A hypothalamic circuit links high-fat diet exposure to anxiety-like behavior and anxiety-associated hyperphagia

Peng Wang; Jun Li; Xin Cheng; Wei Chen; Zhengyu Li; Zheng Fang; Dijia Wang; Bin Mei; Hu Liu; Shaohua Hu; Zhilai Yang; Jiqian Zhang; Gaolin Qiu; Xin Qing; Xuesheng Liu

doi:10.1038/s41467-026-77749-w

. 2026 Sep 16;17:10892. doi: 10.1038/s41467-026-77749-w

A hypothalamic circuit links high-fat diet exposure to anxiety-like behavior and anxiety-associated hyperphagia

Peng Wang ^1,^#, Jun Li ^1,^#, Xin Cheng ^1,^2,^#, Wei Chen ¹, Zhengyu Li ¹, Zheng Fang ¹, Dijia Wang ¹, Bin Mei ¹, Hu Liu ¹, Shaohua Hu ², Zhilai Yang ¹, Jiqian Zhang ¹, Gaolin Qiu ^1,^✉, Xin Qing ^1,^✉, Xuesheng Liu ^1,^2,^✉

PMCID: PMC13677739 PMID: 42893998

Abstract

Obesity and anxiety disorders are frequently comorbid, yet the neural circuit mechanisms linking metabolic and affective states remain incompletely understood. Here, in male mice exposed to long-term high-fat diet, we identify a hypothalamic circuit comprising arcuate nucleus agouti-related protein neurons and paraventricular nucleus corticotropin-releasing hormone neurons that regulates anxiety-like behavior under metabolic stress, with downstream recruitment of lateral hypothalamic area glutamatergic neurons to modulate feeding in a state-dependent manner. In vivo fiber photometry and ex vivo electrophysiology show that high-fat diet enhances activity and feeding-locked recruitment of this circuit, associated with increased anxiety-like behavior and feeding. Circuit-specific inhibition attenuates both phenotypes, whereas corticotropin-releasing hormone receptor 2 signaling in the lateral hypothalamic area selectively mediates high-fat diet-associated hyperphagia without affecting anxiety-like behavior. In this work, we show a disease-state-dependent hypothalamic circuit differentially recruited under chronic high-fat diet to coordinate anxiety-related behavior and feeding output, providing a framework for metabolic-affective interactions.

Subject terms: Emotion, Feeding behaviour

Obesity and anxiety often co-occur, but their neural link is unclear. Here, authors show that an Arc-PVN-LHA hypothalamic circuit coordinates HFD-associated anxiety-like behavior and feeding in a state-dependent manner.

Introduction

Obesity has reached alarming proportions as a global health crisis and is projected to affect half of all adults and millions of children worldwide^1,2. Epidemiological studies have demonstrated significant associations between obesity and multiple neuropsychiatric disorders³, with particularly strong bidirectional links between obesity and anxiety disorders^4,5. However, although previous studies have investigated overlapping mechanisms underlying obesity and anxiety comorbidity^5–7, the circuit-level processes through which metabolic and affective states interact remain incompletely understood. One clinically relevant manifestation of this interaction is anxiety-associated feeding, in which negative affective states bias the consumption of palatable, energy-dense foods independent of homeostatic hunger⁸. Current mechanistic studies, including those focused on mesolimbic dopamine signaling^9,10, have largely addressed how anxiety or stress enhances the reward value of high-fat foods. Less is known about how chronic high-fat diet (HFD)-associated metabolic perturbation recruits hypothalamic circuits to generate anxiety susceptibility and how this anxiety-related state, in turn, becomes functionally coupled to feeding behavior.

The arcuate nucleus (Arc) of the hypothalamus is a critical node in neural circuits controlling energy balance and body weight regulation¹¹. It serves as a central integrator of peripheral energy-state signals¹² and contains neuronal populations that are both necessary and sufficient for driving feeding behavior^13–17. Beyond rodents, studies in primates and humans also support the Arc as an important hypothalamic region involved in obesity-related metabolic regulation^18,19. However, whether chronic HFD exposure alters Arc-centered hypothalamic circuits in a manner that links metabolic perturbation to anxiety-related behavioral states remains unclear.

Agouti-related protein (AgRP) neurons in the Arc are a canonical orexigenic population that integrates peripheral energy-state signals and rapidly promotes food-seeking and food intake^12,14. Previous studies have established that activation of AgRP neurons is sufficient to induce robust feeding^14,20,whereas inhibition or loss of AgRP neuronal activity suppresses feeding-related motivation and intake^21–24. In addition to their local role in energy homeostasis, Arc^AgRP neurons send divergent projections to multiple hypothalamic and extra-hypothalamic targets^15,25–30, allowing metabolic-state information to be routed into broader behavioral circuits. Among these downstream targets, the paraventricular nucleus of the hypothalamus (PVN) is of particular interest because it participates not only in energy balance but also in neuroendocrine stress responses and anxiety-related behaviors^31–33. Thus, although the orexigenic role of Arc^AgRP neurons and their projections to PVN-related hypothalamic circuits are well established, whether this metabolic circuit is recruited under chronic HFD conditions to regulate anxiety susceptibility remains unknown. Given that PVN corticotropin-releasing hormone (CRH) neurons are key regulators of stress-related endocrine and behavioral responses, we hypothesized that an Arc^AgRP-PVN^CRH pathway may provide a circuit-level mechanism through which chronic HFD-associated metabolic perturbation is linked to anxiety-like behavior.

The lateral hypothalamic area (LHA) is another major hypothalamic hub for feeding regulation and reward-related motivated behavior and receives dense inputs from multiple upstream nuclei, including the PVN³⁴. Cell-type-specific studies have revealed that LHA neuronal populations exert bidirectional control over feeding, with LHA^Glu neurons generally constraining food intake and LHA^GABA neurons promoting appetitive and consummatory feeding behaviors^35,36. These observations suggest that the LHA may serve as a downstream node through which anxiety-related PVN signals are translated into altered feeding output. Importantly, the established anorexigenic role of LHA^Glu neurons does not necessarily conflict with a role for the PVN^CRH-LHA^Glu pathway in anxiety-associated hyperphagia. Rather, the behavioral effect of LHA^Glu neurons may depend on the direction and physiological context of upstream synaptic control. Under basal or physiological feeding conditions, LHA^Glu neurons may be primarily driven by excitatory inputs that enhance their feeding-suppressive function. In contrast, under chronic HFD-associated anxiety-susceptible conditions, inhibitory input from PVN^CRH neurons could reduce the activity of LHA^Glu neurons and thereby weaken their anorexigenic output, resulting in disinhibition of food intake. This framework raises the possibility that a PVN^CRH-defined LHA^Glu circuit may participate in anxiety-associated feeding by suppressing, rather than activating, the canonical feeding-inhibitory function of LHA^Glu neurons.

Corticotropin-releasing hormone receptor 2 (CRHR2), a receptor for CRH-related peptides, is highly represented within hypothalamic circuits and has been implicated in stress adaptation, energy homeostasis, and feeding regulation^37,38. CRHR2 has been shown to regulate meal size and stress-related feeding responses³⁹, and CRHR2 signaling in hypothalamic nuclei such as the ventromedial hypothalamus contributes to the control of feeding and lipid metabolism⁴⁰. Recent evidence further suggests that hypothalamic CRHR2 signaling links PVN^CRH activity to stress-related metabolic responses, supporting the possibility that CRHR2 may couple affective-state signals to metabolic output⁴¹. Given the anatomical and functional position of the LHA in feeding control, we hypothesized that CRHR2 signaling in LHA neurons may act downstream of PVN^CRH activity to regulate anxiety-associated feeding under chronic HFD conditions.

In the present study, we combined circuit tracing, in vivo fiber photometry, in vitro whole-cell patch-clamp recordings, optogenetics, chemogenetics, and pharmacological or genetic manipulation of CRHR2 signaling to determine how chronic HFD exposure links metabolic perturbation to anxiety-like behavior and anxiety-associated feeding. Although AgRP, PVN^CRH, and LHA^Glu neurons have established roles in feeding and stress regulation, how these nodes are dynamically integrated under chronic HFD-associated anxiety states remains unclear. We identify an Arc^AgRP-PVN^CRH pathway involved in HFD-associated anxiety-like behavior and further show that PVN^CRH neurons engage LHA^Glu neurons to regulate anxiety-associated hyperphagia. Moreover, blockade or cell-type-specific knockdown of CRHR2 signaling in the LHA attenuates HFD-associated hyperphagia, suggesting that LHA-CRHR2 signaling functions as a downstream effector mechanism linking anxiety-related hypothalamic activity to feeding output. Together, these findings support a disease-context-dependent Arc^AgRP-PVN^CRH-LHA^Glu circuit model that contributes to state-dependent functional coupling between metabolic state, anxiety susceptibility, and feeding behavior under chronic HFD conditions.

Results

Chronic HFD reveals an anxiety-susceptible subgroup with increased feeding and body weight gain

To investigate the effects of chronic HFD exposure on anxiety-like behavior and subsequent feeding in mice, we fed mice an HFD for 12 weeks to establish a diet-induced obesity model (Fig. 1a). Compared with mice fed a chow diet (CD), HFD-fed mice exhibited significantly larger body size (Fig. 1b), markedly greater weight gain (Fig. 1c), and higher levels of high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), total cholesterol (TC) and triglyceride (TG) (Fig. 1d–g). These findings confirm the successful induction of obesity via HFD.

We next assessed anxiety-like behaviors using the elevated plus maze (EPM) and open field test (OFT) (Fig. 1a). Compared with CD-fed mice, HFD-fed mice displayed significant anxiety-like behaviors, as evidenced by a reduced percentage of time spent in the open arms of the EPM and decreased time spent in the central area of the OFT. No significant difference was observed in the total distance traveled between the CD and HFD groups, indicating that HFD feeding did not impair general locomotor activity (Fig. 1h–k). To further determine whether these behavioral alterations were accompanied by a stress-related physiological response, we measured serum corticosterone levels (Fig. 1a). Serum corticosterone levels were significantly increased in HFD-fed mice compared with CD-fed mice (Fig. 1l). These results provide physiological evidence that chronic HFD exposure is associated with an enhanced stress-axis state, consistent with the anxiety-like behaviors observed in the EPM and OFT.

To further account for individual variability in anxiety-like behavior among HFD-fed mice, we performed an unsupervised hierarchical clustering analysis based on the percentage of time spent in the open arms during the EPM test. This analysis subclassified HFD-fed mice into anxiety-susceptible and anxiety-unsusceptible subgroups. After one week of monitoring food intake and body weight (Fig. 1m), we found that the anxiety-susceptible subgroup showed significantly increased daily food consumption and body weight compared with the anxiety-unsusceptible subgroup (Fig. 1n, o).

To further test whether the anxiety-related state functionally contributes to feeding behavior under HFD conditions, we treated HFD-fed anxiety-susceptible mice with the anxiolytic drug midazolam. Following phenotype stratification, anxiety-susceptible mice received daily intraperitoneal injections of saline or midazolam (0.2 mg/kg) prior to behavioral and feeding assessments (Supplementary Fig. 1a). Compared with saline-treated mice, midazolam-treated anxiety-susceptible mice showed reduced food intake and attenuated body weight gain (Supplementary Fig. 1b, c). These results indicate that pharmacological attenuation of anxiety decreases feeding output in HFD-fed anxiety-susceptible mice.

Collectively, these data demonstrate that chronic HFD exposure induces obesity, anxiety-like behavior, and enhanced stress-axis activity, and further reveal an anxiety-susceptible subgroup characterized by increased feeding and body weight gain. The midazolam experiment provides functional evidence that anxiety-related states can modulate feeding behavior under HFD conditions, supporting the concept of anxiety-associated hyperphagia rather than a generalized or arbitrarily defined emotional eating phenotype.

Arc^AgRP neurons are recruited during HFD-associated anxiety and hyperphagia

To identify hypothalamic neuronal populations associated with HFD-related anxiety and hyperphagia, we first assessed changes in neuronal activation by quantifying c-Fos expression. HFD-fed mice showed a marked increase in c-Fos expression across multiple brain regions, including the Arc, a key hypothalamic region involved in energy homeostasis and body weight regulation (Supplementary Fig. 2a, b). To determine the phenotype of activated Arc neurons, we performed immunofluorescence staining for c-Fos and AgRP (Supplementary Fig. 3a–c). This analysis revealed robust c-Fos expression in HFD-fed mice and showed that a substantial proportion of activated Arc neurons were AgRP-positive (Supplementary Fig. 3d), suggesting recruitment of Arc^AgRP neurons under chronic HFD conditions.

To directly monitor Arc^AgRP neuronal activity in vivo, we expressed GCaMP6m selectively in AgRP-Cre mice and performed fiber photometry recordings (Fig. 2a–c and Supplementary Fig. 4a). Baseline recordings showed that chronic HFD feeding enhanced Arc^AgRP neuronal activity compared with CD conditions (Supplementary Fig. 4b, c). Because baseline activity alone does not define how Arc^AgRP neurons are dynamically coupled to feeding behavior, we further performed feeding-locked fiber photometry analysis aligned to spontaneous feeding bouts. Arc^AgRP neuronal activity increased around feeding initiation and returned toward baseline after feeding termination (Fig. 2d, e). Notably, this feeding-associated activation was more pronounced in HFD-fed mice than in CD-fed mice, indicating that chronic HFD exposure enhances the dynamic recruitment of Arc^AgRP neurons during feeding behavior (Fig. 2f). These results suggest that Arc^AgRP neurons are not only tonically activated under HFD conditions but are also more strongly engaged during feeding episodes.

Fig. 2 — a Viral injection strategy. b Representative AgRP-GCaMP6m expression in the Arc. Scale bars, 100 µm (left) and 10 µm (right). c Quantification of GCaMP6m/AgRP colocalization (87%; n = 5 slices from 3 mice). Arc^AgRP GCaMP6m signals at feeding onset (d) and offset (e) in each group. f Fluorescence changes before, during, and after feeding within CD (F _(2,16) = 4.666; P = 0.03 and 0.04 for during vs. pre and post), HFD-anxiety-unsusceptible (F _(2,16) = 44.9; both P < 0.001), and HFD-anxiety-susceptible groups (F _(2,16) = 28.95; both P < 0.001), and during-feeding comparisons among groups (n = 9 trials from 3 mice; F _(2,24) = 6.877; CD vs. HFD-anxiety-unsusceptible: P = 0.05, CD vs. HFD-anxiety-susceptible: P = 0.004). Chemogenetic inhibition strategy (g, h), representative hM4Di-mCherry expression (i; scale bar, 100 µm), and viral spread quantification (j). k Representative Arc^AgRP action potentials in acute slices (Scale bars, 10 s, 20 mV). Representative EPM/OFT trajectories (l), EPM open-arm time (m; F _(3,27) = 11.74; both P < 0.001), OFT center time (n; F _(3,27) = 5.862; P = 0.008 and 0.03), and OFT total distance (o) across treatment groups (n = 10). Hourly (p; F _{(3, 36)} = 14.22, P = 0.009, 0.002, 0.001 for HFD + mCherry + CNO vs. HFD + hM4Di + CNO at 1–3 h; P = 0.01, <0.001, <0.001 for HFD + hM4Di + vehicle vs. HFD + hM4Di + CNO at 1–3 h) and cumulative food intake (q; F _{(3, 36)} = 17.65, P = 0.003, <0.001, <0.001, <0.001 for HFD + mCherry + CNO vs. HFD + hM4Di + CNO at 1–4 h; P = 0.02, <0.001, <0.001, <0.001 for HFD + hM4Di + vehicle vs. HFD + hM4Di + CNO at 1–4 h) (n = 10). Statistics: one-way repeated-measures ANOVA with Bonferroni correction (f, within-group), one-way ANOVA with Bonferroni correction (f, between-group), two-way ANOVA with Bonferroni correction (m–o), and two-way repeated-measures ANOVA with Bonferroni correction (p, q). Data are mean ± SEM. AgRP, agouti-related protein; Arc, arcuate nucleus; CD, chow diet; EPM, elevated plus maze; HFD, high-fat diet; OFT, open-field tests. Created in BioRender. Peng, W. (2026) https://BioRender.com/fsk2rp7 (a, g, h). Source data are provided as a Source Data file.

