Bridging Energy Need and Feeding Behavior: The Impact of eIF2α Phosphorylation in AgRP Neurons

The aberrant accumulation of misfolded or unfolded proteins in the endoplasmic reticulum (ER) causes ER stress, which leads to the activation of multiple signaling pathways in the unfolded protein response (UPR) and ER-associated protein degradation to restore the ER’s protein-folding capacity (1,2). When the UPR detects an insufficiency in the ER protein-folding capacity, UPR activation is triggered to increase it (3,4). During the past decade, much work has been done to clarify the relationships between ER stress–induced UPR activation in multiple peripheral tissues and the hypothalamus and the development of metabolic disorders (5–7).

Eukaryotic translation initiation factor 2α (eIF2α) is a key mediator of ER stress–induced UPR activation and helps alleviate the burden on the ER by suppressing protein translation (8,9). Normally, eIF2α is a subunit of the eIF2 complex (α, β, and γ) that plays a critical role in the control of translation initiation in diverse cellular environments. During translation initiation, eIF2 recruits the initiator methionyl-tRNA and guanosine triphosphate to form a ternary complex that binds to the ribosome. In mammals, eIF2α is phosphorylated on serine 51 in response to various cellular stresses, including ER stress, amino acid deficiency, and oxidative stress (9,10). Although the general function of phosphorylated eIF2α (p-eIF2α) is attenuation of translation initiation to restore cellular homeostasis, p-eIF2α selectively facilitates the translation of specific mRNAs, such as activating transcription factor 4 (ATF4), which is a transcriptional regulator involved in protein synthesis for cellular homeostasis, including autophagy and apoptosis (9–11). It is of particular interest that a deficiency of p-eIF2α in mice increased cellular stresses, such as ER stress and oxidative stress, that cause dysfunctions in cellular metabolism and are associated with metabolic disorders (12–14).

ER stress–induced eIF2α phosphorylation in the hypothalamus is closely related to the control of whole-body energy metabolism in pathological and pharmacological environments (5–7). For example, diet-induced obesity leads to increases in ER stress and ER stress–induced p-eIF2α in the hypothalamus, where p-eIF2α participates in the posttranslational modification of proopiomelanocortin (POMC). During overnutrition, increased ER stress and p-eIF2α in the hypothalamus cause a decrease in prohormone convertase 2, an enzyme that acts in the POMC processing cascade, resulting in a decrease in the anorexigenic α-melanocyte–stimulating hormone (α-MSH). Therefore, ER stress–induced eIF2α phosphorylation is regarded as a pathological cellular event in POMC neurons because it disturbs metabolic regulation during overnutrition (15,16). Although it has been highlighted that hypothalamic eIF2α phosphorylation is closely correlated with obesity pathogenesis and linked to ER stress, the physiological roles of hypothalamic p-eIF2α are largely unknown.

Short-term starvation primarily stimulates the activity of agouti-related peptide (AgRP) neurons in the hypothalamic arcuate nucleus (ARC) (17,18). Selective activation of AgRP neurons in the ARC strongly promotes feeding behavior by counteracting POMC neurons, and thus the dysfunction of AgRP neurons leads to abnormalities in whole-body energy metabolism (19,20). The starvation-induced activation of AgRP neurons requires intracellular machinery, including ER function (7,21). In line with that notion, recent studies have reported that starvation upregulates the UPR specifically in AgRP neurons but not in POMC neurons (7,22), suggesting that the UPR in AgRP neurons is more prominent than the UPR in POMC neurons in promoting feeding behavior during starvation. However, the role of UPR activation in intracellular signaling events in AgRP neurons stimulated by starvation is not fully understood. Thus, it is worth identifying the underlying intracellular mechanisms that occur in AgRP neurons during short-term starvation.

eIF2α is a key molecule in UPR activation and an essential factor for protein translation, which is affected by changes in the phosphorylation state of eIF2α. In this study, we used transgenic mice with a specific impairment of eIF2α phosphorylation in AgRP neurons to investigate the role of p-eIF2α in hypothalamic AgRP neurons in energy-deficit conditions. We found that selective impairment of eIF2α phosphorylation in AgRP neurons affected starvation-induced AgRP neuronal activation, the expression of UPR and autophagy genes in AgRP neurons, and metabolic phenotypes.

