Forkhead Box i2 Transcription Factor Regulates Systemic Energy Metabolism Via Neuropeptide AgRP

Obesity is a major health problem that results from a disequilibrium between energy intake and energy expenditure (1,2). Energy metabolism is tightly controlled by the hypothalamus. The arcuate nucleus (ARC), located on the floor of the third ventricle, is the chief hypothalamic area involved in the control of food intake, and it contains two interconnected groups of neurons that release agouti-related protein (AgRP) and neuropeptide Y (NPY), which stimulate food intake, and the anorexigenic peptides proopiomelanocortin (POMC) and cocaine- and amphetamine-regulated transcript (CART) (3,4). AgRP neurons have emerged as key mediators of appetite behavior. Pharmacogenetic or optogenetic activation of AgRP neurons in mice results in rapid stimulation of food intake, even in caloric-replete states (5). These effects of AgRP neurons on food intake are mediated by the release of AgRP, NPY, and γ-aminobutyric acid (6,7). POMC is the precursor of α-melanocyte–stimulating hormone, and the anorexigenic effect of melanocortin is mediated by two receptors, MC3R and MC4R, on second-order neurons (8). AgRP influences food intake mainly through the competitive antagonism of central melanocortin receptors.

In addition to food intake, AgRP neurons regulate energy expenditure and substrate utilization. Acute activation of AgRP neurons reduces energy expenditure, while mice lacking AgRP neurons have enhanced energy expenditure (9,10), indicating a direct relationship between AgRP neurons and brown fat function.

The forkhead class of transcription factors is characterized by a 110–amino acid monomeric DNA-binding domain (11). The forkhead box i (Foxi) subfamily comprises three members (Foxi1, Foxi2, and Foxi3) (12). Mice with a Foxi1-null mutation exhibit defects in inner ear formation, resulting in hearing and balance impairment (13). Foxi3-mutant mice are not viable and display severe branchial arch–derived facial skeleton defects (14). Foxi2 is mainly expressed in the developing forebrain and has a significant role in embryonic development, especially brain development (15). However, whether and how Foxi2 regulates systemic metabolism remains unknown.

In the current study, we show that Foxi2 exerts its effects on food intake and energy expenditure through induction of AgRP expression. Our study reveals a novel regulator of nutrient sensing and energy homeostasis.

Global Foxi2-mutant mice were obtained from the Cyagen Biosciences (stock no. 270004). AgRP-IRES-Cre mice were purchased from The Jackson Laboratory (stock no. 012899). Wild-type (WT) C57BL/6N mice were purchased from SPF (Beijing) Biotechnology Co., Ltd. All animals were housed in 12-h light and 12-h dark photoperiods. For high-fat diet (HFD) experiments, 6–8-week-old male mice were fed an HFD with 60% kcal from fat (D12492; Research Diets, New Brunswick, NJ) for 3–4 months. All animal experiments were reviewed and approved by the institutional animal care and use committee of Tianjin Medical University.

Glucose intolerance tests (GTTs) and insulin tolerance tests (ITTs) were performed as described previously (16,17). For GTT, mice were fasted overnight and injected intraperitoneally with d-glucose (2 g/kg). For ITT, mice were fasted for 6 h and injected intraperitoneally with human insulin (0.75 units/kg; Solarbio). Blood glucose concentrations were measured from the tail vein at indicated times.

For the daily food intake assay, mice were individually housed in cages for 5 days. Food intake and spillage were measured during the last 4 days (18). For fasting-induced refeeding assays, the single-housed mice were deprived of food overnight, then a preweighed amount of food was placed in cages, and food intake was calculated for 24 h (19).

