Obesogenic effects of warm temperature involve feeding adaptation by preoptic area leptin receptor neurons

Abstract

The preoptic area (POA) is a well-established regulator of body temperature, but its role in feeding behavior remains underexplored. Our study identifies leptin receptor (Lepr)-expressing neurons in the POA (POA^Lepr) as critical component to suppress food intake (FI) and increase satiety in response to warm ambient temperatures. Utilizing chemogenetic activation in mice of both sexes, we demonstrate that selective activation of POA^Lepr neurons mimics the effects of warm temperatures, leading to a significant reduction in FI. POA^Lepr neurons project to the melanocortin pathway, where activation of melanocortin-4 receptors (MC4R) also suppresses FI in a temperature-dependent manner. Our findings suggest that POA^Lepr neurons integrate thermal and metabolic cues, demonstrating that ambient temperature is an integral part of body weight homeostasis by modulating meal size and satiety via POA^Lepr neurons. These results offer new insights into the neurochemical and functional properties of POA functions, expanding the traditional view that the POA is exclusively involved in thermoregulation and underscoring its broader role in energy balance.

Warm-sensing leptin receptor neurons in the preoptic hypothalamus suppress feeding by reducing meal size and increasing satiety. This study shows they integrate temperature and metabolic signals via melanocortin circuits to regulate body weight.

Introduction

With obesity rates continuing to rise globally, there is an urgent need to identify the neural circuits that integrate environmental and metabolic cues to develop more effective strategies for prevention and treatment. Body weight homeostasis depends on a dynamic balance between food intake (FI) and energy expenditure (EE), governed by distinct but interacting neural pathways. Circuits that mediate changes in FI have been extensively studied in the context of energy availability—such as during fasting and refeeding, but far less is known about how the brain adjusts feeding behavior in response to other physiological challenges, including ambient temperature. Thermoregulatory circuits, particularly those driving thermogenesis and EE, have primarily been studied in the context of cold exposure or fever and are known to involve the preoptic area (POA) of the hypothalamus¹. Despite its central role in temperature regulation, the POA has not been traditionally recognized as a regulator of FI or body weight (BW)².

The POA is a hypothalamic structure consisting of several nuclei, including medial and lateral POA, and receive multi-synaptic spinal sensory input from the skin to communicate changes in ambient temperature. Distinct warm- and cold-activated neurons within these regions also intrinsically detect small changes in temperature and cause temperature-dependent changes in neuronal firing^3,4. Recent work has shown that cold- or warm-activated neurons that both express the vesicular-glutamate-transporter-2 (POA^Vglut2)^5,6 and that these POA^Vglut2 neurons influence temperature-dependent feeding behavior. Bulk POA^Vglut2 activation suppresses FI, while projection-specific activation revealed a mix of orexigenic and anorexigenic responses⁵. Similarly, bulk activation lowers body temperature—a warm-adaptive response—yet increases nesting behavior, which is typically associated with cold exposure⁵. These findings suggest that simultaneous activation of warm- and cold-responsive POA^Vglut2 populations produce competing thermoregulatory behaviors⁷, complicating interpretation of how warm-specific circuits regulate FI.

Our lab previously identified a distinct warm-sensing POA population expressing leptin receptors (POA^Lepr)⁸ that respond to circulating leptin. Leptin is secreted by adipose tissue to regulate adiposity by adapting FI and EE, and its effectiveness depends strongly on energy state⁹. POA^Lepr neurons are embedded in classic thermoregulatory circuits that suppress thermogenesis, inhibit FI and promote BW loss^8,10. Importantly, we and others have shown that warm-activated POA neurons, specifically POA^Lepr neurons, are glutamatergic^8,11–13, a notable shift from earlier models viewing warm-sensing POA neurons primarily as GABAergic^14,15. This highlights Lepr expression as an effective marker for isolating a uniform, warm-activated POA population and enables the investigation of warm-induced FI suppression without confounding cold-activated circuits.

Downstream of the POA, the arcuate nucleus (ARC) is a critical hub for energy sensing and BW regulation, integrating melanocortin signals from anorexigenic Pro-opiomelanocortin (POMC) and orexigenic Agouti-Related Protein (AgRP) neurons¹⁶. POMC neurons are activated to promote satiety during refeeding, whereas AgRP neurons are activated during fasting to stimulate hunger and food-seeking behaviors^17,18. Both populations project to second-order nuclei that express melanocortin-4-receptors (MC4R), including paraventricular hypothalamus (PVN) and dorsomedial hypothalamus (DMH), which modulate meal initiation, satiety, and metabolic adaptations¹⁹.

Recent studies have implicated the ARC in temperature-dependent FI, demonstrating that AgRP neurons are required for cold-induced hyperphagia^5,6,20. Activation of POA^Vglut2 > ARC projections predominantly drives cold-induced FI^5,6, likely through the recruitment of AgRP neurons. However, the ARC also contains POMC neurons that express warm-sensitive receptors²¹, and MC4R activation is important for FI suppression at warm ambient temperatures²². Thus, activating mixed warm- and cold-responsive POA^Vglut2 > ARC projections may obscure contributions of warm-sensing circuits, including POMC neurons and downstream MC4R signaling.

In contrast, activation of POA^Vglut2 projections to other melanocortin pathway nodes, including the PVN and DMH, results in FI suppression that is consistent with warm-induced responses⁵. These findings underscore the complex and region-specific ways that the POA engages in the melanocortin system. Whether the POA modulates FI through direct projections to FI-modulating regions, such as the PVN and DMH, or via ARC neurons remains an important and unresolved question.

The current study identifies POA^Lepr neurons as critical mediators of warm temperature-induced FI suppression and defines downstream circuits that contribute to this effect. Through selective activation of POA^Lepr neurons, we demonstrate that they are sufficient to mimic the feeding suppression observed during warm exposure. Our findings further establish that glutamatergic POA^Lepr neurons integrate thermal and metabolic signals to regulate BW and engage melanocortin nodes, where MC4R activation further suppresses FI in a temperature-dependent manner. Together, these results position POA^Lepr neurons as a central link between thermoregulation and feeding pathways that expand their role in FI adaptation and energy balance regulation.

Results

Temperature-dependent adaptation of energy expenditure and food intake induce changes in body weight set point

We first confirmed the robust physiological adaptations of EE and FI elicited by acute changes in ambient temperature by housing mice at cold (10 °C), room (22 °C) or warm temperature (30 °C) for 48 h (Fig. 1a). As shown previously⁸, exposing mice to warm and cold temperature caused significant changes in FI (Fig. 1b; one-way repeated measures ANOVA: F_{(1.570, 18.84)} = 24.44, p < 0.0001) but not in BW (Fig. 1c; one-way repeated measures ANOVA: F_{(1.357, 18.32)} = 1.350, p = 0.2728), due to simultaneous energetic balancing with suppressed EE (Fig. 1d; one-way repeated measures ANOVA: F_{(1.925, 25.99)} = 1058, p < 0.0001). Note, significant FI changes between 22 °C and 30 °C occurred during the light phase, blunting effects over the full 48 h (Fig. 1b, inset).

Fig. 1. Temperature-dependent adaptation of energy expenditure and food intake alters body weight set point.

a Schematic overview of acute 48 h exposure to warm (30 °C), room (22 °C), or cold (10 °C) temperatures in male and female mice (n = 15–16, 7–8 males, 8 females). b 12 h food intake during light phase (inset shows 48 h FI (kcal/gBW)). c 48 h change in body weight (%). d Average energy expenditure during 48 h exposure. e Schematic overview of chronic 4wk exposure to cold (4 °C), room (22 °C), or warm (28 °C) temperatures over four weeks (n = 12, all males). f 24 h food intake at the end of 4wk exposure. g 4wk change in body weight (%). h Change in fat mass accumulation. Data are presented as mean ± SEM, ***p < 0.001, ****p < 0.0001 (one-way repeated measures ANOVA with Bonferroni’s post hoc test for (b–d); one-way ANOVA with Bonferroni’s post hoc test for (f, g); two-way repeated measures ANOVA with Bonferroni’s post hoc test for (h)). Males are represented by black circles, females by white circles. Created in BioRender. Kaiser, L. (2026) https://BioRender.com/zzcd4f1.