Mice with a homozygous eIF2α Ser51/Ala mutation, in which the Ser51 is changed to Ala, have impaired eIF2α phosphorylation, as previously described (12,13,23). Mice were intraperitoneally (ip) or intracerebroventricularly (icv) injected with materials. Food intake was measured for 1 h, 2 h, and 24 h after injection. Details are given in Supplementary Materials.

To analyze AgRP and POMC neuron–specific gene expression, we used Ribo-Tag assays with Rpl22^HA mice, as previously described (24,25). Mice brains were fixed and sliced to 50-μm thicknesses with a vibratome. Coronal brain sections were stained with antibodies. Protein samples were separated by 10% SDS-PAGE and were incubated with antibodies. Metabolic phenotypes of O₂ consumption, CO₂ production, and indirect calorimetry of energy expenditure were measured by metabolic chambers (26). Real-time PCR was performed and mRNA expression was calculated using the comparative cycle threshold method (27). Details of these steps are given in Supplementary Materials.

Statistical analyses were performed in GraphPad Prism 9 software (GraphPad Software, San Diego, CA). All data are expressed as the mean ± SEM. The statistical significance between two groups was analyzed by unpaired Student’s t test. Two-way ANOVA followed by Tukey’s multiple comparison test was used for unequal replications. ANCOVA was used to analyze the correlation between energy expenditure and body weight.

The data and resources generated or analyzed in this study are available from the corresponding author upon reasonable request.

We first determined effect of energy deficit on the p-eIF2α level in the murine hypothalamus. Hypothalamic p-eIF2α was increased by overnight fasting for 18 h (Fig. 1A and B), whereas fasting induced no change in p-eIF2α levels in other brain regions (Supplementary Fig. 1). To identify whether eIF2α phosphorylation in AgRP neurons responds to energy deficit, we analyzed immunoreactive p-eIF2α in the hypothalamic AgRP neurons of fasted mice and found an increase compared with the level in fed mice (Fig. 1C and D). However, no difference in number of AgRP neurons was observed between fed and fasted mice (Fig. 1E), and fasting induced no significant change in immunoreactive p-eIF2α in POMC neurons (Supplementary Fig. 2), suggesting that energy deficit specifically promotes eIF2α phosphorylation in AgRP neurons.

We next examined effect of eIF2α signaling on feeding behavior using guanabenz, which increases p-eIF2α level by inhibiting the dephosphorylation of eIF2α (28). Intracerebroventricular administration of guanabenz resulted in enhanced immunoreactivity of p-eIF2α and c-fos in AgRP neurons (Fig. 1F–I). Additionally, it caused a significant increase in food intake for 1 h and 2 h after injection (Fig. 1J). However, no change in daily body weight was observed (Fig. 1K).

However, guanabenz is also known as an α₂ adrenergic receptor agonist (29). Therefore, we further confirmed effect of eIF2α phosphorylation on food intake using salubrinal, another inhibitor of eIF2α dephosphorylation (30). Food intake was increased for 1 h and 2 h after icv administration of salubrinal (Supplementary Fig. 3A). However, food intake and body weight for 24 h after salubrinal injection did not change (Supplementary Fig. 3B and C). Conversely, we found a decrease in fasting-induced food intake after an injection of integrated stress response inhibitor (Supplementary Fig. 3D), which blocks the phosphorylation of eIF2α by promoting eIF2β activity (31,32). Together, these results suggest that eIF2α phosphorylation plays an important role in AgRP neurons during starvation and, thus, affects feeding behavior.

Because UPR activation occurred in AgRP neurons during short-term starvation (7,22), we next determined alterations in AgRP neuron–specific gene expression after overnight fasting. We used a Ribo-Tag technique that can measure mRNAs in the translational process in a single type of cell (33).

The AgRP neuron–specific Ribo-Tag (AgRP-Cre;Rpl22^HA) mice, which expressed hemagglutinin A (HA)-tagged ribosomal protein Rpl22 in AgRP neurons, had specific expression of HA in hypothalamic AgRP neurons (Fig. 2A). Thus, in the AgRP-Cre;Rpl22^HA mice, Agrp mRNA was enriched in RNA samples immunoprecipitated with HA antibody, compared with input RNAs (Fig. 2B). We analyzed the AgRP neuron–specific expression of UPR genes in fasted AgRP-Cre;Rpl22^HA mice. Compared with the normally fed condition, expression of UPR genes (Atf4, C/EBP homologous protein [Chop], total x-box binding protein 1 [Xbp1t], spliced Xbp1 [Xbp1s], and glucose-regulated protein 94 [Grp94]) was significantly increased by overnight fasting (Fig. 2C). However, in POMC-Cre;Rpl22^HA mice, no significant change in UPR gene expression in POMC neuron–specific mRNAs was induced by fasting (Fig. 2D and E). These observations indicate that energy deficit specifically promotes the expression of UPR genes in AgRP neurons.