Body weight was measured weekly. Body temperature was measured by rectal thermometer. Body composition (fat and lean mass) was determined by MRI (EchoMRI-Combo-700). Indirect calorimetry was performed according to standard methods using metabolic chambers (TSE Systems, Columbus Instruments). Energy expenditure and locomotor activity were monitored for 48 h in metabolic cages. Before data collection, the mice were acclimated to their individual housing and calorimetry chamber for 24 h. Room temperature was ∼23°C, and light/dark cycles were 12 h (18,20). The rate of energy expenditure was calculated using a modified Weir equation (energy expenditure [kcal/h] = 60 × [0.003941 × VO₂ + 0.001106 × VCO₂]) and respiratory quotient (RQ) as VCO₂/VO₂.

The full-length mouse Foxi2 gene containing a Flag-tag at the C-terminus was cloned into pcDNA3.1. The Flag-tagged Foxi2 gene was constructed into pAdM-FH-GFP, and recombinant adenovirus expressing Foxi2 was generated by Vigene Biosciences (Jinan, China). For Foxi2 knockdown, shRNA sequences were synthesized by Tsingke Biotechnology Co., Ltd. (Beijing, China) and constructed into lentivirus plasmids pLKO.1-TRC cloning vector (no. 10878; Addgene); the primer sequences are shown in Supplementary Table 1. Lentiviruses were generated as described previously (21). The concentrated viruses were used to infect GT1-7 cells.

Briefly, GT1-7 cells were infected with Ad-GFP or Ad-Foxi2 for 48 h and then were harvested for chromatin immunoprecipitation (ChIP) assay as described previously (22). Primer sequences for quantitative PCR are listed in Supplementary Table 1.

pGL3-AgRP promoter (−1795 to 255 base pairs [bp]) was constructed with NheI/XhoI sites. A series of 5′ deletion constructs of AgRP (−1795 luciferase [Luc], −545Luc, −345Luc, and −195Luc) were prepared by PCR using −1795Luc as a template. Primer sequences are shown in Supplementary Table 1. Luciferase was measured as described previously (23).

The pAAV-EF1a-DIO-EGFP-WPRE (adeno-associated virus [AAV]-GFP) and pAAV-EF1a-DIO-Foxi2-P2A-EGFP-WPRE (AAV-Foxi2) viruses (Obio Technology Co., Ltd., Shanghai, China) were injected bilaterally into the ARC of AgRP-IRES-Cre male mice as described previously (19). Briefly, using a stereotactic device, AAVs were injected slowly with a glass micropipette at a rate of 30–40 nL/min (∼2 × 10¹³ vg/mL) for 10 min according to the following coordinates: −1.50 mm posterior to the bregma, ±0.25 mm lateral to the midline, and −5.80 mm below the surface of the skull. The injector (P1 Technologies) was left in position for an additional 10 min to permit diffusion. After surgery, mice were individually housed and allowed 2 weeks to recover before the start of studies. Each group contained seven mice. The anatomical accuracy and effectiveness of stereotactic viral injection were verified under fluorescence microscopy (for electrophysiology-related studies) or by immunofluorescence assay and quantitative PCR (for anatomy and in vivo studies). The “missed” mice were excluded from data analyses.

For hematoxylin-eosin (H-E) staining, immunohistochemistry, and oil red O staining, inguinal white adipose tissue (iWAT), epididymal white adipose tissue (eWAT), brown adipose tissue (BAT), and liver tissue were treated as described previously (24). Immunofluorescence staining was performed using Foxi2 (sc-515114; Santa Cruz Biotechnology) and anti-AgRP (ab254558; Abcam) as described previously (19).

Immunoblotting was performed using the following primary antibodies: Foxi2 (sc-515114; Santa Cruz Biotechnology), β-tubulin (A12289; ABclonal), β-actin (AC026; ABclonal), p-AKT (Ser473) (#9271; Cell Signaling Technology [CST]), AKT (#9272; CST), UCP1 (#14670; CST); ATGL (A5126; ABclonal), p-HSL (Ser563) (#4139; CST), HSL (WL02643; Wanleibio), and tyrosine hydroxylase (WL01820; Wanleibio).

