GIRK2 potassium channels expressed by the AgRP neurons decrease adiposity and body weight in mice

It is well known that the neuropeptide Y (NPY)/agouti-related peptide (AgRP) neurons increase appetite and decrease thermogenesis. Previous studies demonstrated that optogenetic and/or chemogenetic manipulations of NPY/AgRP neuronal activity alter food intake and/or energy expenditure (EE). However, little is known about intrinsic molecules regulating NPY/AgRP neuronal excitability to affect long-term metabolic function. Here, we found that the G protein-gated inwardly rectifying K+ (GIRK) channels are key to stabilize NPY/AgRP neurons and that NPY/AgRP neuron-selective deletion of the GIRK2 subunit results in a persistently increased excitability of the NPY/AgRP neurons. Interestingly, increased body weight and adiposity observed in the NPY/AgRP neuron-selective GIRK2 knockout mice were due to decreased sympathetic activity and EE, while food intake remained unchanged. The conditional knockout mice also showed compromised adaptation to coldness. In summary, our study identified GIRK2 as a key determinant of NPY/AgRP neuronal excitability and driver of EE in physiological and stress conditions.


Introduction
The arcuate nucleus of the hypothalamus (ARH) is home to several distinct types of neurons that control energy homeostasis [1].In particular, it is well known that neurons co-expressing neuropeptide Y (NPY) and agouti-related peptide (AgRP) (NPY/AgRP neurons) promote food intake [2,3].NPY/AgRP neurons also decrease energy expenditure (EE), at least in part by suppressing sympathetic tone to the brown adipose tissue (BAT) and inhibiting thermogenesis [4,5].Consistent with these findings, manipulating the activity of NPY/AgRP neurons using exogenous genetic constructs (e.g., channelrhodopsin and designer receptors) resulted in acute changes in food intake and energy utilization [6,7].While these studies provided insight into how NPY/AgRP neuronal activity is translated to in vivo metabolic function, we have little information regarding intrinsic molecules that regulate NPY/AgRP neuronal activity per se.
In many excitable cells, the resting membrane potential (RMP) is maintained largely by K + channels [8].For example, the "classic" inwardly rectifying K + (IRK or Kir2) channels maintain RMP of cardiac myocytes [9] and ATP-sensitive K + (K ATP ) channels silence pancreatic β-cells [10].In neurons, K ATP channels and G protein-gated inwardly rectifying K + (GIRK or Kir3) channels have been reported to open at rest to dampen cellular excitability.For example, K ATP channel activity hyperpolarizes membrane potential of the pro-opiomelanocortin (POMC) neurons of the ARH [11] and the serotonin 2C receptor-expressing neurons of the lateral parabrachial nucleus [12].It was also demonstrated that GIRK channels maintain RMP of arcuate POMC neurons [13] and hippocampal CA1 neurons [14].However, little data is currently available on the identity of K + channels that regulate RMP of NPY/AgRP neurons.
In this study, we utilized multiple approaches to identify specific K + channels that regulate NPY/AgRP neuronal activity.Firstly, we found evidence that GIRK2-containing GIRK channels suppress the activity of NPY/AgRP neurons.We subsequently found that GIRK2 ablation in NPY/AgRP neurons results in increased body weight and adiposity when the mice are fed normal chow diet (NCD).Notably, the observed phenotypes were attributed to decreased sympathetic activity and energy expenditure, rather than an increase of food intake.We also found evidence that GIRK2 expressed by NPY/AgRP neurons has a role in cold-induced thermogenesis.Collectively, our results suggest that GIRK2 dampens excitability of the NPY/AgRP neurons to maintain sympathetic tone and thermogenesis in physiological and some stress conditions, which may serve to keep body weight in control independently of appetite.

