A hypothalamic neuronal population coordinates metabolic adaptations to cold

Summary

Cold exposure increases feeding, thermogenesis, and energy expenditure to maintain body temperature and energy balance, but the underlying neural mechanisms remain unclear. Here, we identify cold-responsive neurons in the dorsal posterior periventricular hypothalamus that coordinate these adaptations. Activation of GABAergic or cold-activated neurons increased food intake, body temperature, and brown adipose tissue thermogenesis, whereas their inhibition attenuated cold-induced hyperphagia and thermogenesis. Single-nucleus RNA sequencing identified Lef1 as a marker of cold-activated neurons in this region. Activation of Lef1⁺ neurons recapitulated key cold responses, whereas silencing attenuated them. Lef1 positively regulated Kcnk2, which encodes the cold-sensitive potassium channel TREK-1. Cold reduced Kcnk2 expression and potassium currents, while pharmacological or genetic Kcnk2 inhibition increased feeding and body temperature. Conversely, Kcnk2 gain-of-function blunted cold responses. Together, these findings identify Lef1⁺ hypothalamic neurons and Kcnk2 modulation as a mechanism linking environmental temperature to neuronal activity and systemic metabolic adaptation.

Graphical Abstract

eTOC Blurb

Liu et al. identify a population of Lef1⁺ hypothalamic neurons that coordinates feeding and thermogenic adaptations to cold. Cold suppresses Kcnk2/TREK-1 function to enhance neuronal activity, revealing a molecular mechanism linking environmental temperature to neural and metabolic responses.

Introduction

Maintaining proper appetite and body temperature in response to various challenges is vital for overall health and survival^1,2. Cold exposure elicits a series of physiological responses that promote survival, such as increased appetite, activated brown adipose tissue (BAT) thermogenesis, and elevated heat production^3–6. Cold exposure has emerged as a promising approach to enhance metabolic health^7–10. Multiple studies have demonstrated that cold exposure can enhance insulin sensitivity in both rodents and humans^11–13. In addition, cold exposure accelerates lipid clearance in diet-induced obese mice¹⁴. However, subjecting individuals to chronic cold conditions for metabolic benefits is impractical. Therefore, it is imperative to understand the intricate neural mechanisms underlying the metabolic benefits of cold exposure.

Several specific neural populations involved in cold defense have been identified, many of which are located in the preoptic areas of the hypothalamus (POA)¹⁵, the dorsomedial nucleus of the hypothalamus (DMH)¹⁶, and the lateral parabrachial nucleus (LPBN)^17,18. Particularly, cold-sensitive neurons expressing bombesin receptor subtype 3 (BRS3) in the POA and DMH^19,20, somatostatin (SST) in the LPBN²¹, and brain-derived neurotrophic factor (BDNF) in the DMH²², play essential roles in mediating the behavioral responses to cold. In addition, agouti-related peptide (AgRP) neurons are activated by cold and mediate cold-induced hyperphagia^4,5. More recently, the xiphoid nucleus in the mouse thalamus was identified as being responsive to increased energy expenditure after prolonged cold exposure, which facilitates the transition between energy conservation and food-seeking under cold conditions²³. However, it remains to be elucidated whether other brain regions can also respond to cold and influence cold-induced behavioral adaptations.

We identified the dorsal posterior periventricular hypothalamic nucleus (dPVp) as a cold-responsive center. Activating GABAergic or cold-activated dPVp neurons increased food intake, body temperature, BAT thermogenesis, and energy expenditure, whereas their inhibition impaired cold adaptation. Chronic activation improved insulin sensitivity and glucose homeostasis despite hyperphagia. Single-nucleus RNA sequencing identified Lef1 as a marker of cold-activated dPVp neurons. Lef1 regulated Kcnk2, which encodes the cold-sensitive potassium channel TREK-1. Kcnk2 inhibition mimicked cold responses, whereas Kcnk2 overactivity reduced cold-induced feeding and BAT thermogenesis. Thus, Lef1⁺ dPVp neurons coordinate cold adaptation through Kcnk2, a potential therapeutic target for metabolic disorders.

Results

GABA^dPVp neurons are activated by cold exposure

To identify potential cold-activated neurons, we first analyzed the expression of the immediate early gene marker c-Fos within the brains of wild-type (WT) mice following a 1-hour cold exposure at 6°C (Figure 1A). Consistent with existing literature²¹, we observed significant increases in c-Fos expression within distinct regions, including the DMH, LPBN, paraventricular thalamus (PVT), and lateral septum (LS) (Figure S1A–B). Remarkably, a robust upregulation of c-Fos expression in the dPVp was detected following cold exposure, while mice housed at room temperature (23°C) exhibited minimal c-Fos expression in this region (Figure 1B–C, S1C–D). Using RNAscope to co-label Fos with Slc32a1 and Slc17a6, markers of GABAergic and glutamatergic neurons, respectively, we found that ~97.3% of Fos⁺ dPVp neurons were GABAergic, while only ~2.7% were glutamatergic (Figure 1D–E, S2A–B), indicating that cold exposure primarily activates GABAergic neurons in this region. Furthermore, the Fos⁺ cells accounted for 56.86% of the GABAergic neurons within the dPVp (hereafter referred to as GABA^dPVp neurons, Figure 1F). We next sought to explore the responsiveness of dPVp neurons to nutritional signals. Towards this, WT mice were food-deprived overnight, and the next day were provided access to food for 1 hour. Interestingly, we detected a notable increase in c-Fos expression within the dPVp region after the refeeding paradigm compared to that observed in food-deprived mice (Figure S3A–D).

Figure 1: Cold exposure activates GABAergic neurons in dPVp.

A. Cold exposure and c-Fos staining scheme. B-C. Representative c-Fos staining and Allen Brain Atlas reference (B) and dPVp c-Fos quantification (C) in male WT mice at 23°C or 6°C for 1 h (n=5/group). D-F. RNAscope for Slc32a1 and Fos after cold exposure (6°C, 40 min) (D) and quantification of Slc32a1⁺/Fos⁺ (E) and Fos⁺/Slc32a1⁺ cells (F) (n=3). G. GCaMP6m expression in GABAergic dPVp neurons of Vgat-Flp mice and representative fiber placement. H-I. Cold-evoked (6°C, 5 s) Ca ²⁺ signals (H) and individual responses (I) in GABA^dPVp neurons (n=8). J-K. GABA^dPVp Ca ²⁺ responses to 22–6°C (J) and AUC quantification (K) (n=4). Data are expressed as mean ± SEM. *p < 0.05, ****p < 0.0001.

To further assess the responsiveness of GABA^dPVp neurons to cold exposure and nutritional cues, we delivered an AAV expressing Flp-dependent calcium sensor GCaMP6m into the dPVp of Vgat-Flp mice (Figure 1G), which enabled the visualization of calcium activity in GABA^dPVp neurons. We noted that the calcium signals increased immediately when mice were exposed to 6°C, with response magnitude scaling inversely with ambient temperature (Figure 1H–K), indicating that these targeted GABAergic neurons were activated by cold stimuli. Moreover, we observed an obvious elevation in calcium signals when fasted mice started to approach or consume food (Figure S3E–I), further supporting the notion that GABA^dPVp neurons are also activated by refeeding. Notably, the calcium signals remained unchanged in response to restraint stress (Figure S3J). These data suggest that GABA^dPVp neurons are specifically activated by cold temperature and refeeding, rather than by stress. To determine whether cold exposure and refeeding activate distinct subsets of neurons within the dPVp, we utilized the Fos^2A-iCreERT2 (TRAP2)/tdTomato mice, which express tamoxifen-inducible Cre recombinase in Fos⁺ cells, therefore allowing us to label activated cells with tdTomato. By administering 4-hydrotamoxifen (4-OHT) and subjecting these mice to cold exposure for 2h, we labeled cold-activated neurons with tdTomato (see Movie S1 for the cleared whole brain mapping). A different cohort of TRAP2/tdTomato mice were allowed to recover for two weeks after the cold exposure and 4-OHT induction. Following an overnight fast, they were given access to food for 2h the next morning. We then examined the co-localization of refeeding-induced c-Fos and cold-induced tdTomato in the dPVp (Figure S3K). Only a minimal degree of co-localization was observed (9.02%, Figure S3L), indicating that the neurons activated by cold and refeeding in the dPVp are largely separate populations. The caudal part of the arcuate nucleus of the hypothalamus (ARH) contains appetite-regulating AgRP and proopiomelanocortin (POMC) neurons. To investigate potential colocalization of these neurons with cold-activated dPVp neurons, we exposed AgRP/tdTomato and POMC-CreER/tdTomato mice to cold conditions and subsequently assessed c-Fos expression in the caudal hypothalamus region. Consistent with previous findings^4,5, cold exposure led to increased c-Fos expression in a subset of AgRP neurons (Figure S3M–N), while POMC neurons remained unaffected (Figure S3O–P). Notably, none of the c-Fos⁺ cells within the dPVp exhibited co-localization with tdTomato (Figure S3M–P), indicating that cold-activated neurons in the dPVp are distinct from both AgRP and POMC neurons.

Acute activation of GABA^dPVp neurons enhances appetite and energy expenditure

To further investigate the function of GABA^dPVp neurons, Vgat-Flp mice received an AAV encoding Flp-dependent excitatory DREADDs hM3Dq into the dPVp (hM3Dq mice), which allowed us to use clozapine N-oxide (CNO) to selectively activate GABA^dPVp neurons (Figure S4A–E). WT mice receiving the same virus but lacking the ability to express hM3Dq served as controls (control mice). When saline was administered, neither food intake nor rectal temperatures were significantly affected in both hM3Dq and control mice (Figure S4F–G). However, upon CNO injection, hM3Dq mice exhibited a dramatic increase in both feeding behavior and rectal temperature (Figure S4F–G). We then employed TSE metabolic cages to further measure energy expenditure, locomotor activity and food consumption. Compared to saline treatment, CNO administration robustly enhanced all of these parameters in hM3Dq mice, but had no such effect on control mice (Figure S4H–K), suggesting that activation of GABA^dPVp neurons results in an augmentation of energy expenditure, physical activity, and appetite. To assess the impact of GABA^dPVp neurons on BAT thermogenesis, we implanted probes beneath the interscapular BAT for temperature measurements. While saline injection had no effect on BAT temperature in either group, CNO administration yielded a profound elevation in BAT temperature in hM3Dq mice (Figure S4L). Such responses were not observed in control mice, indicating that GABA^dPVp neuronal activation exaggerates BAT thermogenesis (Figure S4L). In comparison to control mice, the expression of Ucp1, a well-established marker of thermogenesis, was upregulated in the BAT of hM3Dq mice following CNO injection (Figure S4M). Furthermore, CNO injection led to a discernible rise in serum norepinephrine (NE) levels in hM3Dq mice (Figure S4N), suggesting enhanced sympathetic nerve activity. We next sought to establish the connection between the BAT and the dPVp using a trans-multisynaptic pseudorabies virus PRV152 that carries GFP. Five days after the delivery of PRV152 into the BAT, we detected dense GFP signals in the dPVp (Figure S4O–P), suggesting that the dPVp may indirectly innervate the BAT.

Cold exposure elicits a natural inclination towards seeking warmth²⁴. To test the involvement of GABA^dPVp neurons in this warmth-seeking behavior, both hM3Dq and control mice were subjected to CNO injection and then placed in a two-chamber box for a duration of 6 minutes, with one chamber maintained at room temperature (23°C) while the other chamber was equipped with a heat pad to sustain the temperature at 32°C (Figure S4Q). Compared to control mice, hM3Dq mice exhibited increased locomotor activity and spent significantly more time in the warmer chamber (Figure S4R–S), suggesting that activation of GABA^dPVp neurons promotes mice to seek warmth. To further determine the preferred temperature range (defined as the temperature zone where the mouse spent most time in), we employed a thermal gradient box, which provides a stable temperature gradient across a rectangle-shaped arena (Figure S4T). Following CNO administration, mice were allowed to explore in the box for 30 minutes. While the control mice showed a preference for staying in the zone with a temperature of 30°C, the hM3Dq mice demonstrated a preference for a temperature of around 32°C (Figure S4U–V), supporting that GABA^dPVp neurons promote warmth-seeking behavior.