While these acute adaptations are well-documented, the long-term consequences on BW have received less attention. To address this, we assessed whether chronic housing at cold (4 °C) or warm (28 °C) temperatures over four weeks alters BW homeostasis (Fig. 1e). As expected, sustained warm exposure decreased FI compared to 22 °C and 4 °C (Fig. 1f; one-way ANOVA: F_{(2, 31)} = 237.2, p < 0.0001). Chronic temperature exposure also significantly altered BW (Fig. 1g, one-way ANOVA: F_{(2, 31)} = 25.71, p < 0.0001) and fat mass (Fig. 1h). Cold exposure prevented the normal gain in fat mass seen at 22 °C, whereas warm exposure increased fat mass (Fig. 1h; repeated-measures ANOVA: F_{(8, 124)} = 15.75, p < 0.0001).

These findings establish that ambient temperature is a potent exteroceptive cue that impacts energy balance beyond immediate thermogenic responses. Despite reduced FI, the concurrent EE suppression contributes to BW gain and adiposity, emphasizing the interplay of FI and EE balance in temperature-driven metabolic adaptations. While many studies have focused on EE in thermoregulatory control, the neural circuits responsible for warm-induced suppression of FI remain largely unexplored.

Chemogenetic activation of POA^Lepr neurons mediates temperature-dependent adaptations of feeding behavior

Previous studies showed that POA^Lepr neurons exhibit warm-sensing properties^8,11–13 and modulate EE in a temperature-dependent manner⁸. Moreover, chemogenetic activation of POA^Lepr neurons decrease FI⁸, mimicking the natural effects observed during warm exposure. Based on these findings, we hypothesized that activating POA^Lepr neurons would similarly suppress FI temperature-dependently.

To test this, we selectively expressed DREADD-Gq in POA^Lepr neurons of Lepr^cre mice (Fig. 2a), enabling synthetic activation of these neurons. As shown previously⁸, POA^Lepr neurons are highly sensitive to chemogenetic stimulation and suppress EE, core body temperature, and locomotor activity (LA). We evaluated clozapine N-oxide (CNO; i.p.) across doses and identified 0.01 mg/kg as the lowest dose that robustly reduced EE, with effects waning within 3 h (Fig. S1a; ANOVA time/treatment interaction: F_{(1440, 8640)} = 3.668, p < 0.0001). This CNO dose significantly reduced 6-h LA (Fig. S1b; repeated-measures ANOVA: F_{(5, 30)} = 33.23, p < 0.0001) but LA and FI were not significantly correlated (Fig. S1c; correlation: r₍₈₎ = 0.4495, p = 0.2638). Thus, 0.01 mg/kg CNO was used in subsequent experiments to maximize EE effects while minimizing potential LA confounds.

Fig. 2. Activation of POA^Lepr neurons mediates temperature-dependent feeding adaptations.

a Schematic showing AAV5-hSyn-DIO-hM3D(Gq)-mCherry injection into the POA of Lepr^Cre mice, with a representative histological image confirming viral spread and viral injection maps. Each color represents an individual animal. b Experimental timeline assessing the effects of chemogenetic activation of POA^Lepr neurons with CNO (0.01 mg/kg, i.p.) or PBS during 24-h exposure to cold (10 °C) or warm (30 °C) temperatures (n = 10, 5 males, 5 females). c Total food intake over 6 h. d Change in 6-h food intake, demonstrating temperature-dependent suppression of feeding following POA^Lepr activation. e Locomotor activity (LA) remains unchanged across conditions. Data are presented as mean ± SEM. Different letters above bars indicate statistically significant differences between groups. **p < 0.01 (two-way repeated measures ANOVA with Bonferroni’s post hoc test for (c) and (e), paired t-test for (d)). Males are represented by black circles, females by white circles. Created in BioRender. Kaiser, L. (2026) https://BioRender.com/kprb4r5.

Mice were then exposed overnight to either 10 °C or 30 °C to physiologically silence or activate POA^Lepr neurons, respectively⁸, followed by injection with CNO (0.01 mg/kg, i.p.) or PBS the next morning (Fig. 2b). At 10 °C, CNO significantly suppressed FI to levels comparable to natural suppression at 30 °C (Fig. 2c; ANOVA temperature/treatment interaction: F_{(1, 36)} = 9.902, p = 0.0033). In contrast, when POA^Lepr neurons were already physiologically active at 30 °C, CNO had no significant effect on FI. Consistent with a temperature-dependent effect, net CNO-induced FI changes were greater at 10 °C versus 30 °C (Fig. 2d; paired t-test: t₍₉₎ = 3.832, p = 0.0040). These findings demonstrate that POA^Lepr neurons mediate acute, temperature-dependent suppression of FI.

To rule out the possibility that FI suppression was secondary to reduced LA, we again examined LA responses. Although CNO trended to decrease LA at 10 °C, the treatment × temperature interaction was not significant (Fig. 2e; ANOVA interaction: F_{(1, 36)} = 4.015, p = 0.0527). Thus LA may contribute to the observed FI effects but the lack of significant interaction indicates LA is unlikely to be the primary driver.

Temperature-dependent adaptations in meal patterns

Most studies assess feeding by measuring total food consumed over time. However, feeding is a complex behavior composed of distinct components, such as meal size, frequency, duration, and satiety, and anatomically and functionally distinct neural circuits may regulate these behavioral components. To determine whether ambient temperature selectively alters specific aspects of feeding behavior, we performed detailed meal pattern analysis during cold (10 °C; n = 16) or warm (30 °C; n = 16) exposure (Fig. 3).

Fig. 3. Temperature-dependent adaptations in meal patterns.

a Schematic overview of acute 48 h exposure to warm (30 °C) versus cold (10 °C) ambient temperatures. b Locomotor activity across 24 h. c Total food intake analyzed by light/dark phases and temperature conditions (n = 16, 8 males, 8 females). d Heat map showing feeding events over 24 h at 10 °C and 30 °C, with color-coded meal sizes. e Average meal size. f Average meal duration. g Satiety index (minutes of feeding per gram of food consumed). h Total number of meals. i Average meal-to-meal interval (MMI) = time between the start of one meal and the start of the subsequent meal. Data are presented as mean ± SEM. Different letters above bars indicate statistically significant differences between groups. **p < 0.01, ***p < 0.001 (two-way repeated measures ANOVA with Bonferroni’s post hoc test for (b) and (i), paired t-test for (c–h)). Males are represented by black circles, females by white circles. IMI inter-meal interval, LA locomotor activity. Created in BioRender. Kaiser, L. (2026) https://BioRender.com/m7w0z4r.

Mice were acclimated to assigned temperatures prior to 24 h meal-pattern data collection, 18 h to 42 h of exposure), capturing one complete light–dark cycle (Fig. 3a). Neither temperature condition significantly altered LA (Fig. 3b, repeated-measures ANOVA main effect: F_{(1, 30)} = 1.257, p = 0.2711; Fig. S3b). As expected, mice exhibited robust circadian feeding patterns at both temperatures (Fig. 3c, Fig. S3a), with significantly more food consumed during the dark phase (light vs. dark phase: p < 0.0001), consistent with prior reports at room temperature^23,24. Interestingly, FI suppression at 30 °C was specific to the light phase (p_bonf = 0.0032), with no difference during the dark phase (p_bonf = 0.4305).

