ARTICLE https://doi.org/10.1038/s41467-020-14458-y OPEN A specific prelimbic-nucleus accumbens pathway controls resilience versus vulnerability to food addiction Laura Domingo-Rodriguez 1,12, Inigo Ruiz de Azua2,3,12, Eduardo Dominguez4, Eric Senabre 1, Irene Serra5, Sami Kummer1, Mohit Navandar6, Sarah Baddenhausen 2, Clementine Hofmann2,7, Raul Andero8,9,10, Susanne Gerber6, Marta Navarrete 5, Mara Dierssen 4,11, Beat Lutz2,3,13, Elena Martín-García 1,8,13 & 1234567890():,; Rafael Maldonado 1,11,13* Food addiction is linked to obesity and eating disorders and is characterized by a loss of behavioral control and compulsive food intake. Here, using a food addiction mouse model, we report that the lack of cannabinoid type-1 receptor in dorsal telencephalic glutamatergic neurons prevents the development of food addiction-like behavior, which is associated with enhanced synaptic excitatory transmission in the medial prefrontal cortex (mPFC) and in the nucleus accumbens (NAc). In contrast, chemogenetic inhibition of neuronal activity in the mPFC-NAc pathway induces compulsive food seeking. Transcriptomic analysis and genetic manipulation identified that increased dopamine D2 receptor expression in the mPFC-NAc pathway promotes the addiction-like phenotype. Our study unravels a new neurobiological mechanism underlying resilience and vulnerability to the development of food addiction, which could pave the way towards novel and efficient interventions for this disorder. 1 Laboratory of Neuropharmacology-Neurophar, Department of Experimental and Health Sciences, Universitat Pompeu Fabra (UPF), Barcelona, Spain. 2 Instituteof Physiological Chemistry, University Medical Center of the Johannes Gutenberg University Mainz, Mainz, Germany. 3 Leibniz Institute for Resilience Research, Mainz, Germany. 4 Centre for genomic regulation (CRG), The Barcelona Institute of Science and Technology, Barcelona, Spain. 5 Instituto Cajal, CSIC, Madrid, Spain. 6 Faculty of Biology and Center of Computational Sciences, Johannes Gutenberg University, Mainz, Germany. 7 Focus Program Translational Neuroscience, Johannes Gutenberg University Mainz, Mainz, Germany. 8 Department of Psychobiology and Methodology in Health Sciences, Universitat Autònoma de Barcelona (UAB), Bellaterra, Spain. 9 Instituto de Salud Carlos III, Centro de Investigación Biomédica en Red de Salud Mental, CIBERSAM, Bellaterra, Spain. 10 Unitat de Neurociència Traslacional, ParcTaulí Hospital Universitari, Institut d’Investigació i Innovació ParcTaulí (I3PT), Institut de Neurociències, UAB, Bellaterra, Spain. 11 Hospital del Mar Medical Research Institute (IMIM), Barcelona, Spain. 12These authors contributed equally: Laura Domingo-Rodriguez, Inigo Ruiz de Azua. 13These authors jointly supervised this work: Beat Lutz, Elena Martín-García, Rafael Maldonado. *email: rafael.maldonado@upf.edu NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y F ood addiction and compulsive eating have recently gained food. Importantly, transcriptomic analysis of resilient and sus- attention due to the increasing worldwide prevalence, the ceptible mice revealed an upregulation of the dopamine D2 high socio-economical costs, and the lack of effective and receptor gene (Drd2) in mPFC of food addicted mice. Accord- long-lasting treatments1. This addictive disorder has been closely ingly, virally-mediated overexpression of dopamine D2 receptor linked to obesity and eating disorders1. The concept of food (D2R) specifically in the PL-NAc core projection promoted addiction is still controversial being also under debate if it could compulsive eating behavior. Overall, we reveal that the endo- represent a behavioral addiction or whether specific components cannabinoid and the dopaminergic signaling targeting a specific of food could have intrinsic addictive properties similar to drugs mPFC-NAc pathway control resilience and vulnerability to of abuse2. However, the diagnosis of food addiction is not develop food addiction-like behavior in mice. included in the 5th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), although a validated tool, the Yale Food Addiction Scale (YFAS) is widely accepted among Results the scientific community3. Glu-CB1-KO mice display resilience to food addiction. Given Food addiction shares common neurobiological mechanisms that CB1R plays a crucial role in the modulation of glutamatergic with drug addiction4. Food and drug addiction are complex transmission16, we explored the function of this receptor in food multifactorial chronic brain disorders that result from the inter- addiction-like behavior. Glu-CB1-KO mice16 lacking CB1R in action of multiple genes and environmental factors, leading to dorsal telencephalic glutamatergic neurons and their WT litter- interindividual variability in the development of the addictive mates were exposed to the recently established operant model of process. Nevertheless, the precise neurobiological mechanisms food addiction5. Glu-CB1-KO (n = 58) and WT mice (n = 56) underlying vulnerability or resilience to food addiction have were trained under a fixed ratio (FR) 1 schedule of reinforcement remained elusive. In order to identify mechanisms at circuit and during six sessions followed by 112 sessions under FR5 to obtain molecular level, we employed a recently validated animal model chocolate-flavored pellets as reinforcers (Fig. 1a and Methods). In that mimics behavioral abnormalities associated with food FR1, both genotypes increased the number of reinforcers across addiction in humans, such as loss of control, motivation, com- sessions without significant differences indicating similar levels of pulsive eating and impulsivity5. acquisition of the operant conditioning learning. However, when Both natural rewards and drugs of abuse exert initial reinforcing the effort to obtain one single pellet was increased to FR5, the effects, targeting the brain reward circuit mainly triggered by progressive increase of the number of reinforcers was significantly dopamine release in the nucleus accumbens (NAc). However, reduced in Glu-CB1-KO as compared to WT (repeated measures addiction is a result of the transition from initial hedonic intake to ANOVA, genotype effect, P < 0.001; interaction genotype x ses- loss of control after repeated exposure to the reinforcer, which sions, P < 0.001, Fig. 1b), leading to a reduced number of rein- requires long-lasting adaptations in the reward system and asso- forcers over the entire FR5 period. This result suggests that ciated pathways. Three major interconnected networks are palatable pellets were less reinforcing for the mutants since the involved in the development of drug addiction, the basal ganglia, FR5 early period (U Mann–Whitney, P < 0.01, Supplementary the extended amygdala and the medial prefrontal cortex (mPFC)6. Fig. 1a) and this trait may represent an initial protective factor. The mPFC is involved in the top-down regulation of cognitive Animals that responded <25% of all FR5 sessions and did not flexibility, decision-making and inhibitory control and seems to achieve the acquisition criteria were excluded from the remaining play a crucial role in the transition to and persistence of the experimental sequence (14.7% of Glu-CB1-KO and 6.7% of WT). addictive behavior7–9. Indeed, there are evidences of a hypoactivity In the early period (1–18 sessions), a significant genotype of the mPFC in human addicts10 and in rats with a compulsive difference was only observed in the criterion of compulsivity, cocaine self-administration11. Similarly, positron emission tomo- evaluated by the number of active responses associated with a graphy studies described a negative correlation between metabolic footshock delivery. Indeed, a suppressed response in front of activity in prefrontal regions and body weight12. In the mPFC, the negative consequences was revealed in Glu-CB1-KO mice prelimbic (PL) and the infralimbic (IL) cortex send dense pro- compared to WT mice (U Mann–Whitney, P < 0.05, Supplemen- jections to cortical and subcortical regions, including the NAc, tary Fig. 1b–d). In contrast, non-significant genotype differences which plays a crucial role in behavioral inhibitory control13. The were found in the persistence to response, as evaluated by the mPFC contains excitatory glutamatergic pyramidal neurons and number of non-reinforced active responses during the pellet-free inhibitory GABAergic interneurons, establishing a local network period, and in motivation, defined by the breaking point obtained that is modulated in part by the endocannabinoid and the dopa- during the progressive ratio (PR) schedule (Supplementary minergic signaling systems14,15. Fig. 1b–d). In the late period (95–112 sessions), Glu-CB1-KO The aim of this study was to decipher the neurobiological mice showed significantly less persistence to response, reduced mechanisms underlying the resilience and the vulnerability to motivation, and decreased compulsivity in seeking for highly develop food addiction-like behavior. For this purpose, we used palatable food compared to WT mice, revealing a resilient innovative tools consisting in a mouse operant behavioral model phenotype of these mutants to develop food addiction (U of food addiction with face validity combined with conditional Mann–Whitney, P < 0.01, Fig. 1c–e). Two additional phenotypic glutamatergic CB1R mutant mice (Glu-CB1-KO), electro- traits considered as factors of vulnerability to addiction, physiological ex vivo recordings, genome-wide RNA sequencing, impulsivity and sensitive to aversive associative learning, were chemogenetic interference and adenoviral gene delivery in order also evaluated (Supplementary notes, Supplementary Fig. 1e, f). to characterize the endophenotype of resilience and vulnerability Using the results of three food addiction-like criteria in the late at a genetic, cellular, circuit and behavioral level. We found that period, we individually categorized mice in addicted (covering Glu-CB1-KO mice displayed a resilient phenotype to develop 2–3 criteria) or non-addicted, as previously reported5. A mouse food addiction-like behavior, which was associated with an was considered positive for an addiction-like criterion when its increased excitatory synaptic transmission of glutamatergic neu- score for each behavior was equal or beyond the 75th percentile of rons in PL and in NAc. In reverse, the silencing of PL neurons the distribution of the WT group. Only 6.9% of the Glu-CB1-KO projecting to NAc core using a chemogenetic approach induced a mice achieved 2–3 criteria and were considered addicted phenotype of susceptibility to develop food addiction-like beha- compared with the 25.0% of WT mice (chi-square, P < 0.01, vior, in particular, a compulsive eating behavior for palatable Fig. 1f). Of note, a small percentage of Glu-CB1-KO mice were 2 NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y ARTICLE a 118 sessions b Operant conditioning maintained by chocolate-flavored pellets 40 *** Number of reinforcers in 1 h 1 6 1 18 95 112 FR1 FR5 30 Early period Daily self-administration Late period session (60′) 20 Addiction criteria 25′ 10′ 25′ Addiction criteria 1. Persistence to response 1. Persistence to response 2. Motivation 2. Motivation Pellet Pellet Pellet 10 3. Compulsivity 3. Compulsivity period free period WT period Glu-CB1-KO 0 1 61 5 15 25 35 45 55 65 75 85 95 105 115 FR1 FR5 Sessions Late period c Persistence to response d Motivation e Compulsivity f WT Glu-CB1-KO 60 ** 800 ** 25 ** ** Number of shocks in 50 min active responses in 10 min Addicted 20 Addicted Breaking point in 5 h 600 6.9% 25.0% Non-reinforced 40 15 400 Non-addicted Non-addicted 75.0% 93.1% 10 20 200 5 Non-addicted Addicted 0 0 0 0 crit 1 crit 2 crit 3 crit T O T O T O W W W -K -K -K B1 B1 B1 WT 50.0% 25.0% 19.6% 5.4% -C -C -C lu lu lu G G G Glu-CB1-KO 74.1% 19.0% 5.2% 1.7% g Persistence to response h Motivation i Compulsivity 80 WT Glu-CB1-KO 1000 WT Glu-CB1-KO 25 WT Glu-CB1-KO r = 0.74 r = 0.53 r = 0.73 r = 0.62 r = 0.46 r = 0.73 P < 0.001 P < 0.001 Number of shocks in 50 min P < 0.001 P < 0.001 P < 0.001 P < 0.001 active responses in 10 min 800 20 Breaking point in 5 h 60 Non-reinforced 600 15 40 400 10 20 200 5 0 0 0 0 1 2 3 0 1 2 3 0 1 2 3 Addiction criteria Addiction criteria Addiction criteria Fig. 1 Glu-CB1-KO mice display resilience to food addiction. a Timeline of the experimental sequence of the food addiction mouse model. b Reduced number of reinforcers during 1 h of operant training sessions maintained by chocolate-flavored pellets in Glu-CB1-KO compared to WT mice (mean ± S.E.M, repeated measures ANOVA, genotype effect ***P < 0.001). c–e Glu-CB1-KO mice decreased response in the three addiction-like criteria tests (individual