Since GABA^dPVp neurons are excited by refeeding, we further examined their influence on feeding behavior following an overnight fast. Saline injection did not alter the amount of food consumption between two groups (Figure S4W), however, CNO administration significantly increased food intake in hM3Dq mice (Figure S4X), suggesting that GABA^dPVp neurons play a crucial role in regulating feeding behavior during periods of food-deprivation.

Acute activation of cold-activated dPVp neurons enhances appetite and energy expenditure

Cold exposure only activates 56.86% of GABAergic neurons in the dPVp, we then employed TRAP2 mice to specifically elucidate the functional role of this cold-responsive neuronal population. To evaluate the efficiency of these TRAP2 mice in labeling cold-activated neurons, TRAP2/tdTomato mice received 4-OHT injection followed by an immediate exposure to cold (6°C) for 2 hours, which triggered Cre activation and subsequent tdTomato expression in cold-activated neurons (referred as the cold TRAP procedure, Figure 2A). Following a two-week interval, these mice were subjected to a subsequent cold challenge (6°C) for 1 hour prior to perfusion. Through histological analysis, we detected abundant tdTomato⁺ neurons in the dPVp (which were activated during the initial cold exposure), with 81.33% of them co-localized with c-Fos (which was induced during the second cold exposure, Figure 2B–C). In addition, 64.35% of the c-Fos⁺ neurons in the dPVp exhibited tdTomato expression (Figure 2D). We then used separate TRAP2/tdTomato mice to repeat the cold TRAP procedure with extended cold exposure durations of 4 and 6 hours, and found comparable TRAP efficiency to that of the 2-hour cold TRAP procedure (Figure S5A–D). Through these initial analyses, we concluded that the 2-hour cold TRAP procedure can be used to selectively label cold-activated dPVp neurons.

Figure 2: Acute activation of cold-activated dPVp neurons promotes thermogenesis and feeding.

A. Scheme for cold-induced tdTomato TRAP labeling. B. Cold-TRAPed tdTomato, cold-induced c-Fos, and their colocalization in dPVp of male TRAP2/tdTomato mice. C-D. Percentage of c-Fos⁺tdTomato⁺ cells among tdTomato⁺ (C) or c-Fos⁺ (D) cells (n=3). E. Scheme for cold-induced hM3Dq expression in TRAP2 mice. F-G. Food intake (F) and rectal temperature (G) after saline or CNO in hM3Dq-expressing WT (n=10) and TRAP2 (n=8) mice. H-K. Energy expenditure (H), XY (I) and Z (J) activity, and food intake (K) after saline or CNO in WT (n=8) and TRAP2 (n=5) mice. L. BAT temperature after CNO (n=7/group). M-N. BAT Ucp1 (M) and serum NE (N) 2 h after CNO (WT, n=9; TRAP2, n=8). O-Q. Thermal gradient scheme (O), zone occupancy (P), and preferred temperature (Q) after CNO (n=7/group). R-T. Two-chamber temperature preference scheme (R), distance traveled (S), and warm-chamber occupancy (T) after CNO (n=8/group). U-V. 2-h refeeding after saline (U) or CNO (V) in fasted WT (n=10) and TRAP2 (n=8) mice. W-X. Core (W) and BAT surface (X) temperatures after saline or CNO in TRAP2 mice (n=7). Y-Z. Changes in core and BAT temperatures after saline (Y) or CNO (Z) (n=7).

Data are expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

We next utilized the TRAP2 mice to investigate the function of cold-activated dPVp neurons in feeding and thermal regulation. Briefly, both WT and TRAP2 mice received stereotaxic injections of Cre-dependent AAV expressing hM3Dq into the dPVp (Figure 2E). Two weeks after virus delivery, these mice received 4-OHT injection and were immediately exposed to cold (6°C) for 2 hours (Figure 2E). After a 2-week incubation period, we verified that hM3Dq was successfully expressed in the dPVp of TRAP2 mice, which enabled its activation by CNO (Figure S5E–I). Saline injection had no impact on food intake and rectal temperature between the two groups (Figure 2F–G). Remarkably, CNO administration in TRAP2 mice recapitulated all phenotypic outcomes observed in Vgat-Flp mice with Flp-dependent hM3Dq expression in GABA^dPVp neurons, including heightened appetite, rectal temperature, energy expenditure, locomotor activity, BAT temperature, serum NE levels, and the propensity for warmth-seeking behavior in comparison to WT mice (Figure 2F–V). Importantly, activation of cold-TRAPed dPVp neurons elevated both core body and BAT temperatures (Figure 2W–X), with BAT temperature rising more rapidly than core body temperature (Figure 2Y–Z), suggesting that enhanced BAT thermogenesis contributes, at least in part, to the increase in body temperature induced by activation of these neurons. Moreover, CNO injection increased body temperature even in fasted mice, indicating that dPVp-mediated thermoregulation occurs independently of food intake (Figure S5J). To further confirm that these effects were indeed orchestrated by cold-activated dPVp neurons, an additional group of WT and TRAP2 mice received Cre-dependent hM3Dq viral injections to the dPVp followed by 4-OHT injection. However, this cohort was not exposed to cold conditions (Figure S5K). Subsequent CNO administration failed to yield any notable distinctions in food intake, rectal temperature, locomotor activity, or warmth-seeking behavior between the WT and TRAP2 mice (Figure S5L–P). Consequently, the discerned phenotypes observed in mice that underwent the cold TRAP procedure can be attributed to the selective excitation of cold-activated neurons.

Acute inhibition of dPVp neurons disrupts the cold-induced feeding and warmth-seeking

To ascertain the necessity of GABA^dPVp neuronal activation in regulating feeding and body temperature upon cold exposure, we introduced a Flp-dependent AAV expressing the inhibitory DREADDs hM4Di into the dPVp of Vgat-Flp mice (hM4Di mice), which enabled selective inhibition GABA^dPVp neurons using CNO (Figure S6A–E). WT mice receiving the same virus but lacking the ability to express hM4Di served as controls (control mice). Following saline injection, control and hM4Di mice displayed comparable food intake and rectal temperature in both room temperature and cold conditions (Figure S6F–G). As expected, cold exposure initiates an augmentation in feeding and a reduction in body temperature across both groups (Figure S6F–G). Notably, in comparison to control mice, CNO administration mitigated hyperphagia and exaggerated the drop in body temperature induced by cold in hM4Di mice, although no such effects were observed at room temperature (Figure S6H–I). In addition, CNO administration reduced overall locomotor activity and the time spent in the warm side during the two-chamber temperature preference test (Figure S6J–L). Consistent with the role of GABA^dPVp neurons in promoting feeding, suppression of these neurons resulted in decreased food consumption after an overnight fasting (Figure S6M–N).

Similarly, we examined the contribution of cold-activated dPVp neurons to cold-elicited adaptive behaviors. To this end, TRAP2 mice and their littermate WT mice received a Cre-dependent AAV carrying hM4Di into the dPVp (Figure 3A). After a span of 2 weeks, these mice underwent the cold TRAP procedure to selectively induce hM4Di expression in cold-activated dPVp neurons (Figure 3A–E). Just as we had observed in Vgat-Flp mice expressing hM4Di in GABA^dPVp neurons, inhibition of cold-activated dPVp neurons exhibited a parallel pattern-a reduction in hyperphagia and an acceleration in the decline of body temperature as a response to cold stimuli, with no detectable differences between the two groups in food intake and body temperature in normal housing conditions (Figure 3F–I). Moreover, inhibiting cold-activated dPVp neurons decreased locomotor activity (Figure 3J–K), caused avoidance of ambient warmth (Figure 3L), and attenuated refeeding following overnight food deprivation (Figure 3M–N). Considering the established role of AgRP neurons in cold-induced hyperphagia^4,5, we investigated whether manipulating cold-activated dPVp neurons affects AgRP activity. Activation of dPVp neurons had no effect on Fos expression in AgRP neurons (Figure S6O–P). Likewise, cold exposure still elicited robust Fos induction in AgRP neurons even when dPVp neurons were chemogenetically inhibited (Figure S6Q–R), suggesting that AgRP neurons respond to cold independently of dPVp neuron activity.

Figure 3: Cold-activated dPVp neurons mediate cold adaptations and promote metabolic benefits.

A. Scheme for cold-induced hM4Di expression in TRAP2 mice. B. Representative dPVp hM4Di-mCherry expression. C-E. Action potentials (C), firing frequency (D), and resting membrane potential (E) of hM4Di⁺ cold-activated dPVp neurons before and after CNO (10 μM; n=7). F-I. Food intake (F, H) and rectal temperature (G, I) at RT or cold after saline (F-G) or CNO (H-I) in hM4Di-expressing WT (n=8) and TRAP2 (n=7) mice. J-L. Two-chamber temperature preference scheme (J), distance traveled (K), and warm-chamber occupancy (L) after CNO (WT, n=8; TRAP2, n=7). M-N. Food intake in fasted WT (n=8) and TRAP2 (n=7) mice after saline (M) or CNO (N). O. Scheme for chronic hM3Dq activation in HFD-fed WT and TRAP2 mice. P-R. Cumulative food intake (P), rectal temperature (Q), and body weight (R) during daily CNO treatment (n=8/group). S-V. GTT and AUC after 12 days (S-T) and ITT and AUC after 16 days (U-V) of CNO (n=8/group).

Data are expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001.

Chronic activation of dPVp neurons improves glucose homeostasis

Extended cold exposure has been shown to improve glucose metabolism, at least partially through enhancing heat production^13,25, we next sought to investigate the metabolic outcomes following chronic activation of GABA^dPVp neurons. WT and Vgat-Flp mice received Flp-dependent AAV carrying hM3Dq into the dPVp, after allowing a 3-week period for virus expression, we performed CNO injections twice daily for 10 consecutive days to repetitively activate GABA^dPVp neurons in Vgat-Flp mice (Figure S7A). This manipulation resulted in increased food intake and rectal temperature, yet yielding minimal changes in body weight in chow-fed mice (Figure S7B–D). Remarkably, chronic activation of GABA^dPVp neurons led to improved insulin sensitivity without affecting glucose tolerance (Figure S7E–H). Likewise, we explored the effects of prolonged excitation of cold-activated dPVp neurons on glucose homeostasis by delivering Cre-dependent AAV encoding hM3Dq into the dPVp of WT and TRAP2 mice. Three weeks after the cold TRAP procedure, these mice were then subjected to twice-daily CNO injections over 12 days (Figure S7I). Chronic activation of cold-activated dPVp neurons induced a significant increase in food consumption and body temperature, resulting in insignificant changes in body weight in chow-fed mice (Figure S7J–L). Notably, glucose tolerance and insulin sensitivity were improved when cold-activated dPVp neurons were repetitively activated (Figure S7M–P). We further repeated these studies in a cohort of control and TRAP mice fed a high-fat diet (HFD, Figure 3O), and found that re-activation of cold-activated dPVp neurons produced similar improvements in glucose tolerance and insulin sensitivity associated with increased body temperature, despite chronic hyperphagia and the absence of weight reduction (Figure 3P–V). Collectively, these results indicate that chronically activating dPVp neurons improves glucose homeostasis despite hyperphagia.

