of consumption despite negative consequences in a fraction of mice. As a control, in an independent group of mice that had established resistance or sensitivity to the punishment associated with self-stimulation, nociception was evaluated using the tail-flick assay. No difference in the latency to withdraw the tail immersed in hot water between sensitive and resistant was detected (Figure S3).

We next asked, post hoc, whether any particular feature during the acquisition phase of self-stimulation could have predicted the resistance to punishment. Sensitive and resistant mice made an identical number of active and inactive lever presses during baseline sessions, and all mice reached the maximum of 80 LS (Figures S4A and S4B), in a similar amount of time (Figures S4A and S4C). While the fraction of futile active lever press was again not different in the two sub-populations (Figures 4D and S4D), the number of futile lever presses before the onset of the laser stimulation became significantly higher in resistant mice by the end of the acquisition sessions (Figures 4E and S4E). As this behavior developed during acquisition, it may contribute, along with innate impulsivity (Economidou et al., 2009, Broos et al., 2012, Jentsch et al., 2014), in establishing the resistance to punishment. In addition, a progressive ratio trial was performed at day 11 to quantify the motivation for the optogenetic stimulation (Richardson and Roberts, 1996). Resistant mice exhibited a breakpoint not statistically different to sensitive mice (Figure S4F).

Resistance to Punishment for Cocaine but Not for Sucrose

To test whether the paradigm of consumption despite harmful consequences along with impulse lever pressing could also predict compulsive intake of an addictive drug, a new cohort of mice underwent 12 days of cocaine self-administration. Experimental parameters for cocaine self-administration acquisition were set to a maximum of 80 infusions of cocaine within 4 hr during acquisition and to 40 infusions within 2 hr during the three baseline sessions preceding the four punishment sessions (Figures 5A and S5A). Again, two groups emerged after pairing cocaine reward with electric shocks. Indeed, 5 out of 22 mice were classified as resistant (less than 20% decrease from baseline), while 17 qualified as sensitive (more than 80% decrease) and one animal fell in between (13 infusions on day 19) (Figure 5B). We then looked for behavioral predictors of resistance to punishment. Between the two groups, the number of infusions, the rate of infusion, and the number of active or inactive lever presses were not different (Figures S5B–S5D), and the breaking points were similar (Figure S5E). What differed was the evolution of the distribution in time of the futile presses on the active lever. In the first four sessions, futile lever presses regularly decreased during the time-out periods in both resistant and sensitive mice, while at the end of the acquisition, only sensitive mice maintained this behavior (Figures 5C and 5D and S5F). By contrast, resistant mice tended to increase their total number of futile lever presses (Figures 5C and S5D), especially in the last quarter of the time-out period (Figure 5D). While qualitatively similar to the observation previously made with the optogenetic stimulation of DA neurons (see above), the clustering of the futile presses during the early time-out period was not seen with cocaine, most likely owing to the slower kinetics with which the drug increased DA levels. Nevertheless, similar conclusions could be drawn based on this singular evolution of futile lever press distribution during the short period of time preceding “the internal detection of the DA surge.” Our observations thus suggest that the distribution of the futile active lever presses predicts drug use despite negative consequences.

Finally we repeated the experiment with ad libitum-fed mice that could lever press for a sucrose reward. Once punishment was introduced, all mice stopped self-administering the sucrose (Figure 5E), demonstrating that this schedule suppressed the intake of a non-essential natural reward, but allowed the detection of compulsive intake of an addictive drug or strong DA neuron stimulation.

Taken together, these results demonstrate that VTA DA self-stimulation is sufficient to induce compulsivity, as shown by the resistance to punishment in a subset of mice (68%). Similarly, after cocaine SA, some mice became resistant to punishment (23%), which was never the case after sucrose SA (Figure 5F).

To pinpoint the brain area that may control the decision to persevere in self-administration despite negative consequences, we first monitored generic “neuronal activity” by counting the number of neurons in which the punishment session triggered the expression of the immediate early gene cFos in 15 different regions. Mice were intracardially perfused with PFA 90 min after the end of the last punishment session. The control groups included naive animals, as well as mice yoked to sensitive or resistant mice in order to control the possibly confounding effect of the number of shocks received.