Previous reports described a mouse model with a homozygous eIF2α Ser51Ala (A/A) mutation that impaired eIF2α phosphorylation; the mice were rescued from their lethal phenotype by a floxed wild-type transgene that expressed eIF2α flanked by LoxP sites, and those mice were designated as A/A;fTg/Tg mice (12,13). To investigate function of eIF2α phosphorylation in AgRP neurons, heterozygous S/A mice were crossbred with AgRP-Cre mice expressing Cre recombinase in AgRP neurons, which resulted in the generation of S/A;AgRP-Cre mice. Then, A/A;fTg/Tg mice were crossed with the S/A;AgRP-Cre mice to generate A/A;fTg/0;AgRP-Cre (AgRP^eIF2αA/A) mice bearing an AgRP neuron–specific deletion of the transgene (Supplementary Fig. 4A). Therefore, the AgRP^eIF2αA/A mice were specifically deficient in eIF2α phosphorylation in AgRP neurons.

Because the Cre recombinase deletes the transgene and thus coordinates expression of EGFP in AgRP neurons, the AgRP^eIF2αA/A mice displayed EGFP expression in the hypothalamic AgRP neurons, but the control A/A;fTg/0 mice did not (Supplementary Fig. 4B). To further confirm the specific deletion of the transgene in AgRP neurons, the AgRP^eIF2αA/A mice were crossed with Ai14 reporter mice, which labeled the AgRP neurons with tdTomato signals (AgRP^eIF2αA/A;Ai14 mice). EGFP expression was found in 96% of the AgRP neurons labeled with tdTomato signals (Supplementary Fig. 4C).

Energy deficit activates hypothalamic AgRP neurons and is accompanied by the regulation of energy homeostasis through control of feeding and energy expenditure (17,18). To determine whether eIF2α phosphorylation in AgRP neurons correlates with energy deficit–induced AgRP neuronal activation, we analyzed c-fos immunoreactivity in the AgRP neurons of the AgRP^eIF2αA/A;Ai14 mice after overnight fasting. Interestingly, the AgRP^eIF2αA/A;Ai14 mice had decreased c-fos immunoreactivity in AgRP neurons, compared with the control (AgRP-Cre;Ai14) mice, after fasting (Fig. 3A and B). However, no difference in number of AgRP neurons was observed between the control and AgRP^eIF2αA/A;Ai14 mice (Fig. 3C). Furthermore, AgRP^eIF2αA/A mice showed no changes in the quantity or intensity of glial fibrillary acidic protein–positive astrocytes and ionized calcium-binding adapter molecule 1–positive microglia (Supplementary Fig. 5).

AgRP is a key regulator of starvation-induced feeding behavior via inhibiting neurons expressing melanocortin (MC) receptors in the paraventricular nucleus (PVN) of the hypothalamus (34,35). Therefore, we measured immunoreactive AgRP fibers in the PVNs. The AgRP^eIF2αA/A mice displayed decreased intensity and number of AgRP fibers in the PVN, compared with those in the control A/A;fTg/0 mice, after fasting (Fig. 3D–F). The AgRP^eIF2αA/A mice also showed a decrease in fasting-induced food intake, compared with the control A/A;fTg/0 mice (Fig. 3G). Moreover, the AgRP^eIF2αA/A mice had decreased sensitivity to ghrelin (Supplementary Fig. 6), which promotes appetite through the activation of AgRP neurons during starvation (36,37), suggesting that p-eIF2α in AgRP neurons participates in starvation-induced AgRP neuronal activation and thus affects ghrelin-induced feeding.