To determine whether Arc^AgRP neuronal activity is required for HFD-associated anxiety-like behavior and hyperphagia, we used a chemogenetic inhibition strategy by bilaterally injecting AAV-DIO-hM4Di-mCherry into the Arc of AgRP-Cre mice (Fig. 2g–i). We further quantified the spatial distribution of viral expression and mapped injection sites to verify that chemogenetic manipulations were predominantly restricted to the Arc (Fig. 2j and Supplementary Fig. 4d). To strengthen the causal interpretation and exclude potential confounding effects of viral expression or CNO administration, we expanded the behavioral paradigm by incorporating additional control groups. Specifically, mice were assigned to a full factorial design including: HFD + mCherry + vehicle, HFD + mCherry + CNO, HFD + hM4Di + vehicle, HFD + hM4Di + CNO, as well as CD + mCherry + vehicle, CD + mCherry + CNO, CD + hM4Di + vehicle, and CD + hM4Di + CNO (Fig. 2l–q and Supplementary Fig. 6a–h). This design allowed us to independently evaluate the effects of diet, viral expression, and CNO treatment. CNO administration did not alter anxiety-like behavior or feeding in mCherry-expressing mice, and hM4Di expression in the absence of CNO was also insufficient to affect these behavioral outcomes. In HFD-anxiety-susceptible mice, chemogenetic inhibition of Arc^AgRP neurons with CNO (Fig. 2k) significantly alleviated anxiety-like behaviors (Fig. 2l–o), reduced food intake (Fig. 2p, q), and decreased serum corticosterone levels (Supplementary Fig. 5a–c), indicating that Arc^AgRP neuronal activity is required for both behavioral and physiological features associated with the HFD-anxiety-susceptible state. In CD-fed mice, chemogenetic inhibition of Arc^AgRP neurons did not induce anxiety-like behavior, but led to a significant reduction in food intake (Supplementary Fig. 6a–h), indicating a state-dependent requirement of Arc^AgRP neurons for feeding regulation. Together, these results confirm that the behavioral effects observed in HFD-anxiety-susceptible mice are specifically attributable to chemogenetic inhibition of Arc^AgRP neurons under HFD conditions, rather than nonspecific effects of CNO or viral expression.

Conversely, chemogenetic activation of Arc^AgRP neurons by AAV-DIO-hM3Dq-mCherry in AgRP-Cre mice promoted hyperphagia after CNO administration, but did not significantly alter anxiety-like behavior (Supplementary Fig. 7a–i). This dissociation is consistent with the classical orexigenic role of AgRP neurons and indicates that Arc^AgRP activation is sufficient to drive feeding, but is not sufficient by itself to recapitulate the anxiety-like component under normal metabolic conditions.

Collectively, these findings demonstrate that Arc^AgRP neurons are dynamically recruited during feeding and are functionally required for HFD-associated hyperphagia and anxiety-like behavior in anxiety-susceptible mice. However, AgRP neuronal activation alone primarily promotes feeding without inducing anxiety-like behavior, suggesting that the anxiety-related function of Arc^AgRP neurons emerges in a chronic HFD-dependent state rather than representing a constitutive property of this neuronal population.

Arc^AgRP neurons provide input to PVN^CRH neurons

To identify downstream targets of the Arc that may mediate feeding-related changes in anxiety, we performed anterograde tracing by injecting AAV-DIO-EGFP into the Arc of AgRP-Cre mice (Fig. 3a, b and Supplementary Fig. 8a, b). After a three-week period for viral expression and axonal transport, we detected EGFP-positive fibers in multiple brain regions (Supplementary Fig. 8c). Notably, robust projections were observed in the PVN (Fig. 3c), a key center for the regulation of anxiety behavior. To further confirm the Arc-PVN projection, an anterograde monosynaptic tracking strategy was performed by injecting an AAV-hSyn-Cre-EGFP virus into the Arc (Supplementary Fig. 8d) and an AAV-DIO-mCherry virus into the PVN of C57BL/6 J mice (Fig. 3d). MCherry⁺ neurons were observed in the PVN 3 weeks later, which were largely stained positively with CRH antibody (Fig. 3e, f and Supplementary Fig. 8e–h).

Next, we performed monosynapticrabies-based retrograde tracing in CRH-Cre mice (Fig. 3g and Supplementary Fig. 9a). Cre-dependent helper viruses (a mixture of AAV-EFlα-DIO-TVA-mCherry and AAV- EFlα-DIO-RVG) were infused into the PVN, while an EnvA-pseudotyped RV-ΔG-EGFP rabies virus (RV) was injected into the same site in CRH-Cre mice 3 weeks later (Supplementary Fig. 9b). We observed EGFP-positive neurons in several regions, including the Arc (Supplementary Fig. 9c, d), that colocalized with signals from AgRP antibody (Fig. 3h, i). These results thus verified the presence of a Arc^AgRP-PVN^CRH pathway.

To functionally validate the monosynaptic Arc^AgRP-PVN^CRH pathway, we combined optogenetic stimulation with whole-cell patch-clamp recordings (Fig. 3j, n). In brain slices, photostimulation (473 nm) of ChR2 AgRP axon terminals within the PVN reliably evoked excitatory postsynaptic currents (EPSCs) and inhibitory postsynaptic currents (IPSCs) in recorded CRH neurons. The monosynaptic nature of this connection was confirmed by applying TTX, which blocked synaptic transmission, and subsequent co-application of the potassium channel blocker 4-AP, which restored the light-evoked EPSCs or IPSCs by bypassing action potential dependence (Fig. 3k, l, o, p). To further explore the cellular basis of this mixed synaptic transmission, we performed immunofluorescence labeling for neurotransmitter markers. A subset of Arc^AgRP neurons exhibited co-expression of both glutamatergic (Glutamate) and GABAergic (GABA) markers (Supplementary Fig. 10a–f), suggesting potential neurotransmitter co-release. This anatomical heterogeneity provides a structural basis for the observed coexistence of excitatory and inhibitory synaptic responses in PVN^CRH neurons.

Finally, optogenetic stimulation of AgRP terminals (Fig. 3q) and chemogenetic activation of Arc^AgRP neurons (Supplementary Fig. 11a–c) both increased PVN^CRH neuronal activity measured by whole-cell patch-clamp recording (Fig. 3m) or fiber photometry (Fig. 3r–t and Supplementary Fig. 11d, e), indicating that the net effect of Arc^AgRP input results in functional recruitment of PVN^CRH neurons in vivo.

Collectively, these data demonstrate that Arc^AgRP neurons form a monosynaptic but functionally heterogeneous projection to PVN^CRH neurons, comprising both excitatory and inhibitory components, and that the net circuit output results in state-dependent recruitment of PVN^CRH activity rather than a purely unidirectional synaptic effect.

Arc^AgRP-PVN signaling links HFD-associated anxiety-like behavior to hyperphagia

Whole-cell patch-clamp recordings in brain slices revealed that the excitability of the Arc^AgRP-PVN^CRH pathway was significantly increased in HFD-anxiety-susceptible mice compared to CD mice, while no significant change was observed in HFD-anxiety-unsusceptible mice (Fig. 4a-c). In vivo fiber photometry further demonstrated that the activity of the Arc^AgRP-PVN^CRH pathway was markedly elevated in HFD-anxiety-susceptible mice under both baseline conditions and during feeding episodes (Fig. 4d–h and Supplementary Fig. 12a–c). Notably, feeding-locked analysis revealed that pathway activity was dynamically increased around feeding initiation and returned toward baseline after meal termination, with a significantly greater amplitude in HFD-anxiety-susceptible mice compared with CD mice (Fig. 4f–h). These results indicate that chronic HFD exposure selectively enhances the behavioral recruitment of the Arc^AgRP-PVN^CRH pathway during feeding in anxiety-susceptible animals.

Fig. 4 — a Representative Arc^AgRP-PVN projection action potentials in each group. Scale bars, 80 ms, 20 mV. b, c Quantification of action potential properties (n = 12, 13, and 11 neurons from 3 mice in the CD, HFD-anxiety-unsusceptible, and HFD-anxiety-susceptible groups, respectively; 20 pA: F _{(2, 33)} = 13.46, CD vs. HFD-anxiety susceptible: P < 0.001, HFD-anxiety unsusceptible vs. HFD-anxiety susceptible: P = 0.001; 65 pA: F _{(2, 33)} = 9.504, P = 0.001 and 0.003 for the same comparisons; Aps: F _{(2, 462)} = 129.9, P < 0.001). Viral strategy (d) and representative AgRP-GCaMP6m expression in the Arc (e). Scale bars, 100 µm (left), 10 µm (right). Arc^AgRP-PVN-GCaMP6m signals at feeding onset (f) and offset (g) in each group. h Fluorescence changes before, during, and after feeding within (F _(2,16) = 45.52; both P < 0.001) and among groups (F _(2,24) = 15.71; both P < 0.001) (n = 9 trials from 3 mice). Chemogenetic inhibition strategy (i, j), PVN cannula site (k; scale bar, 100 µm), and representative Arc^AgRP-PVN projection neuron action potentials in acute slices (l; scale bars, 10 s, 20 mV). Representative EPM/OFT trajectories (m), EPM open-arm time (n; t ₍₁₈₎ = 4.828, P < 0.001), OFT center time (o; t ₍₁₈₎ = 4.519, P < 0.001), and OFT total distance (p) (n = 10). Hourly (q; F _{(1, 18)} = 13.18, P = 0.03, 0.003, 0.01 for HFD + mCherry vs. HFD + hM4Di at 1–3 h) and cumulative (r; F _{(1, 18)} = 19.43, all P < 0.001 for HFD + mCherry vs. HFD + hM4Di at 2–4 h) food intake after chemogenetic inhibition (n = 10). Statistics: one-way ANOVA with Bonferroni correction (b, h, between-group), one-way repeated-measures ANOVA with Bonferroni correction (h, within-group), two-tailed unpaired Student’s t test (n–p) and two-way repeated-measures ANOVA with Bonferroni correction (c, q, r). Data are mean ± SEM. AgRP, agouti-related protein; Arc, arcuate nucleus; Aps, action potentials; CD, chow diet; EPM, elevated plus maze; HFD, high-fat diet; OFT, open-field tests. Created in BioRender. Peng, W. (2026) https://BioRender.com/fsk2rp7 (d, i, j). Source data are provided as a Source Data file.

To further determine the causal contribution of the Arc^AgRP-PVN^CRH pathway in HFD-associated anxiety and hyperphagia, we employed a projection-specific chemogenetic strategy. Retrograde AAV-DIO-FLP was injected into the PVN, and AAV-fDIO-hM4Di-mCherry was delivered into the Arc of HFD-fed AgRP-Cre mice to selectively inhibit Arc^AgRP neurons projecting to the PVN (Supplementary Fig. 12d–f). This circuit-specific manipulation significantly alleviated HFD-associated anxiety-like behaviors (Supplementary Fig. 12g–j), serum corticosterone levels (Supplementary Fig. 12k), and reduced food intake (Supplementary Fig. 12l, m), indicating a functional requirement of this defined projection in the expression of the HFD-associated behavioral phenotype.

Importantly, to minimize potential confounds arising from collateral projections or off-target systemic effects, we further implemented terminal-level manipulation strategy (Fig. 4i). In this approach, AAV-DIO-hM4Di-mCherry was injected into the Arc, and guide cannulas were implanted above the PVN of HFD-fed AgRP-Cre mice to allow localized pharmacological inhibition of Arc^AgRP axon terminals via site-specific CNO administration (Fig. 4j–l). This terminal-specific inhibition of the Arc-PVN projection recapitulated the effects of somatic inhibition, leading to significant reductions in both anxiety-like behaviors (Fig. 4m–p) and food intake (Fig. 4q, r) in HFD-fed mice.

Together, these complementary strategies-projection-specific somatic inhibition and PVN-restricted terminal inhibition-provide convergent evidence that the Arc^AgRP-PVN^CRH pathway is functionally required for the expression of HFD-associated anxiety-like behavior and hyperphagia under chronic HFD conditions.

Conversely, selective activation of this Arc^AgRP-PVN^CRH pathway in AgRP-Cre mice using the same projection-specific strategy (Supplementary Fig. 13a–c) induced anxiety-like behaviors (Supplementary Fig. 13d–g) and increased food intake (Supplementary Fig. 13h, i), phenocopying key aspects of the HFD-induced state. These findings suggest that recruitment of this defined projection is sufficient to drive both anxiety-associated behavioral changes and feeding enhancement under controlled conditions. At the level of downstream circuitry, immunofluorescence staining demonstrated significant activation of PVN^CRH neurons in HFD-anxiety-susceptible mice (Supplementary Fig. 14a–d). Chemogenetic inhibition of PVN^CRH neurons (Supplementary Fig. 14e–g) alleviated HFD-induced anxiety-like behaviors (Supplementary Fig. 14h–k) and reduced food intake (Supplementary Fig. 14l, m), whereas activation of these neurons induced anxiety-like behaviors and increased feeding (Supplementary Fig. 15a–i), recapitulating the HFD-associated phenotype. Collectively, these results demonstrate that the Arc^AgRP-PVN^CRH pathway is both necessary and sufficient for HFD-associated anxiety-like behavior and anxiety-associated hyperphagia, and is dynamically recruited under chronic HFD conditions in a state-dependent manner.

PVN^CRH neurons inhibit feeding-suppressive LHA^Glu neurons

We next investigated potential downstream targets of the Arc^AgRP-PVN^CRH pathway that might mediate this response. Following the injection of AAV-DIO-mCherry into the PVN of CRH-Cre mice (Fig. 5a, b and Supplementary Fig. 16a, b), abundant mCherry-positive fibers were clearly observed in many brain regions, including the LHA (Fig. 5c and Supplementary Fig. 16c), a region critically involved in feeding behavior. The PVN-LHA projection was confirmed using an anterograde monosynaptic tracing strategy, by co-injecting AAV-hSyn-Cre-EGFP into the PVN (Supplementary Fig. 16d) and AAV-DIO-mCherry into the LHA (Fig. 5d). Three weeks post-injection, the vast majority of mCherry-labeled neurons in the LHA were immune-positive for glutamate (Fig. 5e, f). Furthermore, monosynaptic retrograde tracing using a Cre-dependent helper virus and RV (rabies virus) injected into the LHA of VgluT2-Cre mice (Fig. 5g and Supplementary Fig. 16e–h) resulted in the detection of EGFP⁺ neurons in the PVN that exhibited colocalization with CRH immunoreactivity (Fig. 5h, i).