data with median and interquartile range, U Mann–Whitney, **P < 0.01): c Persistence to response. d Motivation. e Compulsivity. The dashed horizontal line indicated the 75th percentile of distribution of WT mice, it is used as the threshold to consider a mouse positive for one criterion. Addicted mice in gray filled circles for WT and blue for Glu-CB1-KO mice. f Reduced percentage of mice categorized as addicted in Glu-CB1-KO compared to WT mice at the late period (chi-square, **P < 0.01). g–i Pearson correlations between individual values of addiction-like criteria and g non-reinforced active responses in 10 min, h breaking point in 5 h and i number of shocks in 50 min (n = 56 for WT mice and n = 58 for Glu-CB1-KO mice; see also Supplementary Figs. 1 and 2; statistical details are included in Supplementary Table 1). able to become addicted, but the loss of this gene strongly body weight changes when were fed ad libitum17. Our results decreased the likelihood to develop food addiction. Positive show decreased palatable pellets intake by Glu-CB1-KO mice correlations were found between the number of criteria in both without food restriction (U Mann–Whitney, P < 0.001, Fig. 1b). mutant and WT mice and the intensity of each criterion To study if the resilient phenotype of Glu-CB1-KO mice was (Fig. 1g–i), and classified addicted mice showed the highest influenced by the body weight variable, we measured the values (Supplementary Fig. 1g–k). Thus, the deletion of the CB1R evolution of the body weight during the whole experimental in dorsal telencephalic glutamatergic neurons could represent a sequence of 24 weeks, the total average of body weight separately protective factor to food addiction-like behavior but, as expected in the early and in late period and the correlations between the in a multifactorial disease, the mutation of one single gene is not body weight and the three addiction-like criteria depending on enough to totally stop the addiction process. the genotype (Supplementary Fig. 2a–e). The lack of significant Previous studies reported that the deletion of CB1R from dorsal differences between genotypes in the body weight in the early telencephalic glutamatergic neurons has a marked impact on food period (Supplementary Fig. 2b), when mutants showed significant intake in fasting conditions17. However, these mice did not show increased compulsivity, suggested that the body weight is not a NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y predisposing factor in the development of the addictive-like In Glu-CB1-KO mice, EPSCs amplitude was still slightly behavior. In the late period, body weight differences between diminished by the WIN55,212-2 application, but in a significantly genotypes emerged, although the correlations between body lower extension than in WT mice (paired t-test, P < 0.01, Fig. 2s, weight and the three addiction-like criteria (Supplementary t). This partial effect of WIN55,212-2 in the mutants could be Fig. 2c–e) were not significant indicating that the body weight explained by the expression of CB1R in glutamatergic projection variable does not explain the lower persistence to response, neurons to the NAc20–22. Overall, we obtained a functional motivation and compulsivity found in the Glu-CB1-KO mice. confirmation of the absence of the CB1R in the Glu-CB1-KO mice For statistical details see Supplementary Table 1. and demonstrated the modulatory effect of CB1R in excitatory glutamatergic transmission in PL and NAc. Synaptic excitatory transmission is increased in Glu-CB1-KO. Considering that the activation of presynaptic CB1R on gluta- Inhibition of PL-NAc core pathway leads to compulsivity. matergic neurons is supposed to suppress vesicular release of Based on the increased excitatory transmission in mPFC of glutamate18, and given the notion that increased neuronal activity resilient Glu-CB1-KO mice, we hypothesized that hypoactivity of of the mPFC may regulate the development of addiction19, we glutamatergic transmission in mPFC would promote addictive- applied ex vivo electrophysiological experiments to uncover the like behavior in WT mice when exposed to the palatable food consequences of CB1R deletion in glutamatergic transmission in addiction model. To this end, we used a chemogenetic approach mPFC. In naive mutant mice, we quantified the frequency of to selectively reduce the activity of all the glutamatergic neurons miniature excitatory postsynaptic currents (mEPSCs) in layer in the PL. Our results revealed that the decreased excitatory 5 (L5) pyramidal neurons of the PL cortex that project to sub- transmission in PL promotes food addiction-like behavior (Sup- cortical regions by performing whole-cell recordings in brain plementary notes, Supplementary Fig. 4a–m and Supplementary slices in the presence of 2 μM tetrodotoxin (TTX). The frequency Fig. 5a–j). Specifically, 42.8% of mice with the inhibition of glu- of mEPSCs is a readout of the number of neurotransmitter vesicle tamatergic PL neurons accomplished the criteria of addiction as released13. We found that mEPSCs frequency was increased in compared to 15.4% of saline-treated mice (chi-square, P < 0.01, Glu-CB1-KO compared to WT mice, suggesting an enhanced Supplementary Fig. 4j), suggesting that a decreased excitability of probability of glutamate vesicle release onto L5 pyramidal PL glutamatergic transmission in PL neurons is involved in the neurons that could be independent of presynaptic voltage Ca2+ development of this addictive behavior towards highly palatable ion channels (t-test, P < 0.01, Fig. 2a–d). Notably, no significant food. However, PL neurons send projections widely to multiple differences were found in the miniature inhibitory postsynaptic brain areas, and thus, we asked which specific projection is currents (mIPSCs) frequency and in amplitude between Glu- involved in the loss of behavioral control. To answer this ques- CB1-KO and WT mice (Fig. 2e–g). In a next step, we evaluated if tion, we adopted a combined chemogenetic and a retrograde the glutamatergic and GABAergic synaptic transmission in NAc AAV approach23 that enables the retrograde tagging of neuronal was also altered by the loss of CB1R in cortical glutamatergic cells. projections. To this end, we injected two AAVs: AAV-hM4Di- Electrophysiological recordings in the NAc showed an increased DREADD (AAV8-hSyn-DIO-hM4D(Gi)-mCherry) into PL and mEPSCs frequency and no differences in the mIPSCs in Glu- AAV-retrograde-Cre (AAVrg-pmSyn-EBFP-Cre) into the NAc CB1-KO mice compared to WT mice (t-test, P < 0.05, Fig. 2h–n). core (Fig. 3a). Thus, hM4Di receptors expression only occurred in Thus, the lack of CB1R in dorsal telencephalic glutamatergic PL neurons that directly projected to NAc core. mCherry and Cre neurons increased synaptic excitatory transmission in PL cortex recombinase were visualized by immunofluorescence to verify the and NAc without affecting inhibitory synaptic transmission in injection site of the AAVs and the retrograde transport of Cre these areas. (Fig. 3b). Whole-cell current-clamp recordings performed in PL Then, we studied the synaptic facilitation in order to determine L5 of visually identified hM4Di-mCherry expressing neurons whether Ca2+-dependent synaptic transmission was affected in Glu- confirmed that clozapine-n-oxide (CNO) activation of hM4Di CB1-KO mice. We applied a paired-pulse facilitation (PPF) protocol receptor inhibited the activity of PL-NAc core pyramidal neurons. with a 50 ms interstimulus interval in mPFC L2/3 glutamatergic Indeed, CNO bath application decreased membrane resistance axons and recorded evoked local field postsynaptic potentials (paired t-test, P < 0.01, Supplementary Fig. 6a, b) and subse- (fPSPs) in L5. We found higher synaptic facilitation (increased PPF quently blocked current-evoked action potential firing frequency ratio) in Glu-CB1-KO than in WT mice, indicating that the lack (paired t-test, P < 0.05, Fig. 3c, d). Furthermore, the current to of CB1R in glutamatergic terminals also produced an augmented evoke one single action potential in the presence of CNO was Ca2+-dependent synaptic transmission, suggesting an augmented higher as compared to the baseline (paired t-test, P < 0.01, Sup- glutamate release (U Mann–Whitney, P < 0.01, Fig. 2o–p). plementary Fig. 6c, d). This effect was hM4Di receptor-specific Additionally, we validated the functional deletion of CB1R because no significant differences in the firing rate, membrane signaling at glutamatergic presynaptic terminals in the mPFC of resistance nor in rheobase were found when CNO was applied in Glu-CB1-KO mice by studying the modulation of the amplitude of mPFC slices of mice not expressing the hM4Di receptors (Fig. 3d single L5 fPSPs evoked by electrical stimulation on L2/3 in the and Supplementary Fig. 6a–d). Importantly, this PL-NAc core presence of the CB1R agonist WIN55,212-2 (5 μM). As expected, projections inhibited the medium spiny neurons (MSNs) of the the fPSPs amplitude evoked was strongly reduced in the WT mice NAc as shown by the mEPSCs recordings, the cumultative after WIN55,212-2 application compared to the baseline before probability of the amplitude of all signals registered and by the treatment, indicating a functional activation of CB1R signaling decreased changes of mEPSCs frequency in the NAc after CNO (paired t-test, P < 0.001, Fig. 2q, r). In contrast, this effect was application in mice expressing hM4Di receptors (Fig. 3e, f). No blunted in conditional knockout mice. As a further demonstration, changes in the amplitude nor in the resting membrane potential we completely blocked WIN55,212-2 inhibitory effect in mPFC were reported (Supplementary Fig. 6e–g). Altogether, these synaptic transmission by applying the selective CB1R antagonist results suggested that the chemogenetic inhibition of the PL-NAc rimonabant (4 μM; Supplementary Fig. 3a, b). Similar results were core neurons leads to a reduced glutamatergic tone in the NAc. obtained evaluating the changes of EPSCs amplitude in the Using this approach, we expected that selective inhibition of NAc. The EPSCs amplitude was strongly reduced in the WT the PL-NAc core projections would decrease the inhibitory mice after WIN55,212-2 application compared to the baseline. control during food operant training, leading to susceptibility to 4 NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y ARTICLE PL mPFC NAc a PL h NAc 30 μm 30 μm b mEPSCs c 12 d 40 i mEPSCs j 12 k 40 WT WT mEPSCs frequency (Hz) mEPSCs frequency (Hz) mEPSCs amplitude (pA) 35 mEPSCs amplitude (pA) 35 10 10 30 30 8 8 25 25 6 20 6 20 Glu-CB1-KO Glu-CB1-KO 15 15 4 4 10 10 2 2 5 5 0 0 0 0 10 pA 10 pA KO KO KO KO T T T T W W W W 1s 1s - - - - B1 B1 B1 B1 -C -C -C -C lu lu lu lu G G G G e mIPSCs f 12 g 40 l mIPSCs m 12 n 40 WT WT 35 35 mIPSCs amplitude (pA) mIPSCs frequency (Hz) mIPSCs frequency (Hz) mIPSCs amplitude (pA) 10 10 30 30 8 8 25 25 6 20 6 20 Glu-CB1-KO Glu-CB1-KO 15 15 4 4 10 10 2 2 5 5 100 pA 0 0 100 pA 0 0 1s O O O O 1s T T T T -K W -K -K -K W W W B1 B1 B1 B1 -C -C -C -C lu lu lu lu G G G G PL mPFC NAc o WT p q WT r s WT t P1 P2 Baseline Changes of EPSCs amplitude (%) 4 WIN55,212-2 200 Baseline 200 Baseline Changes of fPSPs amplitude (%) Baseline WIN55,212-2 WIN55,212-2 WIN55,212-2 3 150 150 PPF ratio (P2/P1) Glu-CB1-KO 2 Glu-CB1-KO 100 Glu-CB1-KO 100 P1 P2 Baseline WIN55,212-2 Baseline WIN55,212-2 50 1 50 0.4 mV 0.2 mV 0 3 ms 0 50 pA 0 5 ms WT Glu-CB1-Ko WT Glu-CB1-Ko O T -K W 50 ms B1 -C lu G Fig. 2 Synaptic excitatory transmission is increased in Glu-CB1-KO. a, h Schematic drawings and infrared differential interference contrast images showing the recorded neuron in the a PL and h NAc. Scale bars, 30 µm. b, e Representative traces of miniature excitatory and inhibitory postsynaptic currents (mEPSCs, mIPSCs) recorded at a holding potential of −70 mV in PL and i, l −75 mV in NAc from WT (black traces) and Glu-CB1-KO (blue traces) mice. c, d mEPSCs frequency and amplitude and f, g mIPSCs frequency and amplitude of WT and Glu-CB1-KO mice in the PL and j, k, m, n in the NAc (mean ± S.E.M; t-test, *P < 0.05, **P < 0.01, n = 9–12 cells from n = 12 mice). o–p Paired pulse facilitation (PPF). o Representative recordings of L5 field postsynaptic potentials (fPSPs) before (P1) and after (P2) stimulating twice in layer 2/3 with an interpulse interval of 50 ms for WT (above) and Glu-CB1-KO (below). p Increased paired-pulse facilitation ratio (P2/P1) in Glu-CB1-KO compared to WT (mean ± S.E.M, U Mann–Whitney, **P < 0.01; n = 12–14 slices from n = 5 animals per genotype). q Representative recordings of the modulation of fPSPs amplitude before and after application of the CB1R agonist WIN55,212-2 (5 µM) compared to baseline in PL for WT (above) and Glu-CB1-KO (below). r Changes of fPSPs amplitude in percentage in PL for WT and Glu-CB1-KO (mean and individual values, paired t-test, ***P < 0.001; n = 6 slices from n = 3 animals per genotype). s Representative synaptic responses showing mean EPSCs (five consecutive EPSCs) before and after CB1R agonist WIN55,212-2 (5 µM) bath application in the NAc for WT (above) and Glu-CB1-KO (below). t Average relative changes of EPSCs amplitude in NAc of WT and Glu-CB1-KO mice (mean and individual values, paired t-test, **P < 0.01, ***P < 0.001; n = 7–10 cells from n = 4 animals per genotype; see also Supplementary Fig. 3; statistical details are included in Supplementary Table 2). develop addiction-like behavior. Therefore, we used the early food reinforcement before AAVs injection and under FR5 (four addiction protocol (Fig. 3g) to train WT mice (n = 32) to self- sessions) after injection to recover the basal levels of responding. administer chocolate-flavored pellets in the operant chambers Then, an osmotic minipump filled with CNO (n = 22) or saline under FR1 (two sessions) and FR5 (three sessions) schedule of (n = 12) was subcutaneously implanted in the back of each NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y PL mPFC a c AAV-control AAV-hM4Di d 40 Baseline Baseline Baseline CNO DREADD (OFF) AAV8-hSyn-DIO- AAVr pmSyn1- hM4D(Gi)-mCherry EBFP-Cre HIP lox2722 loxP 30 AAV Cre Firing rate (Hz) DREADD (ON) * PL CNO CNO 20 VTA 50 mV 50 mV NAc core AMG 10 0 Control hM4Di NAc AAV8-hSyn-DIO- AAVr pmSyn1- b hM4D(Gi)-mCherry EBFP-Cre e AAV-control AAV-hM4Di f 200 Baseline Changes of mEPSCs frequency (%) Baseline CNO ACC PL CNO 150 IL 20 pA 1s 20 pA 1s * 1.0 1.0 100 Cumulative NAc Cumulative probability probability B: 1.98 B: 1.34 0.5 0.5 Basal 50 Basal CNO CNO 0.0 0.0 0 10 20 30 0 10 20 30 0 Control hM4Di Amplitude (pA) Amplitude (pA) g Expressionof hM4Di receptors activated by CNO Weeks 1 2 3 4 5 6 7 8 9 Operant Operant Operant training (FR5) training Surgeries training Osmotic (FR1-FR5) (FR5) minipump Addiction criteria implantation 1. Persistence to response CNO 2. Motivation 3. Compulsivity h Operant conditioning mantained by chocolate-flavored pellets i Persistence to response j Motivation k Compulsivity Osmotic 80 800 14 150 AAV minipump Expression of hM4Di receptors ** Number of reinforcers in 2 h Number of shocks in 50 min active responses in 10 min surgery implantation activated by CNO 12 125 Breaking point in 5 h 60 600 10 Non-reinforced 100 Saline CNO 8 75 40 400 6 50 4 20 200 25 2 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 0 0 0 Saline CNO Saline CNO Saline CNO FR1 FR5 Sessions l Saline CNO m Persistence to response n Motivation o Compulsivity ** * 100 Saline CNO 1000 Saline CNO 15 Saline CNO Addicted r = 0.50 r = 0.57 r = 0.46 r = 0.53 r = 0.60 r = 0.42 Number of shocks in 50 min active responses in 10 min 16.7% n.s. P < 0.01 n.s. P < 0.05 P < 0.05 P = 0.05 80 800 Breaking point in 5 h Non- Addicted addicted Non-reinforced 50.0% 10 50.0% Non-addicted 60 600 83.3% 40 400 5 Non-addicted Addicted 20 200 0 crit 1 crit 2 crit 3 crit 0 0 0 Saline 33.3% 50.0% 16.7% 0.0% 0 1 2 3 0 1 2 3 0 1 2 3 Addiction criteria Addiction criteria Addiction criteria CNO 13.6% 36.4% 45.4% 4.6% mouse. During the chronic CNO exposure (four weeks, 0.25 μl/h) during the operant conditioning maintained by chocolate- with the subsequent inhibition of the glutamatergic PL-NAc core flavored pellets, indicating that the inhibition of PL-NAc core neurons, mice underwent FR5 sessions for four weeks, and the projection did not affect the reinforcing effects of these pellets three food addiction-like criteria were evaluated during the last (Fig. 3h). In contrast, this specific manipulation of the PL-NAc week. No differences were found in the number of reinforcers in projection produced a robust increase in the food addiction the daily training sessions between CNO and saline-treated mice compulsivity criterion (t-test, P < 0.01, Fig. 3i–k). Thus, mice with 6 NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y ARTICLE Fig. 3 Inhibition of PL-NAc core pathway leads to compulsivity. a Scheme of combinatorial viral strategy for selective hM4Di-mCherry expression in PL- NAc core neurons. b Representative immunofluorescence images of Cre-dependent hM4Di-mCherry detected at PL injection site (left) and Cre recombinase at NAc core (right). c–f Chemogenetic inhibition of glutamatergic PL-NAc core neurons induces changes in excitatory transmission in the PL and NAc core. c Representative recordings showing evoked (200 pA) action potential in WT mice injected with AAV-control (left) and AAV-hM4Di (right) in PL layer 5 mCherry visualized neurons at baseline and after CNO (10 µM) application. d Quantification of the firing rate (Hz) (mean and individual values; paired t-test, *P < 0.05; n = 14 cells from n = 4 mice injected with AAV-control and n = 10 cells from n = 4 animals injected with AAV-hM4Di). e Representative traces of miniature excitatory postsynaptic currents (mEPSCs) recorded at a holding potential of −75 mV in NAc from mice injected with AAV-control (left) or AAV-hM4Di (right) in baseline and after CNO (10 µM) bath application (above). Cumulative probability plot of the mEPSCs amplitude from both control (left) and hM4Di (right) expressing mice in baseline conditions and after CNO (10 µM) bath application (below). f Changes of mEPSCs frequency in percentage (mean and individual values; paired t-test, *P < 0.05; n = 8 cells from n = 6 mice injected with AAV-control and n = 8 cells from n = 7 mice injected with AAV-hM4Di). g Timeline of the experimental sequence of the early period of food addiction mouse model. h Number of reinforcers during operant training sessions maintained by chocolate-flavored pellets (mean ± S.E.M). i–k Behavioral tests of the three addiction-like criteria showing increased compulsivity in CNO-treated mice (individual values with the median and the interquartile range, U Mann–Whitney, **P < 0.01). The 75th percentile of distribution of mice treated with saline is indicated by the dashed horizontal line. Addicted mice in gray filled circles for saline- treated mice and red for CNO-treated mice. l Increased percentage of CNO-treated mice classified as food addicted animals (chi-square, ***P < 0.001). m–o Pearson correlations between individual addiction-like criteria and m non-reinforced active responses in 10 min, n breaking point in 5 h, o number of shocks in 50 min (n = 12 for saline-treated mice and n = 22 for CNO-treated mice; PL prelimbic, NAc nucleus accumbens, Amg amygdala, Hip hippocampus, VTA ventral tegmental area, ACC anterior cingulate cortex, IL infralimbic; see also Supplementary Fig. 6; statistical details are included in Supplementary Table 4). chronic inhibition of PL-NAc core projection could not stop Two different clusters were observed for the addicted and non- responding for chocolate-flavored pellets in the shock test, addicted mice (Supplementary Fig. 7b). Upon performing dif- receiving higher number of shocks compared to saline group. ferential gene expression analysis between non-addicted and Persistence to response and motivation for chocolate-flavored addicted mice, 31 genes were significantly upregulated, whereas pellets were not significantly different in these animals (Fig. 3i–k). 70 genes were downregulated (Fig. 4e and Supplementary The increased compulsivity could not be explained by an Tables 7 and 8). Interestingly, genes previously related to the unspecific effect of CNO, since no significant differences of reward system such as Drd2 (dopamine receptor type 2), CNO on the three addiction-like criteria were detected in Adora2A (adenosine receptor 2a), Gpr88 (orphan G-protein chronically treated mice not expressing the inhibitory DREADD coupled receptor 88), and Drd1 (dopamine receptor type 1) (Supplementary Fig. 5a–d). In addition, no side effects of CNO mRNA were found to be upregulated in the addicted mice. were observed on body weight, food intake or locomotor activity Several genes such as Myh11 (myosin heavy chain 11), Acta2 in these mice (Supplementary Fig. 6h–j). We observed that our (actin alpha 2), Cdh1 (cadherin 1), Ptgds (prostaglandin manipulation produced a vulnerable phenotype to develop food D2 synthase), and Fosb (fosb proto-oncogene, AP-1 transcrip- addiction, as shown by the fact that the majority of CNO-treated tion factor subunit) were downregulated, suggesting changes animals (77.3%) were above the 75th percentile of the in neuronal plasticity, prostaglandin synthesis and gene reg- compulsivity criterion. Using the categorization based on the ulation24–27. The four differentially upregulated genes were three criteria, a highly significant percentage of the CNO-treated selected for technical validation by quantitative PCR (qPCR). mice (50.0%) was considered addicted as compared to the control The results confirmed the upregulation of Drd2, Adora2A, animals (16.7%, chi-square, P < 0.001, Fig. 3l). Finally, a Gpr88, and Drd1 mRNAs in addicted mice (P < 0.05, Fig. 4f). significant positive correlation between the number of criteria As expected, no changes were found in housekeeping genes reached and the values of each addiction-like criterion was found between non-addicted and addicted mice samples (Supple- in CNO group in the three criteria and in saline group only in the mentary Fig. 7c–e), validating RNA-seq data analysis. criterion of compulsivity (Fig. 3m–o), and addicted mice showed Furthermore, RNA-seq data analysis also showed differen- higher values in both saline and CNO-treated mice (Supplemen- tially expressed genes between WT and Glu-CB1-KO mice: 18 tary Fig. 6k–m). Thus, we demonstrated that the specific genes were significantly upregulated, whereas 26 genes were inhibition of the PL neurons projecting to the NAc core is a downregulated (Fig. 4g and Supplementary Tables 9 and 10). crucial network that confers vulnerability to develop food Technical validation by qPCR confirmed that Cnr1 (cannabi- addiction. noid type 1 receptor) was downregulated as expected, and Fos (c-fos) was also downregulated in mutants (t-test, P < 0.05, U Drd2 gene expression is upregulated in mPFC of addicted Mann–Whitney, P < 0.001, Fig. 4h). Dnah6, Spaca1, and some mice. To characterize gene expression signatures for food small nucleolar RNAs (Snora68, 78, 70, 8, 31) were upregulated. addiction, we performed whole transcriptome analysis of mPFC Interestingly, c-fos has been extensively used as molecular (Supplementary Fig. 7a) of addicted and non-addicted WT and marker of neuronal activity, implicating an enhanced neuronal Glu-CB1-KO mice, classified on the bases of the performance at activity in mPFC of WT compared to Glu-CB1-KO mice. the late period (n = 4–6). First, we selected both WT and Glu- A non-significant difference was found for Npas4 mRNA (Fig. CB1-KO mice displaying similar extreme values in the three 4h), encoding the transcription factor neuronal PAS containing addiction-like criteria (Fig. 4a–c). In these mice, non-significant protein 4. This result confirmed the observation that Npas4 was differences were observed in pellets intake between addicted and not differentially expressed according to established thresholds non-addicted mice in the last FR5 session, immediately before of RNA-seq analysis (Fig. 4g). tissue collection (Fig. 4d). Therefore, we assume that changes in In summary, transcriptomic data analysis shed new light into gene expression would be related to the addiction phenotype the gene expression signature in mPFC related to food addiction, and not to the amount of pellets intake during the last training suggesting molecular mechanisms associated with the loss of session. To determine overall transcriptional changes in the control over palatable food intake. As we found that Drd2 gene is addicted vs. non-addicted mice, we applied principle component the most significantly upregulated gene in addicted mice, we analysis (PCA), revealing the variation between the samples. hypothesized that this upregulation could play a key role in the NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y a b c d Pellets intake Persistence to response Motivation Compulsivity in the last FR5 session 60 800 && 25 && # + % 80 ++ %% ++ %% $$ @ Number of shocks in 50 min active responses in 10 min Number of reinforcers in 1 h 20 %% Breaking point in 5 h $ 600 60 Non-reinforced 40 15 400 $$ 40 10 20 200 20 5 0 0 0 0 NA A NA A NA A NA A NA A NA A NA A NA A WT Glu-CB1-KO WT Glu-CB1-KO WT Glu-CB1-KO WT Glu-CB1-KO e Non addicted vs. addicted f 1500 Up = 31 Drd2 Non-addicted Ptgds Down = 70 Adora2a Addicted WT 30 Glu-CB1-KO mRNA gene expression Gpr88 (relative expression) 1000 −log10(pval) Drd1 20 Myh11 Acta2 Fosb 10 Cdh1 500 0 −4 −2 