Chronic inhibition of dPVp neurons impairs cold tolerance

To deepen our understanding of the involvement of GABA^dPVp neurons in energy and thermal homeostasis, we infected Vgat-Flp mice with a Flp-dependent AAV encoding Kir2.1, a well-known potassium channel with potent neural silencing capacities, to selectively target GABA^dPVp neurons for chronic inhibition (Kir2.1 mice, Figure S8A–B). WT mice underwent the same viral procedure, serving as the control group. While food intake, body weight and BAT temperature remained indistinguishable between control and Kir2.1 mice at room temperature (Figure S8C–E), cold-induced hyperphagia was largely blocked in Kir2.1 mice (Figure S8F). In addition, control mice effectively maintained a stable BAT temperature during cold exposure, but this ability was compromised in Kir2.1 mice (Figure S8G). Consistently, the TSE metabolic cage studies revealed that silencing GABA^dPVp neurons reduced energy expenditure at room temperature (Figure S8H–I), without affecting physical activity and feeding patterns (Figure S8J–L). Moreover, baseline BAT Ucp1 expression and serum NE levels were decreased in Kir2.1 mice (Figure S8M–N). Remarkably, compared to control mice, the Kir2.1 mice demonstrated reduced locomotion and warmth avoidance in the two-chamber temperature preference test (Figure S8O–Q), along with a tendency to favor lower temperatures in the thermal gradient box test (Figure S8R–T). In line with the pivotal role of BAT thermogenesis in glucose homeostasis, Kir2.1 mice displayed impaired insulin sensitivity and unaltered glucose tolerance (Figure S8U–X). Furthermore, fasting-evoked refeeding was dampened in Kir2.1 mice (Figure S8Y).

To further evaluate the contribution of cold-activated dPVp neurons to the maintenance of thermal and glucose balance, we delivered Cre-dependent AAVs encoding Kir2.1 into the dPVp of WT and TRAP2 mice (Figure 4A–C). Following the cold TRAP procedure, silencing cold-activated dPVp neurons replicated the phenotypes noted in Vgat-Flp mice with Kir2.1 expression in GABA^dPVp neurons. Specifically, inhibition of these neurons attenuated cold-induced hyperphagia, accelerated the decline in both core body and BAT temperature during cold exposure, decreased baseline energy expenditure, lowered BAT Ucp1 and serum NE levels, a preference for cooler temperatures, impaired glucose tolerance and insulin sensitivity, as well as reduced refeeding (Figure 4D–Z). Together, these findings demonstrate that inhibition of dPVp neurons suppresses heat production and attenuates cold defense, highlight their contributing role in thermal and glucose homeostasis.

Figure 4: Chronic inhibition of cold-activated dPVp neurons impairs cold tolerance and glucose homeostasis.

A. Scheme for cold-induced Kir2.1 expression in TRAP2 mice. B. Representative dPVp Kir2.1-tdTomato expression. C. Action potentials, firing frequency, and resting membrane potential of mCherry control (n=7) or Kir2.1⁺ (n=9) cold-activated dPVp neurons. D. Body weight of Kir2.1-expressing WT (n=7) and TRAP2 (n=6) mice. E-H. Food intake (E, G) and BAT temperature (F, H) at room temperature (E-F) or cold (G-H). I-J. Core (I) and BAT surface (J) temperatures during cold exposure (n=7/group). K. Energy expenditure and body-mass regression, with predicted values at 26 g. L-N. XY (L) and Z (M) activity and food intake (N). O-P. BAT Ucp1 (O) and serum NE (P). Q-S. Two-chamber temperature preference scheme (Q), distance traveled (R), and warm-chamber occupancy (S). T-U. Thermal gradient zone occupancy (T) and preferred temperature (U). V. Food intake after overnight fasting. W-Z. GTT and AUC (W-X) and ITT and AUC (Y-Z). Unless indicated, WT n=7 and TRAP2 n=6.

Data are expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Lef1 is a marker and regulator of cold-activated dPVp neurons

To identify the molecular markers of cold-activated dPVp neurons, we exposed male WT mice to cold temperatures for 30 minutes and collected the posterior part of the hypothalamus for single-nucleus RNA sequencing. A total of 10 biological replicates were included, resulting in 18,730 cells expressing 30,607 unique transcripts/features. We first performed cluster analysis and then looked to see if Fos was enriched in any of the clusters. We discovered 3 clusters of Fos⁺ neurons that co-express lymphoid enhancer binding factor 1 (Lef1), Agrp, and T-box transcription factor 19 (Tbx19), respectively (Figure 5A). Notably, the detection of Agrp is consistent with previous reports indicating that AgRP neurons are activated by cold exposure^4,5. However, AgRP neurons are located in the ventral hypothalamus, which is anatomically distinct from the dPVp. By co-staining Lef1 with c-Fos, we found that Lef1 was highly expressed in the dPVp (Figure 5B), with 94.30% of cold-induced c-Fos overlapping with Lef1 (Figure 5C), which accounts for 63.97% of Lef1⁺ neurons in the dPVp (Figure 5D). Consistent with these findings, 94.53% of Lef1⁺ neurons co-localized with Slc32a1, while only 5.47% expressed Slc17a6 (Figure S9A–B), indicating that Lef1 marks a subset of GABAergic neurons in the dPVp. Conversely, in dPVp samples from cold-treated mice, only a minor portion of Fos⁺ cells expressed Tbx19 and vice versa (Figure S9C–E). Together, these findings suggest that Lef1 is a marker for cold-activated neurons in the dPVp. Prodynorphin (PDYN)-expressing neurons in the dorsal medial region of the ventromedial hypothalamus (dmVMH) have been reported to be activated by cold exposure and to promote homeostatic thermogenesis²⁶. We therefore performed RNAscope staining for Pdyn and Lef1 to determine whether these markers label distinct neuronal populations. We found that Pdyn is predominantly expressed in the rostral hypothalamus, whereas Lef1 expression is largely restricted to the caudal hypothalamus (Figure S10A). Importantly, there was minimal overlap between these populations: only 4.28% of Pdyn⁺ cells in the VMH expressed Lef1, and conversely, just 1.3% of Lef1⁺ cells in the dPVp co-expressed Pdyn (Figure S10B–C). Further analysis revealed that 82.28% of cold-activated neurons in the dPVp co-localize with Lef1, but not Pdyn (Figure S10D–E). Thus, Lef1⁺ dPVp neurons represent a previously unrecognized population of cold-responsive neurons.

Figure 5: Lef1⁺ dPVp neurons mediate cold response.

A. snRNA-seq gene expression in the caudal hypothalamus of cold-exposed male WT mice. B-D. Representative Lef1/c-Fos staining at 23°C or 6°C for 1 h (B) and Lef1⁺/c-Fos⁺ colocalization (C-D; n=3). E-F. Food intake (E) and rectal temperature (F) after CNO in WT and Lef1-Dq mice (n=6/group). G-H. Thermal zone occupancy (G) and preferred temperature (H) in control and Lef1-Dq mice (n=6/group). I-L. Food intake (I, K) and body temperature (J, L) in control and Lef1-Kir2.1 mice at room temperature (I-J) or cold (K-L; n=6/group). M-O. Rectal temperature (M), BAT Ucp1 (N), and serum NE (O) in control and Lef1-KD mice. P-Q. Thermal zone occupancy (P) and preferred temperature (Q). R. Cold-induced body temperature change. S-V. GTT/AUC (S-T) and ITT/AUC (U-V). For Lef1-KD studies, control n=7 and Lef1-KD n=8 unless indicated; N, n=5/group; O, n=6/group.

Data are expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001.

We next introduced Lef1-Cre/ERT2 mice to investigate the role of Lef1⁺ dPVp neurons in mediating cold-induced responses. To selectively activate these neurons, Lef1-Cre/ERT2 (Lef1-Dq) mice and their WT littermates received Cre-dependent AAVs expressing hM3Dq into the dPVp. One week later, these mice were administered tamoxifen to induce Cre activity (Figure S11A–B). After a span of 4 weeks to allow for viral expression, CNO was injected. Compared to WT controls, the Lef1-Dq mice exhibited increased food intake and body temperature in both sexes (Figure 5E–F, S11C–D). In the thermal gradient assay, the Lef1-Dq mice displayed a rightward shift in preferred temperature (Figure 5G–H, S11E–F), indicating that activating Lef1⁺ dPVp neurons promotes warmth seeking. To examine the necessity of Lef1⁺ dPVp neurons, we delivered Cre-dependent AAVs expressing Kir2.1 to silence neuronal activity in a separate cohort of Lef1-Cre/ERT2 (Lef1-Kir2.1) and WT mice (Figure S11G–H). At room temperature, food intake and body temperature were comparable between groups (Figure 5I–J). Under cold conditions, the Lef1-Kir2.1 mice showed reduced food intake and a more rapid drop in body temperature compared to WT controls (Figure 5K–L). However, the Lef-Kir2.1 mice still displayed cold-induced hyperphagia, and their body temperature did not fall to severely hypothermic levels, demonstrating that Lef1⁺ dPVp neurons contribute to, but are not indispensable for, adaptive responses to cold exposure.

We speculate that Lef1 is more than a marker for cold-activated dPVp neurons and Lef1 itself may play a role in regulating energy and thermal homeostasis. To test this, we delivered AAVs expressing Cre or GFP into the dPVp of Lef1^flox/flox mice to generate Lef1-KD and control mice (Figure S11I–L). Although food intake and body weight were indistinguishable (Figure S11M–N), the Lef1-KD mice exhibited elevated body temperature, higher BAT Ucp1 and serum NE levels relative to controls (Figure 5M–O). Additionally, the warmth seeking behavior was enhanced in Lef1-KD mice, as revealed by an increase in preferred ambient temperature (Figure 5P–Q). Moreover, disrupting Lef1 in the dPVp enabled better maintenance of body temperature during cold exposure (Figure 5R). While glucose tolerance remained unaffected (Figure 5S–T), insulin sensitivity was improved in Lef1-KD mice (Figure 5U–V), potentially due to enhanced thermoregulation.

Lef1 stimulates Kcnk2 transcription

Lef1 is a transcription factor that plays key roles in various biological processes^27–30. To determine whether Lef1 modulates cold-induced responses through cold-sensitive ion channels, we conducted an analysis of existing single-cell RNA sequencing data in search of potential cold-sensing markers^31,32. Among the candidates investigated, namely Cnga3, Scnn1a, Trpa1, and Kcnk2, we showed that Kcnk2 was abundantly expressed in the dPVp (Figure S12A). Under normal physiological conditions, Kcnk2 channels remain open to facilitate K⁺ currents. However, cold exposure induces the closure of Kcnk2 channels, which leads to a net depolarization of neurons^33,34. Through ChIP assay, we revealed that Lef1 binds to both the promoter and enhancer regions of the Kcnk2 gene, this interaction was attenuated by cold exposure (Figure S11B–C). To assess the functional relevance of these bindings, we inserted a Kcnk2 regulatory fragment upstream of a luciferase reporter gene, thus allowing reporter activity to serve as a proxy for Kcnk2 transcription. Lef1 knockdown in Neuro-2a cells significantly reduced Kcnk2 transcription (Figure S12D), whereas Lef1 overexpression enhanced it (Figure S12E). Notably, deleting the Lef1-binding sites within the Kcnk2 promoter region led to a marked decrease in luciferase activity (Figure S12E), confirming the functional importance of Lef1-Kcnk2 interactions. RNAscope analysis uncovered that 92.77% Lef1⁺ cells are GABAergic, with 67.47% co-localizing with Kcnk2 (Figure S12F–H). Consistently, cold exposure reduced Lef1 and Kcnk2 mRNA and protein levels in the caudal hypothalamus of WT mice (Figure S12I–M). Furthermore, disruption of Lef1 in the dPVp resulted in a decrease of Kcnk2 in the caudal hypothalamus (Figure S12N–O). Collectively, these findings indicate that Lef1 stimulates Kcnk2 transcription, and that cold exposure reduces this Lef1-Kcnk2 transcriptional axis.