GIRK channels maintain RMP of NPY neurons
A previous study reported a transcriptome obtained from AgRP neurons [15], which included mRNA of various K + channels that may contribute to maintenance of RMP.In particular, Mtype K + or M channels, two-pore K + (K2P) channels, K ATP channels, and GIRK channels had significant levels of mRNA expression [15].Thus, we obtained acute hypothalamic slices from the Npy-hrGFP mice and targeted the fluorescence-labeled NPY neurons within the ARH for whole-cell patch clamp recordings (Fig 1A ), where we tested the effects of pharmacological inhibitors of abovementioned K + channels.
GIRK channels were demonstrated to maintain RMP of several types of central neurons [13,16,17].We tested the involvement of GIRK channels and found that bath applications of tertiapin-Q (300 nM), a GIRK channel blocker, depolarized membrane potential in 5 of 11 (approximately 45%) NPY neurons (from −56.3 ± 3.6 mV to −50.6 ± 2.8 mV, n = 5, Fig 1B and  1E, red lines).We applied small hyperpolarizing current steps before and after tertiapin-Q treatments (Fig 1B and 1C) and plotted the amplitudes of voltage responses against the amplitudes of injected currents to obtain a voltage-current (V-I) relationship (Fig 1D).We noted that the depolarizing effects were accompanied by increased input resistance (from 2.49 ± 0.47 GΩ to 2.84 ± 0.50 GΩ, n = 5, Fig 1F, red lines) with a reversal potential (E rev ) of −101.3 ± 14.6 mV (n = 5) (Fig 1D).Changes of membrane potential and input resistance by tertiapin-Q were significant when we included all neurons recorded (Fig 1E and 1F).We also tested lower (100 nM) and higher (500 nM) concentrations of tertiapin-Q and found that the depolarizing effects become more significant at higher concentrations (Fig 1G and 1H).In addition, the response rate increased at higher concentrations (Fig 1I).These results suggested the contribution of GIRK channels to the maintenance of RMP in NPY neurons.
Notably, NPY neurons depolarized by tertiapin-Q had significantly lower action potential (AP) firing frequency and hyperpolarized RMP compared to those not responding to tertiapin-Q (S1A and S1B Fig).These data suggest that NPY neurons depolarized by tertiapin-Q have active GIRK channels and therefore are more stable.Consistent with this idea, NPY neurons depolarized by tertiapin-Q had lower input resistance than nonresponsive neurons, although the difference was not significant (S1C Fig) .We also noted lower AP threshold in neurons depolarized by tertiapin-Q (S1D Fig), which may result from higher availability of voltage-gated Na + channels due to more negative RMP.
Subsequently, we tested the effects of M channel blockers (10 μM linopirdine and 10 μM XE991) and observed depolarizing responses (3 mV and 5 mV) in 2 of 12 cells (approximately 17%) tested (Figs 1J, S2A, and S2E).These effects were accompanied by increased input resistance (from 2.98 GΩ to 3.80 GΩ and from 3.56 GΩ to 4.84 GΩ) and E rev of −94.0 mV and −81.0 mV, which suggested the contribution of M channels in a small subpopulation of NPY neurons.We also tested the effects of PK-THPP (1 μM, a TASK-3 channel blocker), spadin (1 μM, a TREK-1 channel blocker), and tolbutamide (100 μM, a K ATP channel blocker), but none of these blockers caused significant changes in NPY neuronal membrane potential (Figs 1J, S2B-S2D, and S2F-S2H).Since we included 2 mM of ATP in pipette solutions (see Materials and methods), which may inhibit K ATP channels [18], we also tested the effects of tolbutamide using ATP-free pipette solutions but found that RMP still remains unchanged (from −41.1 ± 1.1 mV to −40.9 ± 1.1 mV, p = 0.623, n = 10).Thus, it appears that neither K2P channel nor K ATP channel plays a measurable role to maintain RMP of NPY neurons.

Arcuate AgRP neurons preferentially express Girk2 over Girk1
Neuronal GIRK channels contain one or both of GIRK1 and GIRK2 subunits [19], and both Girk1 and Girk2 mRNAs were found in the transcriptome of AgRP neurons [15].Therefore, we characterized the expression of Girk1 and Girk2 by arcuate AgRP neurons with fluorescence in situ hybridization (FISH) experiments (RNAscope) targeting Agrp, Girk1, and Girk2 mRNA in wild-type mice.As shown in Fig 2A and 2B, Agrp-expressing neurons (white) expressed both Girk1 (green) and Girk2 (magenta) at mRNA levels within the ARH.

GIRK2-containing GIRK channels are dispensable for GABA B -activated K + currents in NPY neurons
GIRK channels are known to mediate slow synaptic inhibition by the stimulation of GABA B receptors [20].Thus, we performed voltage clamp experiments to determine whether GIRK channels contribute to GABA B -activated currents in NPY neurons.We applied baclofen, a GABA B receptor agonist, to NPY neurons from the Npy-hrGFP transgenic mice using a local perfusion system (see Materials and methods) to record GABA B -activated GIRK currents.The I-V relationship of I Bac showed inward rectification with E rev close to E K (−88.5 ± 0.7 mV, n = 12), consistent with GIRK channel activation.We also calculated the rectification index (I - 120 mV /I -60 mV ), the ratio of absolute values of currents at −120 mV (I -120 mV ) and −60 mV (I -60 mV ) of I-V curve.The average rectification index was 2.5 ± 0.2 (n = 12, S4C Fig).
We also examined whether GIRK2-containing GIRK channels have a role in GABA Binduced hyperpolarization of NPY neuronal membrane potential.We noted that treatments of NPY G2WT neurons with CGP54626 (2 μM), a GABA B receptor antagonist, do not affect RMP and input resistance of NPY G2WT neurons (S6 Fig), which suggested that GABA B receptors are not active at rest to affect the membrane potential of NPY G2WT neurons.Subsequently, we found that application of 10 μM baclofen hyperpolarized NPY G2WT neurons by −14.1 ± 1.9 mV (n = 12 of 14 cells) (Fig 3D and Table 1).The hyperpolarizing effects were accompanied by decreased input resistance and E rev of −104.3 ± 5.3 mV (n = 12) based on the V-I relationship calculated from voltage responses to current steps pulses before and after baclofen perfusion (Fig 3E and 3F).We also tried lower (1 μM) and higher (30 μM and 100 μM) concentrations of baclofen and noted dose-dependent effects (Fig 3H and Table 1), where E rev was comparable across all concentrations tested.We conducted the same series of experiments with NPY G2KO neurons and found that NPY G2KO neurons showed significantly augmented hyperpolarization (−20.9 ± 2.4 mV, n = 8, p = 0.035) by 10 μM baclofen (Fig 3G and 3H and Table 1).The observed augmentation of baclofen-induced hyperpolarization is likely due to increased input resistance together with unchanged GABA B -activated GIRK currents in NPY G2KO neurons (S4D-S4G Fig).