While in most of the chosen regions, the number of cFos-positive neurons was highest in slices from resistant mice compared to naive mice slices, two types of responses emerged, of which the prelimbic cortex (PL) and the lateral OFC are examples. In the PL we found a similar increase of cFos-positive cells in resistant mice and their yoked controls, while in the OFC this increase was only apparent in the resistant and not the corresponding yoked mice (Figures 6A and 6B ). To quantify this difference, all data were first normalized to expression levels in naive animals. Then, the ratio was calculated between the resistant over sensitive divided by the yoked to resistant over yoked to sensitive (Ratio_cfos = (R/S) / (YR/YS), Figure 6B). This procedure identified the cingulate cortex, the OFC, and VTA as the regions that are activated in resistant but not in sensitive mice and where there was little difference in both groups of yoked controls (similar low cFos-positive neurons in yoked, in fact). Finding the VTA is not surprising, as it is the region of laser-stimulated neurons. This is in line with a previous report showing that ChR2 stimulation triggers cFos activation (Lobo et al., 2010, Van den Oever et al., 2013). A low ratio_cfos was found in regions where the activation was similar in sensitive and resistant (such as CeA and PAG). The ratio_cfos was also low when the activation was paralleled by a high difference in the yoked controls (such as the PL, Figure 6C for summarized ratio_cfos data). A similar cFos expression in resistant and yoked resistant mice was therefore most likely driven by the number of foot shocks and had little to do with the resistance to punishment. Taken together, the high ratio_cfos in the OFC suggests that neural activity in this region is associated with resistance to punishment and may thus favor the transition to addiction.

To identify the substrate of the increased neuronal activity in the OFC in the mice resisting punishment, we prepared slices of the PL and L-OFC 24 hr after the last punishment session to test for intrinsic excitability. The two regions were chosen because of their very distinct pattern of c-Fos expression in the previous experiments. The neuronal excitability was quantified by counting the number of action potentials (APs) elicited by the injection of increasing amounts of current (from 0 to 600 pA) in whole-cell recordings. These recordings revealed a sustained hypo-excitability in pyramidal neurons of the PL of resistant mice (and their yoked control) when compared to sensitive or naive mice (Figure 7A). The resting membrane potential (RMP) of recorded neurons was not different between the experimental groups (Figure 7B). These results strongly suggest that the excitability of neurons in the PL directly correlates with the number of shocks received, and maybe not with the decision to resist punishment. This most likely reflects a negative feedback adaptation triggered by neuronal excitation elicited by the foot shocks the day before. By contrast, neurons from L-OFC were more excitable only in resistant mice. Excitability of neurons from yoked mice was not different than excitability of neurons from naive mice, ruling out an effect of the foot shock itself (Figures 7C and 7D). This increased activity of OFC neurons likely underlies the cFos expression and may drive the resistance to punishment.

To test for causality between enhanced OFC neuron excitability and resistance to punishment, we expressed the inhibitory DREADD (designer receptors exclusively activated by designer drugs: CamKIIα-hM4D) in pyramidal neurons of the OFC of DAT-Cre+ mice (Figure 8A). In acute slices from the OFC, bath application of CNO (clozapine-N-oxide) induced a slow outward current, most likely mediated by GIRK channels, that was reversed by barium (Ba²⁺), a non-specific blocker of potassium channels (Figure 8B). The CNO also shifted the input/output curve to the right (Figure 8C). The DAT-Cre+ mice infected with AAV1/CamKIIα-hM4D-mCherry in the OFC (Figure 8D) acquired DA neuron self-stimulation paradigm followed by two successive blocks with the punishment schedule, the first in the presence of CNO and the second without CNO. The two blocks were interrupted by 6 days without punishment (Figure 8E). At the end of the first punishment block, in the presence of CNO, only 5 of 16 mice were resistant (Figure 8F, left panel). In contrast, without OFC inhibition, during the second punishment period, 14 out of 16 were classified as “resistant” (Figures 8F, right panel, and 8G). In other words the fraction of resistant mice was significantly lower in the presence of CNO compared to the first cohort of 34 mice previously tested in the same conditions (between-group comparison, Figure 8H) and became similar to the first cohort without CNO (within-group comparison). Finally, for the nine mice that changed from sensitive to resistant, CNO did not modify the tail-flick latency upon immersion into hot water (Figure 8I).