Fig. 5 — a Schematic of viral injection. b Representative AAV-DIO-mCherry expression and CRH colocalization in the PVN (right). Scale bars, 100 µm (left), 10 µm (right). c Representative image of mCherry⁺ fibre expression in the LHA. Scale bar, 200 µm. d Schematic of viral injection. e Representative LHA injection site with mCherry/Glu colocalization. Scale bars, 100 µm (left), 20 µm (right). (f) Quantification showing that 93% of mCherry⁺ cells were Glu⁺ neurons. (n = 9 slices from 3 mice). g Schematic of viral injection. h Representative RV-EGFP/CRH colocalization in the PVN. Scale bars, 100 µm (left), 20 µm (right). i Quantification showing that 86% of EGFP⁺ cells were CRH⁺ neurons. (n = 6 slices from 3 mice). Optogenetic electrophysiological verification of the PVN^CRH-LHA pathway: experimental configuration (j), representative light-evoked currents before and after TTX or 4-AP (k; scale bars, 20 ms, 50 pA), and quantification (l; n = 6 neurons from 3 mice per group, ACSF vs. TTX: t ₍₅₎ = 10.65, P < 0.001; TTX vs. 4-AP: t ₍₅₎ = 7.775, P < 0.001). Fiber photometry during optogenetic activation of the PVN^CRH-LHA projection: recording strategy (m), representative LHA-Glu^GCaMP6m traces (n), trial heatmaps aligned to photostimulation onset (o; 473 nm, 1 s), and averaged ΔF/F (p; n = 9 trials from 3 mice per group, t ₍₁₆₎ = 7.463, P < 0.001). Statistical significance was determined by two-tailed paired Student’s t test (l) and two-tailed unpaired Student’s t test (p). Data are shown as the mean ± SEM. 4-AP, 4-aminopyridine; ACSF, artificial cerebrospinal fluid; CRH, corticotropin releasing hormone; Glu, glutamate; LHA, lateral hypothalamic area; oEPSC, optically evoked excitatory postsynaptic current; PVN, paraventricular nucleus of hypothalamus; TTX, tetrodotoxin. Created in BioRender. Peng, W. (2026) https://BioRender.com/fsk2rp7 (a, d, g, m) and Microsoft PowerPoint (j). Source data are provided as a Source Data file.

To functionally validate the monosynaptic PVN^CRH-LHA^Glu pathway, we optogenetically stimulated (473 nm) the axon terminals of CRH-expressing neurons within the LHA in brain slices (Fig. 5j). This reliably elicited EPSCs in recorded Glu neurons. The monosynaptic nature of this connection was confirmed by application of TTX followed by the potassium channel blocker 4-AP (Fig. 5k, l). To further investigate the impact of PVN^CRH activation on LHA^Glu neurons, we combined fiber photometry with either optogenetic (Fig. 5m) or chemogenetic (Supplementary Fig. 17a–c) activation for additional validation. The results demonstrated that specific optogenetic or chemogenetic activation of the PVN^CRH-LHA^Glu pathway significantly reduces the activity of LHA^Glu neurons (Fig. 5n–p and Supplementary Fig. 17d, e), further confirming the inhibitory nature of this projection. These results together indicate the presence of a PVN^CRH-LHA^Glu pathway.

PVN^CRH-LHA signaling links HFD-associated anxiety-like behavior to hyperphagia

In vivo fiber photometry recordings (Fig. 6a, b and Supplementary Fig. 18a) revealed that the excitability of the PVN^CRH-LHA^Glu pathway was significantly increased in HFD-anxiety-susceptible mice compared to CD controls, whereas no significant changes were observed in HFD-anxiety-unsusceptible mice (Supplementary Fig. 18b, c). To further characterize the dynamic recruitment of this pathway during feeding behavior, we additionally performed feeding-locked fiber photometry analysis. PVN^CRH-LHA^Glu pathway activity was rapidly elevated around feeding initiation and returned toward baseline following meal termination, with a markedly greater response amplitude observed in HFD-anxiety-susceptible mice (Fig. 6c–e). These results suggest that chronic HFD exposure enhances the behavioral coupling between anxiety-associated state and feeding-related recruitment of the PVN^CRH-LHA^Glu pathway.

To further determine the causal role of this pathway in HFD-associated anxiety and feeding alterations, we specifically expressed chemogenetic inhibitory viruses in the PVN-LHA projection to selectively suppress circuit activity (Supplementary Fig. 18d–f). This manipulation significantly alleviated HFD-associated anxiety-like behaviors (Supplementary Fig. 18g–j), serum corticosterone levels (Supplementary Fig. 18k), and reduced food intake (Supplementary Fig. 18l, m), indicating that the PVN^CRH-LHA^Glu pathway is functionally required for the expression of the HFD-associated behavioral phenotype. Importantly, to minimize potential confounds arising from collateral projections or off-target systemic effects, we further implemented a terminal-level manipulation strategy (Fig. 6f). In this approach, AAV-DIO-hM4Di-mCherry was injected into the PVN, and guide cannulas were implanted above the LHA of HFD-fed CRH-Cre mice to allow localized pharmacological inhibition of PVN^CRH axon terminals via site-specific CNO administration (Fig. 6g–i). This terminal-specific inhibition of the PVN-LHA projection recapitulated the effects of somatic inhibition, leading to significant reductions in both anxiety-like behaviors (Fig. 6j–m) and food intake (Fig. 6n, o) in HFD-fed mice.

Similarly, selective activation of this pathway in CRH-Cre mice using the same projection-specific strategy (Supplementary Fig. 19a–c) induced anxiety-like behaviors (Supplementary Fig. 19d–g) and increased food intake (Supplementary Fig. 19h, i), recapitulating key features of the HFD-associated state.

Notably, immunofluorescence analysis performed after feeding revealed that LHA^Glu neurons exhibited reduced c-Fos expression in HFD-anxiety-susceptible mice compared with controls (Supplementary Fig. 20a–d), suggesting state-dependent suppression rather than baseline floor effects. Chemogenetic activation of LHA^Glu neurons (Supplementary Fig. 20e–g) reduced HFD-induced food intake without affecting anxiety-like behavior (Supplementary Fig. 20h–m), whereas their inhibition increased feeding without altering anxiety-like behavior (Supplementary Fig. 21a–i), suggesting that LHA^Glu neurons selectively regulate the feeding component of the behavioral phenotype.

To further clarify the molecular mechanism underlying this dissociation, we examined CRHR signaling in LHA^Glu neurons. Immunofluorescence staining demonstrated that CRHR2, but not CRHR1, is specifically expressed in LHA^Glu neurons (Supplementary Fig. 21j–m). Based on this observation, we microinjected the CRHR2 antagonist Astressin 2B into the LHA and found that inhibition of CRHR2 signaling significantly reduced HFD-induced hyperphagia without affecting anxiety-like behaviors (Supplementary Fig. 21n–s). To further confirm the cell-type-specific role of CRHR2, we selectively knocked down CRHR2 in LHA^Glu neurons using a Cre-dependent shRNA approach in VgluT2-Cre mice (Supplementary Fig. 22a–e). Consistent with pharmacological results, CRHR2 knockdown reduced HFD-induced excessive feeding while leaving anxiety-like behaviors unchanged (Supplementary Fig. 22f–j).

Together, these findings indicate that the PVN^CRH-LHA^Glu pathway is dynamically recruited under chronic HFD conditions and contributes to both anxiety-like behavior and feeding regulation. Importantly, within this circuit, LHA^Glu neurons selectively mediate the feeding component without directly governing anxiety-related behavior, supporting a model in which metabolic and emotional states are differentially distributed across downstream hypothalamic nodes within a shared circuit framework.

A long-range Arc-PVN-LHA circuit mediates state-dependent coupling between anxiety and feeding

To better visualize the connectivity within the Arc-PVN-LHA circuit, we employed a triple retrograde tracing strategy (Fig. 7a and Supplementary Fig. 23a). A retrograde AAV-hSyn-Cre was injected into the LHA, while a Cre-dependent helper virus was co-injected into the PVN of C57 mice. Three weeks later, RV-ΔG-EGFP was injected into the same PVN site (Fig. 7b), resulting in the observation of EGFP⁺ neurons in the Arc and co-localization with AgRP immunoreactivity (Fig. 7c, d and Supplementary Fig. 23b, c). In a complementary approach, AAV-hSyn-Cre-mCherry (Supplementary Fig. 23d) was injected into the Arc and Retro-DIO-FLP was injected into the LHA, followed by injection of AAV-fDIO-EGFP into the PVN (Fig. 7e). This strategy further demonstrated the Arc-PVN-LHA connection by revealing EGFP signals in the PVN that co-localized with CRH immunoreactivity (Fig. 7f). In addition, we injected AAV-DIO-ChR2-mCherry into the Arc and a retrograde AAV-CRH-EGFP into the LHA of AgRP-Cre mice (Fig. 7g). In brain slices prepared from these mice, optogenetic stimulation (473 nm) of axon terminals in the PVN reliably elicited EPSCs in recorded CRH neurons. This connection was confirmed as monosynaptic (Fig. 7h, i). Furthermore, combined fiber photometry and optogenetic (Fig. 7j)/chemogenetic (Supplementary Fig. 23e–g) showed that Arc-PVN projection activation significantly suppressed LHA^Glu neuron activity (Fig. 7k–m and Supplementary Fig. 23h, i), confirming the functional connectivity of this triple circuit. These data thus demonstrated a direct connection in the Arc^AgRP-PVN^CRH-LHA^Glu circuit.

We next investigated the role of the Arc-PVN-LHA circuit in HFD-induced anxiety-associated hyperphagia. To monitor the dynamic activity of this long-range circuit, we injected AAV-hSyn-Cre into the Arc, Retro-DIO-FLP into the LHA, and AAV-fDIO-GCaMP6m into the PVN of C57 mice to selectively express a calcium indicator within this defined pathway (Fig. 8a, b and Supplementary Fig. 24a). In vivo fiber photometry recordings revealed that, beyond elevated baseline activity (Supplementary Fig. 24b, c), this circuit also exhibited robust feeding-locked dynamics. Specifically, circuit activity was rapidly increased around feeding initiation and returned to baseline following meal termination, with significantly greater amplitude observed in HFD-anxiety-susceptible mice compared with both CD and HFD-anxiety-unsusceptible groups (Fig. 8c–e). These results indicate that chronic HFD exposure enhances the behavioral coupling between anxiety-associated state and feeding-related recruitment of the Arc-PVN-LHA circuit.

Fig. 8 — Viral strategy targeting the Arc-PVN-LHA pathway (a) and representative CRH-GCaMP6m expression in the PVN (b; scale bars, 100 µm and 20 µm). Arc-PVN-LHA GCaMP6m signals at feeding onset (c) and offset (d) in each group. e Fluorescence changes before, during, and after feeding within (F _(2,16) = 77.77; both P < 0.001 for during vs. pre and post) and among groups (F _(2,24) = 28.93; CD vs. HFD-anxiety-susceptible: P < 0.001, HFD-anxiety-unsusceptible vs. HFD-anxiety-susceptible: P < 0.001) (n = 9 trials from 3 mice). Chemogenetic inhibition strategy (f, g) and representative LHA cannula site (h; scale bar, 200 µm). Representative EPM/OFT trajectories (i), EPM open-arm time (j; t ₍₁₈₎ = 4.221, P < 0.001), OFT center time (k; t ₍₁₈₎ = 4.378, P < 0.001), and OFT total distance (l) in HFD+mCherry and HFD+hM4Di mice (n = 10). **m-n** Hourly (m; F _{(1, 18)} = 18.66, P = 0.04, 0.007, and 0.02 for HFD+mCherry vs. HFD+hM4Di at 1–3 h) and cumulative (n; F _{(1, 18)} = 21.13, all P < 0.001 for HFD+mCherry vs. HFD+hM4Di at 2–4 h) food intake after chemogenetic inhibition of the Arc-PVN-LHA pathway (n = 10). Statistical significance was determined by one-way repeated-measures ANOVA with Bonferroni correction (e, within-group), one-way ANOVA with Bonferroni correction (e, between-group), two-tailed unpaired Student’s t test (j–l), and two-way repeated-measures ANOVA with Bonferroni correction (m, n). Data are shown as the mean ± SEM. ANOVA, analysis of variance; Arc, arcuate nucleus; CD, chow diet; CNO, clozapine N-oxide; CRH, corticotropin-releasing hormone; EPM, elevated plus maze; HFD, high-fat diet; LHA, lateral hypothalamic area; OFT, open-field tests; PVN, paraventricular nucleus of hypothalamus. Created in BioRender. Peng, W. (2026) https://BioRender.com/fsk2rp7 (a, f, g). Source data are provided as a Source Data file.

To determine the causal contribution of this circuit to HFD-associated behavioral alterations, we employed a projection-defined chemogenetic strategy to selectively inhibit the Arc-PVN-LHA pathway (Supplementary Fig. 24d–f). This manipulation significantly alleviated HFD-induced anxiety-like behaviors (Supplementary Fig. 24g–j), serum corticosterone levels (Supplementary Fig. 24k), and reduced food intake (Supplementary Fig. 24l, m), indicating that this long-range circuit is functionally required for the expression of the HFD-associated phenotype.

Importantly, to minimize potential confounds arising from collateral projections, we further performed complementary terminal-level manipulation by implanting guide cannulas into the LHA and locally administering CNO to selectively inhibit projection terminals within this region (Fig. 8f, h). This localized inhibition of the Arc-PVN-LHA circuit recapitulated the effects of somatic manipulation, leading to significant reductions in both anxiety-like behaviors (Fig. 8i–l) and food intake (Fig. 8m, n), thereby strengthening the circuit specificity of the observed effects.

Conversely, selective activation of this circuit in C57 mice using the same projection-specific strategy (Supplementary Fig. 25a–c) was sufficient to induce anxiety-like behaviors (Supplementary Fig. 25d–g) and increase food intake (Supplementary Fig. 25h, i), recapitulating key features of the HFD-associated state.

Collectively, these results identify a long-range Arc-PVN-LHA circuit that is dynamically recruited under chronic HFD conditions and functionally contributes to anxiety-associated hyperphagia through state-dependent coupling of metabolic and emotional processes, thereby promoting the maintenance of obesity.

Discussion

Substantial efforts have been devoted to understanding how hypothalamic neurons regulate obesity, feeding behavior, and emotional comorbidities^42–45. However, how established hypothalamic feeding and stress-related circuits are dynamically linked to modulate food intake specifically during anxiety-like states remains incompletely understood. An important point clarified by the present study is that our findings should be interpreted in the context of previously established hypothalamic circuit functions. AgRP neurons are canonical feeding-promoting neurons, and the Arc^AgRP-PVN pathway has been shown to contribute to hunger-related feeding control^14,15,20. Similarly, PVN^CRH neurons are well recognized for their role in stress-related behavioral and endocrine responses³¹, and LHA^Glu neurons have been reported to negatively regulate feeding and reward-related consumption³⁵. Therefore, the main advance of our study is not the isolated identification of these neuronal populations or projections, but the demonstration that these known circuit elements are functionally coupled under HFD-induced anxiety conditions to produce anxiety-associated hyperphagia.

In the present study, we demonstrate a key context-dependent feature of AgRP neurons in regulating feeding and anxiety-related behaviors. In chow-fed mice, chemogenetic activation of AgRP neurons robustly increased food intake without affecting anxiety-like behaviors, whereas in HFD-fed mice, inhibition of AgRP neurons reduced both hyperphagia and anxiety-like behaviors. This dissociation indicates that AgRP neurons primarily drive feeding under normal metabolic conditions, consistent with their classical orexigenic role^15,20,21, but acquire the capacity to influence anxiety-related behaviors only under chronic HFD exposure. Importantly, this context-dependent function is not mediated uniformly at the population level but is instead implemented through projection-specific circuit segregation. We identified an Arc^AgRP-PVN^CRH pathway that is selectively engaged in HFD-fed mice exhibiting anxiety-susceptible phenotypes. Chemogenetic inhibition of this pathway significantly alleviated both anxiety-like behavior and anxiety-driven feeding, indicating that it represents a key substrate linking metabolic perturbation to affective behavioral output. This projection-specific mechanism provides a circuit-level explanation for how AgRP neurons acquire anxiety-modulating capacity only under pathological metabolic conditions.