Total RNA was extracted from tissues or cells with TRIzol reagent (Invitrogen). Real-time PCR was performed as described previously (25). The related primers are shown in Supplementary Table 1.

Serum triacylglycerol (TG) and total cholesterol (TC) levels were determined using an automated Monarch device (Tianjin Medical University General Hospital). Hepatic concentrations of TG and TC were quantified with enzymatic assays (Applygen Technologies, Beijing, China) according to the manufacturer’s instructions. Serum insulin levels were determined using an ELISA kit (#EZRMI-13K; EMD Millipore).

Mice fed a standard chow diet were fasted 24 h, then sacrificed for electrophysiological recordings as described previously (26). Briefly, mice were anesthetized with isoflurane, and brains were removed and submerged in ice-cold artificial cerebrospinal fluid (ACSF) containing 120 mmol/L NaCl, 2.5 mmol/L KCl, 2 mmol/L CaCl₂ ⋅ 2H₂O, 1.25 mmol/L KH₂PO₄, 2 mmol/L MgSO₄ ⋅ 7H₂O, 26 mmol/L NaHCO₃, and 10 mmol/L d-glucose saturated in 95% O₂/5% CO₂. Coronal sections containing the ARC (300 μm) were cut using a Leica VT1200S vibratome and bathed in oxygenated ACSF (33°C) for 30 min, then slices were incubated in oxygenated ACSF at room temperature for at least 1 h before recording. Slices were removed to a recording chamber and bathed in oxygenated ACSF at a flow rate of 1.5–2 mL/min at room temperature. The pipette solution for whole-cell recording contained 140 mmol/L potassium gluconate, 2 mmol/L MgCl₂, 10 mmol/L HEPES, 10 mmol/L KCl, 5 mmol/L EGTA, 1 mmol/L CaCl₂, and 2 mmol/L Mg ATP (pH 7.3). The osmolarity of the internal solution was adjusted to ∼290 mmol/kg. Recording electrodes had resistances of 3–5 mol/LΩ when filled with potassium gluconate internal solution. We determined the location of AAV-infected AgRP neurons in ARC visualized by fluorescence microscope (40× infrared lens) (BX51WI; Olympus) with blue excitation block, at which time the light source was switched to bright field to record neuronal discharge. For current-clamp experiments, the resting membrane potential was measured immediately after achieving whole-cell configuration, and at least 5 min was allowed for equilibration between pipette solution and cell to permit membrane conductance to stabilize. Current injection started at −10 pA and was gradually increased to 40 pA to assess neuronal excitability and membrane properties. The current injection step was held at 10 pA, and the action potential (AP) recording time was 500 ms. For spontaneous excitatory postsynaptic current (sEPSC) experiments, cells were held in voltage-clamp mode at −70 mV for 5 min in the presence of 100 μmol/L picrotoxin to eliminate ionotropic GABAergic transmission. Electrophysiological signals were recorded using an Axopatch 700B amplifier (Molecular Devices), low-pass filtered at 2–5 kHz, and analyzed offline on a personal computer running the pCLAMP program (Molecular Devices). The series resistance in whole-cell patch-clamp recording was <30 mol/LΩ and remained stable during the course of the experiments.

All data are presented as mean ± SEM of three independent experiments. Results were analyzed using Microsoft Excel and/or GraphPad Prism software. Statistical significance was determined by unpaired two-tailed Student t test and one-way ANOVA followed by Tukey multiple comparison test. For the curves of body weight and tolerance tests, a t test was used to compare differences between groups of mice at each time point. For energy expenditure, an ANCOVA was performed using SPSS software. ANCOVA with fat and lean mass as well as body weight as covariates was performed for energy expenditure to statistically adjust for body composition differences between groups. Adjusted marginal means and partial η² values as approximations of effect size were calculated. P < 0.05 was considered significant.

No data sets or applicable resources were generated or analyzed during the current study.