GIRK2 ablation, but not GIRK1 ablation, results in a persistent increase of AgRP neuronal activity
Given the higher expression of Girk2 mRNA than Girk1 mRNA (Fig 2 ) as well as the contribution of GIRK2 subunits to the RMP (Fig 3), we assumed that the GIRK2-containing GIRK channels may play a more important role than GIRK1-containing GIRK channels to maintain AgRP neuronal activity.To test this idea, we labeled AgRP neurons with tdTomato reporter using Agrp-ires-Cre::tdTomato (Agrp tdTomato ) mice and performed immunohistochemistry (IHC) experiments to measure Fos expression level in arcuate AgRP neurons.We found that 56.0 ± 3.2% (n = 6) of AgRP neurons express Fos when the mice were fasted overnight for 18 h (Fig 4A and 4B).

Deletion of GIRK2 subunits in AgRP neurons increases adiposity and body weight independently of food intake
In order to delineate the metabolic function of GIRK2 subunits expressed by AgRP neurons, we measured body weight and food intake of GIRK2 AgRP-KO and GIRK2 WT mice once a week and found that GIRK2 AgRP-KO mice gained more body weight than GIRK2 WT    increased size of adipocytes within the inguinal white (IGW) and perigonadal white (PGW) fat tissues of GIRK2 AgRP-KO mice (Fig 5E).We noted that differences in food consumption do not explain the increased adiposity, since cumulative food intake was not different between GIRK2 WT mice and GIRK2 AgRP-KO mice (Fig 5F).We also found that food intake was not influenced by GIRK2 deletion when the mice (21-to 22-week-old) were refed after overnight fasting (Fig 5G ).

GIRK2-containing GIRK channels expressed by AgRP neurons are required for normal sympathetic activity and BAT function
Given no changes in food intake, we hypothesized that the body weight gain observed in GIR-K2 AgRP-KO would be caused by decreased energy expenditure.To test this idea, we measured oxygen consumption (VO 2 ) and carbon dioxide production (VCO 2 ) with an indirect calorimetry from 20-week-old GIRK2 WT and GIRK2 AgRP-KO mice.We observed significantly decreased VO 2 and VCO 2 in GIRK2 AgRP-KO mice compared to GIRK2 WT  AgRP neurons were shown to regulate anxiety level [24], but our OFT results demonstrated similar levels of anxiety regardless of genotypes, based on their comparable preference to the center zone and the outer zone in the chamber (S9E-S9G Fig).
Thus, the decreases in EE observed in GIRK2 AgRP-KO mice are likely due to reduced basal metabolic rate.Decreased BAT thermogenesis is often a major cause of reduced basal metabolic rate and energy expenditure [25].Indeed, we noted increased adiposity and triacylglycerol level in the BAT from the GIRK2 AgRP-KO mice by HE and oil red O staining (Fig 6B , top and middle).In addition, uncoupling protein-1 (UCP-1) immunoreactivity was markedly decreased in the BAT of GIRK2 AgRP-KO mice (Fig 6B , bottom).Since BAT thermogenesis is regulated by sympathetic tone [26] and NPY/AgRP neurons are known to decrease sympathetic activity [5,27,28], we predicted that increased activity of NPY/AgRP neurons would result in decreased sympathetic activity of GIRK2 AgRP-KO mice.To test this idea, we performed IHC experiments and measured Fos levels in the cholinergic sympathetic preganglionic neurons of the intermediolateral column (IML) of T1 to T6 spinal cords.We found in GIRK2 AgRP-KO mice a significantly lower percentage (32.4± 2.6%, n = 4, p = 0.005) of choline acetyltransferase (ChAT)positive IML neurons expressing Fos compared to observations in the GIRK2 WT mice (52.4 ± 3.9%, n = 6) (Fig 6C -6E) at 8 to 12 weeks of age.Together, these results suggest that decreased sympathetic activity and BAT thermogenesis lead to decreased energy expenditure and body weight gain in GIRK2 AgRP-KO mice.