Kcnk2 mediates the response of dPVp neurons to cold

To study the role of Kcnk2, we first injected spadin, a selective Kcnk2 inhibitor, into WT mice. Spadin administration led to a profound upregulation of c-Fos expression within the dPVp (Figure S13A–B). To record the response of cold-activated dPVp neurons to spadin, we employed the cold TRAP procedure to label these neurons with tdTomato in TRAP2/tdTomato mice. Following a 2-week interval, we noticed that spadin amplified the overall firing frequency and resting membrane potential of tdTomato⁺ dPVp neurons (Figure S13C–E). Additionally, the depolarization of dPVp neurons evoked by spadin persisted in the presence of TTX and synaptic blockers (CNQX, D-AP5, bicuculline, Figure S113C, F), indicating that inhibition of Kcnk2 directly depolarizes cold-activated dPVp neurons. To explore the involvement of Kcnk2 in mediating the response of dPVp neurons to cold, we designed two Cre-dependent AAV vectors. One vector carries a wild-type Kcnk2, while the other expresses a gain-of-function mutant Kcnk2 protein (Kcnk2-E306A) that remains constitutively active (Figure S13G)^35,36. These AAVs, alongside Cre-dependent GCaMP6m, were selectively delivered into the dPVp of TRAP2 mice. At the same time, an optic fiber was placed above the dPVp to monitor calcium activity (Figure 6A). TRAP2 mice that received Cre-dependent GCaMP6m and GFP, but not the Kcnk2/Kcnk2-E306A vector, served as controls. Remarkably, the calcium signals in dPVp neurons increased robustly in the control mice upon cold exposure (Figure 6B). However, the expression of Kcnk2 reduced the response of dPVp neurons to cold (Figure 6C), suggesting that Kcnk2 dampens their activation in response to cold stimuli. Strikingly, when Kcnk2-E306A was expressed, it completely diminished dPVp neurons’ response to cold (Figure 6D–E), further supporting the crucial role of Kcnk2 in modulating the cold response of dPVp neurons.

Figure 6: Kcnk2 in the dPVp modulates cold response.

A. Scheme for GCaMP6m and GFP/Kcnk2/Kcnk2-E306A expression in dPVp of TRAP2 mice. B-D. Cold-evoked (6°C, 15 s) Ca ²⁺ signals in GFP (B), Kcnk2 (C), and Kcnk2-E306A (D) mice (n=11/group). E. AUC for B-D. F-L. Food intake and BAT temperature at room temperature (F-G) or cold (H), BAT Ucp1 and serum NE (I), thermal preference (J), fasting-induced feeding (K), and GTT/ITT (L) in control (n=6), Kcnk2 (n=7), and E306A (n=6) mice. M. Scheme for dPVp Kcnk2-E306A expression and cannulation. N-Q. Food intake (N), rectal temperature (O), and thermal preference (P-Q) after dPVp ACSF or Spadin in WT and TRAP2 mice (n=7/group). R. Scheme for dPVp Cas9/sgKcnk2 injection and representative GFP expression. S-Z. Food intake (S), rectal/BAT temperature (T), BAT Ucp1 (U), serum NE (V), thermal preference (W), body weight (X), GTT (Y), and ITT (Z) in control and sgKcnk2 mice (n=8/group).

Data are expressed as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001.

To further interrogate the role of Kcnk2 in mediating the behavioral adaptations to cold, TRAP2 mice received Cre-dependent AAV encoding GFP, Kcnk2 or Kcnk2-E306A into the dPVp (referred as control, Kcnk2 or E306A mice respectively, Figure 6F, S13H). After being subjected to the cold TRAP procedure, no discernible differences emerged regarding baseline food intake and BAT temperature (Figure 6F–G). However, cold-induced hyperphagia was reduced in Kcnk2 mice and more robustly decreased in E306A mice (Figure 6H). In addition, the ability to defend BAT temperature under cold conditions was impaired in Kcnk2 mice and further exacerbated in E306A mice (Figure 6H). Although Kcnk2 and E306A mice exhibited similar reductions in BAT Ucp1 expression and serum NE levels (Figure 6I), the preferred temperature shifted towards lower values in Kcnk2 mice, which was further decreased in E306A mice (Figure 6J, S13I–K). Food deprivation-evoked refeeding was reduced in Kcnk2 mice, with a greater reduction observed in E306A mice (Figure 6K). Moreover, E306A mice displayed disrupted glucose tolerance and insulin sensitivity, whereas Kcnk2 mice showed impaired insulin sensitivity with normal glucose tolerance (Figure 6L, S13L–M). To investigate whether Kcnk2 inhibition in the dPVp elicits metabolic adaptations similar to those caused by cold exposure, we implanted cannula over the dPVp of WT mice, which enabled us to deliver spadin (0.2 μg) directly into the dPVp (Figure S13N). Compared to ACSF control, spadin treatment evoked an increase in food intake and body temperature (Figure S13O–P), as well as an inclination towards seeking warmth (Figure S13Q–R). To rule out potential off-target effects of spadin, we repeated the spadin delivery study in a cohort of cold-TRAPed WT and TRAP2 mice that received Cre-dependent Kcnk2-E306A into the dPVp (Figure 6M, S13S). Kcnk2-E306A expression in the dPVp blocked spadin-induced behavioral changes (Figure 6N–Q), confirming the important role of dPVp Kcnk2 in mediating the action of spadin. Intriguingly, cold exposure decreased Kcnk2 mRNA levels in GABA^PVp neurons (Figure S13T–U). To investigate whether Kcnk2 downregulation recapitulates the metabolic adaptations induced by cold, we constructed an AAV vector carrying a single guide RNA targeting the mouse Kcnk2 locus to create indel mutations (AAV-sgKcnk2-DIO-GFP). We then stereotaxically injected AAV-sgKcnk2-DIO-GFP and AAV-DIO-saCas9 into the dPVp of TRAP2 and WT littermates to generate Kcnk2 knockdown mice and their controls (Figure 6R). After the cold TRAP procedure, Kcnk2 knockdown mice demonstrated increased food intake, higher rectal and BAT temperature, elevated BAT Ucp1 and serum NE levels, and a preference for ambient warmth, while maintaining body weight comparable to controls under room temperature conditions (Figure 6S–X). Additionally, these mice showed improved insulin sensitivity with no change in glucose tolerance (Figure 6Y–Z). In line with these findings, selective reduction of Kcnk2 expression in Lef1⁺ dPVp neurons by injecting AAV-sgKcnk2-DIO-GFP into the dPVp of Lef1-CreER mice produced similar metabolic phenotypes (Figure S14A–L). Overall, these data provide evidence that Kcnk2 in the dPVp plays a pivotal role in modulating the behavioral response to cold.

dPVp neurons exhibit intrinsic cold sensitivity

To test the thermal sensitivity of cold-activated dPVp neurons, we employed whole-cell patch-clamp recordings to monitor their response to temperature changes in cold-TRAPed TRAP2/tdTomato mice. While tdTomato⁻ cells were unresponsive to temperature changes (Figure S15A–C), tdTomato⁺ neurons depolarized with temperature decrease and hyperpolarized with temperature increase (Figure 7A–C), even in the presence of TTX and synaptic blockers (Figure 7D–E). Importantly, these tdTomato⁺ neurons exhibited an average thermal coefficient of −0.5 impulses/s/°C and a Q10 value below 1 (Figure 7F, S15D), supporting their intrinsic sensitivity to temperature changes. Pretreatment of brain slices with spadin blocked this temperature sensitivity (Figure 7G–I). In addition, the expression of Kcnk2-E306A in cold-activated dPVp neurons completely blocked their response to temperature fluctuations (Figure 7J–K). These results demonstrate that Kcnk2 is involved in modulating dPVp neuronal activity under cold conditions. It is also worth noting that the spontaneous firing was completely abolished by Kcnk2-E306A (Figure 7K), indicating a critical role of Kcnk2 in regulating action potentials in these neurons. Consistently, we observed reduced potassium currents in cold-activated dPVp neurons following a temperature decrease (Figure 7L); this effect was blocked by spadin (Figure 7M), suggesting that Kcnk2 currents contribute to cold-induced reductions in potassium currents. Remarkably, Kcnk2-E306A also blocked temperature fluctuation-induced changes in potassium currents (Figure 7N), further confirming the role of Kcnk2 in temperature sensing.

Figure 7: Cold-activated dPVp neurons are inherently cold-sensitive.

A-C. Action potentials (A), firing frequency (B), and resting membrane potential (C) of tdTomato⁺ dPVp neurons during 34→30°C or 30→34°C shifts (n=11). D-E. Action potentials (D) and resting membrane potential (E) during temperature shifts with TTX and synaptic blockers (n=11). F. Thermal coefficient during 34→30°C cooling (n=11). G-I. Action potentials (G), firing frequency (H), and resting membrane potential (I) during temperature shifts with spadin (n=11). J-K. Action potentials (J) and resting membrane potential (K) during temperature shifts in Kcnk2-E306A-expressing neurons (n=11). L-N. K⁺ currents during temperature shifts in control (L), spadin-treated (M), or Kcnk2-E306A-expressing (N) neurons.

Data are expressed as mean ± SEM and individual data points. *p < 0.05, **p < 0.01, ***p < 0.001.

dPVp neurons are positioned within the established thermoregulatory network

To map the downstream targets of cold-activated dPVp neurons, we injected a Cre-dependent AAV encoding ChR2-GFP (AAV-DIO-ChR2) into the dPVp of TRAP2 mice (Figure 8A). Three weeks after cold exposure and 4-OHT administration, we observed dense GFP-labeled cell bodies within the dPVp, along with extensive projections to the LS, diagonal band of Broca (DBB), POA, lateral hypothalamus (LH), DMH, PVT, lateral periaqueductal gray (LPAG), dorsal Raphe nucleus (DRN), LPBN, peri-locus coeruleus (periLC), rostral raphe pallidus (RPa), and nucleus of the solitary tract nucleus (NTS) (Figure 8A). To identify upstream inputs to cold-activated dPVp neurons, we used a monosynaptic retrograde tracing approach based on a modified rabies virus³⁷. Briefly, TRAP2 mice received injections of a Cre-dependent helper virus (AAV-DIO-GTB) into the dPVp to express the TVA receptor and rabies glycoprotein required for viral entry and trans-synaptic spread. After two weeks of expression, mice underwent the cold TRAP procedure to induce Cre recombination. Following an additional two-week incubation, EnvA-ΔG-Rabies-GFP was injected into the dPVp (Figure 8B). One week later, GFP⁺ neurons were detected in the dPVp, POA, and DMH (Figure 8C), indicating that dPVp neurons receive direct monosynaptic inputs from the POA and DMH. Together, these results demonstrate that the dPVp is anatomically embedded within a well-defined thermoregulatory network.

Figure 8: Outputs and inputs of cold-activated dPVp neurons.

A. Scheme for Cre-dependent ChR2 injection into the dPVp and representative fluorescence images showing ChR2-GFP expression in the dPVp and fibers in the LS, DBB, POA, LH, DMH, PVT, LPAG, DRN, periLC, LPBN, NTS, and RPa of TRAP2 mice. B. Scheme for retrograde Rabies tracing from cold-activated dPVp neurons. C. Rabies-GFP expression in the dPVp, POA and DMH.

