A specific prelimbic-nucleus accumbens pathway controls resilience versus vulnerability to food addiction

The study addresses why some individuals develop loss of control over palatable food intake (food addiction) while others remain resilient. Food addiction, closely linked to obesity and eating disorders, shares neurobiological mechanisms with drug addiction, involving the basal ganglia, extended amygdala, and medial prefrontal cortex (mPFC). Evidence points to mPFC hypoactivity in addiction and a negative correlation between prefrontal metabolic activity and body weight. The prelimbic (PL) and infralimbic (IL) cortices project densely to the nucleus accumbens (NAc), a key node for inhibitory control. The mPFC circuitry of excitatory glutamatergic pyramidal neurons and inhibitory interneurons is modulated by endocannabinoid and dopaminergic systems. The authors hypothesize that specific circuit- and molecular-level mechanisms within the PL–NAc pathway govern resilience versus vulnerability to food addiction. They combine a validated operant mouse model of food addiction with conditional CB1 receptor deletion in glutamatergic neurons, ex vivo electrophysiology, RNA sequencing, chemogenetics, and projection-specific D2 receptor overexpression to dissect these mechanisms.

Prior work shows overlapping neurocircuitry between food and drug addiction, with adaptations in mPFC function contributing to compulsive seeking. Human imaging studies report mPFC hypoactivity in addiction and lower prefrontal activity with higher body weight. PL and IL mPFC subregions regulate decision-making and inhibitory control, projecting to the NAc core, which modulates GO/NO-GO behaviors. Endocannabinoid CB1 receptors modulate synaptic transmission and have been implicated in addiction-related behaviors; dopamine receptors, notably D1 and D2, shape corticostriatal signaling and inhibitory control. These insights frame the investigation into PL–NAc circuit modulation by CB1R and dopaminergic signaling in food addiction.

• Animal model and operant training: Male mice (WT and conditional Glu-CB1-KO lacking CB1R in dorsal telencephalic glutamatergic neurons) underwent operant self-administration of chocolate-flavored pellets. Training involved FR1 (acquisition; 1–6 sessions) followed by long-term FR5 (112 sessions) for the long protocol (1 h/day) or a shorter early protocol (2 h/day). Acquisition criteria required stable responding, ≥75% active lever responses, and ≥5 reinforcers/session.
• Addiction-like criteria: Three tests captured DSM-related hallmarks: (1) Persistence to response: non-reinforced active responses during a 10-min pellet-free, illuminated period; (2) Motivation: progressive ratio schedule (up to 5 h) with escalating response requirements; (3) Compulsivity: number of footshocks (0.18 mA, 2 s) received during a 50-min session where responding yielded shocks, with every fourth active response punished (no pellet) and fifth punished plus pellet. Mice were classified addicted if meeting ≥2 criteria, defined as scores ≥75th percentile of the control distribution.
• Electrophysiology: Ex vivo whole-cell recordings in brain slices from PL layer 5 pyramidal neurons and NAc medium spiny neurons measured miniature EPSCs/IPSCs (with TTX) and evoked synaptic responses. Paired-pulse facilitation (PPF; 50 ms interval) assessed presynaptic release probability. CB1R function was probed with WIN55,212-2 (5 µM) and rimonabant (4 µM). For DREADD and D2R experiments, current-clamp characterized firing, membrane resistance, and rheobase; quinpirole (2 µM) or dopamine (10 µM) tested D2R function.
• Chemogenetics (PL and PL–NAc): AAV8-hSyn-DIO-hM4D(Gi)-mCherry (or control) was injected bilaterally into PL of Nex-Cre mice to silence glutamatergic neurons. For projection-specific silencing, WT mice received PL AAV-hM4Di and NAc core AAV-retrograde Cre (AAVrg). Chronic CNO delivery via subcutaneous osmotic minipumps (0.25 µl/h for 4 weeks) activated hM4Di during FR5 sessions and addiction-criteria testing. Viral expression and projection specificity were verified by fluorescence and anti-Cre immunostaining; electrophysiology confirmed CNO-induced inhibition.
• Transcriptomics: mPFC was dissected immediately after the last FR5 session from addicted and non-addicted WT and Glu-CB1-KO mice (n=4–6/group). Total RNA underwent rRNA depletion and NextSeq 500 sequencing. Reads were aligned (TopHat), counted (HTSeq), and analyzed with DESeq (FDR 0.1); genes with >1.5-fold change, p<0.01, and mean counts >40 were called differentially expressed. qPCR validated selected genes (Drd2, Adora2A, Gpr88, Drd1; and genotype-dependent genes including Cnr1 and Fos).
• Projection-specific D2R overexpression: To mimic Drd2 upregulation in addicted mice, WT mice received PL AAV-hSyn-DIO-D2R(L)-mVenus (or flox-stop-GFP control) and NAc core AAVrg-Cre. ISH confirmed endogenous Drd2 is low in PL; qPCR verified ~40-fold Drd2 overexpression in mPFC. Electrophysiology assessed quinpirole effects on PL-NAc neurons and NAc mEPSCs. Behavioral testing followed the early protocol; addiction-like criteria were measured.
• Statistics: Group comparisons used t-tests or Mann–Whitney U; within-group paired t-tests or Wilcoxon; repeated-measures ANOVA for session evolution; two-way ANOVA where appropriate; chi-square for proportions of addicted animals; Pearson correlations between number of criteria and individual criterion scores. Alpha=0.05; sample sizes (12–22/group) provided 73–90% power.