GIRK2 expressed by AgRP neurons is necessary for prompt adaptation to a cold temperature
Our results suggested that GIRK2 subunits expressed by AgRP neurons contribute to maintain body weight by promoting EE in non-stress conditions.To explore if GIRK2 subunits also have a role in stress conditions, we intraperitoneally (i.p.) injected 10-week-old GIRK2 WT and GIRK2 AgRP-KO mice with ghrelin (0.4 mg/kg) and measured food intake for 4 h after injections.We expected that ghrelin produces hunger-induced stress, but found that ghrelininduced increase of food intake was similar between GIRK2 WT and GIRK2 AgRP-KO mice (S10 Fig) .GIRK2 WT and GIRK2 AgRP-KO mice used for this experiment weighed 27.7 ± 1.0 g (n = 4) and 28.1 ± 1.2 g (n = 4), respectively (p > 0.5 by unpaired t test).
In a different set of experiments, we exposed 10-week-old GIRK2 WT and GIRK2 AgRP-KO mice to a cold environment (5˚C) to challenge the mice with cold stress.There was no significant difference in body weight or body composition between the genotypes (Fig 7A -7D).When the temperature dropped from 25˚C to 5˚C, GIRK2 WT mice showed a prompt increase of VO 2 and VCO 2 , which reached a new steady state after approximately 4 h (Fig 7E and 7F).GIRK2 AgRP-KO mice also showed increase of VO 2 and VCO 2 in response to the cold exposure, but there was a significant delay in the rising phase of VO 2 and VCO 2 (Fig 7E and 7F).The calculated EE were also significantly different in the rising phase between the genotypes (Fig 7G).We noted no significant differences in ambulatory movement or rearing activity of GIRK2 WT and GIRK2 AgRP-KO mice (Fig 7H and 7I), suggesting that the increases of VO 2 and VCO 2 are likely from increased BAT thermogenesis.Taken together, we propose that GIRK2 expressed by AgRP neurons is dispensable for ghrelin-induced feeding, but is necessary for prompt adaptation to a cold environment.

Discussion
In this study, we found evidence that GIRK2 subunits are key to regulating the long-term baseline activity of arcuate NPY/AgRP neurons.In agreement, GIRK2 ablation in arcuate NPY/ AgRP neurons resulted in increased adiposity and body weight.This phenotype was associated with decreased sympathetic activity and reduced energy expenditure, but not changes in food intake.We also demonstrated that GIRK2 ablation in arcuate NPY/AgRP neurons delays the initial phase of cold-induced thermogenesis.Together, our findings identified GIRK2 as a regulator of arcuate NPY/AgRP neuron activity that maintains sympathetic activity and burns fat, which should help to maintain homeostasis in physiological (normal caloric or non-stressed) and some stressed conditions.

In vivo metabolic effects of AgRP neuron activity
Previous studies demonstrated that the optogenetic or chemogenetic activation of AgRP neurons results in increased food intake and/or decreased energy expenditure [6,7,29], which occurred within hours.The activation of AgRP neurons using the "capsaicin-Trpv1" system also resulted in rapid decreases of energy expenditure and thermogenesis [30].On the other hand, chemogenetic inhibition of AgRP neurons for 14 days led to decreased body weight, decreased food intake, and fat burning [31].It was also shown that chemogenetic inhibition of AgRP neurons reverses diabetes-induced hyperphagia and hyperglycemia within a few hours [32].Therefore, available data suggested that modulation of AgRP neuronal activity can significantly affect food intake and energy expenditure in time frames of hours to days.
Since acute activation of AgRP neurons resulted in rapid metabolic effects regardless of activating methods, we may expect that long-term activation of AgRP neurons would also produce similar phenotypes.A recent study overexpressed bacterial sodium channel (NachBac) or Kir2.1 channel selectively in AgRP neurons to achieve long-term activation and inhibition of neuronal activity, respectively [33].The authors reported that NachBac overexpression resulted in massive obesity accompanied by increased food intake but no changes of energy expenditure, but that Kir2.1 overexpression did not produce any phenotype.In this study, we genetically deleted GIRK2 subunits selectively in the AgRP neurons, which presumably increased AgRP neuronal activity for longer periods (approximately 5 months).While we noted significantly increased body weight in GIRK2 AgRP-KO mice, cumulative food intake was not different between genotypes.We also noted that fasting-and ghrelin-induced feeding were not different between genotypes.This finding was quite surprising given the prominent role of AgRP neurons in the regulation of food intake.Instead, O 2 consumption and CO 2 production were significantly decreased in GIRK2 AgRP-KO mice, which were associated with decreased activity of sympathetic preganglionic neurons and decreased UCP-1 expression by BAT.GIR-K2 AgRP-KO mice also showed a significant delay in cold-induced increase of energy expenditure compared to GIRK2 WT mice ( Fig 7), which further suggested that thermogenic response is compromised in GIRK2 AgRP-KO mice.
It is not clear why persistently increased activity of AgRP neurons decreased energy expenditure but did not regulate food intake in our study.One hypothesis is that GIRK2-expressing AgRP neurons preferentially regulate energy expenditure, like leptin receptor-expressing POMC neurons [34].In this scenario, food intake and energy expenditure are regulated by distinct subpopulations of AgRP neurons, as previously suggested for POMC neurons [35].An alternative possibility is that GIRK2-expressing AgRP neurons also regulate food intake, but this effect is discernable only in short-term time frames.In other words, AgRP neurons can regulate both food intake and energy expenditure, but a compensatory anorexia (due to decreased energy expenditure) may develop over time to mask increased food intake.In either case, the inhibition of energy expenditure by GIRK2-ablated AgRP neurons looks large enough.It is also important to note that we deleted GIRK2 subunits before birth.As GIRK channels can be assembled in 5 different compositions [36,37], other GIRK channel subunits may take the role of GIRK2 subunits in AgRP neurons in GIRK2 AgRP-KO mice.Therefore, we may need a strategy to delete GIRK2 subunits postnatally to delineate the actual function of GIRK2 subunits expressed by AgRP neurons.Indeed, a previous study suggested that a compensatory mechanism may develop before birth to overcome the severe anorexia observed in AgRP-ablated mice [38].We suggest that future studies are directed to delineate the metabolic functions designated to individual AgRP neurons, which will help to understand variable phenotypes obtained from AgRP neuron-specific conditional knockout mouse models.