Building on this, PVN^CRH neurons emerge as a critical integrative node. As major regulators of the hypothalamic-pituitary-adrenal axis³², PVN^CRH neurons are highly sensitive to metabolic and stress-related signals. Consistent with previous reports that HFD induces HPA axis hyperactivity⁴⁶, we observed elevated serum corticosterone levels and enhanced PVN^CRH-related circuit activity. Given their dense input from AgRP neurons and broad projections to downstream emotional regulation centers such as the LHA, PVN^CRH neurons are well positioned to integrate metabolic state and stress signals. Within this framework, we propose a sequential model in which chronic HFD exposure first induces metabolic and neuroendocrine “priming” of PVN^CRH and downstream LHA glutamatergic neurons, thereby sensitizing this axis to AgRP input. Under this primed state, AgRP neuron activity is no longer restricted to feeding regulation but is capable of engaging stress-related circuits to drive anxiety-associated feeding. In contrast, under normal metabolic conditions, AgRP neuron activity remains confined to feeding regulation without affecting anxiety, reflecting a circuit that is functionally gated by metabolic state. Together, these findings refine the functional framework of AgRP neurons by demonstrating that their influence on emotional feeding is not intrinsic or constitutive, but emerges only when downstream hypothalamic stress circuits are metabolically and neuroendocrinologically primed by chronic HFD exposure.

One of the most unexpected findings of the present study is that optogenetic stimulation of Arc^AgRP terminals evoked monosynaptic excitatory responses in PVN^CRH neurons. At first glance, this observation appears inconsistent with the classical view that AgRP neurons are predominantly GABAergic and inhibit downstream targets to promote feeding behavior. Indeed, extensive studies have established that AgRP neurons release GABA, neuropeptide Y (NPY), and AgRP peptide, and that inhibition of PVN neurons constitutes a major mechanism through which AgRP neurons regulate energy homeostasis^15,28,29,47. However, accumulating evidence suggests that AgRP neurons are not a homogeneous neuronal population. Single-cell transcriptomic analyses have revealed substantial molecular diversity among arcuate AgRP neurons, including distinct neurotransmitter-related gene expression profiles and projection-specific subpopulations¹⁶. Consistent with this concept, our immunofluorescence analysis demonstrated that while all traced AgRP neurons expressed GABAergic markers, approximately 30% additionally expressed glutamatergic markers, suggesting the existence of a previously underappreciated AgRP subpopulation with the potential for GABA/glutamate co-transmission. Neurotransmitter co-release has increasingly emerged as a common organizational feature of neural circuits involved in motivation, stress, and feeding. Several neuronal populations previously considered purely inhibitory or excitatory have subsequently been shown to co-release multiple neurotransmitters depending on projection target and physiological state. Accordingly, it is plausible that a subset of AgRP neurons projecting to PVN^CRH neurons may possess mixed neurotransmitter phenotypes, allowing both inhibitory and excitatory synaptic transmission. Consistent with this possibility, our electrophysiological recordings detected not only monosynaptic oEPSCs but also monosynaptic oIPSCs in PVN^CRH neurons following stimulation of AgRP terminals, suggesting that both forms of transmission coexist within this pathway. Importantly, the functional impact of this projection cannot be inferred solely from single-synapse polarity. Recent studies indicate that AgRP-derived signaling can produce net excitation through circuit-level mechanisms, including local disinhibition. For example, AgRP neurons regulate stress-related behaviors by suppressing inhibitory BNST inputs to the PVN, thereby increasing PVN^CRH neuronal activity despite GABA release at the synaptic level⁴⁸. In addition, under conditions of altered intracellular chloride homeostasis, GABAergic transmission may become depolarizing rather than inhibitory⁴⁹. Thus, the net excitation of PVN^CRH neurons observed in our study likely reflects a complex integration of multiple, potentially overlapping mechanisms at both synaptic and circuit levels. Notably, our observations are not entirely isolated from existing literature. While many studies report inhibitory regulation of PVN neurons by AgRP neurons, several recent investigations have documented activation of PVN^CRH neurons by AgRP-related pathways under specific physiological and HFD-associated state conditions, suggesting that the functional relationship between these neuronal populations may be more complex than previously appreciated^48,50. Therefore, rather than contradicting the established framework, our findings extend it by revealing a previously underappreciated layer of functional heterogeneity within the AgRP-PVN network.

We next sought to determine how anxiety-related signals from the PVN ultimately drive feeding output. We focused on the LHA because it is a crucial hypothalamic center for feeding regulation⁵¹ and receives dense inputs from the PVN. The established anorexigenic role of LHA^Glu neurons initially appears to raise a conceptual paradox in the interpretation of our findings. Previous studies have shown that activation of LHA glutamatergic neurons suppresses feeding, whereas genetic ablation or inhibition of these neurons increases caloric intake, supporting the view that LHA^Glu neurons act as a feeding-suppressive population³⁵. Our data do not contradict this model. Instead, they suggest that the functional output of LHA^Glu neurons depends on the direction and context of upstream synaptic control. Under basal or physiological feeding conditions, LHA^Glu neurons may be primarily driven by excitatory inputs, leading to increased LHA^Glu activity and suppression of food intake. However, in the chronic HFD-induced anxiety-susceptible state, PVN^CRH input to LHA^Glu neurons appears to be predominantly inhibitory. This inhibitory input would be expected to decrease the activity of LHA^Glu neurons and thereby attenuate their feeding-suppressive function. Thus, the observed increase in feeding can be interpreted as a consequence of disinhibition of food intake through suppression of an anorexigenic LHA^Glu population, rather than as evidence that LHA^Glu neurons intrinsically promote feeding. This interpretation reconciles our findings with previous work by suggesting that PVN^CRH-LHA^Glu signaling promotes anxiety-associated hyperphagia under HFD conditions by inhibiting, rather than activating, the canonical feeding-suppressive function of LHA^Glu neurons.

CRHR2 may provide an important molecular mechanism through which PVN^CRH-related anxiety signals are translated into altered LHA-dependent feeding output. CRHR2 is widely distributed in hypothalamic regions and participates in the regulation of stress adaptation, energy metabolism, and appetite³⁷. In the present study, we found that CRHR2, but not CRHR1, was expressed in LHA^Glu neurons. Pharmacological blockade of CRHR2 signaling in the LHA alleviated HFD-induced hyperphagia without affecting anxiety-like behavior. Furthermore, cell-type-specific viral knockdown of CRHR2 in LHA^Glu neurons produced a similar behavioral effect, reducing HFD-induced excessive feeding while leaving anxiety-like behavior largely unchanged. These results indicate that LHA-CRHR2 signaling does not directly drive the anxiety-like state itself, but instead acts as a downstream effector mechanism that links anxiety-related hypothalamic activity to excessive feeding behavior. In contrast, inhibition of the upstream Arc^AgRP-PVN^CRH-LHA^Glu pathway ameliorated both anxiety-like behavior and anxiety-driven feeding, supporting the idea that the upstream circuit conveys anxiety-related information, whereas LHA-CRHR2 signaling participates more selectively in the feeding-output component of this process.

Accordingly, we interpret the Arc^AgRP-PVN^CRH-LHA^Glu pathway not as an entirely unexplored circuit, but as a disease-context-specific functional coupling of established hypothalamic feeding and stress circuits. This interpretation more accurately delineates what was previously known and what our study adds: chronic HFD exposure links AgRP-related metabolic signaling, PVN^CRH stress circuitry, and LHA^Glu feeding-control mechanisms to sustain an obesity-anxiety functional coupling. Importantly, our additional pharmacological experiment provides functional insight into the interaction between affective state and feeding behavior under HFD conditions. We found that pharmacological attenuation of anxiety using midazolam in HFD-fed anxiety-susceptible mice resulted in a significant reduction in food intake, indicating that anxiety-related states can contribute to the regulation of feeding behavior in the context of metabolic stress. However, we interpret this effect as supporting a modulatory interaction rather than a fixed causal hierarchy between obesity and anxiety-like behavior, as it suggests that affective state can bias feeding output when metabolic circuits are challenged by chronic HFD exposure, without providing evidence for a closed or self-reinforcing loop between the two processes. Taken together, these findings are consistent with a model in which metabolic and emotional states interact through shared hypothalamic circuitry in a state-dependent manner, whereby anxiety-related signals selectively modulate feeding behavior under conditions of metabolic perturbation. Within this framework, a key contribution of our study is the identification of how established hypothalamic circuit nodes are dynamically recruited under HFD-induced anxiety conditions, and how CRHR2 signaling in LHA^Glu neurons specifically contributes to the feeding component of anxiety-associated hyperphagia, thereby refining the circuit-level mechanism underlying this behavioral coupling.

Several limitations of the present study should be considered. First, all experiments were performed in male mice. Although this design reduced biological variability related to the estrous cycle, it limits the extent to which our findings can be generalized across sexes. Sex is now widely recognized as a key biological variable in preclinical neuroscience research, and sex-dependent differences may influence stress responsiveness, anxiety-like behavior, metabolic regulation, and hypothalamic circuit function^52,53. Therefore, future studies including female mice are needed to determine whether the Arc^AgRP-PVN^CRH-LHA^Glu circuit identified here is conserved across sexes or modulated by sex-specific hormonal and neural mechanisms. Second, circadian regulation may influence the interpretation of feeding-related outcomes. Feeding behavior is strongly organized by the circadian system, and hypothalamic clocks contribute to the temporal regulation of food intake and metabolic state⁵⁴. Moreover, high-fat diet exposure can disrupt behavioral and molecular circadian rhythms, while restricting food access to specific circadian windows can markedly alter metabolic consequences under HFD conditions^55,56. In the present study, mice were maintained under a standardized 12-h light/dark cycle, and behavioral testing, pharmacological or chemogenetic manipulation, blood collection, and food-intake measurements were performed within consistent time windows across experimental groups to minimize circadian variability. However, we did not directly assess circadian phase, clock-gene expression, or day-night feeding patterns. Thus, future studies incorporating time-resolved feeding analysis and circadian-phase-specific neural recordings will be important to determine how circadian timing interacts with the Arc^AgRP-PVN^CRH-LHA^Glu circuit during HFD-induced anxiety-associated hyperphagia. Third, while our circuit mapping integrates multiple anatomical and functional approaches, we acknowledge that AAV2/1-hSyn-Cre-mediated anterograde transsynaptic labeling has inherent methodological limitations. In particular, this approach may be influenced by viral spread beyond the injection site, non-specific uptake, and activity-dependent labeling biases, and therefore cannot be interpreted as definitive evidence of monosynaptic connectivity when used in isolation. Accordingly, in this study we did not rely on AAV2/1-based tracing as a standalone proof of synaptic connections. Instead, it was used as a complementary anatomical screening tool, and all circuit conclusions were independently validated using monosynaptic rabies tracing, optogenetic-assisted electrophysiology under TTX/4-AP conditions, and convergent functional manipulations. This multi-layered validation strategy minimizes potential confounds associated with AAV2/1-based labeling and strengthens the robustness of our circuit-level conclusions. Fourth, although our electrophysiological and anatomical data support the existence of both excitatory and inhibitory monosynaptic components within the Arc^AgRP-PVN^CRH pathway, the precise molecular mechanisms underlying these responses remain to be determined. Future studies combining projection-specific transcriptomics, spatial transcriptomics, neurotransmitter sensor imaging, and electron microscopy will be required to definitively establish whether individual AgRP neurons co-release glutamate and GABA at the same synapse or whether distinct AgRP subpopulations mediate the observed bidirectional transmission. Such investigations will further clarify how AgRP neurons dynamically regulate stress and feeding circuits under physiological and HFD-associated state conditions. Fifth, we acknowledge that emotional feeding behavior may not exclusively represent a pathological process, but can also serve as an adaptive or compensatory response to acute or chronic stress. In the present study, we primarily focused on the circuit mechanisms underlying anxiety-associated feeding under high-fat diet conditions, and did not directly assess whether suppression of feeding behavior is accompanied by alterations in underlying emotional state or stress resilience. Therefore, it remains unclear whether modulation of the Arc-PVN-LHA circuit leads to long-term improvements in affective behavior or whether compensatory emotional adaptations may emerge. Future studies incorporating chronic manipulation paradigms and longitudinal behavioral assessments will be necessary to determine the sustained emotional and metabolic consequences of targeting this circuit.

In summary, this study identifies a disease-context-specific mechanism by which chronic HFD exposure functionally couples hypothalamic feeding and stress circuits to promote anxiety-associated hyperphagia. Rather than representing a novel linear pathway, our findings reveal that established hypothalamic nodes-including Arc^AgRP, PVN^CRH, and LHA^Glu neurons-are dynamically integrated under metabolic stress conditions to generate coordinated behavioral outputs. Within this framework, AgRP neurons primarily regulate feeding under normal metabolic states, whereas chronic HFD exposure induces functional reorganization of downstream PVN^CRH and LHA circuits, enabling stress-related signals to exert a stronger influence on feeding behavior. CRHR2 signaling in LHA^Glu neurons further refines this process by selectively contributing to the feeding component of anxiety-associated behavioral responses, without directly governing anxiety-like states (Fig. 9).

Fig. 9 — Arc, arcuate nucleus; CRH, corticotropin-releasing hormone; CRHR2, corticotropin-releasing hormone receptor 2; EPM, elevated plus maze; HFD, high-fat diet; HFD-NS, HFD anxiety-unsusceptible; HFD-S, HFD anxiety-susceptible; LHA, lateral hypothalamic area; PVN, paraventricular nucleus of hypothalamus. Figure9 was created by Adobe Illustrator.

Together, these results support a model in which metabolic and emotional processes converge on shared hypothalamic circuitry in a state-dependent manner, leading to functional coupling between anxiety and feeding behavior under conditions of chronic metabolic stress.

This framework emphasizes a dynamic circuit-level interaction in which behavioral phenotypes emerge through context-dependent recruitment of pre-existing neural pathways. Given the growing clinical recognition that anxiety and depressive symptoms are associated with adverse health outcomes, understanding the biological coupling between affective and metabolic states may have broader translational relevance⁵⁷. Our findings therefore provide a conceptual basis for understanding how metabolic dysfunction and affective state interact at the circuit level, and identify potential nodes for intervention aimed at modulating this interaction in obesity-associated emotional disorders.

Methods

Animals

All procedures involving animals in this research were conducted in strict compliance with institutional animal care policies and relevant national regulations, specifically adhering to the National Standard of the People’s Republic of China (GB/T 35892-2018; Laboratory animal-Guideline for ethical review of animal welfare). The experimental protocols received formal approval from the Animal Ethics Committee of Anhui Medical University (Approval No: LLSC20211529). The mouse strains employed in this study included AgRP-Cre, VgluT2-Cre, CRH-Cre (sourced from the Jackson Laboratory) and male C57BL/6 J mice (aged 8 weeks), which were obtained from GemPharmatech Co., Ltd. (Nanjing, China). Animals were housed under controlled environmental conditions: a 12-h light/dark cycle (lights on from 8:00 a.m. to 8:00 p.m.), ambient temperature of 24 ± 2 °C, and relative humidity of 55 ± 5%. Mice were group-housed with no more than five individuals per cage and had free access to food and water throughout the study. Experiments were performed with male mice.