0 2 4 0 log2FoldChange Drd2 Adora2A Gpr88 Drd1 g WT vs. Glu-CB1-KO h ∞ Up = 18 500 Spaca1 WT Down = 26 40 Dnah6 Glu-CB1-KO 400 Non-addicted mRNA gene expression (relative expression) Addicted 30 −log10(pval) 300 20 Cnr1 Fos 200 *** * Npas4 Snora68 10 Snora78 Snora70 Snord8 100 Snora31 0 0 −6 −4 −2 0 2 4 6 Crn1 Fos Npas4 log2FoldChange Fig. 4 Drd2 gene expression is upregulated in mPFC of addicted mice. a–c Behavioral tests of the three addiction-like criteria during the late period for those mice selected for RNA-seq in each of the four groups, addicted (A) and non-addicted (NA) mice in both genotypes (individual values and bars with median and the interquartile range; n = 5 WT NA, n = 6 WT A, n = 5 Glu-CB1-KO NA, n = 4 Glu-CB1-KO A). a Persistence to response. b Motivation. c Compulsivity. d Pellets intake in the last FR5 session before sample collection. e, g Volcano plot of the RNA-seq data analysis. The cutoff of 1.5-fold change and differentially expressed genes are highlighted in orange. e Differentially expressed genes in addicted mice compared to non-addicted mice. g Differentially expressed genes in Glu-CB1-KO compared to WT mice. f, h Quantitative real-time PCR of selected genes. (+ P < 0.05, + + P < 0.01 WT NA vs. WT A, #P < 0.05 WT NA vs. Glu-CB1-KO NA, &&P < 0.01 WT NA vs. Glu-CB1-KO A, %P < 0.05, %%P < 0.01 WT A vs. Glu-CB1-KO NA, @P < 0.05, WT A vs. Glu-CB1-KO A, $P < 0.05, $$P < 0.05 Glu-CB1-KO NA vs. Glu-CB1-KO A; ★P < 0.05 non-addicted vs. addicted; *P < 0.05, ***P < 0.001 WT vs. Glu-CB1-KO; see also Supplementary Fig. 7; statistical details are included in Supplementary Table 5). development of food addiction-like behavior, irrespective of the selective overexpression of Drd2 in the PL-NAc core projections presence or absence of CB1R. induces the loss of inhibitory control for palatable food self- administration. Using these experimental conditions, we aimed at Drd2 overexpression in PL-NAc core pathway promotes com- mimicking the upregulation of the Drd2 gene observed in pulsivity. Based on the above findings, we tested whether the addicted mPFC after long-term exposure to highly palatable food 8 NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y ARTICLE operant training. First, we confirmed under basal conditions low 30.8% compared to 8.3% in control mice (chi-square, P < 0.01, endogenous Drd2 mRNA expression in PL as compared to NAc Fig. 5l). A positive correlation between the number of criteria and caudate putamen by in situ hybridization (ISH) (Supple- reached and the values obtained in each criterion was found in mentary Fig. 8a). all the groups and addicted animals showed extreme values For specific overexpression in PL-NAc core projections, we (Fig. 5m–o, Supplementary Fig. 10a–c). No differences were used a dual viral vector approach with an Cre-dependent AAV- found in additional variables such as body weight, food intake D2R (AAV-hSyn-DIO-D2L-mVenus, n = 13) and AAV-control and locomotor activity after D2R overexpression (Supplemen- (AAV-Syn1-Stop-GFP, n = 12) injected into PL, and an AAV- tary Fig. 10d–f). In summary, we revealed that overexpression of retrograde-Cre (AAV-pmSyn-EBFP-Cre) injected into the NAc Drd2 allowed that dopamine via D2Rs decreased the excitability core (Fig. 5a). We first verified the AAV injection site by of PL-NAc core projections, conveying the vulnerability to immunofluorescence against mVenus and Cre recombinase develop food addiction. (Fig. 5b). We also confirmed by ISH that retrogradely expressed Cre recombinase mRNA is present in PL and, importantly, approximately 50% of endogenous Drd2 mRNA-positive cells Discussion revealed co-expression with Cre mRNA in PL (Supplementary The easy access to hypercaloric and palatable foods in the western Fig. 8b, c). Quantitative real-time PCR showed 40-fold increased lifestyle is a major contributing factor for developing food levels of Drd2 gene expression in mice overexpressing D2R as addiction and obesity. However, it is still unclear why some compared to control mice in the mPFC (Supplementary Fig. 8d). individuals get loss of control over food intake and develop food Additionally, immunohistochemical experiments revealed over- addiction, whereas others are resilient. In this study, we describe a expressed D2Rs in the neuropil of the PL cortical neurons novel neurobiological mechanism underlying the resilience and (Supplementary Fig. 8e–r). The functional consequence of Drd2 vulnerability to food addiction-like behavior targeting excitatory overexpression was first investigated by electrophysiology. We glutamatergic transmission in PL-NAc core projection, which performed in vitro whole-cell recordings in brain slices using the appears to be modulated by the endocannabinoid and the D2R selective agonist quinpirole to confirm that the over- dopaminergic signaling systems. expression of D2R decreased the excitability of PL-NAc core We first studied the phenotype of food addiction in condi- projection neurons. Quinpirole (2 μM) application significantly tional Glu-CB1-KO mice and their control littermates using a increased rheobase and reduced membrane resistance and firing behavioral animal model with high-translational face validity to rate in response to a 150 pA current square pulse (paired t-test, P human addiction5. Here, we mimicked the transition to addic- < 0.05, Fig. 5c, d and Wilcoxon test, P < 0.01, Supplementary tion after repeated seeking of palatable food in long operant Fig. 9a–d). No differences were observed in control PL L5 training. Our findings revealed that the lack of CB1R in dorsal pyramidal neurons, suggesting that the D2R overexpression was telencephalic glutamatergic neurons induced a strong resilience responsible of this inhibitory effect despite the fact that D2R is to food addiction as revealed by the significantly reduced per- also endogenously expressed in mPFC of control mice (Fig. 5c, d centage (6.9%) of addicted mice in the mutant group. Glu-CB1- and Supplementary Fig. 9a–d). The lack of effect after quinpirole KO mice were characterized by less perseverance, reduced application in control animals could be explained by the low motivation and decreased compulsivity for highly palatable levels of endogenous D2R in mPFC according to RNA-seq data. food. Moreover, a phenotype of improved learning in a cocaine In agreement, our ISH experiments showed endogenous Drd2 self-administration paradigm28, reduced exploratory behavior29, mRNA expression in this type of cells (L5), but at very low levels increased neophobia, high passive fear response after con- close to the limit of detection. Furthermore, we investigated ditioning, and decreased food intake after fasting was described whether this reduced excitability in the PL-NAc core neurons in these mutants17,30. A previous study reported an effect modulates the synaptic glutamatergic transmission in the NAc. on food intake when CB1R was deleted from dorsal telencephalic Whole-cell recordings in the NAc confirmed a reduction in the glutamatergic neurons but exclusively after a 24 h fasting changes of mEPSCs frequency accompanied by a sustained period17. Our results suggest a decreased reinforcing effects of difference in the cumulative probability of the amplitude of all the chocolate-flavored pellets in the mutants, which could con- signals registered (Fig. 5e, f). No changes in the amplitude nor in tribute to the protective phenotype to develop the multifactorial the resting membrane potential were reported (Supplementary food addiction disorder. Notably, Glu-CB1-KO mice did not Fig. 9e–f). Additionally, dopamine application (10 μM) in the PL show changes in body weight in the early period similarly to L5 neurons showed a reduction in membrane resistance, firing what was previously reported when mice were fed ad libitum rate and increased rheobase (Wilcoxon test, P < 0.05, P < 0.01, with regular chow in a 12 weeks period of time17. In the late paired t-test, P < 0.001, Supplementary Fig. 9g–l). Therefore, these period, mutants showed lower body weight than WT mice but electrophysiological results reveal that the overexpression of D2R this variable did not affect the resilient phenotype since no at glutamatergic PL neurons projecting to the NAc has an effect correlations between the three addiction-like criteria and body in both areas, the PL and NAc. weight were revealed. For the behavioral analysis, we used a food addiction In our study, Glu-CB1-KO mice showed an enhanced inhibi- procedure duringthe early period similar to experiments shown tory control to palatable food operant seeking that strongly above (Fig. 5g). We found that overexpression of D2R in PL- reduced the transition from controlled to compulsive seeking for NAc core projection neurons produced compulsive behavior highly palatable food. This resilient phenotype was mediated by towards highly palatable food, despite harmful consequences in increased synaptic glutamatergic transmission in mPFC and in the shock test. This manipulation only affected the compulsivity NAc. Specifically, our electrophysiological recordings of mEPSCs addiction criterion, since no differences were found in the frequency in Glu-CB1-KO mice revealed an enhanced excitatory persistence to response, motivation nor reinforcement (t-test, synaptic transmission onto pyramidal glutamatergic neurons in P < 0.05, Fig. 5h–k). A subset of 57.1% of manipulated mice were L5 of the PL cortex and in their targeted MSNs in the NAc. above the 75th percentile threshold of the control group in the This effect was specific of glutamatergic transmission, since the compulsivity criterion, and this result was not found in the other synaptic GABAergic transmission in these mutants was not addiction-like criteria. Finally, the percentage of D2R over- altered in both brain regions. Additionally, the PPF results expressing mice that achieved 2–3 addiction-like criteria was revealed that the deletion of presynaptic excitatory CB1R NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y PL mPFC a c AAV-control AAV-D2R d 25 Baseline Quinpirole Baseline Baseline AAV2/1-hSyn- AAVr pmSyn1- Drd2 (OFF) 20 * DIO-Drd2(L)-IRES- EBFP-Cre loc2722 HIP Firing rate (Hz) mVenus loxP AAV Cre Drd2 (ON) 15 PL Quinpirole Quinpirole 10 VTA 50 mV 50 mV NAc core AMG 5 0 Control D2R NAc b AAV2/1-hSyn-DIO-D2R(L)- e AAV-control AAV-D2R f IRES-mVenus AAVr-pmSyn1-EBFP-Cre 200 Changes of mEPSCs frequency (%) Baseline Baseline Quinpirole Quinpirole ACC 150 20 pA 20 pA 1s 1s PL 1.0 1.0 ** 100 Cumulative Cumulative probability IL probability NAc 0.5 0.5 Basal Basal Quinpirole 50 Quinpirole B: 1.98 B: 1.34 0.0 0.0 0 10 20 0 10 20 0 Amplitude (pA) Amplitude (pA) Control D2R g Expression of D2R Weeks 1 2 3 4 5 6 7 8 9 Operant Operant Operant training (FR5) training Surgeries training (FR1-FR5) (FR5) Addiction criteria 1. Persistence to response 2. Motivation 3. Compulsivity h Operant conditioning mantained by chocolate-flavored pellets i Persistence to response j Motivation k Compulsivity 125 80 600 100 AAV Behavioral tests after * Number of shocks in 50 min active responses in 10 min surgery 4 weeks post-surgery Number of reinforcers in 2 h 100 80 Breaking point in 5 h 60 Non-reinforced 400 75 60 40 50 40 200 25 Control 20 20 D2 R 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 0 0 0 FR1 FR5 Control D2R Control D2R Control D2R Sessions l Control D2R m Persistence to response n Motivation o Compulsivity Addicted 80 Control D 2R 1000 Control D2R 100 Control D2R 8.3% ** r = 0.34 r = –0.03 r = 0.66 r = 0.40 r = 0.06 r = 0.40 Number of shocks in 50 min active responses in 10 min n.s. n.s. P < 0.05 n.s. n.s. n.s. 800 80 Breaking point in 5 h Addicted 60 Non-reinforced 30.8% Non- addicted 600 60 Non-addicted 69.2% 40 91.7% 400 40 20 Non-addicted Addicted 200 20 0 crit 1 crit 2 crit 3 crit 0 0 0 Saline 33.4% 58.3% 8.3% 0.0% 0 1 2 3 0 1 2 3 0 1 2 3 Addiction criteria Addiction criteria Addiction criteria D2 R 15.4% 53.8% 30.8% 0.0% produced an enhancement in glutamate vesicle release in the PL or genetic deletion showed decreased addiction-like behavior5, cortex that is action potential dependent. Altogether, these results but the precise pathways and cell types involved have not been highlight the involvement of the glutamatergic PL-NAc projec- characterized. In our study, we revealed that the main cells tions in the behavioral phenotype of inhibitory control observed involved are the dorsal telencephalic glutamatergic neurons that in Glu-CB1-KO mice. According to this protective phenotype, project to the NAc. Although CB1R on glutamatergic neurons animal studies targeting the CB1R by pharmacological blockade are less abundant than in GABAergic neurons31, the signal 10 NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y