• CB1R deletion confers resilience: Glu-CB1-KO mice earned fewer reinforcers under FR5 versus WT (repeated-measures ANOVA, genotype effect P<0.001), indicating reduced reinforcing value of palatable pellets. In the late period (sessions 95–112), mutants showed significantly reduced persistence, motivation, and compulsivity (U Mann–Whitney, P<0.01). Only 6.9% of Glu-CB1-KO were classified addicted (≥2 criteria) vs 25.0% of WT (chi-square, P<0.01). Body weight differences did not account for behavioral phenotypes.
• Enhanced excitatory transmission in Glu-CB1-KO: mEPSC frequency increased in PL L5 pyramidal neurons (t-test, P<0.01) and in NAc MSNs (t-test, P<0.05); mIPSCs were unchanged. PPF ratio was higher in mutants (U Mann–Whitney, P<0.01), indicating increased Ca2+-dependent release. CB1R agonist WIN55,212-2 reduced PL fPSP and NAc EPSC amplitudes in WT but effects were blunted/absent in Glu-CB1-KO, confirming functional CB1R deletion at glutamatergic terminals.
• PL–NAc core inhibition induces compulsivity and vulnerability: Projection-specific chemogenetic silencing (PL hM4Di + NAc retro-Cre) reduced PL neuron excitability and decreased NAc mEPSC frequency upon CNO. Behaviorally, CNO-treated mice showed increased compulsivity (U Mann–Whitney, P<0.01) without changes in persistence or motivation; 50.0% classified addicted vs 16.7% saline (chi-square, P<0.001). No changes in reinforcement rate, body weight, intake, or locomotion.
• mPFC transcriptomic signature of addiction: Addicted vs non-addicted mice showed 31 upregulated and 70 downregulated genes in mPFC (≥1.5-fold, p<0.01). Upregulated: Drd2, Adora2A, Gpr88, Drd1 (qPCR confirmed, P<0.05). Downregulated included Myh11, Acta2, Cdh1, Ptgds, Fosb. Genotype comparison confirmed reduced Cnr1 and Fos in Glu-CB1-KO.
• D2R overexpression in PL–NAc promotes compulsivity: Projection-specific Drd2 overexpression in PL–NAc neurons decreased PL neuron excitability upon quinpirole and reduced NAc mEPSC frequency. Behaviorally, compulsivity increased (t-test, P<0.05) with no changes in persistence, motivation, or reinforcement; 30.8% of D2R-overexpressing mice met 2–3 criteria vs 8.3% controls (chi-square, P<0.01).

The findings delineate a circuit- and receptor-specific mechanism for resilience versus vulnerability to food addiction. Enhanced excitatory transmission in the PL–NAc core pathway, as seen in Glu-CB1-KO mice, supports inhibitory control and resilience against developing compulsive intake. Conversely, reducing PL–NAc activity—either by chemogenetic silencing or by increasing D2R-mediated inhibition of PL projections—selectively disrupts inhibitory control under punishment, producing compulsive food seeking. Transcriptomics implicate upregulation of Drd2 (and Adora2A, Gpr88, Drd1) in mPFC of addicted mice; projection-specific D2R overexpression causally reproduced the compulsive phenotype, indicating that elevated D2R signaling can diminish PL neuron excitability and glutamatergic drive to NAc, biasing behavior toward loss of control. These data integrate endocannabinoid and dopaminergic modulation within a defined PL–NAc circuit to explain interindividual variability in addiction-like behavior, highlighting a top-down mechanism likely influencing D2-MSN indirect pathway engagement and NO-GO control.

This work identifies the prelimbic cortex to nucleus accumbens core glutamatergic projection as a critical substrate governing resilience versus vulnerability to food addiction-like behavior. Deletion of CB1R in cortical glutamatergic neurons enhances PL–NAc excitatory transmission and confers resilience, while chemogenetic inhibition or D2R overexpression in PL–NAc induces compulsive food seeking under punishment. A distinct mPFC transcriptional signature, with Drd2 as the most upregulated gene in addicted mice, supports a mechanistic role for D2R in reducing PL excitability and weakening inhibitory control. These insights suggest that targeting endocannabinoid and dopaminergic signaling in PL–NAc pathways could inform prevention or therapeutic strategies for compulsive eating and related disorders. Future work should map other PL projections (e.g., to amygdala, VTA), test the role of D1R upregulation in mPFC, and translate findings to human-relevant interventions.

• The mouse operant model, while translational, may not capture the full complexity of human food addiction and eating disorders.
• A small fraction of Glu-CB1-KO mice still became addicted, underscoring multifactorial etiology beyond single-gene effects.
• Chemogenetic manipulations showed residual viral expression in adjacent dorsal mPFC regions (e.g., anterior cingulate), so off-target regional contributions cannot be fully excluded.
• Projection specificity was focused on PL–NAc core; other PL outputs (e.g., to amygdala, hippocampus, VTA) were not directly tested and may contribute to other addiction-like endophenotypes (persistence, motivation).
• Transcriptomic analyses were limited to mPFC tissue bulk; cell-type-specific resolution was not assessed.
• Although chronic CNO showed no detected side effects in controls, potential off-target effects or back-metabolism cannot be entirely ruled out.