Experimental model

Eight-week-old male C57BL/6 J mice were randomly allocated via a computer-based system to receive either a standard chow diet (CD) or an HFD. The HFD group was subsequently administered a commercially available high-fat feed (XTHF60, Jiangsu Xietong Pharmaceutical Bio-engineering Co., Ltd.) for 12 weeks to establish a diet-induced obesity (DIO) model.

Hierarchical cluster analysis

Hierarchical cluster analysis was used to stratify HFD-fed mice into anxiety-susceptible and anxiety-unsusceptible subgroups based on individual behavioral performance in the EPM. This method was selected because hierarchical clustering is an unsupervised, distance-based approach that identifies naturally occurring similarity structures among individual cases without requiring predefined group assignments^58,59. The analysis was performed in SPSS Statistics (version 27.0, IBM SPSS) as a case-based agglomerative hierarchical clustering procedure. Each HFD-fed mouse was entered as an individual case and labeled by its sample ID, and the time spent in the open arms of the EPM was entered as the clustering variable⁶⁰.

Before clustering, open-arm residence time was standardized using Z-score transformation to minimize scale-dependent effects on distance calculation. Similarity between individual mice was quantified using squared Euclidean distance, and clusters were merged using the between-groups linkage method. The agglomeration schedule was generated to document the stepwise merging of individual cases and the corresponding linkage distance at each clustering step. A dendrogram was then used to visualize the hierarchical structure of the HFD-fed cohort and to identify the major cluster solution.

The dendrogram revealed two principal clusters among HFD-fed mice. Mice in the cluster characterized by shorter open-arm residence time were operationally defined as HFD-anxiety-susceptible mice, whereas mice in the cluster characterized by longer open-arm residence time were defined as HFD-anxiety-unsusceptible mice. This stratification was performed before subsequent behavioral, circuit, electrophysiological, and fiber photometry analyses.

Behavioral tests

For all behavioral tests, mice were transferred to the testing room and allowed to acclimate for at least 1 h before testing. Food and water were available ad libitum during the acclimation period. Behavioral experiments were conducted under controlled low-light conditions, with the illumination intensity maintained at approximately 15–20 lux at the center of the testing apparatus. After each trial, the apparatus was thoroughly cleaned with 75% ethanol and allowed to dry completely before the next animal was tested to eliminate residual olfactory cues. Each mouse was used only once in the same behavioral experiment.

Open-field test (OFT): To evaluate general locomotor activity and anxiety-like behaviors, mice underwent open-field testing. Each animal was gently placed in the center of a square arena (50 × 50 × 50 cm) and permitted to explore freely for a 5-min session. The arena was conceptually divided into a 3 × 3 grid. All trials were recorded and analyzed automatically using the Any-Maze video tracking system (Stoelting Co., USA).

EPM test: To assess anxiety-like responses in mice, the EPM paradigm was employed. Following the open field test, each animal was transferred to the EPM apparatus and positioned in the central zone facing a randomly selected open arm. The subject was then permitted to freely explore the maze for a 5-minute duration. Behavioral performance was monitored and quantified using the Any-Maze video tracking system (Stoelting Co.).

Food intake and body weight monitoring

After a minimum three-day acclimation period in individual home cages (25 × 15 × 10 cm) with ad libitum access to food and water, daily body weight and food consumption were measured for each mouse. Food was weighed using a precision electronic balance (Sartorius CPA224S), with a readability (0.001 g). To minimize measurement error caused by food spillage, mice were individually housed during the monitoring period, and both remaining food pellets and visible spilled fragments or crumbs were carefully collected and weighed at each measurement time point. Food intake was calculated as the weight of food initially provided minus the weight of recovered food, including residual pellets and spilled fragments. In experiments involving chemogenetic or pharmacological interventions, chronic treatment was initiated three days prior to testing. This included either intraperitoneal injections of Clozapine-N-oxide (CNO) administered on days 19–21 following viral injection, or continuous delivery of receptor antagonists. Baseline food intake was assessed between 9:00 and 10:00 a.m., after which measurements continued for four consecutive hours post-CNO/receptor antagonist administration. For extended monitoring, cumulative 24-h food intake and body weight were recorded at 9:00 a.m. each day, and these parameters were tracked over seven consecutive days.

Drugs

CNO (Enzo Life Sciences) was administered via intraperitoneal injection at a dose of 2 mg/kg for systemic chemogenetic manipulation. For local chemogenetic manipulation through implanted guide cannulas, CNO was dissolved in sterile artificial cerebrospinal fluid (ACSF) to a final concentration of 1 mM and bilaterally microinfused into the target brain region at a volume of 300 nL per side. The infusion was delivered slowly through an internal injector connected to a microsyringe at a rate of 100 nL/min. After infusion, the injector was left in place for an additional 5 min to allow diffusion and minimize backflow along the injection tract.

Astressin 2B (catalog number: 2391, Tocris) was injected into the LHA via a microcatheter at a dose of 3.3 ng /μL (500 nL/side).

Midazolam was administered via intraperitoneal injection at a dose of 0.2 mg/kg.

ELISA

Serum biomarkers, including HDL-C, LDL-C, TC, TG, and corticosterone, were measured using ELISA kits. Briefly, blood samples were collected at a consistent time window across groups to minimize potential circadian effects on corticosterone levels. Samples were allowed to clot at room temperature and then centrifuged to obtain serum. Before the assays, ELISA kit components and serum samples were equilibrated to room temperature for at least 30 min. HDL-C, LDL-C, TC, and TG levels were measured using UpingBio ELISA kits, and serum corticosterone levels were measured using the Cusabio ELISA kit following the manufacturer’s protocol.

Stereotaxic viral injection

Mice were anesthetized via intraperitoneal administration of 1% sodium pentobarbital (100 mg/kg). Upon loss of the righting reflex, each animal was seregulationd in a stereotaxic instrument using ear bars. The skull position was carefully aligned by adjusting the X, Y, and Z-axis micromanipulators of the stereotaxic device. After leveling the skull surface, burr holes were created with a dental drill at targeted locations. A virus-loaded glass micropipette was attached to a 10-μL syringe, and viral suspensions were injected in volumes of 100-300 nL, depending on the titer and expression efficiency of the virus. Injections were performed at a constant rate of 30 nL/min. To minimize backflow, the pipette remained in situ for an additional 10 minutes post-infusion. Stereotaxic coordinates for each injection site were defined in three planes relative to anatomical landmarks: anterior-posterior (AP) from bregma, mediolateral (ML) from the midline, and dorsoventral (DV) from the pial surface. The following coordinates were used: Arc: AP, −1.35 mm, ML, ±0.12 mm, DV, −6.00 mm; PVN: AP, −0.83 mm, ML, ±0.20 mm, DV, −4.65 mm; LHA: AP, −1.50 mm, ML, ±1.20 mm, DV, −4.80 mm.

For monosynaptic anterograde tracing, C57BL/6 J male mice (8 weeks old) received stereotaxic injections of rAAV-hSyn-Cre-EGFP-WPRE-hGH-pA (AAV2/1 serotype, titer: 1.1 × 10¹³ vg/mL; 100 nL) into the Arc or PVN. This allowed anterograde transport of Cre recombinase to downstream neuronal cell bodies. In the same session, rAAV-Eflα-DIO-mCherry-WPRE-pA (AAV2/9, 5.35 × 10¹² vg/mL; 150 nL) was delivered into either the PVN or LHA. Following a three-week incubation period to permit transgene expression, the mice were deeply anesthetized and subjected to transcardial perfusion. Brain sections were then collected for subsequent detection of mCherry fluorescence. Additionally, immunohistochemical co-staining was performed using antibodies against CRH, arginine vasopressin (AVP), and oxytocin (OXT) in the PVN, or against glutamate in the LHA.

For retrograde monosynaptic tracing, CRH-Cre or VgluT2-Cre male mice (8 weeks old) received stereotaxic delivery of a mixture of helper viruses into the PVN or LHA, respectively. The mixture consisted of rAAV-Eflα-DIO-RVG-WPRE-pA (AAV2/9, 4 × 10¹² vg/mL) and rAAV-Eflα-DIO-mCherry-2a-TVA-WPRE-pA (AAV2/9, 4 × 10¹² vg/mL) at a 1:2 ratio, with a total injection volume of 200 nL. After three weeks, the same animals were injected with the rabies virus RV-EnvA-ΔG-EGFP (2 × 10⁸ IFU/mL, 300 nL) into the corresponding regions using identical coordinates and surgical conditions. One week following the rabies virus injection, the mice were deeply anesthetized with sodium pentobarbital (administered intraperitoneally) and transcardially perfused. Brain sections were then prepared for detection of EGFP-positive signals. Immunohistochemical staining was also conducted using an AgRP antibody in the Arc or a CRH antibody in the PVN.

To perform triple-circuit tracing from the Arc to PVN and then to LHA, adult male C57BL/6 J mice (8 weeks old) were initially injected with retro-AAV-hSyn-Cre-WPRE-pA (AAV2/retro, 6.0 × 10¹² vg/mL, 200 nL) into the LHA. Concurrently, a Cre-dependent helper virus mixture was delivered into the ipsilateral PVN. After a three-week incubation period, RV was administered into the same region. In a separate experimental group, rAAV-hSyn-Cre-mCherry-WPRE-hGH-pA (AAV2/1, 1.1 × 10¹³ vg/mL; 100 nL) was delivered into the Arc, rAAV-Eflα-fDIO-EGFP-WPRE-pA (AAV2/9, 8.0 × 10¹² vg/mL; 200 nL) into the PVN, and retro-AAV-hSyn-DIO-FLP-WPRE-pA (AAV2/retro, 6.0 × 10¹² vg/mL, 200 nL) into the LHA. After virus expression was completed, animals were deeply anesthetized with pentobarbital (20 mg/kg, i.p.) and transcardially perfused. Brain sections were then collected for EGFP signal detection, and immunohistochemical co-staining was conducted using an AgRP antibody in the Arc or a CRH antibody in the PVN.

For chemogenetic experiments, mice were assigned to the corresponding viral groups before behavioral stratification and received stereotaxic injections of either control virus or chemogenetic virus according to the experimental design. AAV2/9 viruses encoding either hM4D(Gi)-mCherry or hM3D(Gq)-mCherry under Cre- or FLP-dependent promoters (rAAV-EF1α-DIO/fDIO-hM4D(Gi)-mCherry-WPRE-pA, 2.5 × 10¹³ vg/mL, 300 nL; rAAV-EF1α-DIO/fDIO-hM3D(Gq)-mCherry-WPRE-pA, 2.8 × 10¹³ vg/mL, 300 nL) were stereotaxically delivered into the target brain regions. Control animals received the corresponding Cre- or FLP-dependent mCherry virus (rAAV-EF1α-DIO/fDIO-mCherry-WPRE-pA, 8.5 × 10¹² vg/mL, 300 nL) following the same experimental timeline. After viral expression and 12 weeks of HFD feeding, mice within each viral group were subjected to EPM testing and stratified by hierarchical cluster analysis based on open-arm residence time. HFD-anxiety-susceptible and HFD-anxiety-unsusceptible subgroups were therefore defined within each viral condition. Subsequent chemogenetic analyses were performed in mice classified as HFD-anxiety-susceptible within their respective viral groups. Behavioral assessments and food intake measurements were conducted three weeks after viral injection. One hour before testing, mice received chemogenetic activation or inhibition either by intraperitoneal injection of CNO (2 mg/kg, Enzo Life Sciences, USA) for systemic manipulation or by local microinfusion through implanted guide cannulas for region-restricted manipulation. For local infusions, CNO was dissolved in sterile ACSF to a final concentration of 1 mM and bilaterally delivered at 300 nL per side at a rate of 100 nL/min. Injectors were left in place for an additional 5 min to allow diffusion and minimize backflow.

Electrophysiological measurements of neuronal excitability were performed on visually identified cells. In studies of the Arc^AgRP-PVN pathway, this was achieved by injecting rAAV-Eflα-DIO-ChR2-mCherry-WPRE-pA (AAV2/9, 1.0 × 10¹³ vg/mL, 150 nL) into the Arc of AgRP-Cre mice, allowing for precise targeting of AgRP neurons during recording.

Fiber photometry was employed to monitor neuronal excitability under different experimental conditions. To assess the activity of Arc^AgRP neurons, AgRP-Cre mice received injections of rAAV-Eflα-DIO-GCaMP6m-WPRE-pA (AAV2/9, 2.53 × 10¹² vg/mL, 200 nL) into the Arc. For examining the Arc^AgRP-PVN pathway, retro-AAV-hSyn-DIO-FLP-WPRE-pA (AAV2/retro, 6.0 × 10¹² vg/mL, 200 nL) was delivered into the PVN, while rAAV-Eflα-fDIO-GCaMP6m-WPRE-pA (AAV2/9, 2.8 × 10¹² vg/mL, 200 nL) was administered into the Arc of AgRP-Cre mice. To evaluate excitability in the PVN^CRH-LHA circuit, CRH-Cre mice were injected with retro-AAV-hSyn-DIO-FLP-WPRE-pA (AAV2/retro, 6.0 × 10¹² vg/mL, 200 nL) in the LHA and rAAV-Eflα-fDIO-GCaMP6m-WPRE-pA (AAV2/9, 2.8 × 10¹² vg/mL, 200 nL) in the PVN. Finally, for investigating the polysynaptic Arc-PVN-LHA pathway, C57 mice were used: retro-AAV-hSyn-DIO-FLP-WPRE-pA (AAV2/retro, 6.0 × 10¹² vg/mL, 200 nL) was injected into the LHA, rAAV-Eflα-fDIO-GCaMP6m-WPRE-pA (AAV2/9, 2.8 × 10¹² vg/mL, 200 nL) into the PVN, and rAAV-hSyn-Cre-WPRE-hGH-pA (AAV2/1, 1.1 × 10¹³ vg/mL; 100 nL) into the Arc.

To enable optogenetically guided electrophysiological recording across distinct neural circuits, stereotaxic injections of Cre-dependent Channelrhodopsin-2 (ChR2) vectors were performed in specific cell populations. In AgRP-Cre mice, rAAV-Eflα-DIO-ChR2-mCherry-WPRE-pA (AAV2/9, 1.0 × 10¹³ vg/mL, 150 nL) was delivered into the Arc, followed three weeks later by combined optogenetic-electrophysiological recordings in the PVN. In a separate group of CRH-Cre mice, the same ChR2 construct was injected into the PVN, with subsequent electrophysiological experiments conducted in the LHA after an equivalent expression period. Additionally, AgRP-Cre mice received co-injections of retro-AAV-hSyn-CRH-EGFP-WPRE-pA (AAV2/retro, 5.0 × 10¹² vg/mL, 200 nL) into the LHA and rAAV-Eflα-DIO-ChR2-mCherry-WPRE-pA (AAV2/9, 1.0 × 10¹³ vg/mL, 150 nL) into the Arc. Following a three-week incubation to allow for sufficient transgene expression, all groups underwent visualized patch-clamp recordings in their respective target regions to assess synaptic connectivity and functional properties.