To investigate how POMC neurons respond to altered AgRP neuronal activity in AgRP^eIF2αA/A mice during short-term calorie restriction, we assessed c-fos immunoreactivity in POMC neurons. The AgRP^eIF2αA/A mice exhibited increased c-fos immunoreactivity in POMC neurons compared with the control mice (Fig. 4A and B). We then examined the projection of POMC neurons in the mice PVNs after fasting by measuring the immunosignals of α-MSH fibers from POMC neurons. We observed an increase in the intensity and number of α-MSH fibers in the PVNs of AgRP^eIF2αA/A mice, compared with the control mice, after fasting (Fig. 4C–E). Because neurons expressing MC receptors in the PVN are activated by α-MSH (35), we measured c-fos activity in the PVNs of the mice after overnight fasting. In the AgRP^eIF2αA/A mice, fasting induced an increase in c-fos immunoreactivity in the PVN, compared with the control mice (Fig. 4F and G). The AgRP^eIF2αA/A mice also had increased responsiveness to the MC3/4 receptor agonist melanotan II (Supplementary Fig. 7), which causes a strong inhibition of food intake in animals (38), suggesting that the decreased release of AgRP, an endogenous antagonist to MC3/4 receptors, could cause increased responsiveness to the agonist in AgRP^eIF2αA/A mice.

In agreement with our finding of increased α-MSH fibers in the PVNs of AgRP^eIF2αA/A mice, we also found that those mice had enhanced responsiveness to leptin. Although icv administration of a low dose (1 μg) of leptin did not inhibit fasting-induced food intake in the control A/A;fTg/0 mice, that dose of leptin caused significantly suppressed food intake in the AgRP^eIF2αA/A mice (Fig. 4H). In parallel, leptin treatment in the AgRP^eIF2αA/A mice led to an increase in ARC cells that were positively labeled with p-STAT3, compared with the control mice (Fig. 4I and J). Collectively, these findings suggest that a deficiency of eIF2α phosphorylation in AgRP neurons induces an increase in α-MSH expression by decreasing AgRP release, which affects leptin sensitivity and the activation of neurons expressing MC receptors.

To further identify the physiological relevance of eIF2α phosphorylation in AgRP neurons, we investigated the metabolic phenotype of AgRP^eIF2αA/A mice and found a decrease in body weight, compared with that of the control A/A;fTg/0 mice, during the observation period (Fig. 5A). The AgRP^eIF2αA/A mice also had less food intake than did the control mice (Fig. 5B). In parallel, the fat tissues of the AgRP^eIF2αA/A mice weighed less than those of the control mice (Fig. 5C). In addition, the AgRP^eIF2αA/A mice had increased energy expenditure, O₂ consumption, and CO₂ production, compared with the control mice, though the respiratory exchange rate did not differ between groups (Fig. 5D–L). However, no correlation was observed between energy expenditure and body weight in both groups of mice (Fig. 5F and Supplementary Fig. 10E), suggesting that the assessed energy expenditure can be considered relatively independent of variables related to body weight. Higher locomotor activity was also observed in the AgRP^eIF2αA/A mice than in the control mice (Fig. 5M).

We then determined whether eIF2α phosphorylation in AgRP neurons affects body temperature and glucose homeostasis. The AgRP^eIF2αA/A mice had increases in body temperature and uncoupling protein 1 expression in the brown adipose tissue (Supplementary Fig. 8), compared with the control mice. No difference in the glucose tolerance test and insulin tolerance test results was observed between the control and AgRP^eIF2αA/A mice (Supplementary Fig. 9).

Female AgRP^eIF2αA/A mice had phenotypes similar to those of male AgRP^eIF2αA/A mice in most metabolic parameters (Supplementary Fig. 10), suggesting that eIF2α phosphorylation deficiency in AgRP neurons does not affect sexual differences in energy metabolism. Collectively, these findings indicate that a deficiency of eIF2α phosphorylation in AgRP neurons causes reductions in body weight by affecting food intake and energy expenditure in male and female mice.

To further verify whether the UPR is altered during short-term starvation in AgRP neurons with impaired eIF2α phosphorylation, we measured the fasting-induced expression of UPR genes in AgRP neurons from AgRP^eIF2αA/A;Rpl22^HA mice. The AgRP^eIF2αA/A;Rpl22^HA mice displayed a decrease in the fasting-induced expression of UPR genes (Atf4, Chop, Xbp1t, Xbp1s, and Grp94) in AgRP neurons, compared with the control AgRP-Cre;Rpl22^HA mice (Fig. 6A), suggesting that eIF2α phosphorylation is required for the fasting-induced UPR gene expression in AgRP neurons.