Discussion

Temperature regulation is a fundamental and vital function of the central nervous system, which involves thermal-sensitive neural populations located in the POA, DMH and LPBN^18,38. A few neural subsets have been identified as being activated by cold, such as BRS3 neurons in the POA and DMH, BDNF neurons in the DMH, SST neurons in the LPBN, PDYN neurons in the dmVMH, and the xiphoid nucleus in the thalamus^19–23,26. However, the neural populations involved in cold defense beyond these regions remain largely unknown. The dPVp caught our attention as it is an intriguing and largely underexplored brain region located in the posterior hypothalamus. Unlike neighboring VMH neurons, which are predominantly glutamatergic, the dPVp is predominantly composed of GABAergic neurons. Furthermore, in contrast to GABAergic neurons from the ARH, which are inhibited by refeeding^39–41, the dPVp is activated by food consumption after a fasting challenge. These distinctive features of the dPVp set it apart from adjacent neurons and suggest a unique role for this nucleus in the regulation of energy and thermal homeostasis. We identified Lef1 as a molecular marker of cold-activated dPVp neurons. Although Lef1 and Pdyn are located in neighboring hypothalamic regions, they display distinct spatial distributions: Pdyn is enriched in the VMH, whereas Lef1 is concentrated in the caudal hypothalamus, with minimal co-localization. These findings indicate that Lef1-expressing and Pdyn-expressing neurons represent segregated cold-responsive populations. Pdyn neurons in the VMH have been shown to promote BAT thermogenesis and increase body temperature²⁶. Our data demonstrate that activation of Lef1 neurons in the dPVp enhances appetite and elevates body temperature, whereas silencing these neurons partially attenuates cold-induced responses. This partial effect suggests that additional cold-responsive populations, including those in the POA, DMH, LPBN, and VMH, contribute to maintaining homeostasis when dPVp activity is reduced. Moreover, we found that the dPVp receives inputs from key thermoregulatory centers, including the DMH and POA, and projects to other temperature-regulating regions such as the LPBN, RPa, and POA. However, the precise mechanisms by which dPVp neurons interact with other thermoregulatory circuits remain to be elucidated.

Proper temperature sensation is crucial to maintaining core body temperature and essential biological functions⁴². While changes in ambient temperature are thought to be detected by sensory nerve endings in the skin, core temperature variations can be detected by thermosensitive neurons in the brain^18,43. Remarkably, recent research has revealed a close correlation between brain temperature and core body temperature, both of which are influenced by ambient temperature⁴³. Hence, the brain is subjected to temperature fluctuations, highlighting the importance of internal temperature sensors. Within the brain, temperature sensing is orchestrated by distinct types of ion channels, including the transient receptor potential (TRP) channels, epithelial sodium channels and K2P (encoded by Kcnk2)^32,44. While TRPV1 and TRPM2 have been proposed as warm sensors^43,45, TRPM8, TRPA1 and Kcnk2 are well-established cold sensors^46–50. We found that GABA^dPVp neurons express Kcnk2 and possess intrinsic cold sensitivity. Kcnk2 gain-of-function effectively blocks the cold response of dPVp neurons, while its inhibition recapitulates the behavioral response induced by cold. We speculate that acute cold exposure directly inhibits Kcnk2 current, leading to rapid neuronal activation. However, direct measurement of dPVp temperature will be necessary to further test this possibility. At the molecular level, cold exposure downregulates Kcnk2 expression in the dPVp via suppression of Lef1. Disruption of Kcnk2 in cold-activated dPVp neurons increases food intake and body temperature, mimicking the physiological response to cold. We therefore propose a two-phase mechanism in which acute closure of the Kcnk2 channels drives the rapid activation of dPVp neurons, while reduced Kcnk2 expression further contributes to the long-term behavioral adaptations to cold exposure.

In summary, we discovered a Lef1⁺ neuronal population in the dPVp that exhibits profound activation in response to cold exposure. These neurons play a pivotal role in maintaining the body’s temperature homeostasis by modulating Kcnk2, a potential cold-sensing ion channel. Remarkably, chronic activation of dPVp neurons led to significant improvements in insulin sensitivity and glucose homeostasis. These findings advance our understanding of the neural mechanisms that underline the response to ambient cold and offer promising cellular (dPVp neurons) and molecular (Lef1 and Kcn2) targets for developing therapeutic strategies against type 2 diabetes.

Resource Availability:

Lead Contact:

Request for further information and resources should be directed to and will be fulfilled by the lead contact Yong Xu (yongxu@usf.edu).

Materials Availability:

Reagents generated in this study will be made available on request.

Data and Code Availability:

All data generated or analyzed during this study are included in this published article. Single-cell RNAseq data were previously published³¹ and are available at GSE146692. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the Lead Contact upon request.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Animals

8–16 weeks old adult male or female wild-type, TRAP2 mice (Jackson Laboratory, #030323), AgRP-Cre (Jackson Laboratory, #012899), POMC-CreER⁵⁶, Vgat-2A-Flp mice (Jackson Laboratory, #029591), Lef1-CreERT2 (Jackson Laboratory, #016236), and Lef1^flox/flox (Jackson Laboratory, #030908) mice were used in this study. We crossed TRAP2 mice with Rosa26-LSL-tdTomato mice (Jackson Laboratory, #007905) to generate TRAP2/Rosa26-LSL-tdTomato mice. AgRP/tdTomato and POMC-CreER/tdTomato mice were generated by crossing AgRP-Cre or POMC-CreER mice with Rosa26-LSL-tdTomato mice. Mice were housed in a temperature-controlled environment using a 12-hour light and 12-hour dark cycle. Mice were individually housed at least 1 week before the study. The mice were fed a standard chow (5% fat, 5V5R, LabDiet) or high-fat (60% fat, #D12492, Research Diets) diet. Water was provided ad libitum. All animal procedures were approved by the Institutional Animal Care and Use Committees of Baylor College of Medicine and the University of South Florida.

METHOD DETAILS

sgKcnk2 construction

The sgRNAs (for saCas9) targeting mouse Kcnk2 was designed using the CRISPR tools (https://portals.broadinstitute.org/gppx/crispick/public) with minimal potential off-target effects. A total of three sgRNAs were designed with 2 targeting exon 2 and 1 targeting exon 3. All 3 sgRNAs were screened for on-target activity using T7E1 assay. Briefly, the sgRNAs were first cloned into the plasmid px601 (a gift from Yuet Wai Kan, Addgene plasmid # 84040) and transfected into mouse cell line Neuro2A. The genomic DNA was extracted using Monarch^® Genomic DNA Purification Kit (NEB) and amplified with primer pair: for sgRNA targeting exon 2: 5’-AGAAAGCTGCTGGGTGAAGT-3’ and 5’-TGGTGCCCTAGTCTGCTCTT-3’, and for sgRNA targeting exon 3: 5’-AGCCATCCATGCTAGAGAACA-3’ and 5’-AGTTTGCTGAGTCCCTCCAT-3’. The genome editing efficiency was assessed using EnGen^® Mutation Detection Kit (NEB). The sgRNA#1 (TCCACCCGGGAAGCAAGCACG) and sgRNA#2 (AATAGTGGCAGCAATAAACGC) were selected to target exon 2 and exon 3 of the Kcnk2 gene, respectively, due to their relatively high on-target activity and low off-target potentials. Then the AAV-CMV-DIO-GFP-U6-sgRNA#1-U6-sgRNA#2 vector were constructed and verified by full sequencing. The virus was packaged by Baylor IDDRC Neuroconnectivity Core.

Stereotaxic surgery

Mice received buprenorphine and meloxicam injection 1 hour before surgery. During the surgery, mice were anesthetized (with 2% isoflurane) and placed in a stereotaxic instrument. Artificial eye ointment was applied to prevent corneal drying, and a heat pad was used to hold body temperature at 37°C. To chemogenetically activate or inhibit GABA^dPVp neurons, we injected AAV8-EF1α-fDIO-hM3Dq-mCherry (Addgene, #154868, titer: 4.16×10¹² GC/ml, 0.2 μl/injection) or AAV8-EF1α-fDIO-hM4Di-mCherry (Addgene, #154867, titer: 3.1×10¹² GC/ml, 0.2 μl/injection), respectively, into the dPVp (AP: −2.4 mm; ML: ±0.15 mm; DV: −5.65 mm) of Vgat-Flp mice. To chronically inhibit GABA^dPVp neurons, we injected AAV8EF1α-F-FLEX-Kir2.1-tdTomato into the dPVp of Vgat-Flp mice (Addgene, #60661, titer: 6.58×10¹⁰ GC/ml, 0.2 μl/injection). To record the response of GABA^dPVp neurons to cold or refeeding, Vgat-Flp mice received AAV8-EF1α-fDIO-GCaMP6 injection into the dPVp (Addgene, #105714, titer: 5×10¹² GC/ml, 0.2 μl). During the same surgery, an optic fiber (MFC_400/430–0.66, Doric) was placed over the dPVp (AP: −2.4 mm; ML: 0.15 mm; DV: −5.5 mm). To chemogenetically activate or inhibit cold-sensitive dPVp neurons, we injected AAV8-hSyn-DIO-hM3Dq-mCherry or AAV8-hSyn-DIO-hM4Di-mCherry (Addgene, #44362, titer: 5×10¹² GC/ml, 0.2 μl), respectively, into the dPVp of TRAP2 mice. To chronically inhibit cold-sensitive dPVp neurons, we injected AAV9-EF1α-DIO-Kir2.1-dTomato into the dPVp of TRAP2 mice (kindly provided by Dr. Mingshan Xue, titer: 3.46×10¹¹ GC/ml, 0.2 μl/injection).

To study the role of Kcnk2 in mediating the adaptive response of dPVp neurons to cold, TRAP2 mice received synapsin-driven Cre-dependent AAV9 carrying GFP (Addgene, #100043, titer: 2×10¹¹ GC/ml, 0.2 μl/injection) or Kcnk2 (Baylor IDDRC Neuroconnectivity Core, titer: x10¹¹ GC/ml, 0.2 μl/injection) or Kcnk2-E306A (Baylor IDDRC Neuroconnectivity Core, titer: 2×10¹¹ GC/ml, 0.2 μl/injection) into the dPVp. To explore the effects of Kcnk2-E306A on dPVp neurons’ response to cold, TRAP2 mice received Cre dependent GCaMP6m (AAV9-Syn-DIO-GCaMP6m, Addgene, #100838, titer: 5×10¹² GC/ml, 0.2 μl), along with GFP or Kcnk2 or Kcnk2-E306A, into the DPVp. During the same surgery, an optic fiber (MFC_400/430–0.66, Doric) was placed over the DPVp (AP: −2.4 mm; ML: 0.15 mm; DV: −5.5 mm). To disrupt Kcnk2 in cold-activated or Lef1⁺ dPVp neurons, Cre-dependent sgKcnk2 (Baylor IDDRC Neuroconnectivity Core, titer: 1×10¹² GC/ml, 0.1 μl/injection) and saCas9 (AAV-CMV7-DIO-saCas9, Vector Biolabs, #7122, titer: 1×10¹² GC/ml, 0.1 μl/injection) were co-injected into the dPVp of WT and TRAP2 or Lef1-CreERT2 mice. To study the effects of spadin, we implanted a cannula over the dPVp (AP: −2.4 mm; ML: 0.15 mm; DV: −5.4 mm) of WT or TRAP2 mice infected with Kcnk2-E306A.

To selectively activate or inhibit Lef1⁺ dPVp neurons, we injected AAV8-hSyn-DIO-hM3Dq-mCherry or AAV9-EF1α-DIO-Kir2.1-dTomato into the dPVp of Lef-CreERT2 mice (0.2 μl/site). One week after surgery, tamoxifen (MCE, #HY-13757A, 0.2g/kg) was administered to induce Cre activity. To knockdown Lef1 in the dPVp, AAV8-hSyn-Cre-GFP (Vector Biolabs, #7062, titer: 1×10¹² GC/ml, 0.2 μl/injection) or AAV8-hSyn-GFP (Vector Biolabs, #7061, titer: 1×10¹² GC/ml, 0.2 μl/injection) were injected into the dPVp of Lef1^flox/flox mice.

To map the projections of cold-sensitive dPVp neurons, we injected AAV9-EF1α-DIO-ChR2-GFP (Addgene, #20298, titer: 5×10¹² GC/ml, 0.2 μl) into the dPVp of TRAP2 mice. To identify the upstream inputs of cold-activated dPVp neurons, TRAP2 mice received Cre-dependent AAV expressing GTB (AAV8-EF1α-Flex-GTB, Addgene, #26197, titer: 2×10¹² GC/ml, 0.2 μl/injection) into the dPVp. Four weeks later, G-deleted rabies virus (EnvA G-deleted Rabies-GFP, Salk, #32635, titer: 1×10⁸ GC/ml, 0.2 μl/injection) was delivered into the same region.