By combining chemogenetics with optical fiber recording, we verified the function of circuits. To investigate the Arc-PVN neural pathway, AgRP-Cre mice were used. The PVN was injected with retro-AAV-hSyn-DIO-FLP-WPRE-pA (AAV2/retro, 6.0 × 10¹² vg/mL, 200nL) and rAAV-Eflα-CRH-GCaMP6m-WPRE-pA (AAV2/9, 4.0 × 10¹² vg/mL, 200 nL), while the Arc received rAAV-Eflα-fDIO-hM3D(Gq)-mCherry-WPRE-pA (2.5 × 10¹³ vg/mL, 300 nL). For the PVN-LHA circuit, CRH-Cre mice were administered retro-AAV-hSyn-DIO-FLP-WPRE-pA (AAV2/retro, 6.0 × 10¹² vg/mL, 200 nL) and rAAV-Eflα-VgluT2-GCaMP6m-WPRE-pA (AAV2/9, 4.0 × 10¹² vg/mL, 200 nL) into the LHA, along with rAAV-Eflα-fDIO-hM3D(Gq)-mCherry-WPRE-pA (2.5 × 10¹³ vg/mL, 300 nL) delivered into the PVN. To assess the polysynaptic Arc-PVN-LHA pathway, AgRP-Cre mice underwent triple viral injections: retro-AAV-hSyn-DIO-FLP-WPRE-pA (AAV2/retro, 6.0 × 10¹² vg/mL, 200 nL) was targeted to the PVN, rAAV-Eflα-fDIO-hM4D(Gi)-mCherry-WPRE-pA (2.5 × 10¹³ vg/mL, 300 nL) to the Arc, and rAAV-Eflα-VgluT2-GCaMP6m-WPRE-pA (AAV2/9, 4.0 × 10¹² vg/mL, 200 nL) to the LHA.

By combining optogenetics with optical fiber recording, we verified the function of circuits again. To investigate the Arc-PVN neural pathway, AgRP-Cre mice were used. The PVN was injected with retro-AAV-hSyn-DIO-FLP-WPRE-pA (AAV2/retro, 6.0 × 10¹² vg/mL, 200 nL) and rAAV-Eflα-CRH-GCaMP6m-WPRE-pA (AAV2/9, 4.0 × 10¹² vg/mL, 200 nL), while the Arc received rAAV-Eflα-fDIO-ChR2-mCherry-WPRE-pA (AAV2/9, 1.0 × 10¹³ vg/mL, 150nL). For the PVN-LHA circuit, CRH-Cre mice were administered retro-AAV-hSyn-DIO-FLP-WPRE-pA (AAV2/retro, 6.0 × 10¹² vg/mL, 200 nL) and rAAV-Eflα-VgluT2-GCaMP6m-WPRE-pA (AAV2/9, 4.0 × 10¹² vg/mL, 200 nL) into the LHA, along with rAAV-Eflα-fDIO-ChR2-mCherry-WPRE-pA (AAV2/9, 1.0 × 10¹³ vg/mL, 150 nL) delivered into the PVN. To assess the polysynaptic Arc-PVN-LHA pathway, AgRP-Cre mice underwent triple viral injections: retro-AAV-hSyn-DIO-FLP-WPRE-pA (AAV2/retro, 6.0 × 10¹² vg/mL, 200 nL) was targeted to the PVN, rAAV-Eflα-fDIO-ChR2-mCherry-WPRE-pA (AAV2/9, 1.0 × 10¹³ vg/mL, 150 nL) to the Arc, and rAAV-Eflα-VgluT2-GCaMP6m-WPRE-pA (AAV2/9, 4.0 × 10¹² vg/mL, 200 nL) to the LHA.

For the specific knockdown of CRHR2 in LHA^Glu neurons, rAAV-CMV-DIO-mCherry-shRNA (CRHR2)-WPRE-pA (AAV2/9, 5.0 × 10¹² vg/mL, 300 nL) and the corresponding control virus rAAV-CMV-DIO-mCherry-shRNA(scramble)-WPRE-pA (AAV2/9, 5.0 × 10¹² vg/mL, 300 nL) were stereotaxically injected into the LHA of VgluT2-Cre mice.

For specific labeling of Arc^AgRP-PVN projection neurons combined with immunofluorescence staining, AgRP-Cre mice received stereotaxic injections of retro-AAV-hSyn-DIO-FLP-WPRE-pA (AAV2/retro, 6.0 × 10¹² vg/mL, 200 nL) into the PVN, while rAAV-Eflα-fDIO-EGFP-WPRE-pA (AAV2/9, 8.0 × 10¹² vg/mL, 200 nL) was injected into the Arc.

All viral vectors used in this study were obtained from BrainVTA (Wuhan, China). Following deep anesthesia, mice were perfused transcardially with chilled 0.9% saline and subsequently with 4% paraformaldehyde. Fluorescence imaging was performed on a ZEISS LSM 980 confocal microscope (Germany) to visualize signal expression. Animals exhibiting inaccurate injection sites were excluded from subsequent analysis.

In vivo fiber photometry recording

Following a three-week period for recovery and habituation, fiber photometry (THINKERTECH) was performed to record neural activity. A dual-color photometry system was employed to simultaneously capture the calcium-sensitive fluorescence signal (excited at 480 nm) and an isosbestic reference signal (excited at 405 nm). The 405 nm channel served as an internal control to correct for potential motion artifacts and background noise. Analog voltage signals were digitized at a sampling rate of 100 Hz using a dedicated data acquisition (DAQ) module. Data acquisition was controlled by the FiberPhotometry_dual_color_v2.0.vi software (THINKERTECH). During recordings, the laser power was maintained between 40 and 70 μW at the fiber tip. For the analysis of calcium signals, the change in fluorescence (ΔF/F) was calculated. In experiments examining neuronal excitability (AUC-based analysis): The recorded signals were analyzed using the dual-color multichannel analysis software scripts provided by ThinkerTech based on MATLAB (2023b, MathWorks, Cambridge, UK) software. Each mouse was imaged for 5–10 min at a time to minimize phototoxicity and bleaching. Five-minute recording sessions were analyzed for each mouse. Fluorescence signal change intensity (ΔF/F₀) was calculated as ΔF/F₀ = (V_signal − F₀)/F₀, where F₀ is defined as the baseline fluorescence signal averaged over a 50 s-long control time window and V_signal is the peak value of the fluorescence signal, and signals were normalized to a range between 0 and 1 using the following formula: V_signal = (ΔF/F-min(ΔF/F))/(max(ΔF/F)-min(ΔF/F)). A 50% threshold was applied to identify events, and the area under the curve (AUC) of the ΔF/F trace was calculated for quantitative comparisons⁶¹. Combining chemogenetics or optogenetics with fiber photometry recordings, peri-event photometry analyses synchronized to feeding bouts: The formula used was ΔF/F = (F − F₀)/(F₀ − F_baseline), where F represents the real-time fluorescence value recorded after the photometry system was connected to the implanted optical fiber. F₀ denotes the average fluorescence signal during a defined baseline period after connection. F_baseline corresponds to the mean fluorescence value measured before the system was connected to the mouse, establishing a pre-connection baseline³².

In vitro electrophysiological recordings

Brain slice preparation: Adult mice were deeply anesthetized and transcardially perfused with an ice-cold, oxygenated cutting solution composed of N-Methyl-D-glucamine-based ACSF (NMDG-ACSF). The NMDG-ACSF (300–305 mOsm/kg, pH 7.3–7.4) contained the following (in mM): 93 NMDG, 20 HEPES, 25 glucose, 2.5 KCl, 0.5 CaCl₂, 1.2 NaH₂PO₄, 30 NaHCO₃, 10 MgSO₄, 5 Na-ascorbate, 3 Na-pyruvate, 3 glutathione, and 2 thiourea. Using a vibrating microtome (Leica VT1200s, Germany), 300 μm thick coronal slices containing the Arc, PVN, or LHA were prepared. These slices were first recovered in oxygenated HEPES-ACSF recovery solution at 28 °C for a minimum of 1 h. The HEPES-ACSF (300–305 mOsm/kg, pH 7.3–7.4) was composed of (in mM): 92 NaCl, 30 NaHCO₃, 20 HEPES, 25 glucose, 2.5 KCl, 2 CaCl₂, 2 MgSO₄, 1.2 NaH₂PO₄, 5 Na-ascorbate, 3 Na-pyruvate, 3 glutathione, and 2 thiourea. For electrophysiological recordings, slices were subsequently transferred to a submersion chamber (Warner Instruments, USA) and continuously superfused at a rate of 3 ml/min with oxygenated standard ACSF maintained at 32 °C (300-310 mOsm/kg, pH 7.3–7.4). The standard ACSF contained (in mM): 129 NaCl, 20 NaHCO₃, 3 KCl, 10 glucose, 2.4 CaCl₂, 1.3 MgSO₄, 1.2 KH₂PO₄, and 3 HEPES.

Whole-cell patch-clamp recordings: Whole-cell patch-clamp recordings were initiated by locating the target brain area under a 10x objective, followed by switching to a 40× objective for detailed access. Borosilicate glass pipettes (Sutter Instrument; BF150-86-10) were fabricated on a P-97 puller (Sutter Instrument) to final resistances of 3–5 MQ. For most recordings, the pipette internal solution consisted of (in mM): 135 K-gluconate, 10 NaCl, 10 HEPES, 0.5 EGTA, 10 Na-phosphocreatine, 4 Mg-ATP, and 0.4 Na2-GTP, with pH adjusted to 7.3 using KOH and osmolality maintained at 290–300 mOsm. Neuronal signals were acquired using a Multiclamp 700B amplifier (Molecular Devices) coupled with a DigiData 1440 A digitizer (Molecular Devices). The signals underwent low-pass filtering at 10 kHz and were digitized at a sampling rate of 10 kHz via a MICRO3 1401 unit (Cambridge Electronic Design). All acquired data were stored and analyzed using Clampfit 10.6 software (Cambridge Electronic Design). To ensure data quality and consistency, only neurons fulfilling the following criteria were included in the final analysis: series resistance (Ra) below 25 MQ, input resistance exceeding 300 MQ, and leak currents under 50 pA. Any cells failing to meet these standards were discarded. Neurons were selected in an unbiased, randomized manner for recording.

Light-evoked response: For recording light-evoked postsynaptic currents (473 nm wavelength, 20 ms pulses, 2 Hz stimulation), a cesium-based internal solution was used, containing (in mM): 135 CsMeSO₃, 10 HEPES, 0.5 EGTA, 2 QX-314, 4 Mg-ATP, 0.4 Na₂-GTP, and 10 Na₂-phosphocreatine, adjusted to pH 7.2 with CsOH and an osmolality of 290–300 mOsm. Neurons were voltage-clamped at a holding potential of -70 mV to isolate oEPSCs, and at 0 mV (close to the chloride reversal potential) to isolate oIPSCs, respectively. To verify the monosynaptic nature of the optogenetically-evoked responses, tetrodotoxin (TTX; 1 µM) was first applied via the ACSF to block voltage-gated sodium channels and abolish polysynaptic activity. This was followed by co-application of 4-aminopyridine (4-AP; 100 µM), a potassium channel blocker, to restore and enhance transmitter release from the directly stimulated optogenetic terminals.

Optogenetic and chemogenetic manipulation

Three weeks after viral transduction, mice were acclimated to the testing environment for approximately 1 h. Under brief isoflurane anesthesia, chronically implanted optical fibers (200 µm OD, 0.37 NA, RWD Life Science) were connected to a laser source via ceramic sleeves. Following recovery in their home cage for at least 30 minutes, optogenetic stimulation was delivered using either blue light (473 nm, 5-8 mW, 15 ms pulses, 20 Hz), controlled by a Master-8 pulse stimulator (A.M.P.I.). For chemogenetic experiments, mice received an intraperitoneal injection of CNO under light isoflurane anesthesia, and behavioral testing commenced after a 30-minute rest period. Control animals underwent identical experimental procedures. Upon completion of all behavioral assays, mice were euthanized, and histological verification of viral expression and fiber placement was performed. Data from subjects with off-target injections or incorrect fiber localization were excluded from subsequent analysis.

Western blot

The LHA was dissected and homogenized in ice-cold RIPA lysis buffer (Beyotime, China) containing protease and phosphatase inhibitor cocktails (Beyotime, China). The homogenates were incubated on ice for 30 min, then centrifuged at 12,000 × g for 15 min at 4 °C. The supernatant was collected, and protein concentration was determined using a BCA assay kit (Thermo Fisher Scientific, USA). Equal amounts of protein (20–40 μg) were denatured by boiling in SDS sample buffer at 95 °C for 5 minutes, then separated on a 10% SDS-PAGE gel. Proteins were transferred onto PVDF membranes (Millipore, USA) using a semi-dry transfer system. Membranes were blocked with 5% non-fat milk in TBST (Tris-buffered saline with 0.1% Tween-20) for 1 h at room temperature, then incubated overnight at 4 °C with primary antibodies against CRHR2 (1:1000, catalog number: 236982, Abcam, UK) and GAPDH (1:10000, catalog number: 60004-1-Ig, Proteintech, USA; loading control). After three washes with TBST, membranes were incubated with HRP-conjugated secondary antibodies (1:5000, catalog number: 7074, CST, USA) for 1 h at room temperature. Signals were detected using an enhanced chemiluminescence (ECL) kit (Thermo Fisher Scientific, USA) and visualized with a chemiluminescence imaging system (Amersham Imager 600, USA). Band intensities were quantified with ImageJ software. Data are presented as relative CRHR2 expression levels to controls.

Immunofluorescence

Brain tissues were fixed in 4% paraformaldehyde (PFA, BL539A; Biosharp, China) at 4 °C for 24 h and then transferred to 15% and 30% sucrose solution for dehydration until the tissue was completely dehydrated and sank to the bottom. Afterward, the tissues were embedded in optimal cutting temperature compound (OCT, 4583; SAKURA Tissue-Tek, USA) and coronally sectioned at a thickness of 40 μm. These sections were stored at −20 °C for subsequent experimental procedures. Prior to immunostaining, the sections were rinsed three times (10 min each) in phosphate-buffered saline (PBS, pH 7.4, ZLI-9062; Biosharp, China) to remove residual embedding medium. Antigen retrieval was performed by heating the sections in citrate buffer at 95 °C for 15 min and then allowing them to cool naturally to room temperature. To minimize nonspecific antibody binding, the sections were incubated in a blocking solution consisting of 5% bovine serum albumin (BSA, 4240GR100; Biosharp, China) and 0.3% Triton X-100 (BS084; Biosharp, China) in PBS for 1 h at room temperature. An overnight incubation was performed at 4 °C with the specified primary antibodies, which were diluted in an antibody dilution buffer consisting of PBS supplemented with 5% BSA and 1% Triton X-100: Goat anti-CRHR1 (1:500; catalog number: 77686, Abcam); Mouse anti-CRHR2 (1:500; catalog number: 1205, Immunoway); Mouse anti-Agouti-Related Peptide (1:200; catalog number: H-003-57, Phoenix Pharmaceuticals, Inc.); Rabbit anti-Glutamate (1:200; catalog number: 6642, Sigma-Aldrich); Rabbit anti-CRH (1:200; catalog number: 272391, Abcam); Rabbit anti-Vasopressin (1:500; catalog number: 213708, Abcam); Rabbit anti-Oxytocin-neurophysin 1 (1:500; catalog number: 212193, Abcam); Mouse anti-c-Fos (1:100; catalog number:166940, Santa Cruz Biotechnology). Post-incubation with the primary antibody, sections were allowed to equilibrate at room temperature for 30 minutes. After extensive washing with PBS (three times, 10 min per wash), a fluorescent-dye-conjugated secondary antibody, prepared in a blocking buffer of PBS with 5% BSA and 0.3% Triton X-100, was applied, and the sections were incubated for 1 h at room temperature, protected from light. The secondary antibodies used were: Alexa Fluor 488-conjugated goat anti-rabbit IgG (1:500; catalog number: 150077, Abcam); Alexa Fluor 488-conjugated donkey anti-goat IgG (1:500; catalog number: 15029, Abcam); Alexa Fluor 488-conjugated goat anti-Mouse IgG (1:500; catalog number: 150113, Abcam); Alexa Fluor 568-conjugated goat anti-Rabbit IgG (1:500; catalog number:175471, Abcam); Alexa Fluor 568-conjugated goat anti- Mouse IgG (1:500; catalog number:175473, Abcam). Immunofluorescence imaging was conducted on a confocal laser scanning microscope (Zeiss LSM 980, Germany) with identical laser and configuration settings for all samples. Subsequent colocalization assessments were carried out utilizing the ImageJ software (Fiji distribution, NIH).