ARTICLE Fig. 5 Drd2 overexpression in PL-NAc core pathway promotes compulsivity. a Scheme of combinatorial viral strategy for selective D2R-mVenus expression in PL-NAc core projecting neurons. b Representative immunofluorescence images showing Cre-induced D2R-mVenus protein at PL injection site (left) and Cre recombinase protein at NAc core injection site (right). c–f Overexpression of D2R in PL-NAc core neurons induces changes in excitatory transmission in the PL and NAc core. c Representative recordings showing evoked (150 pA) action potential in WT mice injected with AAV-control (left) and AAV-D2R (right) in PL layer 5 mVenus visualized neurons at baseline and after quinpirole (2 µM) application. d Quantification of the firing rate (Hz) (mean and individual values; Wilcoxon test, *P < 0.05, n = 9 cells from n = 3 animals injected with AAV-control and n = 9 cells from n = 3 animals overexpressing D2R). e Representative traces of miniature excitatory postsynaptic currents (mEPSCs) recorded at a holding potential of −75 mV in NAc from mice injected with AAV-control (left) or AAV-D2R (right) in baseline and after quinpirole application (above). Cumulative probability plot of the mEPSC amplitudes from both control (left) and D2R (right) expressing mice in baseline conditions and after quinpirole bath application (below). f Changes of mEPSCs frequency in percentage (mean and individual values; Wilcoxon test, **P < 0.01, n = 7 cells from n = 5 animals injected with AAV-control and n = 10 cells from n = 7 animals overexpressing D2R). g Timeline of the experimental sequence of the early period of food addiction mouse model. h Number of reinforcers during operant training sessions maintained by chocolate-flavored pellets. i–k Behavioral tests of the three addiction-like criteria in the early period showed increased compulsivity in mice overexpressing D2R (individual values with the median and the interquartile range, t-test, *P < 0.05). The 75th percentile of distribution of control mice is indicated by the dashed horizontal line. Addicted mice in gray filled circles for control and green for D2R mice. l Increased percentage of mice overexpressing D2R classified as food addicted animals (chi-square, **P < 0.01). m–o Pearson correlations between individual addiction-like criteria and m non-reinforced active responses in 10 min, n breaking point in 5 h, o number of shocks in 50 min (n = 12 for control mice and n = 13 for D2R mice; PL prelimbic, NAc nucleus accumbens, Amg amygdala, Hip hippocampus, VTA ventral tegmental area, ACC anterior cingulate cortex, IL infralimbic; see also Supplementary Figs. 8–10; statistical details are included in Supplementary Table 6). transduction processes appear to be more efficient in glutama- corresponds to the dorsolateral PFC in humans37, an area tergic than in GABAergic neurons32. classically involved in response inhibition34. However, it remains Furthermore, an additional experiment of electrophysiology still unknown, which is the specific downstream target of PL ex vivo in the mPFC and NAc with the application of the CB1R projections involved in food addiction. Pyramidal glutamatergic agonist WIN55,212-2 in mutants and in WT mice has been neurons of the PL cortex project to different brain areas, such as performed. With this experiment, we validated at functional level hippocampus, ventral tegmental area (VTA), amygdala and NAc, the deletion of CB1R at glutamatergic presynaptic terminals both among others, conferring to the mPFC a complex connectivity in PL and in NAc of Glu-CB1-KO. As it was expected, the role as a central hub of communications34. Considering that the amplitude of fPSPs in the mPFC and the EPSCs in the NAc were PL area preferentially projects to the core part of the NAc13, and strongly reduced in WT mice after WIN55,212-2 application the functional role of this pathway38, we specifically targeted the compared to the baseline, which indicates a functional activation PL-NAc core pathway using a dual viral vector approach. We of presynaptic CB1R. In the mutants, the fPSPs amplitude was not injected an AAV expressing a Cre-dependent inhibitory diminished in the mPFC confirming the absence of the CB1R at DREADD in the PL and a retrograde AAV-variant expressing presynaptic glutamatergic terminals. On the other, in the NAc, Cre recombinase into the NAc core. Using this dual viral vector the EPSCs amplitude was slightly decreased after WIN55,212-2 approach in combination with a chronic delivery of CNO ligand, application, but in a significantly lower extend than in WT mice we found that silencing the PL-NAc core projections enhanced possibly due to the presence of CB1R in other cell-types or other specifically the compulsive eating behavior, triggering that ani- glutamatergic projecting neurons to the NAc20–22. PL pyramidal mals could not stop palatable food self-administration despite neurons projecting to the NAc showed increased excitatory negative consequences. The other addiction-like criteria, persis- transmission due to the lack of the CB1R and we therefore pos- tence to response and motivation, were not modified by tulate that they stimulate the GABAergic D2-MSNs indirect chemogenetic interference. These criteria represent different pathway, thereby facilitating the avoidance behavior (NO GO endophenotypes from compulsivity and consequently, the neu- response) (Supplementary Fig. 11a). This top-down mechanism ronal pathways recruited could be different. Persistence to involving PL-NAc projections seems crucial in the resilient response reflects the difficulty of mice to stop food seeking. This phenotype of the mutants and was further investigated in our behavior is related to a persistent desire or unsuccessful efforts to study. Importantly, addiction-like behavior is associated with an cut down the response due to a habit formation or disruption of alteration in the synaptic excitatory transmission of prefrontal extinction learning and has been reported to involve dorsal cortical areas that normally provide inhibitory control striatal- striatal and hippocampal pathways39. In turn, motivation is mediated behavior33. apparently directly related to reward processing and has been Previous data pointed out that the PL subregion of the mPFC is associated with VTA-NAc pathways4. The successful silencing of particularly involved in drug addiction34. Indeed, inhibition of PL-NAc core cortical pyramidal neurons was demonstrated by the PL cortex by in vivo optogenetics increased compulsive patch-clamp experiments, showing a decreased firing rate and cocaine intake in rats11, and the repetitive transcranial magnetic membrane resistance in neurons expressing hM4Di receptors stimulation applied to the human dorsolateral PFC, equivalent to exclusively in the presence of CNO, confirming that the driving the rodents PL cortex, reduced cocaine use and craving35. compulsive food seeking was underlined by a reduced PL-NAc Therefore, we decreased the glutamatergic transmission of the core activity. Future studies should be performed to elucidate the PL region by using a Cre-dependent inhibitory chemogenetic role of other pathways from the PL to other downstream areas, approach in Nex-Cre mice. CNO-induced silencing of glutama- such as the amygdala, in the development of food addiction-like tergic neurons in hM4Di-injected mice increased the percentage behavior. of addicted animals. These results revealed a crucial role of this A small percentage of Glu-CB1-KO mice became addicted after cortical region in the development of food addiction-like beha- the prolonged highly palatable food exposure. Therefore, the lack vior. We cannot discard the possible involvement of the anterior of a single gene was not enough to totally block the transition to cingulate cortex in this behavior, since we also observe residual addiction, as expected for a multifactorial disease. This evidence viral expression in this brain region. This area is anatomically highlights the complex and polyfactorial nature of food addiction grouped with the PL region to the dorsal mPFC36, that and prompted us to study the transcriptomic changes underlying NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y the resilient and vulnerable phenotype to develop food addiction- indirect pathway47. Additionally, the inhibition of the D2- like behavior. The comparison of the transcriptomic profiles in MSNs in the NAc core enhanced motivation for cocaine in a the mPFC between addicted and non-addicted mice revealed a self-administration paradigm, but not for standard food48. remarkable upregulation of the Drd2, Adora2A, Gpr88, and Drd1 Therefore, our results provide a new mechanism of the loss of genes in addicted mice independently of the genotype. All these inhibitory control for food seeking behavior involving D2R in PL genes encode G-protein-coupled receptors and are known to be cortical projections to NAc core. In particular, overexpression of involved in the neurobiological pathways recruited in addiction40. D2R diminished the excitability of the pyramidal neurons in the D2R and A2AR, encoded by Drd2 and Adora2A, respectively, are PL projecting to NAc core. According to our model (Supple- colocalized in the mPFC glutamatergic neurons projecting from mentary Fig. 11b), we predict that reduced glutamatergic trans- the L2/3 to L5 at presynaptic terminals41. Here, receptor activa- mission in the NAc core will decrease the activation of D2-MSN tion synergistically inhibit synaptic glutamatergic transmission41. indirect pathway, thereby suppressing the avoidance behavior This is contrary to the well stablished postsynaptic antagonistic (NO GO response) and promoting the loss of control towards interaction in the GABAergic MSNs of the striatum42. Regarding palatable food consumption. Gpr88, inactivation of this gene enhanced the excitability of both In summary, we elucidated the crucial role of the glutamatergic D1 and D2 MSNs in the striatum43. Therefore, we could speculate PL-NAc core pathway modulated by CB1R and D2R as a critical that upregulation of Gpr88 in the mPFC of addicted mice could mechanism for the loss of inhibitory control for palatable food also contribute to the alteration of cortico-limbic system in these seeking and consumption. An increase in the activity of this mice. On the other hand, D1R (encoded by Drd1) in the striatum pathway plays a key role in resilience to develop food addiction. conforms the direct pathway (GO), which promotes the approach Our results provided new mechanistic underpinnings of food to addiction-like behavior opposite to the D2R stimulating the addiction, which might be valued for other psychiatric disorders indirect pathway (NO GO)6. In contrast, D1R stimulation in the with alterations in compulsive behavior due to the transdiagnostic mPFC decreases release of glutamate onto L5 pyramidal cells44 nature of this concept. In addition, our findings pave the way for similar to D2R does41. Thus, the increased Drd1 mRNA levels in prevention measures, by identifying mechanisms required for PFC of the addicted mice could also contribute to the decreased strengthening the resilient phenotype. excitability of the glutamatergic mPFC neurons projecting to NAc. Therefore, it could be of interest to explore in future studies Methods whether overexpression of D1R in mPFC can also modify the Behavioral experiments. Self-administration session: The beginning of each self- development of food addiction-like behavior. administration session was signaled by turning on a house light placed on the Notably, the most differentially expressed gene found in our ceiling of the chamber during the first 3 s. Daily self-administration sessions maintained by chocolate-flavored pellets lasted 1 h in the long food addiction study was the Drd2. To our knowledge, this is the first study protocol and 2 h in the short food addiction protocol to increase the exposure of revealing an increased expression of the gene encoding for D2R the palatable pellets on each day to ensure the development of the addiction-like in the mPFC in the context of addiction. In contrast, a reduced phenotype. The self-administration sessions were composed by two pellet periods levels of D2R in the striatum have been already implicated (25 min and 55 min) separated by a pellet-free period (10 min). During the pellet periods, pellets were delivered contingently after an active response paired with a in this disorder45. Thus, neuroimaging studies reported a stimulus light (cue light). A time-out period of 10 s was established after each pellet downregulation of D2R in the striatum, which correlated with an delivery, where the cue light was off, and no reinforcer was provided after hypofunction of the