At end of experiments, all mice were perfused with 10% formalin. Brain sections were collected and sectioned at 30 μm. The expression of virus was checked, only those with accurate targeting were included for data analyses.

TRAP induction

50 mg of 4-hydroxytamoxifen (4-OHT, Sigma, H6278) was dissolved in 2.5 mL ethanal to make a concentration at 20 mg/mL, then stored at −80 °C. Before use, 4-OHT was dissolved by shaking at 37°C for 10 minutes, then sunflower seed oil and castor oil (4:1) was added for a final concentration of 10 mg/mL. After evaporating the ethanol in a vacuum (3000 rpm, 15 minutes), the final 4-OHT solution was injected intraperitoneally at a dose of 50 mg/kg. For labeling cold-activated neurons, TRAP2 mice received 4-OHT injection and were subjected to cold exposure (6°C) immediately for 2 hours.

Chemogenetics

After allowing enough time for virus expression, mice that received hM3Dq or hM4Di injection were singly housed at least 1 week before they were subjected to metabolic studies. For acute measurements that were done at fed conditions, food was removed from cages 0.5h before saline or CNO (i.p., 3mg/kg, #16882, Cayman) injection. After saline or CNO injection, pre-weighed regular chow was put back to the cages. Food intake, rectal and BAT temperature were monitored at various time points as indicated in figures. In the fast-refeeding paradigm, CNO was injected 0.5h before food was added to the cages. To chronically activate GABAergic or cold-sensitive dPVp neurons, CNO was injected twice daily (9am and 9pm) for 10 or 12 consecutive days.

Behavioral assays

For temperature preference in the two-chamber box test, one side of the chamber was set at room temperature, the other side was pre-heated using a heat pad, with temperature controlled at 32°C. Mice were allowed to acclimate for 2 minutes, then explored in the box for 6 minutes. For mice that were infected with hM3Dq or hM4Di, CNO was injected 1 hour before they were put in the chamber. The time spent in each chamber was recorded and analyzed for chamber preference using the Ethovision XT software.

For temperature preference in the thermal gradient box test (Bio-TGT2, Bioseb), mice were placed in a linear chamber (125 × 10 × 18 cm3) with a floor temperature gradient and were allowed to freely explore the chamber for 1 hour to select the optimal-temperature position. For mice that were infected with hM3Dq, CNO was injected 1 hour before they were put in the chamber. One end of the chamber was set at 10 °C and the other end was set at 55 °C to establish a continuous temperature gradient from 15.2 °C to 46.1 °C. The box was divided into 20 different temperature zones, each was assigned a specific temperature: 15.2, 17.3, 18.2, 20.7, 21.6, 22.8, 23.7, 24.6, 25.5, 26.6, 27.6, 28.8, 30.3, 31.3, 32.8, 34.3, 36.4, 39.2, 43.2, and 46.1°C. These values represent the average temperatures of the respective zones. The time spent in each zone was recorded for the last 30 minutes. The preferred temperature was defined as the temperature of the zone where the mice spent the most time during the testing period. Data were analyzed using Thermal Gradient Test software (Bioseb). The preferred temperature was defined as the temperature zone where mice spent the most time during the recording period.

For the cold tolerance test, food was removed from the cage 1 hour before the test. For mice that were infected with hM4Di, CNO was injected 0.5 hour before the start of cold exposure. Afterwards, baseline BAT temperature was measured, pre-weighed food was provided to the cage, then mice were put into the cold room for 6 hours. Food intake and BAT temperature were measured at various time points.

BAT temperature and energy expenditure measurements

To measure BAT temperature, a miniature telemetric transponder (IPTT300, BioMedic Data Systems) was implanted beneath the interscapular BAT of anesthetized mice. After recovering for at least one week, mice were subjected to various treatments as mentioned in the manuscript, BAT temperature was measured using a wireless scanner (DAS-8007, BioMedic Data Systems). For mice that were infected with hM3Dq, CNO was injected 0.5 hour before the measurements. Energy expenditure measurements were performed in temperature-controlled (23 °C) cabinets containing 16 TSE PhenoMaster metabolic cages. Mice were acclimatized to these cages for at least 2 days. For the mice that received hM3Dq infection, saline and CNO were given at 8am on separate days.

Core body temperature and BAT surface temperature measurements

To monitor core body temperature, mice were anesthetized with 2% isoflurane and placed in dorsal recumbency on a heating pad. All surgical procedures were performed under sterile conditions. A small midline abdominal incision was made, and a telemetry temperature probe (G2 E-Mitter, STARR Life Sciences) was carefully implanted into the abdominal cavity while avoiding damage to the visceral organs. The abdominal musculature and skin were then closed separately using non-absorbable sutures. Following surgery, mice were singly housed and monitored daily until the surgical wound had completely healed. Animals were allowed to recover for at least one week before experimentation. The hair over the interscapular BAT region was shaved 24–48 h before testing. On the day of the experiment, mice were placed individually in cages positioned on telemetry receiver plates to continuously record core body temperature and locomotor activity. BAT surface temperature was simultaneously measured using a thermal imaging camera (FLIR). Core body temperature was recorded at 1 Hz, and BAT surface temperature was acquired every 10 s. Baseline values were defined as the 30-min period immediately preceding CNO administration or cold exposure (placement of the cage in an ice-water bath), depending on the experimental paradigm.

Serum norepinephrine assay

Blood was collected at the end of study and processed to measure serum norepinephrine levels using the mouse norepinephrine ELISA kit (KA1891, Abnova). For mice that were infected with hM3Dq, CNO was injected 2 hours before the collection of blood.

Single-nucleus RNA-seq analysis

8-week-old male WT mice were exposed to cold (6°C) for 30 minutes, then the brains were isolated, and the posterior part of the hypothalamus was collected. The tissue was put into 1.5ml RNAase free Eppendorf tubes, flash-frozen using liquid nitrogen, and stored at −80°C. For each sample, brain tissues from five mice were pooled into a single tube for single-nucleus suspensions. Single-nucleus suspensions were prepared following the protocol we described previously⁵⁷. Nuclei were stained by Hoechst-33342 (1:1000; >5min). Next, we collected nuclei using Fluorescence activated cell sorting (FACS). We used the BD Aria III sorter for collecting nuclei. Hoechst+ nuclei were collected during sorting into a 1.5ml tube with 200ul 1x PBS with 0.5% BSA as the receiving buffer (RNase inhibitor added). For each 10x Genomics run, 80k–150k nuclei were collected. Nuclei were spun down for 10 min at 800g at 4°C, and then resuspended using 30ul of 1x PBS with 0.5% BSA (RNase inhibitor added). 2ul nucleus suspension was used for counting the nuclei with hemocytometers to calculate the concentration. When loading to the 10x controller, we always try to target 10k nuclei for each run. We observed that loading 2 folds more could allow us to recover about 10k cells after sequencing. For example, if we try to target 10k nuclei, we will load 20k into a 10x controller. Next, 10x Genomics sequencing libraries were prepared following the standard protocol from 10x Genomics 3’ v3.1 dual index kit with following settings. All PCR reactions were performed using the Biorad C1000 Touch Thermal cycler with 96-Deep Well Reaction Module. Cycle numbers were used as the 10x protocol recommended for cDNA amplification and sample index PCR. As per 10x protocol, 1:10 dilutions of amplified cDNA and final libraries were evaluated on Agilent TapeStation. Libraries were then shipped on dry ice to Novogene (Sacramento, CA) for sequencing. Fastqs were downloaded from Novogene’s website. Fastqs were aligned to Mus musculus genome assembly GRCm39 using STAR solo 2.7.10b using options soloFeatures set to GeneFull_Ex50pAS and soloCellFilter set to EmptyDrops_CR. Output was then loaded into R 4.4.0. Ambient RNA was estimated and removed using SoupX 1.6.2 and using adjustCounts method set to “multinomial”⁵⁸. Potential doublets were identified and removed using scDblFinder 1.17.3⁵⁹. Cells with extreme values for counts, mitochondrial percent, ribosomal protein percent, and hemoglobin counts were filtered out using mean + 4 s.d. as a threshold. Data were analyzed using Seurat 5.0.3⁶⁰. Counts were normalized using SCTransform function with do.scale and do.center set to false. Mitochondrial, ribosomal proteins, and hemoglobin genes were excluded from variable genes. Principal component (PC) analysis was performed on the variable features. T-distributed stochastic neighbor embedding was estimated for the 1st 30 PCs. Clusters were determined using FindNeighbors and FindClusters functions with resolution set to 0.8. Marker genes were then used to annotate cell-types. One to two genes with the largest log2 fold-change were used to annotate the cell-types. Fos and select marker genes were then plotted with the function FeaturePlot.

RNAscope

To check the co-localization of cold-induced Fos with Slc32a1 and Slc17a6, 8-week WT mice were placed in a cold room (6°C) for 40 minutes before perfusion. To evaluate the effect of cold exposure on Kcnk2 expression and to characterize the distribution of Lef1 and Pdyn, as well as their responsiveness to cold, 8-week WT mice were housed at room temperature (23°C) or challenged with cold (6°C) for 40 minutes before perfusion. Food was removed before cold exposure. To determine whether AgRP neuronal activity is influenced by activation of the dPVp, TRAP2 mice were injected with a Cre-dependent hM3Dq virus in the dPVp. Two weeks after cold TRAP, mice received CNO (3 mg/kg) 30 minutes prior to perfusion. To examine whether inhibition of the dPVp alters the response of AgRP neurons to cold, TRAP2 mice were injected with a Cre-dependent hM4Di virus in the dPVp. Two weeks after cold TRAP, mice received CNO (3 mg/kg) 30 minutes prior to a 40-minute cold exposure (6 °C), followed by perfusion. Mice were anesthetized and perfused transcardially with 0.9% saline followed by 10% formalin. Brains were removed and post fixed in 10% formalin for 16 h at 4°C and cryoprotected in 30% sucrose for 48 hours. Brains were frozen, sectioned at 20 μm using the cryostat, washed in DEPC-treated phosphate buffered saline (PBS) for 10 minutes. Sections were mounted on DEPC-treated charged slides, dried for 0.5 hour at room temperature and stored at −80°C. On the day of the RNAScope assay, the slides were thawed and rinsed 2 times in 1X PBS and placed in an oven for 30 minutes at 60°C. After that, slides were post fixed in 10% formalin for 20 minutes at 4°C. Slides were then gradually dehydrated in ethanol (50, 70 and 100%, 5 minutes each) and underwent target retrieval for 5 minutes at 100°C. Slides were incubated in protease III (#322337, ACDBio) for 30 minutes at 40°C. Slides were then rinsed in distilled water and incubated in RNAscope probes for Slc32a1 (#319191-C3, ACDBio), or Slc17a6 (#319171, ACDBio), or Fos (#584741-C2/C3, ACDBio) or Kcnk2 (#440421-C2, ACDBio) or Lef1 (#441861/C3, ACDBio) or Tbx19 (#484741-C2, ACDBio) or Pdyn (#318771, ACDBio), or Agrp (#537161, ACDBio) for 2 hours at 40°C. Sections were then processed using the RNAscope Multiplex Fluorescent Reagent Kit v2 (#323100, ACDBio) according to the manufacturer instructions. Slides were cover-slipped and analyzed using a fluorescence microscope.