Statistics and reproducibility

All of the sample sizes were selected according to previous research using similar experimental paradigms. No statistical power calculations were used to predetermine sample size. Animals were excluded only according to predefined technical criteria, including inaccurate viral targeting, off-target viral expression, incorrect fiber placement, or failed histological verification. All animals that met both inclusion and histological verification standards were retained for final analyses. No animals were excluded on the basis of behavioral, electrophysiological, or fiber photometry outcomes. Animals were assigned to treatment groups by means of a random number table and tested in ascending order of their identifiers. Experimenters were blinded to the specific drug treatments administered throughout the study to minimize potential bias. All statistical analyses were performed using GraphPad Prism (version 9.5.1) or MATLAB R2023. For representative images and statistical data, each experiment was repeated in at least three mice, with multiple brain slices in the manuscript, with consistent results. Data normality was assessed using the D’Agostino-Pearson or Shapiro-Wilk tests. Homogeneity of variances was verified with F tests, and when variances were unequal between groups, the Sattherthwaite approximation was applied to determine the effective degrees of freedom. For comparisons between two groups, a two-tailed Student’s t-test was employed. Experiments involving a single factor with more than two groups were analyzed by one-way analysis of variance (ANOVA) followed by Bonferroni post hoc tests. For experimental designs incorporating two factors, two-way ANOVA was performed, with subsequent multiple comparisons adjusted using the Bonferroni correction. For datasets with repeated measurements from the same subjects, including body weight trajectories, daily food intake, hourly feeding after pharmacological or chemogenetic manipulations, electrophysiological recordings with escalating injected currents, and anxiety-like behavior or serum corticosterone measurements at multiple time points under HFD or CD conditions, statistical analyses were performed using two-way repeated-measures ANOVA followed by Bonferroni post hoc tests. For fiber photometry recordings during feeding behavior, calcium signals from the same mice across different behavioral epochs, including pre-feeding, feeding, and post-feeding periods, were analyzed using one-way repeated-measures ANOVA followed by Bonferroni post hoc tests.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary information^{(24.8MB, pdf)}

Peer Review File^{(21.7MB, pdf)}

Reporting Summary^{(94.8KB, pdf)}

Source data

Source Data^{(1MB, xlsx)}

Acknowledgements

We especially thank the Center for Scientific Research of Anhui Medical University for providing support with the use of research equipment.

Author contributions

These authors contributed equally: P.W., J.L., and X.C. X.L., X.Q., and G.Q. supervised the project. P.W., J.L., and X.C. conceived the experiments; W.C., Z.L., Z.F., S.H., Z.Y., J.Z., and D.W. carried out the experiments; B.M. and H.L. helped with data interpretations. P.W., J.L., and X.C. wrote the manuscript. All authors discussed the results and commented on the paper.

Peer review

Peer review information

Nature Communications thanks Shinsuke Ishigaki, Sora Shin and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by the National Natural Science Foundation of China (82371288, 82571376 to X. L.; 82571450 to G.Q.).

Data availability

The experimental data generated in this study are available within the Article and Supplementary Information. All data underlying this study are available from the corresponding author upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Peng Wang, Jun Li, Xin Cheng.

Contributor Information

Gaolin Qiu, Email: gaolinqiu@fy.ahmu.edu.cn.

Xin Qing, Email: yfy1023275@fy.ahmu.edu.cn.

Xuesheng Liu, Email: liuxuesheng@ahmu.edu.cn.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77749-w.

References

1.Collaborators, G. B. D. O. et al. Health effects of overweight and obesity in 195 countries over 25 years. N. Engl. J. Med.377, 13–27 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Heymsfield, S. B. & Wadden, T. A. Mechanisms, pathophysiology, and management of obesity. N. Engl. J. Med.376, 254–266 (2017). [DOI] [PubMed] [Google Scholar]
3.Wang, R. Z. et al. Body weight in neurological and psychiatric disorders: a large prospective cohort study. Nat. Ment. Health2, 41–51 (2024). [Google Scholar]
4.Tsai, S. F. et al. Stress aggravates high-fat-diet-induced insulin resistance via a mechanism that involves the amygdala and is associated with changes in neuroplasticity. Neuroendocrinology107, 147–157 (2018). [DOI] [PubMed] [Google Scholar]
5.Xia, G. et al. Reciprocal control of obesity and anxiety-depressive disorder via a GABA and serotonin neural circuit. Mol. Psychiatry26, 2837–2853 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Alonso-Caraballo, Y., Hodgson, K. J., Morgan, S. A., Ferrario, C. R. & Vollbrecht, P. J. Enhanced anxiety-like behavior emerges with weight gain in male and female obesity-susceptible rats. Behav. Brain Res.360, 81–93 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Zhang, H. et al. Multiple mediation of the association between childhood emotional abuse and adult obesity by anxiety and bulimia - a sample from bariatric surgery candidates and healthy controls. BMC Public Health24, 653 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
8.You, I. J., Bae, Y., Beck, A. R. & Shin, S. Lateral hypothalamic proenkephalin neurons drive threat-induced overeating associated with a negative emotional state. Nat. Commun.14, 6875 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Ip, C. K. et al. Critical role of lateral habenula circuits in the control of stress-induced palatable food consumption. Neuron111, 2583–2600.e2586 (2023). [DOI] [PubMed] [Google Scholar]
10.Fujioka, Y. et al. Stress-impaired reward pathway promotes distinct feeding behavior patterns. Front. Neurosci.18, 1349366 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Berthoud, H. R., Seeley, R. J. & Roberts, S. B. Physiology of energy intake in the weight-reduced state. Obesity29, S25–S30 (2021). [DOI] [PubMed] [Google Scholar]
12.Andermann, M. L. & Lowell, B. B. Toward a wiring diagram understanding of appetite control. Neuron95, 757–778 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Zhang, X. & van den Pol, A. N. Hypothalamic arcuate nucleus tyrosine hydroxylase neurons play orexigenic role in energy homeostasis. Nat. Neurosci.19, 1341–1347 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Krashes, M. J. et al. Rapid, reversible activation of AgRP neurons drives feeding behavior in mice. J. Clin. Investig.121, 1424–1428 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Atasoy, D., Betley, J. N., Su, H. H. & Sternson, S. M. Deconstruction of a neural circuit for hunger. Nature488, 172–177 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Fenselau, H. et al. A rapidly acting glutamatergic ARC-->PVH satiety circuit postsynaptically regulated by alpha-MSH. Nat. Neurosci.20, 42–51 (2017). [DOI] [PMC free article] [PubMed]
17.Jais, A. et al. PNOC(ARC) Neurons Promote Hyperphagia and Obesity upon High-Fat-Diet Feeding. Neuron106, 1009–1025.e1010 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Lei, Y. et al. Region-specific transcriptomic responses to obesity and diabetes in macaque hypothalamus. Cell Metab.36, 438–453.e436 (2024). [DOI] [PubMed] [Google Scholar]
19.Tadross, J. A. et al. A comprehensive spatio-cellular map of the human hypothalamus. Nature639, 708–716 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Aponte, Y., Atasoy, D. & Sternson, S. M. AGRP neurons are sufficient to orchestrate feeding behavior rapidly and without training. Nat. Neurosci.14, 351–355 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Betley, J. N. et al. Neurons for hunger and thirst transmit a negative-valence teaching signal. Nature521, 180–185 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Vardy, E. et al. A new DREADD facilitates the multiplexed chemogenetic interrogation of behavior. Neuron86, 936–946 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Gropp, E. et al. Agouti-related peptide-expressing neurons are mandatory for feeding. Nat. Neurosci.8, 1289–1291 (2005). [DOI] [PubMed] [Google Scholar]
24.Luquet, S., Perez, F. A., Hnasko, T. S. & Palmiter, R. D. NPY/AgRP neurons are essential for feeding in adult mice but can be ablated in neonates. Science310, 683–685 (2005). [DOI] [PubMed] [Google Scholar]
25.Betley, J. N., Cao, Z. F., Ritola, K. D. & Sternson, S. M. Parallel, redundant circuit organization for homeostatic control of feeding behavior. Cell155, 1337–1350 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Burnett, C. J. et al. Hunger-driven motivational state competition. Neuron92, 187–201 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Dietrich, M. O., Zimmer, M. R., Bober, J. & Horvath, T. L. Hypothalamic Agrp neurons drive stereotypic behaviors beyond feeding. Cell160, 1222–1232 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Garfield, A. S. et al. A neural basis for melanocortin-4 receptor-regulated appetite. Nat. Neurosci.18, 863–871 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Krashes, M. J. et al. An excitatory paraventricular nucleus to AgRP neuron circuit that drives hunger. Nature507, 238–242 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Steculorum, S. M. et al. AgRP neurons control systemic insulin sensitivity via myostatin expression in brown adipose tissue. Cell165, 125–138 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Fuzesi, T., Daviu, N., Wamsteeker Cusulin, J. I., Bonin, R. P. & Bains, J. S. Hypothalamic CRH neurons orchestrate complex behaviours after stress. Nat. Commun.7, 11937 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Qiu, G. et al. Dexmedetomidine inhibits paraventricular corticotropin-releasing hormone neurons that attenuate acute stress-induced anxiety-like behavior in mice. Anesthesiology140, 1134–1152 (2024). [DOI] [PubMed] [Google Scholar]
33.Milanick, W. J., Polo-Parada, L., Dantzler, H. A. & Kline, D. D. Activation of alpha-1 adrenergic receptors increases cytosolic calcium in neurones of the paraventricular nucleus of the hypothalamus. J. Neuroendocrinol.31, e12791 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Stuber, G. D. & Wise, R. A. Lateral hypothalamic circuits for feeding and reward. Nat. Neurosci.19, 198–205 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Stamatakis, A. M. et al. Lateral hypothalamic area glutamatergic neurons and their projections to the lateral habenula regulate feeding and reward. J. Neurosci.36, 302–311 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Jennings, J. H. et al. Visualizing hypothalamic network dynamics for appetitive and consummatory behaviors. Cell160, 516–527 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Steckler, T. & Holsboer, F. Corticotropin-releasing hormone receptor subtypes and emotion. Biol. Psychiatry46, 1480–1508 (1999). [DOI] [PubMed] [Google Scholar]
38.Deussing, J. M. & Chen, A. The corticotropin-releasing factor family: physiology of the stress response. Physiol. Rev.98, 2225–2286 (2018). [DOI] [PubMed] [Google Scholar]
39.Tabarin, A. et al. Role of the corticotropin-releasing factor receptor type 2 in the control of food intake in mice: a meal pattern analysis. Eur. J. Neurosci.26, 2303–2314 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Chao, H., Digruccio, M., Chen, P. & Li, C. Type 2 corticotropin-releasing factor receptor in the ventromedial nucleus of hypothalamus is critical in regulating feeding and lipid metabolism in white adipose tissue. Endocrinology153, 166–176 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Liu, L. et al. Hypothalamus-sympathetic-liver axis mediates the early phase of stress-induced hyperglycemia in the male mice. Nat. Commun.15, 8632 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Waterson, M. J. & Horvath, T. L. Neuronal regulation of energy homeostasis: beyond the hypothalamus and feeding. Cell Metab.22, 962–970 (2015). [DOI] [PubMed] [Google Scholar]
43.Berthoud, H. R., Munzberg, H. & Morrison, C. D. Blaming the brain for obesity: integration of hedonic and homeostatic mechanisms. Gastroenterology152, 1728–1738 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Herzog, H. Integrated pathways that control stress and energy homeostasis. Nat. Rev. Endocrinol.16, 75–76 (2020). [DOI] [PubMed] [Google Scholar]
45.Timper, K. & Bruning, J. C. Hypothalamic circuits regulating appetite and energy homeostasis: pathways to obesity. Dis. Model. Mech.10, 679–689 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Kurhe, Y., Mahesh, R. & Devadoss, T. QCM-4, a 5-HT(3) receptor antagonist ameliorates plasma HPA axis hyperactivity, leptin resistance and brain oxidative stress in depression and anxiety-like behavior in obese mice. Biochem. Biophys. Res. Commun.456, 74–79 (2015). [DOI] [PubMed] [Google Scholar]
47.Kakizawa, K. et al. A novel GABA-mediated corticotropin-releasing hormone secretory mechanism in the median eminence. Sci. Adv.2, e1501723 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Douglass, A. M. et al. Neural basis for fasting activation of the hypothalamic-pituitary-adrenal axis. Nature620, 154–162 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Kim, J. S. et al. Chronic hyperosmotic stress converts GABAergic inhibition into excitation in vasopressin and oxytocin neurons in the rat. J. Neurosci.31, 13312–13322 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Fernandes, A. C. A. et al. Arcuate AgRP, but not POMC neurons, modulate paraventricular CRF synthesis and release in response to fasting. Cell Biosci.12, 118 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Teitelbaum, P. & Epstein, A. N. The lateral hypothalamic syndrome: recovery of feeding and drinking after lateral hypothalamic lesions. Psychol. Rev.69, 74–90 (1962). [DOI] [PubMed] [Google Scholar]
52.Clayton, J. A. & Collins, F. S. Policy: NIH to balance sex in cell and animal studies. Nature509, 282–283 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
53.Shansky, R. M. & Murphy, A. Z. Considering sex as a biological variable will require a global shift in science culture. Nat. Neurosci.24, 457–464 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
54.Bechtold, D. A. & Loudon, A. S. Hypothalamic clocks and rhythms in feeding behaviour. Trends Neurosci.36, 74–82 (2013). [DOI] [PubMed] [Google Scholar]
55.Kohsaka, A. et al. High-fat diet disrupts behavioral and molecular circadian rhythms in mice. Cell Metab.6, 414–421 (2007). [DOI] [PubMed] [Google Scholar]
56.Hatori, M. et al. Time-restricted feeding without reducing caloric intake prevents metabolic diseases in mice fed a high-fat diet. Cell Metab.15, 848–860 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
57.Chung, J. et al. Anxiety and depression in non-cardiac surgical patients and associations with adverse outcomes: a systematic review and meta-analysis. J. Anesth. Transl. Med.4, 207–219 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Bosch-Bouju, C., Larrieu, T., Linders, L., Manzoni, O. J. & Laye, S. Endocannabinoid-mediated plasticity in nucleus accumbens controls vulnerability to anxiety after social defeat stress. Cell Rep.16, 1237–1242 (2016). [DOI] [PubMed] [Google Scholar]
59.Li, Y. et al. The role of gut microbiota in chronic itch-evoked novel object recognition-related cognitive dysfunction in mice. Front. Med.8, 616489 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
60.Walf, A. A. & Frye, C. A. The use of the elevated plus maze as an assay of anxiety-related behavior in rodents. Nat. Protoc.2, 322–328 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
61.Ding, W. et al. The endocannabinoid N-arachidonoyl dopamine is critical for hyperalgesia induced by chronic sleep disruption. Nat. Commun.14, 6696 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary information^{(24.8MB, pdf)}

Peer Review File^{(21.7MB, pdf)}

Reporting Summary^{(94.8KB, pdf)}

Source Data^{(1MB, xlsx)}

Data Availability Statement

[CR1] 1.Collaborators, G. B. D. O. et al. Health effects of overweight and obesity in 195 countries over 25 years. N. Engl. J. Med.377, 13–27 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR2] 2.Heymsfield, S. B. & Wadden, T. A. Mechanisms, pathophysiology, and management of obesity. N. Engl. J. Med.376, 254–266 (2017). [DOI] [PubMed] [Google Scholar]