Metabolic function of K + channels expressed by NPY/AgRP neurons
A few previous studies reported the role of K + channels expressed by the AgRP neurons in the regulation of energy balance.For example, a study demonstrated that down-regulation of the small conductance Ca 2+ -activated K + (SK) channels contribute to fasting-induced activation of AgRP neurons [39].It was noted that SK3 channel deletion resulted in increased firing rate of AgRP neurons without changes in RMP, which makes sense given the role of SK channel in the regulation of afterhyperpolarization.AgRP neuron-specific deletions of SK3 channels resulted in a transient obesity in NCD-fed mice, but profoundly exacerbated HFD-induced obesity.The increased susceptibility to HFD was associated with increased food intake and decreased EE, while the locomotive activity remained unchanged.More recently, it was shown that CRISPR knockdown of Kcnq3, an M channel subunit, in NPY/AgRP neurons did not affect RMP but increased input resistance and decreased rheobase current, which suggested increased response to external stimuli [40].However, it was noted that in NCD conditions KCNQ3 deficiency resulted in no changes of food intake and body weight, while locomotive activity was decreased.Similar results were obtained in HFD-fed mice except that there was an increase of abdominal fat mass.Together, it is suggested that M channels expressed by NPY/ AgRP neurons are largely dispensable for the control of energy balance whether the mice are given NCD or HFD.
In this study, we found a predominant role for GIRK channels to maintain RMP of NPY/ AgRP neurons (Fig 1).Notably, knockout of GIRK2 resulted in a depolarized RMP (approximately 3.5 mV), amplitude of which was comparable to the depolarizing effects of the GIRK channel blocker (approximately 4 mV).We also noted a few cells depolarized by M channel blockers, but a majority of NPY/AgRP neurons did not respond (S2 Fig), which is consistent with the limited in vivo function of KCNQ3 expressed by NPY/AgRP neurons [40].While SK3 and KCNQ3 expressed by the AgRP neurons were largely dispensable in NCD-fed mice [39,40], GIRK2 ablation in AgRP neurons led to significantly increased body weight in NCDfed mice (Fig 5).Together, it appears that GIRK2-containing GIRK channels have a unique role in the physiological NCD-fed conditions to regulate NPY/AgRP neuronal excitability and energy balance.

Functional GIRK channel subunit composition in NPY/AgRP neurons
GIRK1/GIRK2 heterotetramers are the prototypes of neuronal GIRK channels, and loss of either GIRK1 or GIRK2 is usually sufficient to eliminate most or all of GIRK channel activity in central neurons [36,37].A notable exception is the GIRK channel of midbrain dopaminergic neurons.They lack GIRK1 subunits, and GIRK2 homotetramers and/or GIRK2/GIRK3 heterotetramers are believed to be the subunit composition of GIRK channels in these neurons [36,37].In this study, we found that GIRK2 is the major GIRK channel subunit expressed by arcuate AgRP neurons (Figs 2 and S3).Analyses of our FISH data suggest that GIRK2 homotetramer (and GIRK2/GIRK3 heterotetramer) may constitute a majority of functional GIRK channels in AgRP neurons (S3 Fig) .Therefore, GIRK channel subunit composition of AgRP neurons appear to be similar to that of midbrain dopaminergic neurons.
We unexpectedly found that ablation of GIRK2 did not affect GABA B -activated K + currents in NPY/AgRP neurons (S4 Fig) .On the other hand, GIRK2-containing GIRK channels contributed to long-term control of NPY/AgRP neuronal excitability and in vivo energy balance (Figs 4-6).Together, these data may suggest that GIRK2-containing GIRK channels are open at rest to regulate in vivo metabolic function but are not functionally coupled to GABA B receptor stimulation.In this case, GIRK1/GIRK3 and/or GIRK1/GIRK4 heterotetramers may be responsible for GABA B -activated K + currents.However, we also need to consider the possibility that the loss of GIRK2 subunits and the contribution to GABA B -activated K + currents was replaced by other GIRK subunits through a compensatory mechanism in our model.This is especially so given that almost all NPY/AgRP neurons generate outward currents by baclofen (S4 Fig) but Girk1/Girk3 (7.7 ± 0.7%, n = 3) and Girk1/Girk4 (8.1 ± 0.2%, n = 3) co-expression levels in NPY/AgRP neurons are quite low (S3B and S3C Fig) .It remains to be tested whether GIRK2-containing GIRK channels are coupled to other G i/o protein-coupled receptors.Overall, more investigations are necessary to delineate the subunit composition of functional GIRK channels of NPY/AgRP neurons.