Fiber photometry

Mice that received GCaMP6 injection into the dPVp were allowed enough time for recovery, then these mice were acclimatized to investigator handling for 1 week before experiments. To monitor the response to cold, mice were put into pre-cold cages (6°C) after recording the baseline signals for 5 minutes. To record the response to chow food, mice were fasted overnight. The next morning, mice were given 5 minutes to acclimate to the tethered patchcord prior to recording. Chow was provided back to the cages after recording the baseline signals for 5 minutes. Continuous <20 μW blue LED at 465 nm and UV LED at 405 nm served as excitation light sources, driven by a multichannel hub (Doric Lenses), modulated at 211 Hz and 330 Hz, respectively. The light was delivered to a filtered minicube (FMC5, Doric Lenses) before connecting through optic fibers to a rotary joint (FRJ 1 × 1, Doric Lenses) to allow for movement. GCaMP6 calcium GFP signals and UV autofluorescent signals were collected through the same fibers back to the dichroic ports of the minicube into a femtowatt silicon photoreceiver (2151, Newport). The digital signals were then amplified, demodulated, and collected through a lock-in amplifier (RZ5P, Tucker-Davis Technologies). The fiber photometry data was collected using Synapse 2.0 (Tucker-Davis Technologies) and sampled down to 8 Hz. We derived the values of GCaMP6 fluorescence change (ΔF/F) by calculating (F₄₆₅−F₀)/F₀, where F₀ is the baseline fluorescence of the F₄₆₅ channel signal 5 seconds prior to the onset of cold or food exposure. The F₄₀₅ channel was used as an isosbestic fluorescence channel; we derived the values of isosbestic fluorescence change (ΔF/F₀) by calculating (F₄₀₅−F₀)/F₀, where F₀ is the baseline fluorescence of the F₄₀₅ channel signal 5 seconds prior to the onset of cold or food exposure.

Glucose tolerance test (GTT) and Insulin sensitivity test (ITT)

For GTT, after an overnight fast, mice received intraperitoneal (i.p.) injections of 1g/kg D-glucose (G8270, Sigma) at 9:00 am. Blood glucose was measured from tail blood using a glucometer (OneTouch Ultra) at the serial time points as indicated in figures (0, 15, 30, 60 and 120 min). For ITT, after a 4-h fast to empty the stomach, mice received i.p. injections of insulin (0.75 U/kg). Blood glucose was measured at 0, 15, 30, 60 and 90 min.

Real-time PCR

At the end of study, mice were sacrificed and the caudal hypothalamus or the interscapular brown adipose tissues were isolated. For mice that were infected with hM3Dq, CNO was injected 2 hours before they were sacrificed. RNA from these tissues were extracted using the RNeasy Lipid Tissue Mini Kit (#166050808, QIAGEN) according to the manufacturer’s instruction. 2 μg mRNA was reversed-transcribed using the qScript cDNA SuperMix (#66179183, Quanta bio). Quantitative PCR reactions were performed by a CFX384 Real-Time System (Bio-Rad) using SsoAdvanced Universal SYBR Green SuperMix (#1725274, Bio-Rad). Gene expression levels were determined using 36B4 (for BAT) or β-actin (for hypothalamus) as the housekeeping gene, and relative quantification was achieved using the ΔΔCT method. Primer sequences were listed below: 36B4 (F-GAGGAATCAGATGAGGATATGGGA, R-AAGCAGGCTGACTTGGTTGC), Ucp1 (F-AGGCTTCCAGTACCATTAGGT, R-CTGAGTGAGGCAAAGCTGATTT), β-actin (F-GGCTGTATTCCCCTCCATCG, RCCAGTTGGTAACAATGCCATGT), Lef1 (F-TGTTTATCCCATCACGGGTGG, R-CATGGAAGTGTCGCCTGACAG), Kcnk2 (F-CCGAGGCTCTCATTCTCCTCA, R-AGGACGACCACCAGGAAAATC).

Western blotting

Eight-week-old male WT mice were housed at room temperature (23°C) or exposed to cold (23°C) for 4 hours. The caudal hypothalamus was then collected and lysed in RIPA buffer (J63306, Alfa Aesar) supplemented with a protease inhibitor cocktail (#11697498001, Millipore Sigma). Brain tissues were homogenized for 1 minute, and lysates were centrifuged at 12,000 × g for 15 minutes at 4 °C. The resulting supernatants containing protein extracts were subjected to SDS-PAGE and immunoblot analysis. Proteins were separated on a 10% SDS-polyacrylamide gel and transferred to a polyvinylidene difluoride (PVDF) membrane. Membranes were incubated overnight at 4 °C with primary antibodies against Lef1 (#2230, Cell Signaling Technology), Kcnk2 (APC-047, Alomone Labs), or β-actin (AC004, ABclonal). After washing, membranes were incubated for 1 hour with Alexa Fluor Plus 680-conjugated (A32729, Invitrogen) or 800-conjugated (A32735, Invitrogen) secondary antibodies. Fluorescent signals were detected using the Odyssey Infrared Imaging System (LI-COR Biotechnology), and band intensities were quantified using ImageJ software. Lef1 and Kcnk2 protein levels were normalized to β-actin in each sample.

Immunohistochemistry

To check cold-induced c-Fos expression, 8-week-old male WT mice or AgRP/tdTomato or POMC-CreER/tdTomato mice were housed at room temperature (23°C) or challenged with cold (6°C) for 1 hour before perfusion. POMC-CreER/tdTomato mice received tamoxifen (i.p., 0.2g/kg) injection 1 week prior to cold exposure. To assess the TRAP efficiency, TRAP2/tdTomato mice were trapped with cold exposure to induce tdTomato expression in cold-activated neurons. Two weeks later, these mice were subjected to a second cold exposure (6°C) for 1 hour. To ascertain whether cold exposure and refeeding activate the same or different neural subsets, TRAP2/tdTomato mice were subjected to cold trap procedure. After a two-week interval, these mice were fasted overnight and then given access to food for 2h the following morning. Mice were transcardially perfused with saline, followed by 10% formalin. The brain sections were cut at 30 μm and collected into five consecutive series. One series of the sections were blocked (3% Normal donkey serum) for 1 hour, incubated with Rabbit anti-c-Fos antibody (1:1,000, 226008, Synaptic Systems) on shaker at 4°C for overnight, followed by donkey anti-rabbit AlexaFluor 488 (1:200, A21206, Invitrogen) for 2 h. To check the colocalization of Lef1 with cold-induced c-Fos, brain sections from control and cold treated mice were incubated with Rabbit anti-Lef1 (1:500; 2230S, Cell Signaling Technology) and Mouse anti-c-Fos (1:1,000, ab208942, abcam) on shaker at 4°C for overnight, followed by donkey anti-rabbit AlexaFluor 488 and donkey anti-mouse AlexaFluor 594 (1:200, A21203, Invitrogen) for 2 h. To validate the knockdown of Lef1, brain sections from control and Lef1-KD mice were incubated with Rabbit anti-Lef1 (1:500; 2230S, Cell Signaling Technology) at 4°C for overnight, followed by donkey anti-rabbit AlexaFluor 594 (1:200, A21207, Invitrogen) for 2 h. Slides were cover-slipped and analyzed using a fluorescence microscope. To validate the knockdown of Lef1, For whole brain c-Fos mapping, the brain sections were pretreated with 0.3% H₂O₂ for 30 min, then blocked with 3% Normal goat serum for 1 h, incubated with mouse anti-c-Fos antibody (1:1,000; ab208942, Abcam) on shaker at room temperature for overnight, followed by biotinylated goat anti-mouse secondary antibody (1:1,000; BA-9200, Vector) for 2 h. Sections were then incubated in the avidin-biotin complex (1:500, ABC; PK-6101, Vector) and incubated in 0.04% 3,3′-diaminobenzidine and 0.01% hydrogen peroxide. After dehydration, the slides were covered and imaged using a bright-field microscope.

Whole-brain imaging

8-week-old male TRAP2/tdTomato mice received 4-OHT injection (50mg/kg) and were subjected to cold exposure (6°C) for 2 hours. Four weeks later, mice were perfused with saline followed by 10% formalin, brains were harvested and fixed overnight in 10% formalin at 4°C. The fixed whole brains were processed with SHIELD protocol and the Delipidation protocol according to the manufacturer’s instructions (LifeCanvas Technologies, Cambridge, MA). Then, brain samples were incubated overnight in 50% EasyIndex (RI = 1.52, LifeCanvas Technologies) at 37 °C, followed by 1-day incubation in 100% EasyIndex for refractive index matching. After that, the samples were imaged with ZEISS Lightsheet 7 microscope using Detection Optics 20x/1.0 and Illumination 10X/0.2 and lasers at 561 nm, followed by image post-processing.

Slice electrophysiology

Electrophysiology recordings were performed as previously described⁶¹. Mice were deeply anesthetized with isoflurane and transcardially perfused with a modified ice-cold sucrose-based cutting solution (pH 7.3) containing 10 mM NaCl, 25 mM NaHCO₃, 195 mM Sucrose, 5 mM Glucose, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 2 mM sodium pyruvate, 0.5 mM CaCl₂, and 7 mM MgCl₂, bubbled continuously with 95% O₂ and 5% CO₂. The mice were then decapitated, and the entire brain was removed and immediately submerged in the cutting solution. Coronal brain slices (220 μm) containing the dPVp or LPAG were cut with a Microm HM 650 V vibratome (Thermo Scientific) in oxygenated cutting solution. Slices were then incubated in oxygenated artificial CSF (aCSF; 126 mM NaCl, 2.5 mM KCl, 2.4 mM CaCl₂, 1.2 mM NaH₂PO₄, 1.2 mM MgCl₂, 11.1 mM glucose, and 21.4 mM NaHCO₃, balanced with 95% O₂/5% CO₂, pH7.4) to recover ~25 min at 34°C and subsequently for 1 hour at room temperature before recording. Slices were transferred to a recording chamber and allowed to equilibrate for at least 10 minutes before recording. The slices were superfused at 34°C in oxygenated aCSF at a flow rate of 1.8–2 ml/min. GFP or mCherry or tdTomato-labeled neurons were visualized using epifluorescence and IR-DIC imaging on an upright microscope (Eclipse FN-1, Nikon) equipped with a movable stage (MP-285, Sutter Instrument). Patch pipettes with resistances of 3–5 MΩ were filled with intracellular solution (pH 7.3) containing 128 mM K-Gluconate, 10 mM KCl, 10 mM HEPES, 0.1 mM EGTA, 2 mM MgCl₂, 0.3 mM Na-GTP and 3 mM Mg-ATP. Recordings were made using a MultiClamp 700B amplifier (Axon Instrument), sampled using Digidata 1440A and analyzed offline with pClamp 10.3 software (Axon Instruments). Series resistance was monitored during the recording, and the values were generally <10 MΩ and were not compensated. The liquid junction potential was +12.5 mV and was corrected after the experiment. Data were excluded if the series resistance increased dramatically during the experiment or without overshoot for action potential. Currents were amplified, filtered at 1 kHz, and digitized at 20 kHz. Current clamp was engaged to test neural firing frequency and resting membrane potential at the baseline or in response to CNO (Cayman, #16882, 10 μM) or spadin (Tocris, #5594/1, 1 μM) or 34°C → 30°C and 30°C → 34°C temperature changes. Synaptic blockers (TTX (1 μM), CNQX (30 μM), D-AP5 (30 μM), and bicuculline (50 μM)) were added to the aCSF in to determine whether the responses to spadin and temperature changes are direct. Thermal sensitivity was quantified using both the thermal coefficient (slope of firing vs. temperature)⁶² and Q₁₀ analysis. Q₁₀ values were calculated using responses measured at 34°C and 30°C according to the standard formula (Q₁₀=(R2/R1)^10/(T2−T1))⁶³. To isolate potassium current in dPVp neurons, TTX (1 μM) was added to the aCSF to block sodium channel and CaCl₂ was removed from the aCSF to exclude calcium current, with NaCl added to maintain equal osmolarity. A 600 ms depolarization pulse from −60 mV to +50 mV was applied in the voltage-clamp mode to induce potassium currents in dPVp neurons at different temperatures. The selective Kcnk2 current blocker spadin was used to identify Kcnk2 current changes following temperature fluctuations. Blue light stimulation (20 Hz) was utilized for the light-evoked IPSC recordings, the internal recording solution contained: 140 mM CsCl; 5 mM MgCl2; 1 mM EGTA; 10 mM HEPES; 5 mM (Mg) ATP; 0.3 mM (Na) GTP, 3 mM QX-314 (pH 7.3 with CsOH). Bicuculline (50 μM; a GABA receptor antagonist) was used to confirm the GABAergic inputs. 4-AP and TTX were used to confirm the evoked IPSC currents are monosynaptic currents.

