Scientists uncover brain circuit that turns cravings into eating behavior

Scientists uncovered how the brain’s bed nucleus of the stria terminalis acts as a master switchboard, merging reward and need signals to control eating, a discovery that could guide new treatments for obesity and illness-related weight loss.

Study: A brain center that controls consummatory responses

In a recent study published in the journal Cell, a group of researchers tested whether the bed nucleus of the stria terminalis (BNST) integrates central amygdala (CEA) prodynorphin (Pdyn) and hypothalamic agouti-related peptide (AGRP) signals to control consumption, with downstream consequences for body weight.

Background

Cravings seem simple—see sugar, want sugar—but the choice to actually consume is a brain-wide negotiation. People feel it when a dessert is irresistible while hungry, yet forgettable after lunch. Public health feels it amid obesity, appetite-blunting cancer therapies, and satiety drugs. A question follows: where do “this tastes good” and “my body needs it” meet? The BNST is a switchboard that integrates sensory valence with internal state to regulate consumption. Knowing how this works could help counter illness-related weight loss and refine anti-obesity treatments. Further research is needed to map input specificity and state-dependent rules across nutrients.

About the study

The investigators first identified CEA neurons encoding sweet valence using Act-seq single-cell ribonucleic acid sequencing (scRNA-seq) and immediate-early gene Fos co-expression to identify sweet-responsive CEA neurons co-expressing Fos and Pdyn. They recorded activity from these neurons with adeno-associated virus (AAV)-delivered circularly permuted green fluorescent protein-calmodulin-M13 peptide calcium indicator 6s (GCaMP6s) and fiber photometry while delivering tastants via an intraoral fistula. To test causality, they expressed channelrhodopsin-2 (ChR2) or the inhibitory anion channelrhodopsin GtACR2 in Pdyn neurons and coupled activation or silencing to licking or lever-press tasks. Anterograde tracing mapped projections from CEA Pdyn neurons to the BNST. Responses within the BNST were dissected by targeting the vesicular gamma-aminobutyric acid transporter (VGAT) population and performing fiber photometry and optogenetics.

The convergence of internal-state signals was tested using monosynaptic rabies tracing from BNST VGAT neurons and by optogenetically stimulating agouti-related peptide (AGRP) neurons in the arcuate nucleus (ARC) that express ChrimsonR. Microendoscopic calcium imaging with a gradient refractive index (GRIN) lens quantified ensemble coding across fed, hungry, and sodium-depleted states. Finally, chemogenetics (designer receptors exclusively activated by designer drugs (DREADDs), hM3Dq; clozapine N-oxide (CNO)) and a pharmacologically selective actuator module 4–4-glycine receptor (PSAM4-GlyR) plus its ligand ultrapotent Pharmacologically Selective Effector Molecule 792 (uPSEM792) tested whether BNST modulation alters food intake and body weight in cisplatin-treated or diet-induced-obese mice.

Study results

Pdyn-positive neurons in the CEA encoded attraction to sweet: optogenetic activation made otherwise neutral water highly sought, and silencing abolished preference for both artificial sweetener and sugar while leaving fat preference intact, demonstrating modality specificity. Dense projections from these CEA Pdyn neurons to the BNST were mapped; within BNST, VGAT neurons responded robustly and selectively to sweet stimulation. Optogenetic stimulation of Pdyn terminals in the BNST increased licking to water, establishing a causal link between amygdala sweet coding and consumption; conversely, inhibiting BNST VGAT neurons suppressed sweet intake. Importantly, the Pdyn to BNST circuit did not drive dry-licking when no fluid was available, indicating consummation rather than pure reward seeking, whereas activating Pdyn somata in the CEA did evoke dry-licking self-stimulation by engaging additional reward-related targets.

Internal state powerfully reweighted BNST processing. Hunger amplified sweet-evoked responses in BNST and boosted consumption by approximately 250–300%, while TRPM5 knockout mice did not show this effect; stimulating AGRP inputs to BNST in sated mice mimicked this enhancement, recruiting previously silent sweet-responsive BNST neurons and supporting convergence of taste and need signals within BNST. Retrograde monosynaptic tracing corroborated direct inputs from the CEA and the ARC to BNST VGAT neurons. Beyond hunger, sodium depletion selectively increased BNST responses to sodium chloride (NaCl) by ~300% and shifted behavior toward salt seeking, revealing state-specific gating for need.

At the population level, microendoscopic single-cell calcium imaging revealed that internal states recruit additional BNST neurons without significantly altering individual response amplitudes, thereby expanding the ensemble representing sweet during hunger. A softmax decoder trained on BNST activity separated stimulus identity (sweet vs. salt) and internal state (fed, hungry, sodium-depleted), achieving ~80% accuracy overall and showing that BNST ensembles jointly encode “what” and “need.”

Causally, the BNST behaved as a general consumption controller: optogenetic activation of BNST VGAT neurons drove broad intake, including solid food, sweet, high-sodium water, normally aversive bitter substances, and even fictive food pellets, whereas silencing suppressed consummatory behavior irrespective of the stimulus or state. Translational tests underscored clinical relevance. In a cisplatin model that induces cachexia-like weight loss, chemogenetic activation of BNST VGAT neurons preserved body weight. Conversely, chemogenetic inhibition of BNST induced weight loss in diet-induced obese mice, an effect that reversed when inhibition ceased. A glucagon-like peptide-1 receptor (GLP-1R) agonist, semaglutide, induced Fos selectively in protein kinase C delta (PKCδ)-positive BNST neurons, suggesting that these neurons are a potential site contributing to the anti-obesity drug action. Together, these data define the BNST as a unified “consumption dial” that flexibly integrates appetitive value and internal need to drive consummatory responses.

Conclusions

To summarize, this work positions the BNST as a control hub that turns liking into eating by merging sensory valence from CEA Pdyn neurons with need signals from AGRP neurons. Clinically, a bidirectional lever over consumption suggests ways to counter cancer-therapy-related wasting or, oppositely, to promote weight loss by inhibition. The association of GLP1R agonism with PKCδ–positive BNST neurons highlights a mechanistic waypoint for anti-obesity drugs.