PFC in cocaine abusers45. In agreement, responding on the active lever. Responses on the active lever and all the responses a specific Drd2 polymorphism is associated with the “Reward performed during the time-out period were recorded. During the pellet-free period, Deficiency Syndrome”, consisting in a hypodopaminergic state no pellet was delivered and this period was signaled by the illumination of the entire self-administration chamber. In the operant conditioning sessions, mice due to the altered D2R function46. were under fixed ratio 1 (FR1) schedule of reinforcement (one lever-press resulted In addition, several genes in the mPFC were downregulated in in one pellet delivery) followed by an increased FR to 5 (FR5) (five lever-presses addicted mice independently of the genotype. Interestingly, genes resulted in one pellet delivery) for the rest of the sessions. As previously descri- related to neuronal plasticity (Myh11, Acta2, and Cdh1)24,25, bed49, the criteria for the achievement of the operant responding were acquired when all of the following conditions were met: (1) mice maintained a stable prostaglandin synthesis (Ptgds)26 and gene regulation (Fosb)27 responding with <20% deviation from the mean of the total number of reinforcers were differentially expressed. The transcriptomic analysis between earned in three consecutive sessions (80% of stability); (2) at least 75% responding WT and Glu-CB1-KO mice revealed a downregulation in the on the active lever; and (3) a minimum of 5 reinforcers per session. After each Cnr1 gene as expected in the mPFC of the mutants confirming session mice were returned to their home cages. Three addiction-like criteria: Three behavioral tests were used to evaluate the the genetic deletion of the CB1R in this area. Additionally, Fos food addiction-like criteria as recently described5 and adapted from cocaine gene, encoding for c-fos protein, was less expressed in the addiction-like in rats50. These three criteria summarized the hallmarks of addiction mutants implicating an enhanced neuronal activity in mPFC of based on DSM-IV50, specified in DSM-5 and now included in the food addiction WT compared to Glu-CB1-KO mice. diagnosis through the YFAS 2.03. Persistence to response: Non-reinforced active responses during the pellet-free In agreement to our transcriptomic data analysis, we predicted period (10 min), when the box was illuminated and signaling the unavailability of that the upregulation of Drd2 gene expression in PL-NAc core pellet delivery, were measured as a persistence of food- seeking behavior. On the 3 pathway could have a critical role in promoting the vulnerability consecutive days before the progressive ratio mice were scored. to develop food addiction. Importantly, ISH experiments Motivation: The progressive ratio schedule of reinforcement was used to demonstrated endogenous Drd2 mRNA expression in the PL of evaluate the motivation for the chocolate-flavored pellets. The response required to earn one single pellet escalated according to the following series: 1, 5, 12, 21, 33, 51, naive WT mice, although at very low levels close to the limit of 75, 90, 120, 155, 180, 225, 260, 300, 350, 410, 465, 540, 630, 730, 850, 1000, 1200, detection. Therefore, we assume that addicted mice may have a 1500, 1800, 2100, 2400, 2700, 3000, 3400, 3800, 4200, 4600, 5000, and 5500. The significant increase in endogenous Drd2 mRNA levels based on maximal number of responses that the animal performs to obtain one pellet was gene expression data. Then, we aimed at mimicking the upre- the last event completed, referred to as the breaking point. The maximum duration of the progressive ratio session was 5 h or until mice did not respond on any lever gulation of Drd2 gene expression found in the mPFC of addicted within 1 h. mice by overexpressing D2R selectively in the PL-NAc core Compulsivity: Total number of shocks in the session of shock test (50 min) projections. Our results revealed that these mice showed performed after the PR test, when each pellet delivered was associated with a enhanced compulsive eating behavior despite the aversive punishment, were used to evaluate compulsivity-like behavior, previously described as resistance to punishment5,50. Mice were placed in a self-administration chamber consequences. Previous studies have demonstrated that the without the metal sheet with holes and consequently with the grid floor exposed PL-NAc core projections are regulated by presynaptic D2R and (contextual cue). In this shock-session, mice were under a FR5 schedule of directly innervate the D2R expressing MSNs (D2-MSNs) of the reinforcement during 50 min with two scheduled changes: at the fourth active 12 NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y ARTICLE lever-response mice received only an electric footshock (0.18 mA, 2 s) without pellets. The three addiction-like criteria (1) persistence to response (2), motivation pellet delivery and at the fifth active lever-response, mice received another electric (3), and compulsivity were evaluated at two different time points in each mouse. footshock with a chocolate-flavored pellet paired with the cue light. The schedule The first time point was the early period (sessions 1–18 of FR5) and the second was reinitiated after 10 s pellet delivery (time-out period) and after the fourth time point was the late period (sessions 95–112 of FR5). For the in vitro electro- response if mice did not perform the fifth response within a min. physiological recordings, we used naive Glu-CB1-KO mice (n = 6) and their WT Attribution of the three addiction-like criteria: After performing the three littermates (n = 6) (see section In vitro electrophysiology in brain slices). behavioral tests to measure the food addiction-like behavior, mice were categorized For the inhibition of the glutamatergic transmission in PL subregion in addicted or non-addicted animals depending on the number of positive criteria (Supplementary Fig. 4e), mice followed the same behavioral procedure described that they had achieved. An animal was considered positive for an addiction-like for the early period in the first experiment with some variations due to the surgical criterion when the score of the specific behavioral test was above the 75th AAV injection. In particular, Nex-Cre mice were trained to acquire the operant percentile of the normal distribution of the chocolate control group. Mice that conditioning maintained by chocolate-flavored pellets under FR1 (two sessions) achieved two or three addiction-like criteria were considered addicted animals and and FR5 (two sessions) schedule of reinforcement followed by the surgery for mice that achieved 0 or 1 addiction-like criteria were considered non-addicted injecting Cre-dependent AAVs carrying the DREADD (DREADD approach). After animals. bilateral intracranial injection of the AAV-hM4Di in the PL, the expression of the AAV was allowed during the period of four weeks. At the beginning of this period, mice were under FR5 (four sessions) to recover the basal levels of operant Surgery and virus vector microinjection. General surgical procedures: Mice were responding. At the end of these four weeks, an osmotic minipump filled with CNO anesthetized as reported in the drugs section (Supplementary Information) and or saline was subcutaneously implanted in the back of each mouse. Subsequently, placed into a stereotaxic apparatus for receiving the adeno-associated virus (AAV) during the chronically CNO-induced activation of the expressed hM4Di receptors, intracranial injections. All the injections were made through a bilateral injection mice were under FR5 scheduled sessions followed by the measurement of the three cannula (33-gauge internal cannula, Plastics One, UK) connected to a polyethylene addiction-like criteria. To verify the viral expression, mice were perfused at the end tubing (PE-20, Plastics One, UK) attached to a 10 μl microsyringe (Model 1701 N of the experiment, and the fluorescent reporter mCherry was visualized in brain SYR, Cemented NDL, 26 ga, 2 in, point style 3, Hamilton company, NV). The slices using a Leica DMR microscope equipped with a digital camera Leica DFC displacement of an air bubble inside the length of the polyethylene tubing that 300FX (10x objectives). In a subset of injected mice, in vitro electrophysiology connected the syringe to the injection needle was used to monitor the micro- recordings were used to verify CNO-induced suppression of neuronal activity injections. The volume [0.2 μl per site in prelimbic (PL), 0.4 μl per site in nucleus (see section In vitro electrophysiology in brain slices). accumbens core (NAc core)] was injected at a constant rate of 0.05 μl/min (PL) or For the specific inhibition of the projecting neurons from PL to NAc core 0.1 μl/min (NAc core) by using a microinfusion pump (Harvard Apparatus, Hol- (Fig. 3g), mice followed a similar experimental design as described above with a liston, MA) for 4 min. After infusion, the injection cannula was left in place for an modification in the surgical intervention. In this experiment, a dual viral approach additional period of 10 min to allow the fluid to diffuse and to prevent reflux, and was performed to selectively silence the PL neurons that project to NAc core (retro- then it was slowly withdrawn during 10 additional min. We used the following DREADD approach): bilateral intracranial injection of AAV-hM4Di targeting the coordinates to target our injections according to Paxinos and Franklin:51 (PL) PL and of AAV-retrograde-Cre targeting the NAc core. To verify viral expression, AP + 1.98 mm, L ± 0.3 mm, DV − 2.3 mm; (NAc core) AP + 1.34 mm, L ± 1 mm, mice were perfused at the end of the experiment and the fluorescent reporter DV − 4.6 mm. mCherry was visualized in brain slices, as previously described. Cre-recombinase Viral vectors: We used the following vectors: AAV-hM4Di-DREADD (AAV8- expression was detected by immunofluorescence using an anti-Cre recombinase hSyn-DIO-hM4D(Gi)-mCherry, 1.21E + 13 gc/ml), AAV-control-DREADD antibody (see section Immunofluorescence studies in Supplementary Information). (AAV8-hSyn-DIO-mCherry, 1.19E + 13 gc/ml) and AAV-retrograde-Cre-GFP Same as in the previous experiment, in vitro electrophysiology recordings were (AAVrg-Syn1-GFP-Cre; 8.2E + 12 gc/ml) from Viral Vector Production Unit of used in a subset of injected mice to verify CNO-induced suppression of neuronal Universitat Autònoma de Barcelona, AAV-retrograde-Cre-BFP (AAVrg pmSyn1- activity (see section In vitro electrophysiology in brain slices). EBFP-Cre; 6 × 10¹² vg/ml) from Addgene (viral prep # 51507-AAVrg), AAV-D2R For the transcriptomic analysis, Glu-CB1-KO and WT mice were sacrificed (AAV2/1-hSyn-DIO-SF-D2R(L)-IRES-mVenus,1,23E + 13 gc/ml), the plasmid was immediately after the last FR5 session of the food addiction procedure and mPFC a gift from Christoph Kellendonk and Jonathan Javitch’s lab52, and the was extracted by macrodissection to perform RNA sequencing (see section RNA corresponding AAV-control (AAV1/2-hSyn-floxstop-hrGFP, 7,69E + 11 gc/ml) sequencing). was from Beat Lutz’s lab. For the inhibition of glutamatergic neurons in the PL For the overexpression of D2R in PL-NAc core pathway (Fig. 5g), mice followed subregion, a bilateral injection targeting the PL was performed in Nex-Cre mice. the same behavioral and surgical procedure with a dual vector approach similar to Mice received an injection of 0.2 μl per site of the AAV-hM4Di-DREADD or 0.2 μl that described in the previous experiment with slight modifications: (1) AAV-D2R per site of the AAV-control-DREADD. For the specific inhibition of the projecting or AAV-control was injected in the PL, (2) the surgical intervention for the osmotic neurons from PL to NAc core, two bilateral injections were performed in WT minipump filled with CNO was not required, and (3) an immunofluorescence C57BL/6J mice, one targeting the PL and the other the NAc core. Mice received an assay was performed after the perfusion of the mice using an anti-D2R and anti- injection of 0.2 μl per site of the AAV-hM4Di-DREADD into PL and an injection GFP antibody that visualize the overexpressed D2Rs and mVenus reporter of the of 0.4 μl per site of the AAV-retrograde-Cre-EBFP into the NAc core. A subset of AAV injected in PL and against Cre recombinase to visualized the injection site of mice used for electrophysiological recordings in the NAc received the AAV- the retrograde AAV in the NAc core and the retrograde transport to the PL retrograde-Cre-GFP. For the overexpression of D2R in the PL-NAc core projecting (see section Immunofluorescence studies in Supplementary Information). Equally neurons, two bilateral injections were performed in WT C57BL/6J