ChIP PCR

Eight-week-old male WT mice were housed at room temperature (23°C) or exposed to cold (4°C) for 2 hours. The caudal hypothalamus was then collected and homogenized for subsequent use in the ChIP assay (Sigma, 17–611). Anti-Lef1 antibody (ab137872, Abcam) or IgG was used for immunoprecipitation. DNA was then purified using spin columns. Purified DNA was subjected to PCR using primers specific for putative Lef1-binding sites in the promoter (F: AGGCGGCCACGCAATC; R: GTAGGCATTTTCCCCGGCTTT) and enhancer regions (EnhancerAtlas 2.0; http://enhanceratlas.org/;>Brain_E14.5_chr1: 191011600–191012330_ENSMUSG00000037624; F: TAACCCTTCCCCAGTGTAGC; R: CACGGGTAGGCATTTTCCC) of target genes.

Cell culture and luciferase assay

Mouse Kcnk2 promoter (−882-+20) was cloned by PCR (Figure S16A) and was inserted into the PGL-3 basic vector (Promega). Neuro-2a (N2A) cells were cultured in Dulbecco’s Modified Eagle’s Minimum Essential Medium supplemented with 10% (v/v) of fetal bovine serum under 5% CO2 atmosphere at 37oC humidified incubator. To knock down Lef1 in N2A cells, N2A cells were transfected with either Lef1-targeting siRNA (sc-35805, Santa Cruz Biotech) or a control siRNA (sc-37007, Santa Cruz Biotech) using Lipofectamine RNAiMAX (Thermo Fisher) according to the manufacturer’s instructions. To overexpress Lef1 in N2A cells, cells were transiently transfected with pcDNA3.1(+1)-Lef1 plasmid using Lipofectamine 3000 reagent (Thermo Fisher). One day after transfection, N2A cells were passaged and, upon reaching 70% confluency, transfected with the PGL-3-Kcnk2 promoter plasmid using Lipofectamine LTX reagent (Thermo Fisher). Another day after transfection, cells were lysed and assayed using luciferase reagents (Promega). To assess the functional significance of Lef1 binding to the Kcnk2 regulatory region, Lef1-binding sites on the Kcnk2 promoter were mutated from the WT promoter luciferase reporter construct using the Q5^® Site-Directed Mutagenesis Kit (E0554S, New England Biolabs).

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistics

The minimal sample size was determined by the nature of experiments. For behavioral measurements 5–10 different mice per group were included. For histology studies, the same experiment was repeated in at least 3 different mice. For electrophysiological studies, at least 6 different neurons from 3 different mice were included. The data are presented as mean ± SEM or as individual data points. Statistical analyses were performed using GraphPad Prism to evaluate normal distribution and variations within and among groups. Methods of statistical analyses were chosen based on the design of each experiment and are indicated in figure legends. P<0.05 was considered to be statistically significant.

Supplementary Material

1

Supplementary Movie 1: Cold-TRAPed neurons throughout the brain. (related to Figure 1) Whole-brain imaging showing tdTomato expression in cold-TRAPed TRAP2/tdTomato mice.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti-Lef1 antibody	Cell Signaling Technology	Cat# 2230; RRID:AB_823558
Rabbit anti-Lef1 antibody	Abcam	Cat# ab137872; RRID:AB_2892647
Rabbit anti-c-Fos antibody	Synaptic Systems	Cat# 226008; RRID:AB_2891278
Mouse anti-c-Fos antibody	Abcam	Cat# ab208942; RRID:AB_2747772
Alexa Fluor 488 conjugated donkey anti-rabbit IgG	Invitrogen	Cat# A21206; RRID:AB_2535792
Alexa Fluor 594 conjugated donkey anti-mouse IgG	Invitrogen	Cat# A21203; RRID:AB_2535789
Alexa Fluor 594 conjugated donkey anti-rabbit IgG	Invitrogen	Cat# A21207; RRID:AB_141637
Biotinylated goat anti-mouse IgG	Vector Laboratories	Cat# BA-9200; RRID:AB_2336171
Rabbit anti-Kcnk2 antibody	Alomone Labs	Cat# APC-047; RRID:AB_2040136
Mouse anti- β-actin antibody	ABclonal	Cat# AC004; RRID:AB_2737399
Alexa Fluor Plus 680-conjugated goat anti-mouse IgG	Invitrogen	Cat# A32729; RRID:AB_2633278
Alexa Fluor Plus 800-conjugated goat anti-rabbit IgG	Invitrogen	Cat# A32735; RRID:AB_2633284
Bacterial and virus strains
AAV8-EF1α-fDIO-hM3Dq-mCherry	Ulrik Gether Lab	Addgene, 154868
AAV8-EF1α-fDIO-hM4Di-mCherry	Runegaard et al.51	Addgene, 154867
AAVDJ-EF1α-F-FLEX-Kir2.1-tdTomato	Xue et al.52	Addgene, 60661
AAV8-EF1α -fDIO-GCaMP6s	Rylan Larsen Lab	Addgene, 105714
AAV8-hSyn-DIO-hM3Dq-mCherry	Krashes et al.53	Addgene, 44361
AAV8-hSyn-DIO-hM4Di-mCherry	Krashes et al.53	Addgene, 44362
AAV9-EF1α-DIO-Kir2.1-dTomato	Xue et al.52	N/A
AAV9-hSyn-DIO-EGFP	Bryan Roth Lab	Addgene, 50457
AAVDJ-DIO-Kcnk2-GFP	This paper	N/A
AAVDJ-DIO-Kcnk2-E306A-GFP	This paper	N/A
AAV9-Syn-DIO-GCaMP6m	Chen et al.54	Addgene, 100838
AAVDJ-CMV-sgKcnk2-DIO-GFP	This paper	N/A
AAV-CMV7-DIO-saCas9	Vector Biolabs	7122
AAV8-hSyn-Cre-GFP	Vector Biolabs	7062
AAV8-hSyn-GFP	Vector Biolabs	7061
AAV9-EF1α-DIO-ChR2-GFP	Karl Deisseroth Lab	Addgene, 20298
AAV8-EF1α-Flex-GTB	Haubensak et al.55	Addgene, 26197
EnvA G-deleted Rabies-GFP	Salk	32635
Biological samples
Chemicals, peptides, and recombinant proteins
RNAscope Probe-Mm-Slc32a1-C2	ACDBio	319191-C2
RNAscope Probe-Mm-Slc17a6	ACDBio	319171
RNAscope Probe-Mm-Fos-C3	ACDBio	584741-C3
RNAscope Probe-Mm-Fos-C2	ACDBio	584741-C2
RNAscope Probe-Mm-Kcnk2-C2	ACDBio	440421-C2
RNAscope Probe-Mm-Tbx19-C2	ACDBio	484741-C2
RNAscope Probe-Mm-Lef1-C3	ACDBio	441861-C3
RNAscope Probe-Mm-Lef1	ACDBio	441861
RNAscope Probe-Mm-Agrp	ACDBio	537161
RNAscope Probe-Mm-Pdyn	ACDBio	318771
RNAscope Protease III	ACDBio	322337
Clozapine N-oxide	Cayman	16882
4-hydroxytamoxifen	Sigma	H6278
Spadin	Tocris Bioscience	5594
TTX	Tocris Bioscience	1078
CNQX	Tocris Bioscience	0190
D-AP5	Tocris Bioscience	0106
Bicuculline	Tocris Bioscience	0130
4-AP	Tocris Bioscience	0940
Lipofectamine 3000	Thermo Fisher Scientific	L3000008
Lipofectamine RNAiMAX	Thermo Fisher Scientific	13-778-150
DAPI	Vector Laboratories	H-1500-10
cOmplete™ Protease Inhibitor Cocktail	MilliporeSigma	11697498001
SYBR Green SuperMix	Bio-Rad	1725274
Critical commercial assays
RNAscope Multiplex Fluorescent Reagent Kit v2	ACDBio	323100
Q5® Site-Directed Mutagenesis Kit	New England Biolabs	E0554S
VECTASTAIN® Elite ABC-HRP Kit	Vector Laboratories	PK-6101
Vector VIP Substrate Kit, Peroxidase (HRP)	Vector Laboratories	SK-4600
Norepinephrine ELISA Kit	Abnova	KA1891
Monarch® Genomic DNA Purification Kit	New England Biolabs	T3010
EnGen® Mutation Detection Kit	New England Biolabs	E3321S
RNeasy Lipid Tissue Mini Kit	QIAGEN	166050808
Deposited data
Single-nucleus RNA seq	This paper	N/A
Experimental models: Cell lines
N2A	American Type Culture Collection	Cat# CCL-131; RRID: CVCL_0470
Experimental models: Organisms/strains
TRAP2	Jackson Laboratory	030323
AgRP-Cre	Jackson Laboratory	012899
Vgat-2A-Flp	Jackson Laboratory	029591
Lef1-CreERT2	Jackson Laboratory	016236
Lef1flox/flox	Jackson Laboratory	030908
Rosa26-LSL-tdTomato	Jackson Laboratory	007905
POMC-CreER	Berglund et al	N/A
Oligonucleotides
Kcnk2-targeting siRNA	Santa Cruz Biotech	sc-35805
Scramble siRNA	Santa Cruz Biotech	sc-37007
QPCR primers for 36B4: F-GAGGAATCAGATGAGGATATGGGA, R-AAGCAGGCTGACTTGGTTGC	Sigma	N/A
QPCR primers for Ucp1: F-AGGCTTCCAGTACCATTAGGT, R-CTGAGTGAGGCAAAGCTGATTT	Sigma	N/A
QPCR primers for β-actin: F-GGCTGTATTCCCCTCCATCG, R-CCAGTTGGTAACAATGCCATGT	Sigma	N/A
QPCR primers for Lef1: F-TGTTTATCCCATCACGGGTGG, R-CATGGAAGTGTCGCCTGACAG	Sigma	N/A
QPCR primers for Kcnk2: F-CCGAGGCTCTCATTCTCCTCA, R-AGGACGACCACCAGGAAAATC	Sigma	N/A
Primers targeting the Lef1 promoter: F: AGGCGGCCACGCAATC, R: GTAGGCATTTTCCCCGGCTTT	Sigma	N/A
Primers targeting the Lef1 enhancer: F-TAACCCTTCCCCAGTGTAGC; R: CACGGGTAGGCATTTTCCC	Sigma	N/A
Primers for sgRNA targeting Kcnk2 exon 2: 5’-AGAAAGCTGCTGGGTGAAGT-3’ and 5’-TGGTGCCCTAGTCTGCTCTT-3’	Sigma	N/A
Primers for sgRNA targeting Kcnk2 exon 3: 5’-AGCCATCCATGCTAGAGAACA-3’ and 5’-AGTTTGCTGAGTCCCTCCAT-3’	Sigma	N/A
Recombinant DNA
pcDNA3.1(+1)-Lef1	This paper	N/A
PGL-3-Kcnk2	This paper	N/A
Software and algorithms
Graphpad Prism 10		https://www.graphpad.com/scientific-software/prism/
Adobe Photoshop 2022	Adobe	https://www.adobe.com/
Adobe Illustrator 2022	Adobe	https://www.adobe.com/
Other
Optic Fiber	Doric Lenses	B280-4655
Miniature Telemetric Transponder	BioMedic Data Systems	IPTT300
Chow Diet	LabDiet	5V5R
High-fat Diet	Research Diets	D12492

Highlights.

• Cold exposure activates Lef1⁺ hypothalamic neurons.

• Lef1⁺ hypothalamic neurons increase feeding and body temperature.

• Cold exposure suppresses Kcnk2 activity to increase neuronal excitability.

• Manipulating Kcnk2 bidirectionally controls physiological adaptations to cold.