Previous studies reported that eIF2α phosphorylation participates in the activation of autophagy by facilitating the selective translation of ATF4 (9,39), and the induction of autophagy in AgRP neurons is crucial for the maintenance of energy homeostasis during energy deficit (40,41). Therefore, we investigated the possible involvement of eIF2α phosphorylation in fasting-induced autophagy in AgRP neurons. The AgRP^eIF2αA/A;Rpl22^HA mice had less fasting-induced expression of autophagy genes (microtubule-associated protein 1 light chain 3 β [Lc3b], p62, lysosomal-associated membrane protein 1 [Lamp-1], autophagy related 5 [Atg5], Atg7 and Atg12) in AgRP neurons than did the control mice (Fig. 6B).

We next analyzed the immunoreactivity of the autophagy-related marker p62 (42) in the AgRP neurons of mice after fasting. In the AgRP^eIF2αA/A;Ai14 mice, the AgRP neurons had decreased immunoreactivity to p62, compared with those in the control mice after fasting (Fig. 6C and D). In normal-fed conditions and when exposed to a high-fat diet, we observed a reduction in the expression of UPR and autophagy genes in AgRP neurons of AgRP^eIF2αA/A;Rpl22^HA mice (Supplementary Figs. 11 and 12). These findings indicate that the alterations in gene expression by deficiency of eIF2α phosphorylation are not limited to the fasted state. Overall, these results suggest that eIF2α phosphorylation plays a crucial role in regulating UPR and autophagy in AgRP neurons during metabolic shifts.

In this study, we have shown that eIF2α phosphorylation in AgRP neurons is increased by short-term starvation and that AgRP neuron–specific impairment of eIF2α phosphorylation inhibited starvation-induced AgRP neuronal activation and thus altered metabolic phenotypes.

It is well established that ER stress–induced UPR activation in the brain is directly associated with the progression of cellular pathological processes such as inflammation (5–7). In particular, eIF2α phosphorylation was shown to be essential for UPR signaling and ER stress–induced UPR activation, which is closely correlated with the development of metabolic disorders associated with perturbed cellular homeostasis (12). Despite the well-known pathological involvement of eIF2α phosphorylation in the development of metabolic disorders, the physiological functions of eIF2α phosphorylation in metabolic control have remained largely unknown. Intriguingly, we found that starvation increased the p-eIF2α level only in the hypothalamus, not in the hippocampus or the cortex. The hypothalamus is the central apparatus for controlling whole-body energy metabolism based on fuel availability, so an acute change in energy state primarily affects the operation of the hypothalamic circuit (17,43). Therefore, these findings suggest that eIF2α phosphorylation might be a specific cellular event driving homeostatic metabolic responses, including feeding behavior and energy expenditure, rather than a global cellular event that occurs during an energy deficit.

Recent studies have shown that UPR activation specifically occurs in hypothalamic AgRP neurons during periods of starvation (7,22). Consistent with those reports, we also found that overnight fasting stimulated UPR gene expression in AgRP neurons, whereas no difference was detected in POMC neurons. This UPR activation in AgRP neurons is strongly associated with the energy deficit–induced activation of AgRP neurons and is implicated in the regulation of increased appetite and energy conservation during starvation. Interestingly, it has also been observed that expression of UPR-related genes, such as Xbp1s, Atf4, and Atf6, in POMC neurons increases after short-term refeeding of fasted mice, suggesting that UPR activation in AgRP and POMC neurons may be differentially influenced by the energy status (7,44).

However, the fasting-induced expression of UPR and autophagy genes in AgRP neurons and the activation of AgRP neurons was markedly suppressed in AgRP^eIF2αA/A mice bearing an AgRP neuron–specific impairment of eIF2α phosphorylation. Previous studies have shown that ER stress–induced UPR and eIF2α phosphorylation are functionally associated with the stress-induced stimulation of autophagy in different types of cells (39,45,46). Those cellular responses were also revealed in AgRP neurons during short-term starvation and are considered to be involved in maintaining whole-body energy homeostasis (40,41). Therefore, these results together suggest that the energy deficit–induced stimulation of the UPR and eIF2α phosphorylation in AgRP neurons participate in AgRP neuronal activation by regulating the processes of autophagy.