We further quantified and compared FI patterns across the 24 h exposure. A heat map summarizing individual feeding events for each animal is shown in Fig. 3d, providing a visual representation of individual meals, color-coded by size in kcal. This display highlights temporal feeding bouts and relative meal sizes across individual subjects and conditions.

Quantitative analysis revealed that warm exposure significantly reduced meal size (Fig. 3e; paired t-test: t₍₁₅₎ = 4.583, p = 0.0004) and meal duration (Fig. 3f; paired t-test: t₍₁₅₎ = 3.436, p = 0.0037). Warm temperature also increased satiety, reflected by a higher ratio of time between meals to the amount of food eaten (Satiety ratio; Fig. 3g; paired t-test: t₍₁₅₎ = 3.182, p = 0.0062). In contrast, meal numbers (Fig. 3h; paired t-test: t₍₁₅₎ = 0.5005, p = 0.6240) and meal-to-meal intervals (MMI; Fig. 3i; paired t-test: t₍₁₅₎ = 0.4216, p = 0.6793) remain unchanged.

Together, these findings demonstrate that warm ambient temperature selectively alters components of feeding behavior—most notably meal size, duration, and satiety—while preserving meal frequency and timing.

POA^Lepr neuron activation decreases meal size and increases satiety

To determine whether activation of POA^Lepr neurons recapitulates the feeding behavior observed during warm exposure, we performed detailed meal pattern analysis following chemogenetic stimulation of this population. To avoid fluctuations in CNO levels from i.p. bolus injections and to minimize confounding effects from changes in LA, we first tested a range of CNO doses delivered via drinking water over 24 h. We identified a dose of 0.05 mg/kg/day as ideal to effectively suppress EE (Fig. S2a) without altering LA (Fig. S2b) for our experiments.

Mice with POA^Lepr specific DREADD-Gq expression (Fig. 4a) received CNO in the drinking water (0.00025 mg/ml) for 24 h (Fig. 4b), which had no effect on water consumption (Fig. 4c; paired t-test: t₍₄₎ = 2.532, p = 0.0625), but significantly reduced EE (Fig. 4d; repeated-measures ANOVA for the effect of treatment: F_{(47, 376)} = 2.733, p < 0.0001), with no significant effect on LA (Fig. 4e, Fig. S3c; paired t-test: t₍₈₎ = 1.428, p = 0.1913). Heat maps confirmed that CNO-treated mice continued feeding across the 24 h period, even though with overall reduced FI compared to water controls (Fig. 4f). Chemogenetic activation of POA^Lepr neurons significantly reduced 24 h FI (Fig. 4g, Fig. S3d; paired t-test: t₍₄₎ = 4.054, p = 0.0154).

Fig. 4. POA^Lepr activation recapitulates warm-temperature induced satiety.

a Schematic depicting AAV5-hSyn-DIO-hM3D(Gq)-mCherry injection into the POA of Lepr^Cre mice. b Experimental design for continuous POA^Lepr activation. c–e A low-dose CNO regimen (0.05 mg/kg/day in drinking water (amount consumed (c); n = 5, 2 males, 3 females) significantly reduces energy expenditure (d) while maintaining locomotion for feeding by limiting body temperature drop (e) compared to mice given tap water. f Heat map of feeding events and meal sizes over 24 h. g Total food intake over 24 h. h Average meal size. i Average meal duration. j Satiety index. k Total meal count. l Average meal-to-meal interval (MMI) = time between the start of one meal and the start of the subsequent meal. Data are presented as mean ± SEM. *p < 0.05, **p < 0.01 (two-way repeated measures ANOVA with Bonferroni’s post hoc test for (c), paired t-test for (d), (f–k)). Males are represented by black circles, females by white circles. Created in BioRender. Kaiser, L. (2026) https://BioRender.com/s6nrzhy.

Similar to warm exposure, POA^Lepr activation selectively reduced meal size (Fig. 4h; paired t-test: t₍₄₎ = 3.067, p = 0.0374) and increased satiety (Fig. 4j; paired t-test: t₍₄₎ = 2.838, p = 0.0470), while meal duration remained unchanged (Fig. 4i; paired t-test: t₍₄₎ = 1.874, p = 0.1343). Note, meal durations at 22 °C (~10 min) were already shorter than those observed at 10 °C (~12.5 min; see Fig. 3d), which limited the dynamic range for additional reductions. In contrast to warm exposure, POA^Lepr activation also decreased total meal number (Fig. 4k; paired t-test: t₍₄₎ = 5.522, p = 0.0053) and slightly increased MMI (Fig. 4l; paired t-test: t₍₃₎ = 3.365, p = 0.0436).

Together, these results demonstrate that POA^Lepr neuron activation suppresses FI by selectively modulating feeding microstructure, particularly via reduced meal size and increased satiety, supporting their role as warm-sensitive regulators of energy homeostasis.

POA^Lepr neurons project to the melanocortin FI pathway and decrease FI in an energy-state-dependent manner

To selectively trace the projections of warm-activated POA^Lepr neurons, we injected a cre-inducible EGFP reporter virus into the POA of Lepr^Cre mice (Fig. 5a). Labeled axons were present in canonical feeding centers including the DMH, ARC, and PVN (Fig. 5b, c) that are associated with distinct components of feeding regulation: the DMH modulates diurnal feeding rhythms²⁵ with GABAergic Lepr neurons that inhibit AgRP neurons to suppress FI²⁶; the ARC harbors orexigenic POMC neurons involved in satiety, meal size, and duration^27–29; and the PVN integrates POMC and AgRP input via MC4R-expressing neurons to regulate metabolic aspects of FI^30–32.

Fig. 5. WS-POA^Lepr neurons project to satiety-related brain regions and suppress FI following refeeding.

a Example of an injection site in the POA of a Lepr^Cre mouse injected with rAAV5/Efla-DIO-hChR2(H134R)-mCherry. b, c POA^Lepr neurons project to hypothalamic regions involved in satiety regulation, including the lateral arcuate nucleus (lARC) and dorsomedial hypothalamus (DMH) (b) and the paraventricular hypothalamus (PVN) (c). d Experimental timeline assessing the effects of chemogenetic activation of POA^Lepr neurons with CNO (0.01 mg/kg, i.p.) or PBS when mice are either fed (n = 10, 5 males, 5 females) or refed following an overnight fast (n = 10, 5 males, 5 females). e Total food intake over 6 h. f Change in 6-h food intake between PBS and CNO conditions. Data are presented as mean ± SEM. Different letters above bars indicate statistically significant differences between groups. **p < 0.01 (two-way repeated measures ANOVA with Bonferroni’s post hoc test for (e), paired t-test for (f)). Males are represented by black circles, females by white circles.

Given these projection targets, we hypothesized that POA^Lepr neurons modulate FI through the melanocortin pathway. Thus, we examined whether POA^Lepr mediated FI suppression depends on energy state, with enhanced effects when melanocortin signaling is suppressed by fasting. Mice expressing DREADD-Gq in POA^Lepr neurons received CNO (0.01 mg/kg) or PBS either after an overnight fast (just before refeeding) or while continuously fed. FI was measured over 6 h (Fig. 5d). POA^Lepr activation significantly suppressed FI in fasted/refed mice but had minimal effect in continuously fed mice (Fig. 5e; ANOVA food availability/treatment interaction: F_{(1, 36)} = 98.90, p < 0.0001). This energy-state-dependent effect was further supported by a comparison of net CNO-induced FI suppression, which was significantly greater in fasted/refed mice (Fig. 5f; paired t-test: t₍₉₎ = 3.102, p = 0.0127).

These findings indicate that POA^Lepr-mediated anorexia depends on energy state and is most effective when melanocortin signaling is low, supporting a functional interaction between POA^Lepr neurons and the melanocortin pathway in FI regulation.