Ethics statement
All experiments were performed in accordance with the guidelines established by the Korea Advanced Institute of Science and Technology (KAIST) Institutional Animal Care and Use Committee (IACUC) (Protocol No. KA2021-126).KAIST IACUC follows the standard operating guidelines for IACUC established by the Animal and Plant Quarantine Agency and the Ministry of Food and Drug Safety of South Korea.

Mice
All mice used for breeding and experiments in this study were housed in a light-dark (12 h on/ off; lights on at 7:00 AM) and temperature-controlled environment with food and water available ad libitum in the KAIST facilities.Npy-hrGFP mice were obtained from the Jackson laboratory (#006417).For some patch clamp experiments GIRK2 KO mice [41], used with the permission from Dr. Markus Stoffel (ETH Zurich), were crossed with Npy-hrGFP mice.Agrpires-Cre mice (Jackson laboratory, #012899) were crossed with tdTomato reporter mice (Jackson laboratory, #007914), GIRK1 flox/flox mice [22] or GIRK2 flox/flox mice [23] for FISH, ISH, IHC, and in vivo metabolic experiments.Mice were fed standard NCD (Teklad global 18% protein 2018S, ENVIGO).

Electrophysiology
Five-to 13-week-old male Npy-hrGFP mice were used for all patch clamp experiments in order to identify NPY-expressing neurons in the ARH.Npy-hrGFP mice were fasted for 18 h before being killed for experiments.Whole-cell patch clamp recordings from hrGFP-expressing neurons were maintained in acute hypothalamic slice preparations as previously described [42].In brief, mice were deeply anesthetized with isoflurane inhalation and transcardially perfused with a modified ice-cold artificial CSF (ACSF) (described below), in which an equiosmolar amount of sucrose was substituted for NaCl.The mice were then decapitated, and the entire brain was removed from the skull and immediately submerged in ice-cold, carbogensaturated (95% O 2 and 5% CO 2 ) ACSF (123 mM NaCl, 26 mM NaHCO 3 , 2.8 mM KCl, 1.25 mM NaH 2 PO 4 , 1.2 mM MgSO 4 , 2.5 mM CaCl 2 , and 10 mM glucose).A brain block containing the hypothalamus was made.Coronal sections (250 μm) were cut with a Leica VT1200S vibrating microtome and then incubated in oxygenated ACSF at 34˚C for at least 1 h before recording.Brain slices were transferred to the recording chamber and allowed to equilibrate for 10 to 20 min before recording.The slices were bathed in oxygenated ACSF (32˚C to 34˚C) at a flow rate of approximately 2 ml/min.The pipette solution was modified to include an intracellular dye (Alexa Fluor 594) for whole-cell patch clamp recording: 120 mM K-gluconate, 10 mM KCl, 10 mM HEPES, 1 mM CaCl 2 , 1 mM MgCl 2 , 5 mM EGTA, 2 mM Mg-ATP, and 0.03 mM Alexa Fluor 594 hydrazide dye (pH 7.3).Epifluorescence was briefly used to target fluorescent cells, at which time the light source was switched to infrared differential interference contrast imaging to obtain the whole-cell recording (Nikon Eclipse FN1 equipped with a fixed stage and an optiMOS scientific CMOS camera).Recording electrodes had resistances of 3 to 5 MΩ when filled with the K-gluconate internal solution.
In current clamp experiments, input resistance was assessed by measuring amplitudes of voltage deflections in response to hyperpolarizing rectangular current step pulses (500 ms, −25 pA to 0 pA by 5 pA increments or −50 pA to 0 pA by 10 pA increments) which was applied at a stable membrane potential before and after drug application.AP threshold was determined from averaged AP traces of firing neurons.The voltage at the last minimum of dV/dt preceding the spike (within 2 ms preceding 10 V/s) was estimated to be AP threshold, as described previously [43].A drug effect was required to be associated temporally with drug application, and the responses had to be stable within a few minutes.We determined membrane potential before (control) and during drug applications (drug) by averaging membrane potential for 10 s in each condition.A neuron was considered to be depolarized or hyperpolarized if a change in membrane potential was larger than 2 mV in amplitude.Membrane potentials were not compensated for liquid junction potentials (−8 mV).
For voltage clamp experiments, we used the same K-gluconate pipette solutions described above and added 0.5 μM tetrodotoxin (TTX) and synaptic blockers (50 μM picrotoxin and 1 mM kynurenic acid) to bath solutions.We held the membrane potential at −40 mV and locally applied baclofen using micropipettes attached to the Picospritzer III microinjection dispense system (Parker Hannifin).We placed micropipettes 10 to 20 μm away from soma and ejected small volume (15 to 20 pL) of ACSF containing baclofen and the blocker cocktail with a pressure of 16 to 18 psi for 15 s.Baclofen-activated currents (I Bac ) were normalized by cell capacitance.Voltage ramp pulses (from −120 mV to −10 mV, 100 mV/s) were applied before and after baclofen applications from a holding potential of −40 mV to obtain I-V relationships of I Bac .