[CR3] 3.Wang, R. Z. et al. Body weight in neurological and psychiatric disorders: a large prospective cohort study. Nat. Ment. Health2, 41–51 (2024). [Google Scholar]

[CR4] 4.Tsai, S. F. et al. Stress aggravates high-fat-diet-induced insulin resistance via a mechanism that involves the amygdala and is associated with changes in neuroplasticity. Neuroendocrinology107, 147–157 (2018). [DOI] [PubMed] [Google Scholar]

[CR5] 5.Xia, G. et al. Reciprocal control of obesity and anxiety-depressive disorder via a GABA and serotonin neural circuit. Mol. Psychiatry26, 2837–2853 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR6] 6.Alonso-Caraballo, Y., Hodgson, K. J., Morgan, S. A., Ferrario, C. R. & Vollbrecht, P. J. Enhanced anxiety-like behavior emerges with weight gain in male and female obesity-susceptible rats. Behav. Brain Res.360, 81–93 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR7] 7.Zhang, H. et al. Multiple mediation of the association between childhood emotional abuse and adult obesity by anxiety and bulimia - a sample from bariatric surgery candidates and healthy controls. BMC Public Health24, 653 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR8] 8.You, I. J., Bae, Y., Beck, A. R. & Shin, S. Lateral hypothalamic proenkephalin neurons drive threat-induced overeating associated with a negative emotional state. Nat. Commun.14, 6875 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR9] 9.Ip, C. K. et al. Critical role of lateral habenula circuits in the control of stress-induced palatable food consumption. Neuron111, 2583–2600.e2586 (2023). [DOI] [PubMed] [Google Scholar]

[CR10] 10.Fujioka, Y. et al. Stress-impaired reward pathway promotes distinct feeding behavior patterns. Front. Neurosci.18, 1349366 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR11] 11.Berthoud, H. R., Seeley, R. J. & Roberts, S. B. Physiology of energy intake in the weight-reduced state. Obesity29, S25–S30 (2021). [DOI] [PubMed] [Google Scholar]

[CR12] 12.Andermann, M. L. & Lowell, B. B. Toward a wiring diagram understanding of appetite control. Neuron95, 757–778 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR13] 13.Zhang, X. & van den Pol, A. N. Hypothalamic arcuate nucleus tyrosine hydroxylase neurons play orexigenic role in energy homeostasis. Nat. Neurosci.19, 1341–1347 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR14] 14.Krashes, M. J. et al. Rapid, reversible activation of AgRP neurons drives feeding behavior in mice. J. Clin. Investig.121, 1424–1428 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR15] 15.Atasoy, D., Betley, J. N., Su, H. H. & Sternson, S. M. Deconstruction of a neural circuit for hunger. Nature488, 172–177 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR16] 16.Fenselau, H. et al. A rapidly acting glutamatergic ARC-->PVH satiety circuit postsynaptically regulated by alpha-MSH. Nat. Neurosci.20, 42–51 (2017). [DOI] [PMC free article] [PubMed]

[CR17] 17.Jais, A. et al. PNOC(ARC) Neurons Promote Hyperphagia and Obesity upon High-Fat-Diet Feeding. Neuron106, 1009–1025.e1010 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR18] 18.Lei, Y. et al. Region-specific transcriptomic responses to obesity and diabetes in macaque hypothalamus. Cell Metab.36, 438–453.e436 (2024). [DOI] [PubMed] [Google Scholar]

[CR19] 19.Tadross, J. A. et al. A comprehensive spatio-cellular map of the human hypothalamus. Nature639, 708–716 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR20] 20.Aponte, Y., Atasoy, D. & Sternson, S. M. AGRP neurons are sufficient to orchestrate feeding behavior rapidly and without training. Nat. Neurosci.14, 351–355 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR21] 21.Betley, J. N. et al. Neurons for hunger and thirst transmit a negative-valence teaching signal. Nature521, 180–185 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR22] 22.Vardy, E. et al. A new DREADD facilitates the multiplexed chemogenetic interrogation of behavior. Neuron86, 936–946 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR23] 23.Gropp, E. et al. Agouti-related peptide-expressing neurons are mandatory for feeding. Nat. Neurosci.8, 1289–1291 (2005). [DOI] [PubMed] [Google Scholar]

[CR24] 24.Luquet, S., Perez, F. A., Hnasko, T. S. & Palmiter, R. D. NPY/AgRP neurons are essential for feeding in adult mice but can be ablated in neonates. Science310, 683–685 (2005). [DOI] [PubMed] [Google Scholar]

[CR25] 25.Betley, J. N., Cao, Z. F., Ritola, K. D. & Sternson, S. M. Parallel, redundant circuit organization for homeostatic control of feeding behavior. Cell155, 1337–1350 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR26] 26.Burnett, C. J. et al. Hunger-driven motivational state competition. Neuron92, 187–201 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR27] 27.Dietrich, M. O., Zimmer, M. R., Bober, J. & Horvath, T. L. Hypothalamic Agrp neurons drive stereotypic behaviors beyond feeding. Cell160, 1222–1232 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR28] 28.Garfield, A. S. et al. A neural basis for melanocortin-4 receptor-regulated appetite. Nat. Neurosci.18, 863–871 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR29] 29.Krashes, M. J. et al. An excitatory paraventricular nucleus to AgRP neuron circuit that drives hunger. Nature507, 238–242 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR30] 30.Steculorum, S. M. et al. AgRP neurons control systemic insulin sensitivity via myostatin expression in brown adipose tissue. Cell165, 125–138 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR31] 31.Fuzesi, T., Daviu, N., Wamsteeker Cusulin, J. I., Bonin, R. P. & Bains, J. S. Hypothalamic CRH neurons orchestrate complex behaviours after stress. Nat. Commun.7, 11937 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR32] 32.Qiu, G. et al. Dexmedetomidine inhibits paraventricular corticotropin-releasing hormone neurons that attenuate acute stress-induced anxiety-like behavior in mice. Anesthesiology140, 1134–1152 (2024). [DOI] [PubMed] [Google Scholar]

[CR33] 33.Milanick, W. J., Polo-Parada, L., Dantzler, H. A. & Kline, D. D. Activation of alpha-1 adrenergic receptors increases cytosolic calcium in neurones of the paraventricular nucleus of the hypothalamus. J. Neuroendocrinol.31, e12791 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR34] 34.Stuber, G. D. & Wise, R. A. Lateral hypothalamic circuits for feeding and reward. Nat. Neurosci.19, 198–205 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR35] 35.Stamatakis, A. M. et al. Lateral hypothalamic area glutamatergic neurons and their projections to the lateral habenula regulate feeding and reward. J. Neurosci.36, 302–311 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR36] 36.Jennings, J. H. et al. Visualizing hypothalamic network dynamics for appetitive and consummatory behaviors. Cell160, 516–527 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR37] 37.Steckler, T. & Holsboer, F. Corticotropin-releasing hormone receptor subtypes and emotion. Biol. Psychiatry46, 1480–1508 (1999). [DOI] [PubMed] [Google Scholar]

[CR38] 38.Deussing, J. M. & Chen, A. The corticotropin-releasing factor family: physiology of the stress response. Physiol. Rev.98, 2225–2286 (2018). [DOI] [PubMed] [Google Scholar]

[CR39] 39.Tabarin, A. et al. Role of the corticotropin-releasing factor receptor type 2 in the control of food intake in mice: a meal pattern analysis. Eur. J. Neurosci.26, 2303–2314 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR40] 40.Chao, H., Digruccio, M., Chen, P. & Li, C. Type 2 corticotropin-releasing factor receptor in the ventromedial nucleus of hypothalamus is critical in regulating feeding and lipid metabolism in white adipose tissue. Endocrinology153, 166–176 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR41] 41.Liu, L. et al. Hypothalamus-sympathetic-liver axis mediates the early phase of stress-induced hyperglycemia in the male mice. Nat. Commun.15, 8632 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR42] 42.Waterson, M. J. & Horvath, T. L. Neuronal regulation of energy homeostasis: beyond the hypothalamus and feeding. Cell Metab.22, 962–970 (2015). [DOI] [PubMed] [Google Scholar]

[CR43] 43.Berthoud, H. R., Munzberg, H. & Morrison, C. D. Blaming the brain for obesity: integration of hedonic and homeostatic mechanisms. Gastroenterology152, 1728–1738 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR44] 44.Herzog, H. Integrated pathways that control stress and energy homeostasis. Nat. Rev. Endocrinol.16, 75–76 (2020). [DOI] [PubMed] [Google Scholar]

[CR45] 45.Timper, K. & Bruning, J. C. Hypothalamic circuits regulating appetite and energy homeostasis: pathways to obesity. Dis. Model. Mech.10, 679–689 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR46] 46.Kurhe, Y., Mahesh, R. & Devadoss, T. QCM-4, a 5-HT(3) receptor antagonist ameliorates plasma HPA axis hyperactivity, leptin resistance and brain oxidative stress in depression and anxiety-like behavior in obese mice. Biochem. Biophys. Res. Commun.456, 74–79 (2015). [DOI] [PubMed] [Google Scholar]

[CR47] 47.Kakizawa, K. et al. A novel GABA-mediated corticotropin-releasing hormone secretory mechanism in the median eminence. Sci. Adv.2, e1501723 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR48] 48.Douglass, A. M. et al. Neural basis for fasting activation of the hypothalamic-pituitary-adrenal axis. Nature620, 154–162 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR49] 49.Kim, J. S. et al. Chronic hyperosmotic stress converts GABAergic inhibition into excitation in vasopressin and oxytocin neurons in the rat. J. Neurosci.31, 13312–13322 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR50] 50.Fernandes, A. C. A. et al. Arcuate AgRP, but not POMC neurons, modulate paraventricular CRF synthesis and release in response to fasting. Cell Biosci.12, 118 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR51] 51.Teitelbaum, P. & Epstein, A. N. The lateral hypothalamic syndrome: recovery of feeding and drinking after lateral hypothalamic lesions. Psychol. Rev.69, 74–90 (1962). [DOI] [PubMed] [Google Scholar]

[CR52] 52.Clayton, J. A. & Collins, F. S. Policy: NIH to balance sex in cell and animal studies. Nature509, 282–283 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR53] 53.Shansky, R. M. & Murphy, A. Z. Considering sex as a biological variable will require a global shift in science culture. Nat. Neurosci.24, 457–464 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR54] 54.Bechtold, D. A. & Loudon, A. S. Hypothalamic clocks and rhythms in feeding behaviour. Trends Neurosci.36, 74–82 (2013). [DOI] [PubMed] [Google Scholar]

[CR55] 55.Kohsaka, A. et al. High-fat diet disrupts behavioral and molecular circadian rhythms in mice. Cell Metab.6, 414–421 (2007). [DOI] [PubMed] [Google Scholar]

[CR56] 56.Hatori, M. et al. Time-restricted feeding without reducing caloric intake prevents metabolic diseases in mice fed a high-fat diet. Cell Metab.15, 848–860 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR57] 57.Chung, J. et al. Anxiety and depression in non-cardiac surgical patients and associations with adverse outcomes: a systematic review and meta-analysis. J. Anesth. Transl. Med.4, 207–219 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR58] 58.Bosch-Bouju, C., Larrieu, T., Linders, L., Manzoni, O. J. & Laye, S. Endocannabinoid-mediated plasticity in nucleus accumbens controls vulnerability to anxiety after social defeat stress. Cell Rep.16, 1237–1242 (2016). [DOI] [PubMed] [Google Scholar]

[CR59] 59.Li, Y. et al. The role of gut microbiota in chronic itch-evoked novel object recognition-related cognitive dysfunction in mice. Front. Med.8, 616489 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR60] 60.Walf, A. A. & Frye, C. A. The use of the elevated plus maze as an assay of anxiety-related behavior in rodents. Nat. Protoc.2, 322–328 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR61] 61.Ding, W. et al. The endocannabinoid N-arachidonoyl dopamine is critical for hyperalgesia induced by chronic sleep disruption. Nat. Commun.14, 6696 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

A hypothalamic circuit links high-fat diet exposure to anxiety-like behavior and anxiety-associated hyperphagia

Peng Wang

Jun Li

Xin Cheng

Wei Chen

Zhengyu Li

Zheng Fang

Dijia Wang

Bin Mei

Hu Liu

Shaohua Hu

Zhilai Yang

Jiqian Zhang

Gaolin Qiu

Xin Qing

Xuesheng Liu

Abstract

Introduction

Results

Chronic HFD reveals an anxiety-susceptible subgroup with increased feeding and body weight gain

Fig. 1. Chronic HFD induces anxiety susceptibility with hyperphagia and weight gain.

ArcAgRP neurons are recruited during HFD-associated anxiety and hyperphagia

Fig. 2. Feeding-recruited ArcAgRP neurons drive HFD-associated anxiety and hyperphagia.

ArcAgRP neurons provide input to PVNCRH neurons

Fig. 3. ArcAgRP neurons project to PVNCRH neurons.

ArcAgRP-PVN signaling links HFD-associated anxiety-like behavior to hyperphagia

Fig. 4. ArcAgRP-PVNCRH circuit drives HFD-associated anxiety and hyperphagia.

PVNCRH neurons inhibit feeding-suppressive LHAGlu neurons

Fig. 5. PVNCRH neurons project to LHAGlu neurons.

PVNCRH-LHA signaling links HFD-associated anxiety-like behavior to hyperphagia

Fig. 6. PVNCRH-LHAGlu circuit mediates HFD-associated anxiety and hyperphagia.

A long-range Arc-PVN-LHA circuit mediates state-dependent coupling between anxiety and feeding

Fig. 7. Deciphering the Arc-PVN-LHA circuitry.

Fig. 8. Arc-PVN-LHA circuit mediates state-dependent anxiety-feeding coupling under chronic HFD.

Discussion

Fig. 9. Metabolic and emotional states converge on shared hypothalamic circuitry, producing state-dependent functional coupling between anxiety and feeding behavior under chronic high-fat diet conditions.

Methods

Animals

Experimental model

Hierarchical cluster analysis

Behavioral tests

Food intake and body weight monitoring

Drugs

ELISA

Stereotaxic viral injection

In vivo fiber photometry recording

In vitro electrophysiological recordings

Optogenetic and chemogenetic manipulation

Western blot

Immunofluorescence

Statistics and reproducibility

Reporting summary

Supplementary information

Source data

Acknowledgements

Author contributions

Peer review

Peer review information

Funding

Data availability

Competing interests

Footnotes

Contributor Information

Supplementary information

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases

Arc^AgRP neurons are recruited during HFD-associated anxiety and hyperphagia

Fig. 2. Feeding-recruited Arc^AgRP neurons drive HFD-associated anxiety and hyperphagia.

Arc^AgRP neurons provide input to PVN^CRH neurons

Fig. 3. Arc^AgRP neurons project to PVN^CRH neurons.

Arc^AgRP-PVN signaling links HFD-associated anxiety-like behavior to hyperphagia

Fig. 4. Arc^AgRP-PVN^CRH circuit drives HFD-associated anxiety and hyperphagia.

PVN^CRH neurons inhibit feeding-suppressive LHA^Glu neurons

Fig. 5. PVN^CRH neurons project to LHA^Glu neurons.

PVN^CRH-LHA signaling links HFD-associated anxiety-like behavior to hyperphagia

Fig. 6. PVN^CRH-LHA^Glu circuit mediates HFD-associated anxiety and hyperphagia.