Pharmacologically induced satiety via melanocortin-4-receptors is temperature-dependent

Building on our findings that POA^Lepr neuron activation suppresses FI in a temperature-dependent manner and based on their projections to key nodes of the melanocortin pathway, we hypothesized that MC4R-expressing neurons may similarly mediate satiety at warm ambient temperatures. Prior work has shown that melanocortin signaling modulates meal size and satiety^27,29,31, effects that closely resemble those induced by POA^Lepr activation. If MC4R neurons contribute to temperature-dependent FI regulation, central activation of this pathway should suppress FI more effectively at cold than warm ambient temperatures.

We implanted ICV cannulas (Fig. 6a) and verified accurate placement (angiotensin II-induced drinking test, Fig. 6b). Mice were housed overnight at either 10 °C or 30 °C and the MC4R agonist Melanotan II (MTII; 0.3 nmol) or PBS were injected ICV the following morning (Fig. 6c). MTII significantly suppressed 6 h FI at 10 °C, reducing intake to levels comparable to those at 30 °C (Fig. 6d; ANOVA for temperature/treatment interaction: F_{(1, 32)} = 4.872, p = 0.0346). Moreover, the magnitude of MTII-induced FI suppression was greater at 10 °C than at 30 °C (Fig. 6e; paired t-test: t₍₈₎ = 2.826, p = 0.0223). These results support a role for MC4R signaling in temperature-dependent FI suppression and implicate MC4R-expressing neurons as likely downstream effectors of POA^Lepr-mediated satiety, although the precise site of action, e.g., through PVN, DMH, or ARC, remains to be determined.

Fig. 6. Temperature-dependent effects of pharmacologically induced satiety via melanocortin-4 receptor activation.

a Schematic illustrating chronic intracerebroventricular (ICV) cannula implantation in mixed-background mice. b Two-hour water intake following Angiotensin II injection (50 ng, ICV) confirmed cannula placement. c Experimental timeline: MC4R agonist MTII (0.3 nmol, ICV) or PBS was administered during a 24-h exposure to cold (10 °C) or warm (30 °C) temperatures (n = 9, 3 males, 6 females). d 6-h food intake following MTII versus PBS injection. e Change in 6-h food intake following MTII administration at different temperatures. f Experimental timeline assessing the effects of MTII (0.3 nmol, ICV) or PBS administration when mice are either fed or refed following an overnight fast (n = 9, 3 males, 6 females). g Total food intake over 6 h. h Change in 6-h food intake between PBS and MTII conditions. i Representative immunohistochemical images showing cFos (red), MC4R-EGFP (green), and cFos/MC4R-EGFP (merged) in the PVN after 3-h exposure to cold (10 °C; n = 6, 2 males, 4 females) or warm (30 °C; n = 6, 3 males, 3 females) in MC4R reporter mice. Insets show magnified regions. j Total MC4R neuron count in the PVN. k Percentage of cFos+ neurons colocalized with MC4R in the PVN. l Representative immunohistochemical images showing cFos (red), MC4R-EGFP (green), and cFos/MC4R-EGFP (merged) in the DMH after 3-h exposure to cold (10 °C; n = 7, 3 males, 4 females) or warm (30 °C; n = 6, 3 males, 3 females) in MC4R reporter mice. Insets show magnified regions. m Total MC4R neuron count in the DMH. n Percentage of cFos+ neurons colocalized with MC4R in the DMH. Data are presented as mean ± SEM. Different letters above bars indicate statistically significant differences between groups. *p < 0.05, **p < 0.01 (two-way repeated measures ANOVA with Bonferroni’s post hoc test for (d), (g), paired t-test for (e), (h), t-test for (j, k, m, n)). Males are represented by black circles, females by white circles. Created in BioRender. Kaiser, L. (2026) https://BioRender.com/8nw87lt.

We further assessed if MTII effects vary with energy state. Mice received an ICV injection of MTII (0.3 nmol) or PBS either after an overnight fast, just prior to refeeding, or while continuously fed. FI was measured over 6 h (Fig. 6f) with MTII tending to produce greater FI suppression in fasted/refed compared to fed mice, mirroring the energy-state-dependence seen with POA^Lepr activation, even though this did not reach significance (Fig. 6g; ANOVA food/treatment interaction: F_{(1, 32)} = 2.463, p = 0.1264; Fig. 6h; paired t-test: t₍₈₎ = 1.362, p = 0.2104). Together, these findings indicate that MC4R-mediated FI suppression is temperature-dependent and likely engaged by POA^Lepr neurons to adapt feeding to environmental and physiological cues.

Warm ambient temperature decreased neuronal activation in PVN-MC4R neurons

We next examined whether warm exposure modulates the activation of MC4R-expressing populations in downstream hypothalamic targets. MC4R^Cre-GFP reporter mice were exposed to 3 h of 10 °C or 30 °C ambient temperature before perfusion (Fig. 6i). We focused on the PVN and DMH, two hypothalamic regions that receive POA^Lepr projections (Fig. 5), contain dense MC4R-expressing populations, and regulate feeding and energy balance³³. Given our prior findings that MC4R activation suppresses FI in a temperature- and energy-state-dependent manner, we hypothesized that warm exposure would alter recruitment of downstream MC4R neurons.

Quantification of MC4R-expressing neurons in the PVN revealed no significant differences in total MC4R cell counts between cold- and warm-exposed mice (Fig. 6i, j; independent t test: t₍₁₀₎ = 1.080, p < 0.3053). However, cFos analysis revealed significantly greater activation of PVN^MC4R neurons under cold compared to warm conditions (Fig. 6k; independent t test: t₍₁₀₎ = 4.183, p = 0.0019), indicating that ambient temperature dynamically regulates PVN^MC4R neuron recruitment.

In the DMH, total MC4R cell counts and cFos activation of DMH^MC4R neurons were both unchanged between temperature conditions (Fig. 6l, m; independent t test: t₍₁₁₎ = 1.629, p = 0.1315; Fig. 6n; independent t test: t₍₁₁₎ = 0.2303, p = 0.8221). These findings suggest that DMH^MC4R neurons are not differentially recruited by acute changes in ambient temperature and highlight a selective role for PVN^MC4R neurons in mediating temperature-dependent neural and behavioral adaptations.

POA^Lepr neuron activation increases POMC neuron activation in the anterior ARC

The regulation of FI by POMC and AgRP neurons in the ARC is well-established but their interaction with thermoregulatory circuits remains poorly defined. Recent work showed that cold exposure robustly activates AgRP neurons^5,6, whereas POMC neurons appeared responsive to warm temperatures²¹.

We tested whether warm ambient temperature alone activates POMC neurons by comparing cFos expression in POMC neurons following exposure to 10 °C or 30 °C (Fig. 7a). Neither total POMC neurons (Fig. 7b; ANOVA temperature x location: F_{(3, 43)} = 0.1.075, p = 0.3696) nor the percentage of cFos + /POMC+ cells (Fig. 7c; ANOVA temperature x location: F_{(3, 43)} = 0.7723, p = 0.5159) differed between temperatures, suggesting warm exposure alone is insufficient to activate POMC neuron.