In situ hybridization
In situ hybridization experiments were performed using RNAscope or BaseScope assays available from the Advanced Cell Diagnostics Inc. (Hayward, California, United States of America).Briefly, 8-to 12-week-old male mice were deeply anesthetized with isoflurane and transcardially perfused with a DEPC-treated PBS and subsequently with a 4% paraformaldehyde (163-20145, FUJIFILM Wako).Brains were removed from the skull and submerged in cold 4% paraformaldehyde solutions for 24 h (post-fixation).Brains were then transferred to a series of sucrose solutions of gradients (4 h in 10% sucrose, 12 h in 20% sucrose, and 24 h in 30% sucrose) at 4˚C.Brain slices with 10 μm thickness were obtained using a cryostat (Leica) and were stored in a cryo-protectant.We collected coronal sections that contain arcuate nucleus based on the shape of third ventricle, median eminence, and hippocampus referring to the Allen Brain Atlas.Brain slices were transferred to a well plate (20 wells, 4 rows × 5 columns) immediately after sectioning one by one in a rostrocaudal order.We used rows from left to right and columns from up to down.After collecting the first 20 slices, we repeated the same procedure 5 to 7 times so that one well contains 5 to 7 brain slices each of which is spaced by 200 μm.We used 1 column for 1 set of experiments.Brain slices were mounted on glass slides (Superfrost Plus Microscope Slides, Thermo Fisher) for the RNAscope or the BaseScope assays.All reagents for these assays were purchased from Advanced Cell Diagnostics.

RNA extraction and qRT-PCR
Brain blocks were prepared and submerged in ice-cold ACSF.Coronal brain slices (500 μm-1 mm thickness) were obtained using Leica VT1200S vibrating microtome.The brain slices were transferred to DEPC-based PBS on ice, and hippocampal regions containing dentate gyrus and CA1 were punched out using a blunt-end 16 gauge needle (#28110, STEMCELL Technologies) under a stereomicroscope.Hippocampal tissues were transferred to E-tubes on dry ice and preserved at −80˚C.RNA was extracted from the hippocampal tissues by using RNeasy Lipid Tissue Mini Kit (74804, Qiagen).

Immunohistochemistry
Fos activity in the hypothalamus and the spinal cord was detected using IHC experiments.Briefly, 8-to 12-week-old male mice were deeply anesthetized with isoflurane and transcardially perfused with a PBS and subsequently with a 4% paraformaldehyde (163-20145, FUJIFILM Wako).
Brains were obtained from the Agrp tdTomato , Agrp tdTomato /Girk1 KO , and Agrp tdTomato / Girk2 KO mice and submerged in cold 4% paraformaldehyde for 24 h (post-fixation).Brains were then transferred to a series of sucrose solutions (4 h in 10% sucrose, 12 h in 20% sucrose, and 24 h in 30% sucrose) at 4˚C.Brain slices with 15 μm thickness were obtained using a cryostat (Leica) and were prepared for incubation with primary antibodies.Collection of brain slices were performed as described in the methods for in situ hybridization experiments, except that brain slices in 1 well are spaced by 300 μm.Brain slices were mounted on glass slides (Superfrost Plus Microscope Slides, Thermo Fisher) and were treated with PBS containing 10% BSA and 0.3% Triton X-100 (blocking solution).The slices were subsequently incubated with anti-Fos (1:2,000, ab190289, Abcam) antibodies overnight at 4˚C.Brain slices were then washed 3 times with PBS (10 min each) and were incubated with Alexa Fluor 647 donkey anti-rabbit secondary antibodies (Thermo Fisher Scientific) for 1 h at room temperature (RT) for immunofluorescence detection.The slices were incubated with DAPI for 10 min, washed 3 times with PBS (10 min each), and cover-slipped using fluorescence mounting medium (DAKO).Images were obtained with a confocal microscope (LSM 780, Carl Zeiss) and were analyzed with ZEN lite (ZEN Microscopy software) and Image J software.
Thoracic spinal cords obtained from GIRK2 WT and GIRK2 AgRP-KO mice were submerged in cold 4% paraformaldehyde for 12 h (post-fixation).Spinal cords were then transferred to 30% sucrose solutions for 24 h at 4˚C.Spinal cord slices with 40 μm thickness were obtained using a cryostat (Leica).We collected coronal sections beginning from the cervical enlargements to preserve T1 level.Spinal cord slices were transferred to 24-well plates, placing 5 consecutive slices in 1 well.Since thoracic spinal cord (T1-L1) is about 18.2 mm in length [46], we needed 4 plates for 1 mouse spinal cord.We randomly selected 1 slice from 1 well, which allows average spacing of 200 μm between slices.Selected slices were washed 3 times with PBS (10 min each) to be prepared for incubation with primary antibodies.The slices were mounted on adhesive microscope slides (TruBond 380, Electron Microscopy Science) and were washed 3 times with PBS (10 min each).The slices then underwent heat-induced epitope retrieval at 60˚C for 30 min and were treated with PBS containing 5% normal donkey serum (NDS) and 0.3% Triton X-100 (blocking solution).Subsequently, the slices were treated with anti-ChAT (1:100, AB144P, Sigma) antibody diluted with the blocking solution overnight at 4˚C, which was followed by 1 h treatment with Alexa Fluor 488 donkey anti-goat secondary antibodies (Thermo Fisher Scientific) at RT.After that, the slices were treated with the blocking solution for 1 h, and then with anti-Fos (1:500, ab190289, Abcam) antibodies overnight at 4˚C.The slices were then treated with Alexa Fluor 647 donkey anti-rabbit secondary antibodies (Thermo Fisher Scientific) for 1 h at RT.The slices were incubated with DAPI for 10 min and were washed 3 times with PBS (10 min each).Then, the slices were cover-slipped using fluorescence mounting medium (DAKO).Images were obtained with a confocal microscope LSM 780 (Carl Zeiss) and were analyzed with ZEN lite (ZEN Microscopy software) and Image J software.We examined the images to look for red fluorescence-expressing neurons in the IML since the first slice with positive fluorescence is T1.We proceeded starting from that slice to determine the level of each spinal cord section.We selected 48 spinal cord slices (approximately 9.6 mm in length, T1-T6, from each mouse) to be included for analyses.
UCP1 in the BAT was detected using IHC experiments.Briefly, 22-to 23-week-old male mice were deeply anesthetized with isoflurane.BAT was obtained from GIRK2 WT and GIR-K2 AgRP-KO mice and was immediately submerged into formalin (HT501128, Sigma) for at least 24 h.Sections of BAT were treated with recombinant anti-UCP1 antibody (1:1,000, ab234430, Abcam), which was followed by incubation with horseradish peroxidase (HRP) secondary antibodies (Envision kit HRP, DAKO).Images were obtained with a slide scanner (Axio Scan.Z1, Carl Zeiss).