Hypothalamic autophagy depends on the energy state in the body: the processes of autophagy were impaired in the hypothalamic ARC during overnutrition, and the specific loss of autophagy in POMC neurons caused the development of leptin resistance and insulin resistance (47,48). Consequently, those animal models developed diet-induced obesity, suggesting that an autophagy impairment in hypothalamic neurons is a causative factor in obesity pathogenesis and is associated with metabolic disorders (47,48). In addition, it has been well established that autophagy in AgRP neurons is closely associated with the function of those neurons during starvation and thus affects energy intake and consumption (40,41). In support of that evidence, mice with an AgRP neuron–specific deficiency of ATG7, a key autophagy regulator, had a decrease in the starvation-induced AgRP level and food intake, along with an increase in the α-MSH level, which produced a lean phenotype (40,41). A growing body of evidence suggests that p-eIF2α is essential for the stress-induced expression of genes related to autophagy in different types of cells (39,45,49). Phosphorylated eIF2α facilitated the translation of ATF4, which activates autophagy by stimulating the transcription of Atg genes during responses to stress, such as ER stress, suggesting that the p-eIF2α–ATF4 axis plays an important role in the formation and stimulation of autophagy (39,45). In accordance with those previous reports, we here observed that impaired eIF2α phosphorylation in AgRP neurons affected the expression of multiple genes involved in autophagy in AgRP neurons during short-term starvation, which further suggests the importance of AgRP neuron–specific eIF2α phosphorylation in autophagy-induced AgRP neuronal activation.

In this study, we also found that AgRP^eIF2αA/A mice had fewer AgRP-positive fibers and a concomitant increase in POMC-derived anorexigenic α-MSH signals in the PVN, which expresses MC 3 and 4 receptors to mediate α-MSH action on feeding and is an area that retains preganglionic neurons to control sympathetic nerve activity (35). Considering that energy deficit–activated AgRP neurons exert their inhibitory influence on POMC neuronal activation through inhibitory synaptic input (19), it is reasonable to speculate that the increase in α-MSH signals observed in the PVN of AgRP^eIF2αA/A mice results from concurrent activation of POMC neurons triggered by the decreased activity of AgRP neurons in these mutant mice. Furthermore, the AgRP^eIF2αA/A mice also had higher levels of energy expenditure and reduced appetite. The AgRP^eIF2αA/A mice also showed an increased responsiveness to leptin, in addition to the reduction in appetite and activation of POMC neurons (50), but they had a decreased sensitivity to ghrelin, which usually stimulates the appetite and activates AgRP neurons (36). These results further support that p-eIF2α in AgRP neurons plays an important role in maintaining whole-body energy metabolism by regulating AgRP neuronal activity.

Hypothalamic ER stress and ER stress–induced eIF2α phosphorylation are regarded as crucial pathological elements during the overnutrition period (15,16). In this study, we made a compelling observation that AgRP^eIF2αA/A mice exhibited an antiobesity phenotype after short-term high-fat diet feeding (Supplementary Fig. 12A and B). This effect is likely attributed to decreased levels of UPR and autophagy due to eIF2α phosphorylation deficiency, resulting in reduced activity of AgRP neurons. This finding provides substantial evidence supporting the role of eIF2α phosphorylation in orchestrating the necessary homeostatic response for maintaining energy balance. Nevertheless, more investigations are needed to obtain a deeper understanding of the physiological and pathological implications of eIF2α phosphorylation in metabolic regulation.

In summary, we have demonstrated that eIF2α phosphorylation in AgRP neurons is a key mediator of UPR activation and autophagy induction in AgRP neurons during short-term starvation and is thus important in energy deficit–induced AgRP neuronal activation and the regulation of whole-body energy metabolism. These findings collectively suggest that eIF2α phosphorylation is a cellular indicator to determine energy availability in hunger-promoting AgRP neurons and a molecular switch that triggers the homeostatic feeding behavior governed by AgRP neurons.

Funding. This research was supported by the National Research Foundation (NRF) of Korea Priority Research Centers program (NRF-2014R1A6A1030318) and a grant from the Korean government (NRF-2020R1A2C1008080). K.K.K. was supported by the Basic Science Research Program through the NRF of Korea (NRF-2020R1A6A3A01098849). B.S.P. was supported by an NRF grant (NRF-2021R1C1C2005067).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. K.K.K., J.G.K., and B.J.L. designed the experiments, interpreted results, and wrote the manuscript. K.K.K. and T.H.L. performed and analyzed most of the experiments. B.S.P. performed the indirect calorimetry analysis; D.K. performed the Western blot analysis; D.H.K. and B.J. performed the Ribo-Tag system processes and real-time PCR analyses; J.W.K. and H.R.Y. performed the histological analyses; H.R.K. and S.J. performed the feeding behavior analyses. S.H.B. and J.W.P. provided intellectual input. All authors contributed to the article and approved the submitted version. B.J.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.