Fig. 7. POA^Lepr activation increases POMC neuron activation in the anterior ARC.

a Representative immunohistochemical images showing cFos (red), POMC (cyan), and cFos/POMC (merged) after 3-h exposure to cold (10 °C; n = 7, 3 males, 4 females) or warm (30 °C; n = 6, 3 males, 3 females) in MC4R reporter mice. b Total POMC neuron count in the ARC. c Percentage of cFos+ neurons colocalized with POMC in the ARC. d Schematic depicting AAV5-hSyn-DIO-hM3D(Gq)-mCherry injection into the POA of Lepr^Cre mice. e Representative immunohistochemical images showing cFos (red), POMC (cyan), and cFos/POMC (merged) after CNO (0.01 mg/kg; n = 5, 2 males, 3 females) or PBS (n = 6, 4 males, 2 female) injection during 3-h exposure to cold (10 °C). f Total POMC neuron count in the ARC. g Percentage of cFos+ neurons colocalized with POMC in the ARC. h Representative immunohistochemical images showing cFos expression in MC4R reporter mice after 3 h of cold exposure (10 °C; n = 6) and warm exposure (30 °C; n = 10). Insets indicate the medial and lateral ARC regions used for quantification. i–k Quantification of cFos+ neurons in the total ARC (i), medial ARC (j), and lateral ARC (k) under cold and warm conditions. l Representative immunohistochemical images showing cFos expression after CNO (0.01 mg/kg; n = 5) or PBS (n = 6) injection during cold exposure. Insets show medial vs. lateral ARC regions. m–o Quantification of cFos+ neurons in the total ARC (m), medial ARC (n), and lateral ARC (o) following CNO vs. PBS injection during cold exposure. Data are presented as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001 (two-way ANOVA with Bonferroni’s post hoc test for (b), (c), (h–g); t-test for (i–k, m–o)). Males are represented by black circles, females by white circles. Created in BioRender. Kaiser, L. (2026) https://BioRender.com/b00hk7y.

Given POA^Lepr projections to the ARC and that MC4R activation suppresses FI, we next tested whether POA^Lepr neurons drive POMC activation. To maximize the dynamic range to detect activation, mice were cold-exposed (reducing baseline POMC activity) and injected with CNO (0.01 mg/kg, i.p.) 1.5 h before perfusions (Fig. 7d, e). Total POMC neuron counts remained unchanged (Fig. 7f; ANOVA treatment × location: F_{(3, 34)} = 0.4131, p = 0.7447) but chemogenetic activation significantly increased cFos in POMC neurons of the rostral ARC (Fig. 7g; ANOVA main effect: F_{(3, 34)} = 20.80, p < 0.0001; t-test with Bonferroni correction: t₍₃₄₎ = 4.402, p_bonf = 0.0004).

These results support a model in which POA^Lepr neurons engage POMC neurons in the ARC to suppress FI, potentially through downstream activation of MC4R-expressing targets.

POA^Lepr neurons do not suppress medial ARC neurons but stimulate an unknown population of lateral ARC neurons

Because AgRP neurons are activated during cold exposure^5,6, we next investigated whether POA^Lepr neurons affect ARC populations associated with hunger signaling. We assessed cFos in the ARC following cold and warm exposure (Fig. 7h–k), confirming 10 °C induced cFos compared to 30 °C (Fig. 7i; independent t test: t₍₁₁₎ = 2.693, p = 0.0209), and anatomical mapping revealed cFos was located to the medial ARC, where AgRP neurons are enriched³⁴ (Fig. 7j; independent t test: t₍₁₁₎ = 3.750, p = 0.0032), while the lateral ARC showed no significant change (Fig. 7k; independent t test: t₍₁₁₎ = 1.777, p = 0.1033).

We next tested whether chemogenetic activation of POA^Lepr neurons suppressed cold-activated ARC neurons (Fig. 7l–o). Surprisingly, total and medial ARC cFos+ counts were unaffected by POA^Lepr activation (Fig. 7m; independent t test: t₍₉₎ = 2.247, p = 0.0513; Fig. 7n; independent t test: t₍₉₎ = 0.8174, p = 0.4348), indicating that POA^Lepr activation suppress FI even when hunger-promoting AgRP neurons remain active. Instead, POA^Lepr activation significantly increased cFos+ neurons in the lateral ARC (Fig. 7o; independent t test: t₍₉₎ = 2.824, p = 0.0199).

Although POMC neurons are typically located in the lateral ARC our prior mapping indicated that POA^Lepr-induced POMC activation is biased toward the anterior ARC. Thus, these findings suggest that POA^Lepr neurons also recruit an additional, unidentified lateral ARC population that may contribute to warm-induced FI suppression.

Discussion

Our work demonstrates that ambient temperature-induced adaptations of FI and EE are not fully balanced and significantly contribute to body weight changes. Chronic warm exposure is obesogenic, increasing adiposity in mice, whereas chronic cold exposure promotes fat loss. Chemogenetic activation of POA^Lepr neurons mimics warm-suppressed feeding in a temperature- and energy-state-dependent manner, suggesting that leptin signaling influences feeding not only through classical homeostatic pathways (ARC/PVN) but also through thermoregulatory circuits involving the POA, DMH, and raphe pallidus.

MC4R neurons are central regulators of energy-state-dependent adaptations, and our data indicate that MC4R-mediated FI suppression is also temperature- and energy state dependent. This supports a model in which POA^Lepr neurons modulate MC4R neurons to mediate warm-temperature-induced FI suppression. Together, these findings highlight a critical role for POA^Lepr neurons in regulating body weight by modulating meal size and satiety through downstream melanocortin pathways. This suggests that homeostatic “set points” reflect integration of multiple environmental (exteroceptive) inputs.

Leptin is known to suppress homeostatic feeding but its efficiency varies with energy state, being most effective during energy-need states (fasting) when endogenous leptin is low, resulting in enhanced leptin-suppressed FI during refeeding^35,36. Our work extends these classical concepts by demonstrating that POA^Lepr-mediated FI suppression depends on physiological context (thermal and energy states). Although several studies report that cold ambient temperature decreases serum leptin levels in multiple species^37,38, our prior work showed that leptin itself does not acutely alter EE or FI across temperatures³⁹, and leptin-deficient mice maintain an increased level of FI that does not change during warm or cold ambient temperature⁴⁰. Thus, acute POA^Lepr activation during warmth is unlikely to be driven solely by circulating leptin and instead reflects integration of thermal cues with leptin signaling. Consistently, selective Lepr deletion in the POA has minimal baseline effects but impairs adaptive responses under metabolic challenges, in which POA-specific Lepr knockout mice show increased weight gain on high-fat diet and prevention of fasting-induced hypometabolism³⁹. These findings suggest POA^Lepr neurons are recruited during adaptive challenges but have limited impact in unstressed states⁴⁰. Therefore, leptin action in POA^Lepr neurons may contribute to the obesogenic effect during chronic warmth, potentially through rising leptin levels and leptin resistance. This is consistent with prior work showing POA^Lepr neurons primarily mediate adaptive responses under challenge while having limited impact when energy balance is maintained.

Our data further supports thermotherapy (treatment with warm temperature) for patients with anorexia nervosa. Despite concerns that warmth might further reduce FI, our findings indicate that warm temperature promotes overall weight gain, likely due to disproportionate reduction in EE and locomotor activity, which aligns with clinical observations^41,42. Thus, temperature manipulation may have therapeutic relevance not only for anorexia nervosa but also other eating-disorders and weight-loss conditions.

Our findings align with a recent study showing that activation of POA^Vglut2 neurons suppress FI⁵. However, that study did not distinguish cold- vs warm-sensing POA^Vglut2 populations, resulting in mixed adaptive behaviors with projection-specific activation. Notably, POA^Lepr neurons are glutamatergic and exclusively warm-activated⁸, with single-nuclei RNA data supporting a homogenous warm-sensing POA population¹². By focusing on POA^Lepr neurons, we isolate a warm-sensing population that mediates temperature-dependent FI suppression. Whether POA^Lepr neurons are the sole warm-sensing neuronal POA population remains unclear; other markers (TRPM2, PACAP/BDNF, EP3R, OPR5) overlap with POA^Lepr neurons^43–46 and suppress EE and body temperature. The variable potency of body temperature suppression using other POA markers to drive neuronal activation compared to POA^Lepr activation suggests either mixed warm- and cold-activated populations (PACAP/BDNF) or additional warm-sensing populations (TRPM2). Other POA populations, such as sleep-inducing galanin neurons in the ventrolateral POA, do not co-localized with Lepr⁴⁷ but nonetheless promote temperature-dependent EE suppression⁴⁸, and activation of estrogen-receptor-alpha-expressing POA neurons elicit slower-onset EE suppression⁴⁶, indicating that other populations might exist, though their integration in homeostatic circuits remains poorly defined⁴⁹.