Taken together, this experiment demonstrates that the activity of pyramidal neurons of the OFC drives the decision to continue self-stimulation despite negative consequences that represents a key feature of the transition to addiction in rodents.

Discussion 

A recently proposed addiction model distinguishes three steps in the progression of the disease: sporadic recreational drug use, followed by intensified, sustained, escalated drug use, and eventually compulsive use associated with loss of control (Piazza and Deroche-Gamonet, 2013; but see George et al., 2014). Our study demonstrates that stimulation of VTA DA neurons is sufficient to drive this progression with a relatively rapid time course.

By mimicking a naturally occurring burst-firing pattern, an efficient release of DA is evoked in target regions of the VTA, such as the NAc (Bass et al., 2010). DA levels in the NAc therefore likely govern the self-stimulation, just as rodents self-administer the next infusion of cocaine or heroin once the DA concentration drops below threshold (Wise et al., 1995). This is also supported by our observation that cocaine, injected i.p., can occlude self-stimulation. Thus, DA neuron self-stimulation closely resembles drug self-administration, even though its kinetics is certainly faster than any pharmacological substance, including cocaine, as suggested by the different rate of responses observed in the present study.

While we selectively targeted DA neurons of the VTA, their optogenetic self-stimulation may have activated groups of cells with different physiological functions. For example, it has recently been suggested that some DA neurons code for aversive stimuli (Lammel et al., 2012, Gunaydin et al., 2014). These cells project to mPFC, while VTA DA neurons projecting to lateral NAc shell mediate positive reinforcement (Lammel et al., 2012). It would be interesting to assess self-stimulation and progression with selective targeting (Gunaydin et al., 2014). Since our manipulation activated all VTA DA neurons, just as cocaine acts on all DAT-expressing neurons, it is conceivable that some DA neurons would drive reinforcement learning while other DA neurons would drive aversion learning. The net effect would still be a reinforcement of the behavior; however, the “aversion neurons” could contribute to the induction of an opponent process (Koob, 2013, Wise and Koob, 2014).

After forced abstinence, re-exposition to the context induced seeking of the self-stimulation, an established rodent model of drug relapse. Remarkably the underlying neural plasticity is indistinguishable from the one observed after withdrawal from cocaine self-administration (Pascoli, Terrier et al., 2014). This adds to a study that previously reported identical synaptic plasticity in VTA DA neurons evoked by a single session of optical stimulation or a first injection of an addictive drug (Brown et al., 2010). A pattern of synaptic adaptations is emerging that c adaptive behavior common to all addictive drugs.

A striking feature of our study is the dichotomy in the response to an aversive stimulus that is strong enough to disrupt consumption of non-essential natural reward in all animals. In our setting, resistant mice did not show a significantly higher motivation for the reward self-delivery, which contrasts with a study with cocaine in rats (Pelloux et al., 2007). The behavioral predictor for resistance to punishment in mice, however, was futile lever pressing during the 5 s preceding the onset of the DA neuron stimulation. The inability to wait until reward delivery can therefore be seen as a marker of impulsivity (Dalley et al., 2011, Olmstead, 2006, Everitt et al., 2008, Winstanley, 2011, Leyton and Vezina, 2014). We were intrigued by the observation that impulsive taking only developed after several sessions of self-stimulation. This raises the possibility that resistance to punishment (and by extension vulnerability to addiction) may not be fully innate, but develops during the initial phases toward addiction. If this is the case, then the dichotomy observed by us and others (Deroche-Gamonet et al., 2004) may not be solely determined by genetic factors. This would also explain that a similar fraction of individuals becomes addicted in genetically relatively homogeneous mouse strains and genetically certainly more diverse human populations.