mice, one to previous experiments, in vitro electrophysiology recordings were used in a subset targeting the PL and the other the NAc core. Mice received bilateral injections of of injected mice to verify that overexpression of D2R induced the suppression of 0.2 μl of the AAV-D2R or 0.2 μl per site of the AAV-control into PL and bilateral neuronal activity by using D2R agonists, quinpirole and dopamine (see section In injections of 0.4 μl of the AAV-retrograde-Cre-EBFP into the NAc core. vitro electrophysiology in brain slices). To detect the viral expression in all the experiments we visualized each reporter as it is listed below: For the detection of AAV8-hSyn-DIO-hM4D(Gi)-mCherry and AAV8-hSyn- DIO-mCherry in Nex-cre mice, we directly visualized mCherry in the confocal In vitro electrophysiology in brain slices. Animals were sacrificed and brains microscope. mCherry is a bright red monomeric fluorescent protein that was were quickly removed obtaining coronal slices (300 μm) with vibratome (Leica clearly visible in our experimental conditions without performing an VT1200S) upon the presence of low-Na+ cutting solution (composition in mM: immunofluorescence. Sucrose 212, NaHCO3 27, Dextrose 10, CaCl2 2.2, MgSO4 2.2, KCl 2, NaH2PO4 For the detection of AAV8-hSyn-DIO-hM4D(Gi)-mCherry and AAV8-hSyn- 1.5; pH 7.4 when saturated with 95% O2 + 5% CO2). Afterwards, slices were DIO-mCherry in PL, and AAVrg-pmSyn1-EBFP-Cre targeting the NAc, we incubated (40 min/34 °C) in artificial cerebrospinal fluid (ACSF; composition in performed an immunofluorescence immunocytochemistry against Cre to detect the mM: NaCl 124, KCl 2.5, NaHCO3 26, CaCl2 2, MgCl2 1, NaH2PO4 1.25, glucose correct viral infection in the NAc and visualized the labeled Cre and the mCherry 10; pH 7.4 when saturated with 95% O2 + 5% CO2). Visualization of brain slices in the confocal microscope. were performed with an upright microscope (BX51WI, Olympus), outfitted with For the detection of AAV2/1-hSyn-DIO-SF-D2R(L)-IRES-mVenus and x4 lens, x40 water immersion lens, Nomarsky optics and mercury lamp with AAV1/2-hSyn-floxstop-hrGFP in the PL and AAVrg-pmSyn1-EBFP-Cre targeting adequate filters for blue (470–490 nm) and green (533–580 nm) light stimulation. the NAc, we performed an immunofluorescence immunocytochemistry against L5 pyramidal neurons were recognized by their position along the cortical col- mVenus/GFP and against Cre to detect the correct viral infection in both brain umn, soma shape and presence of apical dendrite and electrophysiological areas. We visualized the labeled mVenus or GFP, and the labeled Cre in the properties. confocal microscope. We performed a double immunohistochemistry analysis Voltage-clamp and current-clamp electrophysiological recordings were using anti-D2R and anti-mVenus/GFP antibodies to visualize the expression of performed by using a Multiclamp 700B amplifier (Axon Instruments), filtered at D2R at post- and presynaptic terminals. 1–2 kHz and digitized at 20 kHz with a 16 bits Axon Digidata 1550B (Axon Instruments). Protocols design and data acquisition were performed with pClamp9.2 software (Axon Instruments). Borosilicate patch pipettes (1.5 mm o.d., Experimental design. In the first experiment (Fig. 1a), Glu-CB1-KO mice (n = 58) 0.86 mm i.d., with inner filament; Harvard Apparatus) were used after pulled (P-97, and WT mice (n = 56) were trained under FR1 schedule of reinforcement during Sutter Instrument). Pipette resistance was calculated with pClamp software and six sessions, followed by 112 sessions of FR5 to self-administer chocolate-flavored was estimated among 8–10 MΩ. Electrical DC pulses were applied with a DS3 NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y Isolated Stimulator (Digitimer) using a theta-glass pipette filled with ACSF differential gene expression analysis was performed using R package DESeq55 with solution. All recordings were performed at RT (21–23 °C). FDR rate of 0.1. Variability between the addicted and non-addicted mice was Paired-Pulse facilitation recordings in layer 5 of Glu-CB1-KO mice: Synaptic determined using PCA analysis from “plotPCA” function of DEseq package. For facilitation was achieved by applying consecutive electrical stimulus with a 50 ms PCA analysis, top varying 500 genes were selected. P-value were calculated using interpulse interval in L2/3 of mPFC. The evoked field postsynaptic potentials nbinomTest function from DEseq package. (fPSP) were recorded in the L5 with ACSF filled patch pipette. Once the evoked Differential expression analysis was performed between addicted and non- fPSP were stabilized, at least 50 consecutive responses were recorded. Synaptic addicted mice. Similarly, we performed differential expression analysis between facilitation was estimated as the ratio among second response (P2) respect the first WT and Glu-CB1-KO. All genes with a fold change higher than >1.5-fold, P-value response (P1)41. Electrical stimulus (0.06 Hz) was the 50% of the intensity needed < 0.01 and average read counts > 40 in either condition comparison (i.e., addicted to evoke the maximal fPSP. vs. non-addicted and WT vs. KO mice) were selected as differentially expressed Layer 5 evoked fPSP pharmacological modulation: A single fPSP was evoked on genes. Further, it was visualized with volcano plot using R. The RNA sequencing L5 by stimulating on L2/3 as previously described. Once fPSP was stable, 50 data used in this study is available at GEO under accession number GSE139482. consecutive responses were recorded to stablish the baseline fPSP amplitude. Each pharmacological application was perfused on the recording chamber for 20 min Statistical analysis. All statistical comparisons were performed with SPSS (IBM, while recording evoked fPSP. The effect of WIN55,212-2 5 μM (Sigma-Aldrich, version 25). Comparisons between two groups were analyzed by Student t-test or U Spain) and rimonabant 4 μM (Sanofi-Aventis, Spain) in the fPSP amplitude was Mann–Whitney and within groups by paired t-test or Wilcoxon test depending on the calculated by averaging the last 10 evoked fPSPs. distribution defined by the Kolmogorov-Smirnov normality test. ANOVA with Pharmacological modulation of L5 pyramidal neurons properties of mPFC: repeated measures was used when required to test the evolution over the time. Two- Somatic current-clamp whole-cell recordings were obtained in the presence of way ANOVA by subsequent post hoc analysis (Fisher PLSD) was used for multiple ACSF and pharmacological treatment. Intracellular solution composition (in mM) group comparison. The Pearson correlation coefficient was used to analyze the rela- was: KMeSO4 135, KCl 10, HEPES 10, NaCl 5, ATP-Mg 2.5, GTP-Na 0.3; pH tionship between values in each addiction-like criteria and the final criteria achieved. adjusted to 7.3 by adding KOH. Hyperpolarizing and depolarizing square current The chi-square analyses were performed to compare the percentage of addicted mice pulses were applied (from −200 to 300 pA; Δ25 pA; 1 s duration). Resistance was with the non-addicted ones, we compared the observed frequencies with the fre- obtained from the first depolarizing pulse. Rheobase was calculated by applying a quencies obtained in the control group. Results were expressed as individual values ramp hyperpolarizing current-pulses (from 150 to 300 pA; 1.5 s duration). hM4Di- with the median and the interquartile range or with the mean ± S.E.M specified in the mCherry-positive neurons modulation was estimated after 10 min of perfusing figure legend. A P-value < 0.05 was used to determine statistical significance. The recording chamber with CNO 10 μM (Enzo Life Sciences, NY) diluted in ACSF. sample size was calculated based on the power analysis. The criterion for significance Dopamine hydrochloride 10 μM (Sigma-Aldrich, Spain) and quinpirole (alpha) was set at 0.050 and the statistical test used was two-sample t-test. With the hydrochloride 2 μM (Sigma-Aldrich, Spain) modulation of D2R-mVenus and GFP- sample size of 12–22 mice per group, our studies achieved a power between 73 and positive neurons was performed after 10 min perfusing recording chamber. 90%. Supplementary tables (Supplementary Tables 1–6) provided a complete report of mEPSC and mIPSC detection in PL mPFC and NAc of Glu-CB1-KO mice: the statistical results for the data described in the figures. Patch recording pipettes (4–9 MΩ) were filled with 115 mM cesium See Supplementary Information for extended details for animals, operant methanesulfonate, 20 mM CsCl, 10 mM Hepes, 2.5 mM MgCl2, 4 mM Na2ATP, behavior apparatus, food pellets, drugs, RT-PCR validation, immunofluorescence 0.4 mM Na3GTP, 10 mM sodium phosphocreatine, and 0.6 mM EGTA, pH 7.25. studies and fluorescence in situ hybridization. For mIPSCs recordings, cesium methanesulfonate was replaced with the same concentration of CsCl. Recordings of miniature responses (mEPSC and mIPSC) were performed using Reporting summary. Further information on research design is available in the whole-cell patch clamp technique in neurons at −70 mV in PL and at −75 mV the Nature Research Reporting Summary linked to this article. in NAc in the presence of 1 μM tetrodotoxin, 50 μM picrotoxin, and 100 μM saclofen (for mEPSCs) or 50 μM AP5 and 20 μM CNQX (for mIPSCs). Evoked synaptic responses were evoked with bipolar electrodes using single-voltage Data availability Individual data points are graphed in main and Supplementary Figures. All relevant data pulses (200 μs, up to 20 V). Stimulating electrodes were theta capillaries of that support this study are available from the corresponding author to any interested (4–9 μm tip) filled with ACSF. The stimulating electrodes were placed over NAc core 100–150 μm from the recorded cells. Electrophysiological recordings and data researcher upon reasonable request and at https://doi.org/10.6084/m9.figshare.11366015. acquisition were performed with PC-ONE amplifiers and pClamp software The RNA sequencing data used in this study is available at GEO under accession number (Molecular Devices). Analysis was performed with pClamp software (miniature GSE139482. currents) or custom-made Excel (Microsoft) macros. Pharmacological modulation of MSNs properties of NAc: The effects of Received: 12 June 2019; Accepted: 19 December 2019; pharmacological agents were tested at 10–15 min after bath application of WIN55,212-2, 5 μM, CNO 10 μm, or quinpirole hydrochloride 2 μM, where appropriate, using whole-cell mode and were quantified using 5 min bins. RNA sequencing. RNA extraction: Tissue collection was performed immediately after the last FR5 session. After decapitation, the brains were removed from the References skulls and processed rapidly on ice. The mPFC was isolated by macrodissection 1. Pursey, K. et al. The prevalence of food addiction as assessed by the Yale Food according to the following coordinate from Paxinos and Franklin 200151 (AP + Addiction Scale: a systematic review. Nutrients 6, 4552–4590 (2014). 1.98 mm) and the samples were placed in individual tubes, frozen on dry ice, and 2. Gordon, E. et al. What Is the evidence for “food addiction?” a systematic stored at −80 °C until RNA isolation for the RNA sequencing. The remaining brain review. Nutrients 10, 477 (2018). parts from the same animals were frozen on dry ice and stored at −80 °C. 3. Gearhardt, A. N., Corbin, W. R. & Brownell, K. D. Development of the Yale Total RNA from mPFC was extracted using a miRNeasy Mini kit for Food Addiction Scale version 2.0. Psychol. Addict. Behav. 30, 113–121 (2016). subsequent RNA-seq analysis and RT-PCR validation. Briefly, tissues were 4. Lindgren, E. et al. Food addiction: a common neurobiological mechanism with homogenized in QIAzol Lysis Reagent and, after adding chloroform, the aqueous drug abuse. Front. Biosci. 23, 811–836 (2017). phase was collected and microRNA and total RNA were extracted by using 5. Mancino, S. et al. Epigenetic and proteomic expression changes promoted by miRNeasy Mini kit (Qiagen, 217004). eating addictive-like behavior. Neuropsychopharmacology 10, 2788–800 Library preparation and total RNA sequencing: Further evaluation of the RNA, (2015). RNA library generation, and sequencing were carried out by StarSEQ GmbH 6. Koob, G. F. & Volkow, N. D. Neurobiology of addiction: a neurocircuitry (Mainz, Germany). Sequencing was performed on an Illumina NextSeq analysis. Lancet Psychiatry 3, 760–773 (2016). 500 sequencer with High Output chemistry and minimum output of 40 million 7. Miller, E. K. & Cohen, J. D. An Integrative theory of prefrontal cortex single end reads (75 bp) per sample. function. Annu. Rev. Neurosci. 