We previously showed that POA^Lepr neurons are naturally activated at warm ambient temperature^8,11–13, and that chemogenetic activation at warmth (when these neurons are already endogenously active) has little additional effect on EE⁸. Consistent with this, we now show that POA^Lepr activation robustly suppresses FI during cold but not warm conditions, reaffirming their temperature-dependent role. Moreover, both warm exposure and POA^Lepr activation suppress FI by reducing meal size and duration while increasing satiety, implicating modulation of meal termination^50,51. Thus, these neurons are key mediators of thermoregulatory feeding suppression and integrate with homeostatic feeding adaptations.

Prior studies identified POA projections to the lateral hypothalamus, the bed nucleus of the stria terminalis, and the periaqueductal gray that had no effect on food intake⁵, suggesting they are unlikely mediators of temperature-dependent feeding and were not pursued here. In contrast, projections to the DMH, PVN, and ARC showed temperature-dependent feeding regulation^5,6, which we confirm. POA^Vglut2 > DMH projections suppress FI, but DMH circuits primarily regulate the appetitive stage (food seeking and meal initiation) via GABAergic>ARC^Agrp and >ARC^POMC projections^52,53 and silencing of DMH^Lepr neurons alters diurnal feeding rhythms^8,25,54,55. Notably, POA^Vglut2 > DMH projections suppress FI temperature-independently⁵. In contrast, the PVN is central to hunger regulation and satiety⁵⁶. POA^Vglut2 > PVN projection-specific activation decreases FI in a temperature-dependent manner⁵ and warm-suppressed FI is absent with pharmacologically inhibited MC4R neurons²², implying both direct and indirect POA > PVN routes contribute to warm-induced FI suppression. Our MTII experiments similarly show stronger FI suppression at cold temperatures, suggesting temperature cues modulate melanocortin sensitivity. However, we observed greater cFos expression in PVN^MC4R neurons with cold versus warm exposure, consistent with reports that PVN^MC4R activation increases EE and BAT thermogenesis^57–59 but at odds with work implicating PVN^MC4R neurons exclusively in FI, not EE control³⁰. Thus, while our data support MC4R involvement in temperature-dependent feeding, they do not confirm PVN^MC4R neurons as critical mediators and raise the possibility of MC4R action outside the PVN. Furthermore, additional warm-sensing neurons may contribute to temperature-dependent feeding. Future studies should test if MC4R are required for POA^Lepr-driven changes in FI, EE, and cFos activation.

POA^Lepr neurons are glutamatergic, so the suppression of cold-activated POA^MC4R neurons is unlikely to reflect direct activation of MC4R neurons and a relay via the ARC is plausible. Prior studies show that cold-activated POA^Vglut2 > ARC projections increase AgRP neuronal activity, promote FI^5,6,10,20 and cold-induced ARC^Agrp activation is required for cold-induced FI²⁰. Importantly, we demonstrate that POA^Lepr activation suppresses FI potently during cold exposure while ARC^AgRP activation remains unchanged, indicating POA^Lepr neurons override the orexigenic drive of ARC^AgRP activation downstream of AgRP neurons, possibly via multiple pathways including MC4R signaling, GABAergic and NPY receptor signaling⁶⁰. POA^Lepr activation also induced cFos in anterior ARC^POMC neurons, consistent with direct glutamatergic input and with enhanced satiety mediated through POMC > MC4R pathways. Given that POMC loss increases meal size²⁷, these mechanisms align with the meal-structure changes observed during warm-exposure and POA^Lepr activation.

Although prior reports indicate warm temperature increases POMC activation^21,22,61, we did not detect significant cFos/POMC colocalization after acute warm exposure. POMC neurons require prolonged activation to suppress FI^62–64, POA recruitment increases with extended heat exposure⁴, and POA^Lepr neurons contribute to extreme heat adaptations⁴. Thus, even with very low chemogenetic drive, our manipulation may resembles a hotter than a thermoneutral signal, potentially explaining the lack of POMC activation at 30 °C and the stronger effects with synthetic POA^Lepr activation.

Furthermore, POA^Lepr activation significantly increased cFos in non-POMC neurons within the lateral ARC. The identity of these neurons is unknown and may represent additional FI suppressing populations downstream of POA^Lepr activation^5,65. Future studies should characterize their molecular identity and contribution to thermoregulatory feeding.

Our data investigated acute FI changes across both sexes. Sex differences in energy homeostasis are well-established (body weight, FI, hypothalamic control)^66,67, and POA-mediated metabolic adaptation to thermal and nutritional stress can be sexually dimorphic, with females showing enhanced thermogenic and diet-responsive changes⁶⁸. Leptin action on POA^Lepr neurons has also been linked to reproductive function and luteinizing hormone release⁶⁹. However, our prior work found that POA-Lepr deletion did not affect estrous cycling, and POA^Lepr manipulations did not produce sex-specific changes³⁹. Consistently, feeding suppression, meal-structure adaptations, and downstream melanocortin activation were comparable across sexes in this study, suggesting that POA^Lepr neurons engage thermoregulatory feeding circuits in a sex-invariant manner under the conditions tested, while not excluding sex-specific effects under other paradigms.

Taken together, our findings highlight POA^Lepr neurons as important integrators of environmental temperature cues that modulate feeding behavior and energy homeostasis. These neurons interface thermoregulatory and classical homeostatic feeding circuits, particularly through melanocortin interactions, making POA^Lepr neurons a promising therapeutic entry point for metabolic disease.

Methods

Animal care

Male and female mice from the following strains were used in this study: C57BL/6 J (JAX stock #000664), Lepr^Cre (B6.129-Lepr^tm3(cre)Mgmj/J; JAX stock #032457)⁷⁰, tg(POMC) (C57BL/6J-Tg(Pomc-EGFP)1Low/J; JAX stock #009593)⁷¹, and MC4R^Cre (Mc4r^{tm3.1(cre)Lowl}/J; JAX stock #030759)⁷². Lepr^Cre mice and MC4R^Cre mice were crossed with EGFP-L10a (B6;129S4-Gt(ROSA)26Sor^{tm9(EGFP/Rpl10a)Amc}/J; JAX stock #024750)⁷³ or Ai95(RCL-GCaMP6f)-D (C57BL/6 J) (B6J.Cg-Gt(ROSA)26Sor^{tm95.1(CAG-GCaMP6f)Hze}/MwarJ; JAX stock #028865)⁷⁴ to enable genetic labeling of specific neuronal populations. All experiments included both male and female mice unless otherwise noted. While males exhibited slightly higher baseline body weights than females (Supplementary Fig. 4), percent change in body weight between experimental conditions was equivalent across sexes. Accordingly, data were pooled for analysis, as no sex-dependent differences were detected in the magnitude or direction of treatment or temperature effects. Food intake was expressed as kilocalories per gram of body weight (kcal/gBW) to account for individual differences in body size.

Mice were housed individually in standard ventilated cages with ad libitum access to laboratory rodent diet (#5001, LabDiet) and water unless otherwise noted. Animals were maintained on a 12-h light/dark cycle (lights on at 0600 h) in a temperature- and humidity-controlled vivarium. All cages contained standard bedding and environmental enrichment, and husbandry practices adhered to institutional guidelines.