Body weight and food intake
Body weight and food intake were measured from male mice as indicated in results.Each conditional knockout mouse (GIRK2 AgRP-KO ) had its littermate control mouse (GIRK2 WT ), and 21-to 22-week-old male mice were fasted for 18 h for fast-refeeding experiments.Refeeding started at 10 AM.For some experiments, 10-week-old male mice were i.p. injected with saline or ghrelin (0.4 mg/kg) in fed state (10 AM).

Energy expenditure, physical activity, and body composition
Energy expenditure and physical activity were measured from 20-week-old male mice by an indirect calorimetric chamber (CLAMS 12; Columbus Instruments, Columbus, Ohio, USA).After acclimation for 2 days, O 2 consumption (VO 2 ) and CO 2 production (VCO 2 ) were measured for 2 days to determine the energy expenditure.Simultaneously, physical activity was determined using a multidimensional infrared light beam system with beams installed on bottom and top levels of cage.Ambulatory movement was defined as breaks of any 2 different light beams at bottom level of cage, while rearing was recorded once the mouse broke any light beam at the top level.Body composition was measured by Time Domain (TD) NMR spectrometer (Minispec LF50, Bruker biospin, Rheinstetten, Germany).For some experiments, we exposed mice (10 weeks old) to cold environment by changing the temperature to 5˚C from 25˚C, which occurred at 10:30 AM.The mice were allowed to acclimate for 2 to 3 days at 25˚C before the transition.

Open field test
Single-housed male mice (21 to 22 weeks old) with access to water and food were adapted for 30 min to a custom-made chamber (40 cm × 40 cm × 40 cm) in a ventilated soundproof booth.A camera was installed on the ceiling of the soundproof booths, and mice were allowed to move freely within the chamber for another 30 min.Light intensity was 120 to 140 lux.Light cycle experiment (3:00 to 4:00 PM) and dark cycle experiment (8:30 to 9:30 PM) were performed 4 days apart.A square-shaped area (20 cm × 20 cm) in the center was defined as the center zone, and the remaining area was defined as the outer zone.Locomotion was analyzed by EthoVision XT 15 (Noldus Wageningen, the Netherlands).

Tissue staining
After the end of metabolic and behavioral experiments, all mice (22 to 23 weeks old) were deeply anesthetized with isoflurane and killed to harvest tissues.Liver, brown fat, epididymal fat, and inguinal fat tissues were isolated and fixed in neutralized formaldehyde solution (HT501128, Sigma).Paraffin-embedded tissue sections were stained with HE, or oil red O. Images were obtained with a slide scanner (Axio Scan.Z1, Carl Zeiss).

Fig 5 .
Each tab includes data for individual panels of Fig 5. (XLSX) S6 Data.Original data for the graphs in Figs 6 and S9.Each tab includes data for individual panels of Figs 6 and S9.(XLSX) S7 Data.Original data for the graphs in Figs 7 and S10.Each tab includes data for individual panels of Figs 7 and S10.(XLSX) S8 Data.Original data for

Table 1 . Summary of GABA B -induced hyperpolarization of arcuate NPY neurons.
Changes of membrane potential are presented as mean ± SEM.Numbers in parentheses indicate the number of responsive cells, response rate, and p values (unpaired t test, NPY G2WT neurons vs. NPY G2KO neurons).E rev = reversal potential.The individual numerical data for changes of membrane potential and reversal potential can be https://doi.org/10.1371/journal.pbio.3002252.t001

Table 1 .
Numbers represent individual numerical data for changes of membrane potential and reversal potential in Table 1.(XLSX)