If resistance to punishment reveals the individual vulnerability for addiction, estimated to top 20% in humans even with cocaine (Warner et al., 1995, O’Brien, 1997, George et al., 2014), then the much higher proportion found here could reflect the power of the direct and selective DA neuron stimulation. In other words, selective DA neuron stimulation may be much more addictive than any drug. This may be explained by the non-selective action of pharmacological substances. In the case of cocaine, for example, monoamines other than DA may actually delay the induction of addiction. Indeed, serotonin may oppose DA-dependent adaptive behaviors such as responding for conditioned reward, self-stimulation, and conditioned place preference (Wang et al., 1995, Fletcher and Korth, 1999, Fletcher et al., 2002) by facilitating the association of cues to aversive stimuli (Bauer, 2015, Hindi Attar et al., 2012). Alternatively, the difference may reside in the difference of kinetics between optogenetic self-stimulation and pharmacological induction of extracellular DA increase. Such addictive-potency variation may also exist among different drugs of abuse (George et al., 2014).

While we cannot formally exclude differences in DA release and/or relative signaling to contribute to the establishment of punishment resistance, this scenario is unlikely because the histological validation of the infection of animals included in the study showed eYFP-ChR2 expression in the entire VTA. Moreover, the optogenetic stimulation protocol designed to saturate DA release led to self-stimulation that culminated in unimodally distributed values for the breaking point, reflecting the incentive motivation.

Another surprising result is that the number of electric foot shocks correlated with the excitability of the neurons in the PL. Decreased excitability of pyramidal neurons and increased AMPAR/NMDAR ratio in pyramidal neurons of the same cells has been observed in “addicted rats,” yet these studies did not control for the effect of electric shocks per se (Kasanetz et al., 2010, Kasanetz et al., 2013, Chen et al., 2013). The non-dissociation may therefore be explained by the dual role of the mPFC in both decision processes and fear integration (Peters et al., 2009). For the converse, change in excitability of pyramidal neurons in the infralimbic cortex correlates with foot shocks (Santini et al., 2008). This evidence does not exclude the possibility that mPFC plays a prominent role for the decision of intake pursuit. However, our cFos analysis and observations of intrinsic excitability point to the OFC and cingulate cortex. Furthermore, inhibition of neuronal excitability in the OFC with DREADD prevented resistance to punishment. This causal link represents an important step in understanding the cellular mechanisms responsible for the transition to addiction. Future studies will be needed to test whether this also applies to the whole range of addictive drugs.

Our findings are in line with observations that a dysfunction of the OFC can impair cost-benefit decision making (Seo and Lee, 2010, Walton et al., 2010, Fellows, 2011) and may drive compulsive behaviors (Burguière et al., 2013). In humans, drug abuse has been linked to impaired decision-making and altered OFC function (Lucantonio et al., 2012, Gowin et al., 2013). Taken together, the activity of OFC neurons emerges as a key determinant for the transition to compulsive drug use (Everitt et al., 2007). This does not preclude a role for drug-evoked plasticity at excitatory afferents onto MSNs observed here and in other studies (Kasanetz et al., 2010). It will be interesting to evaluate whether manipulations aiming at controlling the excitability of the OFC affect motivation in addicts.

Here we propose DA neuron self-stimulation as a powerful model to study the stages leading to addiction. We reproduce core components of drug addiction, such as relapse, synaptic plasticity, and perseverance of consumption despite negative consequences. While the model is certainly not suited to study effects specific for a given drug (e.g., compare opioid to psychostimulants), it has several advantages. It allows for a precise temporal control of the reward delivery, it is very specifically activating only the VTA DA neurons, and last but not least, it gives the possibility of studying mice for a much longer time than with drug self-administration. By focusing on the defining commonality of addictive drugs, the hope is to unravel the neural mechanisms underlying also non-substance-dependent forms of addiction (Alavi et al., 2012, Robbins and Clark, 2015) and thus contribute to a general theory of the disease. Optogenetic disease models thus allow a decisive step for a thorough understanding of the neuronal dysfunction involved in late stages of addiction and will guide novel, rational treatments for a disease currently without a cure.

V.P., J.T., and A.H. carried out the behavioral experiments while V.P. did the electrophysiological recordings and coordinated the analysis. The study was designed and written by all authors.

The work was supported by grants from the Swiss National Foundation and the ERC advanced grant (MeSSI), Carigest SA, the Academic Society of Geneva, and the Fondation Privée des Hopitaux Universitaires de Genève. J.T. is a MD-PhD student paid by the Swiss Confederation.

Document S1. Supplemental Experimental Procedures and Figures S1–S6

Table S1. Statistical Analyses