For experiments conducted under standard housing conditions, ambient room temperature was maintained at 22 °C. For thermal challenge paradigms, mice were transferred to temperature-controlled environmental chambers (RIS77SD Rodent Incubator, Power Scientific Inc.; or DB034-LT-GD Laboratory Incubator, Darwin Chambers) set to 4 °C, 10 °C, 28 °C, or 30 °C for durations specified in each experimental protocol. Mice were acclimated to housing and handling for at least one week prior to experimentation. All animal procedures were approved by the Institutional Animal Care and Use Committee of Pennington Biomedical Research Center and complied with NIH guidelines for the care and use of laboratory animals.

Stereotaxic viral injection

We performed stereotaxic viral injection of adenoassociated virus (AAV) as previously described⁵⁵. We used 10–12-week-old Lepr^Cre mice for all surgeries. Mice were deeply anesthetized with 1–4% isoflurane and positioned in a stereotaxic alignment frame (#1900, David Kopf Instruments). Two viral constructs were used in this study: AAV5-hSyn-DIO-hM3D(Gq)-mCherry (a gift from Bryan Roth; Addgene viral prep #44361-AAV5; RRID: Addgene_44361) and rAAV5/Ef1a-DIO-hChR2(H134R)-mCherry (AV5214B; 4.0 × 10¹² vg/mL), the latter obtained from the UNC Vector Core.

Bilateral injections were targeted to the POA using coordinates from the Paxinos and Franklin mouse brain atlas⁷⁵: anteroposterior +0.55 mm, mediolateral ±0.25 mm, and dorsoventral −5.2 mm from Bregma. A total volume of 200–400 nL per side was delivered at a rate of 40 nL/min using a microinfusion system (UltraMicroPump III with MICRO2T, World Precision Instruments). After infusion, the injection needle was left in place for 5 min to minimize reflux. Craniotomies were sealed with bone wax (Lukens #901, Medline Industries), and the scalp was closed using wound clips (#203-1000, CellPoint Scientific).

Postoperative care included administration of warm sterile saline (1.5 mL, i.p.), carprofen (10–15 µL/g, s.c. daily for 72 h), and lidocaine (100 µL, s.c. at the injection site) for analgesia. Mice were individually housed following surgery and allowed to recover for 3–4 weeks to ensure viral expression before experimental procedures.

Only animals with confirmed DREADD-Gq expression in the POA were included in the final analysis (n = 15). Mice with mistargeted or insufficient viral expression (n = 4) were excluded, as these animals failed to exhibit CNO-induced metabolic responses.

Chemogenetic manipulations

For acute chemogenetic activation, mice received a single intraperitoneal injection of phosphate-buffered saline (PBS) or clozapine-N-oxide (CNO; 0.01 mg/kg, i.p.). CNO (freebase; HelloBio #HB1807) was dissolved at 5 mg in 100 µL dimethyl sulfoxide (DMSO) and subsequently diluted in PBS to a working concentration of 0.002 mg/mL for intraperitoneal injection. The resulting final DMSO concentration was less than 0.004% (v/v), a negligible amount with no expected physiological or behavioral effects, therefore, PBS alone was used as the control. Injections were administered following an 18 h fasting period or 18 h of temperature exposure, depending on the experimental paradigm. The selected dose was based on prior studies and pilot data demonstrating robust activation of POA^Lepr neurons with minimal off-target effects (see Supplementary Fig. S1A–C).

In a separate experiment, CNO was administered chronically via drinking water at a dose of 0.00025 mg/ml using water-soluble CNO dihydrochloride (HelloBio #HB6149). Mice were provided with CNO-containing water for a 24-h period during which food intake and metabolic parameters were continuously monitored. Dosage was calculated based on an assumed average body weight of 25 g and estimated daily water consumption of 5 mL per mouse. This delivery approach allowed sustained activation of POA^Lepr neurons while minimizing potential systemic effects (see Supplementary Fig. S2A, B).

Pharmacological manipulations via ICV injection

To manipulate central melanocortin receptor signaling, 10–12-week-old tg(POMC) mice were implanted with unilateral guide cannulas (C315G, Plastics One) targeting the right lateral ventricle. Stereotaxic coordinates relative to Bregma were: anteroposterior −0.3 mm, mediolateral −1.0 mm, dorsoventral −2.1 mm. Mice were anesthetized with 1–4% isoflurane, and cannulas were affixed to the skull with dental cement (C&B Metabond Quick, Parkell). Dummy cannulas (C315DC, Plastics One) were inserted postoperatively to prevent contamination or occlusion. Animals recovered for 3–4 weeks prior to experimental use.

To confirm accurate cannula placement, mice received a test ICV injection of angiotensin II (50 ng in 2 µL sterile saline; BACHEM #4474-91-3), and water intake was measured over 2 h. A positive response was defined as the initiation of robust drinking behavior within 4 min post-injection. This test was repeated daily for up to three trials per animal. Mice that failed to respond were excluded from all subsequent experiments. Animals that passed the placement test were given a minimum of 3 days to recover before beginning the experimental protocol.

For pharmacological manipulation, mice were exposed to either a thermal challenge (10 °C or 30 °C for 24 h) or fasted for 18 h prior to refeeding. At the 18-h time point, mice received a single ICV injection of either PBS or the MC4R agonist melanotan II (MTII; 0.3 nmol in 2 µL sterile PBS; BACHEM #4039778) via the implanted cannula. Injections were performed at 0900 h, and food intake was recorded after 6 h (1500 h).

Environmental temperature manipulations

For both acute and chronic thermal exposure experiments, mice were housed in temperature-controlled chambers (RIS77SD Rodent Incubator, Power Scientific Inc.) set to cold (4 °C or 10 °C), room (22 °C), or warm (28 °C or 30 °C) ambient temperatures, depending on the experimental condition.

Chronic temperature exposures were conducted over a 4-week period, during which mice remained continuously housed at the assigned temperature. Body weight was recorded weekly, and body composition was assessed using nuclear magnetic resonance (NMR) spectroscopy (Minispec LF50/90 TD-NMR System, Bruker). Mice with chronic temperature changes had ad libitum access to food and water and were adapted to mild cold (22 °C) conditions. For mice housed at 4 °C, daily measurements of core body temperature and surface temperature were obtained to monitor thermoregulatory status. In the final week, 24-h food intake was measured to assess long-term effects of thermal environment on energy balance.

Acute thermal manipulations lasted 24 h. Animals were transferred to the temperature-controlled chamber at 1500 h and remained there until 1500 h the following day. These temperature challenges were paired with pharmacological or chemogenetic interventions (CNO or MTII) or used as standalone stimuli prior to tissue collection, as specified in each experimental protocol. For anatomical studies to induce cFos, ad libitum fed mice were placed in chambers at 0900 h for 3 h and perfused at 1200 h for subsequent histological analysis.

Feeding and fasting paradigms

For fasting-refeeding experiments, mice were food-deprived for 18 h beginning at 1500 h. At 0900 h the following morning, animals received an intraperitoneal or intracerebroventricular injection of CNO or MTII, depending on the experimental protocol. Immediately following drug administration, food was returned to the cage to initiate the refeeding phase.

In fed control conditions, mice were maintained on ad libitum access to food prior to treatment. Food intake was measured manually at 6 h post-injection (1500 h) to quantify acute effects on nutrient consumption.

Metabolic phenotyping

Whole-body EE, LA, respiratory exchange ratio (RER), and FI were assessed using the Promethion metabolic phenotyping system (Sable Systems International). Mice were housed individually in sealed Promethion cages equipped with high-resolution temperature, activity, and gas-exchange sensors. Data were collected continuously over a 24- or 48-h period, depending on the experimental protocol.

Temperature-controlled environmental chambers (DB034-LT-GD Laboratory Incubator, Darwin Chambers) housing the Promethion cages were set to either 10 °C or 30 °C to evaluate metabolic responses under cold or warm ambient conditions. In chemogenetic studies, mice were provided ad libitum access to water containing water-soluble CNO or control water during the 24-h monitoring period. All metabolic chambers were calibrated before each trial, and data normalization was performed according to the manufacturer’s guidelines.