24, 167–202 (2001). Total RNA (750–1000 ng/sample) was rRNA depleted by Ribo-Zero rRNA 8. Blakemore, S. -J. & Robbins, T. W. Decision-making in the adolescent brain. Removal Kit (Human/Mouse/Rat) from Illumina. For sequencing library Nat. Neurosci. 15, 1184–1191 (2012). generation, the NEBNext Ultra II Directional RNA Library Kit was used. Resulting 9. Diamond, A. Executive functions. Annu. Rev. Psychol. 64, 135–168 (2013). RNA libraries were size selected to a median insert size of 300 bp. 10. Volkow, N. D., Wang, G. J., Tomasi, D. & Baler, R. D. The addictive RNAseq data analysis: RNA sequencing for WT and Glu-CB1-KO mice were dimensionality of obesity. Biol. Psychiatry 73, 811–818 (2013). performed under addicted and non-addicted conditions. RNA-seq output was 11. Chen, B. T. et al. Rescuing cocaine-induced prefrontal cortex hypoactivity received as raw files in fastq format. To check the quality of individual sequenced prevents compulsive cocaine seeking. Nature 496, 359–362 (2013). sample, we used FASTQC version v0.10.5. Quality check followed by alignment 12. Volkow, N. D. et al. Activity in healthy adults. Obesity 17, 60–65 (2010). using TopHat version v2.153. to the mouse genome (mm9) with default parameters. 13. Riga, D. et al. Optogenetic dissection of medial prefrontal cortex circuitry. Further, mapped reads were considered for read count per gene using HTSeq Front. Syst. Neurosci. 8, 1–19 (2014). version 0.954. HTSeq output (read counts per gene) were normalized and 14 NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y ARTICLE 14. Burns, J. et al. Cannabis addiction and the brain: a review. J. Neuroimmune 43. Quintana, A. et al. Lack of GPR88 enhances medium spiny neuron activity and Pharmacol. 13, 438–452 (2018). alters motor- and cue-dependent behaviors. Nat. Neurosci. 15, 1547–1555 (2012). 15. Hitchcott, P. K., Quinn, J. J. & Taylor, J. R. Bidirectional modulation of goal- 44. Gao, W. J., Krimer, L. S. & Goldman-Rakic, P. S. Presynaptic regulation of directed actions by prefrontal cortical dopamine. Cereb. Cortex 17, 2820–2827 recurrent excitation by D1 receptors in prefrontal circuits. Proc. Natl Acad. (2007). Sci. U.S.A. 98, 295–300 (2001). 16. Monory, K. et al. The endocannabinoid system controls key epileptogenic 45. Volkow, N. D. et al. Decreased dopamine D2 receptor availability is associated circuits in the hippocampus. Neuron 51, 455–466 (2006). with reduced frontal metabolism in cocaine abusers. Synapse 14, 169–177 (1993). 17. Bellocchio, L. et al. Bimodal control of stimulated food intake by the 46. Blum, K. et al. Increased prevalence of the Taq I A1 allele of the dopamine endocannabinoid system. Nat. Neurosci. 13, 281–283 (2010). receptor gene (DRD2) in obesity with comorbid substance use disorder: a 18. Kano, M., Ohno-Shosaku, T., Hashimotodani, Y. & Uchigashima, M. preliminary report. Pharmacogenetics 6, 297–305 (1996). Endocannabinoid-mediated control of synaptic transmission. 309–380 (2009). 47. Cui, Q. et al. Dopamine receptors mediate strategy abandoning via https://doi.org/10.1152/physrev.00019.2008. modulation of a specific prelimbic cortex–nucleus accumbens pathway in 19. Goldstein, R. Z. & Volkow, N. D. Dysfunction of the prefrontal cortex in mice. Proc. Natl Acad. Sci. USA 115, E4890–E4899 (2018). addiction: Neuroimaging findings and clinical implications. Nat. Rev. 48. Bock, R. et al. Strengthening the accumbal indirect pathway promotes Neurosci. 12, 652–669 (2011). resilience to compulsive cocaine use. Nat. Neurosci. 16, 632–638 (2013). 20. Marsicano, G. & Kuner, R. in Cannabinoids and the Brain (ed. Kendall, D. A.) 49. Martín-García, E. et al. New operant model of reinstatement of food-seeking 161–201 (Springer, 2008). behavior in mice. Psychopharmacol. (Berl.). 215, 49–70 (2011). 21. Scofield, M. D. et al. The nucleus accumbens: mechanisms of addiction across 50. Deroche-Gamonet, V., Belin, D. & Piazza, P. V. Evidence for addiction-like drug classes reflect the importance of glutamate homeostasis. Pharmacol. Rev. behavior in the rat. Science 305, 1014–1017 (2004). 68, 816–871 (2016). 51. Paxinos, G. & Franklin, K. B. J. The Mouse Brain in Stereotaxic Coordinates. 22. Stuber, G. D. et al. Excitatory transmission from the amygdala to nucleus (Elsevier, 2001). accumbens facilitates reward seeking. Nature 475, 377–382 (2011). 52. Gallo, E. F. et al. Accumbens dopamine D2 receptors increase motivation by 23. Tervo, D. G. R. et al. A designer AAV variant permits efficient retrograde decreasing inhibitory transmission to the ventral pallidum. Nat. Commun. 9, access to projection neurons. Neuron 92, 372–382 (2016). 1086 (2018). 24. Kneussel, M. & Wagner, W. Myosin motors at neuronal synapses: drivers of 53. Trapnell, C. et al. Transcript assembly and quantification by RNA-Seq reveals membrane transport and actin dynamics. Nat. Rev. Neurosci. 14, 233–247 (2013). unannotated transcripts and isoform switching during cell differentiation. 25. Tan, Z. J., Peng, Y., Song, H. L., Zheng, J. J. & Yu, X. N-cadherin-dependent Nat. Biotechnol. 28, 511–515 (2010). neuron-neuron interaction is required for the maintenance of activity-induced 54. Anders, S., Pyl, P. T. & Huber, W. HTSeq–a Python framework to work with dendrite growth. Proc. Natl Acad. Sci. U.S.A. 107, 9873–9878 (2010). high-throughput sequencing data. Bioinformatics 31, 166–169 (2015). 26. Urade, Y. & Hayaishi, O. Biochemical, structural, genetic, physiological, and 55. Anders, S. & Huber, W. Differential expression analysis for sequence count pathophysiological features of lipocalin-type prostaglandin D synthase. data. Genome Biol. 11, R106 (2010). Biochim. Biophys. Acta 1482, 259–271 (2000). 27. Nestler, E. J. Cellular basis of memory for addiction. Dialogues Clin. Neurosci. 15, 431–443 (2013). 28. Martín-García, E. et al. Differential control of cocaine self-administration by Acknowledgements We thank A. Domi, E. James, E. Cuboni, R. Martín, M. Linares, F. Remmers, C. Maul, A. GABAergic and glutamatergic CB1 cannabinoid receptors. Conrad, and D. Real for their help during the experiments. The support by the IMB Neuropsychopharmacology 1–14 (2015). https://doi.org/10.1038/npp.2015.351. Microscope Core Facility (Mainz, Germany) is gratefully acknowledged. C. Kellendonk 29. Häring, M., Kaiser, N., Monory, K. & Lutz, B. Circuit specific functions of and J. Javitch (Columbia University, New York, USA) provided the AAV-hSyn-DIO- cannabinoid CB1 receptor in the balance of investigatory drive and D2L-mVenus construct. This work was supported by the Spanish Ministerio de Econ- exploration. PLoS ONE 6, e26617 (2011). omía y Competitividad-MINECO (#SAF2017-84060-R-AEI/FEDER-UE), the Spanish 30. Lutz, B., Marsicano, G., Maldonado, R. & Hillard, C. J. The endocannabinoid Instituto de Salud Carlos III, RETICS-RTA (#RD12/0028/0023), the Generalitat de system in guarding against fear, anxiety and stress. Nat. Rev. Neurosci. 16, Catalunya, AGAUR (#2017 SGR-669), ICREA-Acadèmia (#2015) and the Spanish 705–718 (2015). Ministerio de Sanidad, Servicios Sociales e Igualdad, Plan Nacional Sobre Drogas 31. Steindel, F. et al. Neuron-type specific cannabinoid-mediated G protein (#PNSD-2017I068) to R.M., Fundació La Marató-TV3 (#2016/20-30) to E.M-G., the signalling in mouse hippocampus. J. Neurochem. 124, 795–807 (2013). German Research Foundation (#CRC1193 “Neurobiology of Resilience”, TP A05 and 32. Monory, K., Polack, M., Remus, A., Lutz, B. & Korte, M. Cannabinoid CB1 B04) to B.L. and S.G., and the Boehringer Ingelheim Foundation to B.L., S.G. and I.R.A. receptor calibrates excitatory synaptic balance in the mouse hippocampus. J. The work of M.N.A was supported by the Emergent AI Center funded by the Carl-Zeiss- Neurosci. 35, 3842–3850 (2015). Stiftung. NARSAD Young Investigator Award (#22434), MINECO Ramón y Cajal 33. Wang, W. et al. Regulation of prefrontal excitatory neurotransmission by (#RYC- 2014-15784) and (#SAF2016-76565-R) and Fondo Europeo de Desarrollo dopamine in the nucleus accumbens core. J. Physiol. 590, 3743–3769 (2012). Regional (FEDER) to R.A. DIUE Generalitat de Catalunya (#2017 SGR 595), MINECO 34. Moorman, D. E., James, M. H., McGlinchey, E. M. & Aston-Jones, G. (#SAF2016-79956-R), EU (#Era Net Neuron PCIN-2013-060), Fundació La Marató-TV3 Differential roles of medial prefrontal subregions in the regulation of drug (#2016/20-31), CRG acknowledges support of the Spanish Ministry of Economy, seeking. Brain Res. 1628, 130–146 (2015). Industry and Competitiveness (MEIC) to the EMBL partnership, the Center of 35. Terraneo, A. et al. Transcranial magnetic stimulation of dorsolateral Excellence Severo Ochoa (#SEV-2012-0208), the CERCA Programme/Generalitat de prefrontal cortex reduces cocaine use: a pilot study. Eur. Catalunya, the CIBER of Rare Diseases of the ISCIII to M.D. MINECO Ramón y Cajal Neuropsychopharmacol. 26, 37–44 (2016). (#RYC-2016-20414), AGAUR (#2017 SGR 926), (#RTI2018-094887-B-I00) and Fondo 36. Heidbreder, C. A. & Groenewegen, H. J. The medial prefrontal cortex in the Europeo de Desarrollo Regional (FEDER) to M.N. rat: evidence for a dorso-ventral distinction based upon functional and anatomical characteristics. Neurosci. Biobehav. Rev. 27, 555–579 (2003). 37. Vertes, R. P. Interactions among the medial prefrontal cortex, hippocampus and midline thalamus in emotional and cognitive processing in the rat. Author contributions Neuroscience 142, 1–20 (2006). E.M.-G., L.D.-R. and R.M. conceived and designed the behavioral studies with input from 38. McGlinchey, E. M., James, M. H., Mahler, S. V., Pantazis, C. & Aston-Jones, G. B.L. and I.R.A.; B.L. generated the conditional transgenic mice (Glu-CB1-KO); L.D.-R. Prelimbic to accumbens core pathway is recruited in a dopamine-dependent performed the behavioral experiments and the statistical analyses and graphs with the manner to drive cued reinstatement of cocaine seeking. J. Neurosci. 36, supervision of E.M.-G. and R.M.; S.K. collaborated with the extraction of the samples; R.A., 8700–8711 (2016). R.M. and E.M.-G. designed the chemogenetic experiment; E.M.-G. and R.M. set up the 39. Schmitzer-Torbert, N. et al. Post-training cocaine administration facilitates DREADD and retro-DREADD approach; L.D.-R., E.M.-G. and E.S. performed chemoge- habit learning and requires the infralimbic cortex and dorsolateral striatum. netic surgeries; L.D.-R., E.S. and I.R.A. performed immunofluorescence studies and the Neurobiol. Learn. Mem. 118, 105–112 (2015). confocal images supervised by E.M.-G., R.M. and B.L.; E.D., I.S. and M.N. performed and 40. Le Merrer, J. et al. Protracted abstinence from distinct drugs of abuse shows designed electrophysiology experiments supervised by M.D, M.N., R.M. and E.M.-G.; regulation of a common gene network. Addict. Biol. 17, 1–12 (2012). I.R.A. and M.N.A. performed the RNA sequencing and the qRT-PCR experiments and the 41. Real, J. I., Simões, A. P., Cunha, R. A., Ferreira, S. G. & Rial, D. Adenosine A bioinformatic analysis supervised by B.L. and G.S.; C.H., S.B. and I.R.A. performed FISH assay; E.M.-G., L.D.-R., B.L. and R.M. wrote the manuscript and I.R.A provided critical 2A receptors modulate the dopamine D 2 receptor-mediated inhibition of synaptic transmission in the mouse prefrontal cortex. Eur. J. Neurosci. 47, review of the manuscript with inputs from all the other authors. 1127–1134 (2018). 42. Ferre, S. et al. An update on adenosine A2A-dopamine D2 receptor interactions: implications for the function of G protein-coupled receptors. Competing interests Curr. Pharm. Des. 14, 1468–1474 (2008). The authors declare no competing interests. NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications 15 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-14458-y Additional information Open Access This article is licensed under a Creative Commons Supplementary information is available for this paper at https://doi.org/10.1038/s41467- Attribution 4.0 International License, which permits use, sharing, 020-14458-y. adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Correspondence and requests for materials should be addressed to R.M. Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless Peer review information Nature Communications thanks Veronica Alvarez, Daniele indicated otherwise in a credit line to the material. If material is not included in the Piomelli and Ana Joao Rodrigues for their contribution to the peer review of this work. article’s Creative Commons license and your intended use is not permitted by statutory Peer reviewer reports are available. regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ Reprints and permission information is available at http://www.nature.com/reprints licenses/by/4.0/. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. © The Author(s) 2020 16 NATURE COMMUNICATIONS | (2020)11:782 | https://doi.org/10.1038/s41467-020-14458-y | www.nature.com/naturecommunications