Feeding behavior analysis

Feeding events were derived from high-resolution food intake data using MacroInterpreter (Sable Systems International), which analyzes continuous weight measurements of the food hopper to identify discrete intake events. FI bouts were distinguished from non-consumptive fluctuations using the following criteria: a maximum inter-intake interval of 150 s, a maximum bout size of 1 g, and a significance threshold of p < 0.05. Total food consumed, as determined by bout analysis, was validated by correlating it with manually measured changes in hopper weight.

To define meals, intake bouts occurring within 5 min²³ of each other were grouped using a custom R script. The satiety ratio was calculated as the interval between the end of one meal and the onset of the next (i.e., the inter-meal interval), divided by the energy consumed during the preceding meal, as previously described⁷⁶:

$$\frac{timeuntilnextmeal}{kcalconsumedduring\mathrm{a}meal}$$

We quantified the meal-to-meal interval, defined as the elapsed time between the start of one meal and the start of the subsequent meal. This measure reflects the entire meal cycle (consumption plus inter-meal pause) and was chosen to better align with our detection method. FI was converted from grams to kilocalories by multiplying the weight of each meal by 2.89 kcal/g, based on the nutritional content of LabDiet #5001.

Histology and immunohistochemistry

Perfusions and immunohistochemical processing were performed as previously described¹⁰. Mice were deeply anesthetized with isoflurane and transcardially perfused with PBS, followed by 4% paraformaldehyde. Brains were post-fixed, cryoprotected in sucrose, and coronally sectioned at 30 μm using a sliding microtome (SM2000R, Leica).

Free-floating sections were processed for immunohistochemistry to visualize cFos and other molecular markers. Nuclear cFos immunoreactivity was visualized using a diaminobenzidine (DAB) reaction (#34065, Thermo Fisher Scientific) following incubation with a peroxidase-conjugated anti-rabbit IgG reagent (ImmPRESS HRP, Vector Laboratories #30118). All other proteins were detected using fluorophore-conjugated secondary antibodies. The following primary antibodies were used: rabbit anti-cFos (1:1000; Synaptic Systems #226003), rabbit anti-POMC (1:1000; Phoenix Pharmaceuticals #H-029-30), and chicken anti-GFP (1:1000; Abcam #AB13970). Secondary antibodies included donkey anti-rabbit IgG-Alexa Fluor 594 (1:500; Invitrogen #A21207) and donkey anti-chicken IgG-Alexa Fluor 488 (1:500; Life Technologies #532354).

After staining, sections were mounted on glass slides, coverslipped with ProLong™ Gold Antifade Mountant (#P36930, Invitrogen), and imaged using a fluorescence microscope (BX51, Olympus). Images were acquired with a digital camera (DP30BW, Olympus) under appropriate filter sets for fluorophore visualization or bright-field illumination for DAB.

Estimates of cell counts

Quantification of cFos-positive cells was performed in anatomically defined regions of interest based on the Paxinos and Franklin Mouse Brain Atlas⁷⁵. The PVN was analyzed in 2–3 sections spanning −0.83 to −1.07 mm from Bregma; the DMH in 3–4 sections spanning −1.55 to −1.91 mm; and the ARC in 7–8 sections from −1.31 to −2.15 mm. The ARC was further subdivided into four rostro-caudal zones: ARC I (− 1.31 to −1.43 mm), ARC II (− 1.55 to −1.67 mm), ARC III (− 1.79 to −1.91 mm), and ARC IV (− 2.03 to −2.15 mm). In the intermediate ARC (− 1.55 to −1.91 mm), both medial and lateral subdivisions were analyzed separately.

For each animal, cFos-positive nuclei were quantified bilaterally in matched sections using automated spot detection in NIS Elements AR 4.5 software (Nikon). Cytoplasmic markers, including POMC and EGFP, were manually counted due to their diffuse labeling patterns. All quantification was conducted by an experimenter blinded to treatment conditions.

Anatomical tracing

To visualize the axonal projections of POA^Lepr neurons, Lepr^Cre mice received bilateral injections of AAV5-EF1a-DIO-ChR2(H134R)-mCherry (AV4314C; 6 × 10¹² vg/mL; UNC Vector Core) targeted to the POA. Following 12 weeks of viral expression, mice were transcardially perfused, and brains were coronally sectioned at 30 μm using a sliding microtome (SM2000R, Leica).

Native mCherry fluorescence was imaged using a wide-field fluorescence microscope (BX51, Olympus) to visualize ChR2-expressing axons. Projection fields were mapped across brain regions and aligned to neuroanatomical landmarks defined in Paxinos and Franklin Mouse Brain Atlas⁷⁵.

Statistics and reproducibility

All statistical analyses were performed using GraphPad Prism version 10.3.1 (GraphPad Software). Differences between groups were assessed using two-way analysis of variance (ANOVA), including both between‑subjects and repeated‑measures designs, with appropriate post hoc corrections applied as indicated. Depending on the experimental design, either paired or unpaired Student’s t-tests were used to compare two conditions. Data are presented as mean ± standard error of the mean (SEM), and statistical significance was defined as p < 0.05. Full statistical details, including test type, n values, and p-values, are provided in the corresponding results sections. Numerical source data for all graphs in the manuscript can be found in the Supplementary Data 1 file. Numerical source data for all Supplementary figures can be found in the Supplementary Data 2 file.

Schematic figures were created using BioRender (BioRender.com). All graphs were generated in GraphPad Prism version 10.3.1, except for heat maps in Figs. 3 and 4, which were created using ggplot2 (via the Tidyverse package) in RStudio version 2024.12.0 + 467 or later running R version 3.6.0 or later.

To ensure reproducibility, mice were divided into 2–3 separate experimental cohorts consisting of 3–5 mice each. C57BL/6J mice were purchased from Jackson Laboratory. All other experimental mice were bred in house and included littermates from at least 3–4 different litters. No obvious differences were observed between cohorts. Mice were randomly distributed before receiving experimental treatments, always considering balanced sex distribution in all experimental groups. All experiments were done in a cross-experimental design, so that mice underwent all physiological (temperature, fasting/refeeding) and pharmacological (vehicle, CNO and/or MTII) conditions.

Author contributions

L.K.: Experimental planning (All Fig.), data collection (All Fig.), curation, analysis (All Fig.), and manuscript draft. N.L.: Experiment planning (Figs. 2, 5, 6, 7), data collection (Figs. 2, 5, 6, 7), curation (Figs. 2, 5, 6, 7), and manuscript editing/review. K.Z.: Experimental planning (Fig. 7) and data collection (All Fig. 7). C.K.: Manuscript editing/review. J.W.: Histology imaging (Figs. 5, 6, 7), generally advised all histology in the paper, and manuscript editing/review. MS: Data collection (Fig. 6). Manuscript editing/review. R.C.N.: Experiment planning (Fig. 1), data collection (Fig. 1), and manuscript editing/review. S.Y.: Experiment planning (Figs. 1, 5), data collection (Figs. 1, 5), and curation (Figs. 1, 5). H.R.B.: Manuscript editing/review. C.D.M.: Manuscript editing/review. H.M.: Concept, supervision, manuscript writing, editing, and funding.

Peer review

Peer review information

Communications Biology thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editors: Joao Valente. A peer review file is available.

Data availability

Raw images obtained by immunostaining are available upon reasonable request to the Corresponding Author. Code for processing feeding events and generating the related figures is available from 10.5281/zenodo.18437012⁷⁷. Numerical source data for all graphs in the manuscript can be found in the Supplementary Data 1 file. Numerical source data for all Supplementary figures can be found in the Supplementary Data 2 file.

Code availability

Custom code used for the processing and visualization of feeding events is available at 10.5281/zenodo.18437012⁷⁷.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-026-09723-7.