We have previously demonstrated that Foxp1 and Foxq1, two members of the Fox family, regulate hepatic glucose homeostasis (22,23). To further identify potential Fox genes involved in energy metabolism, we next systematically studied the expression of Fox family members in the hypothalamus and other metabolically active tissues under different nutritional statuses by real-time PCR.

Our quantitative PCR and Western blot analyses showed that Foxi2 expression was increased in the hypothalamus of mice with HFD-induced obesity compared with chow-fed control mice (Fig. 1A and C). Furthermore, fasting increased Foxi2 mRNA and protein levels, while refeeding reversed this effect (Fig. 1B and D). Interestingly, the expression patterns of AgRP and NPY in the hypothalamus were similar to that of Foxi2 (Fig. 1E). Of note, Foxi2 was highly expressed in the hypothalamus and rarely expressed in other tissues (Fig. 1F and G). In addition to hypothalamic ARC, Foxi2 was expressed in other brain areas, including the cerebellum and brain stem (Fig. 1H). These data imply that hypothalamic Foxi2 might be involved in systemic energy metabolism.

Given the similar expression patterns of Foxi2, AgRP, and NPY, we further determined whether Foxi2 is expressed in AgRP neurons. Indeed, our data suggest that Foxi2 is localized in AgRP neurons, as revealed by immunofluorescence (Fig. 2A).

We next determined whether Foxi2 affects the expression of the NPY and AgRP genes. AAV vectors expressing Cre-dependent Foxi2 and GFP as a control (AAV-Foxi2 and AAV-GFP, respectively) were bilaterally injected into the ARC of male AgRP-IRES-Cre mice (27) (Supplementary Fig. 1A). Immunofluorescence staining of Foxi2 and AgRP showed that Foxi2 expression was increased in AgRP-Cre mice and that the majority of Foxi2 was localized in AgRP neurons in the mouse ARC (Fig. 2B). In addition, real-time PCR demonstrated that Foxi2 mRNA levels in the ARC were significantly increased in AAV-Foxi2–infected AgRP-Cre mice (henceforth referred to as AgRP-Cre^Foxi2 mice) compared with AgRP-Cre^GFP mice; however, AAV-Foxi2 injection did not alter Foxi2 expression in other brain areas or tissues in the mice (Fig. 2C).

Next, we investigated whether AAV-Foxi2 infection influences the expression of genes encoding AgRP, NPY, POMC, and CART, which are expressed in neurons of the ARC in fasted mice. Our data suggest that AAV-Foxi2 infection increased AgRP gene expression in the ARC of mice but did not significantly alter the expression of other hypothalamic neuropeptides (Fig. 2D). These observations indicate that Foxi2 overexpression specifically increased AgRP expression in vivo.

We also studied the stimulatory effects of Foxi2 on AgRP expression in vitro. We generated a recombinant adenovirus expressing Foxi2 (Ad-Foxi2) and infected hypothalamic GT1-7 cells with this virus. Adenovirus-mediated overexpression of Foxi2 significantly increased the expression of AgRP but did not significantly change the expression of POMC (Fig. 2E). In contrast, Foxi2 knockdown in GT1-7 cells by lenti-shFoxi2 modestly decreased expression of AgRP (Fig. 2F). The above results suggest that Foxi2 is a specific regulator of the neuropeptide AgRP both in vivo and in vitro.

Because Foxi2 belongs to the forkhead family of transcription factors, we wondered whether Foxi2 directly activates AgRP gene transcription. Our luciferase reporter gene assay revealed that overexpression of Foxi2 activated the transcription of pGL3-AgRP promoter regions −1795, −545, and −345 bp, while the stimulatory effects of Foxi2 were abolished when the promoter region was further truncated to −195 bp, suggesting that a Foxi2 binding site was present between −345 and −195 bp (Fig. 2G). Additionally, the occupancy of Foxi2 proteins on the AgRP promoter was also confirmed by ChIP (Fig. 2H). Taken together, these results suggest that Foxi2 is an important regulator of nutrient sensing and can activate AgRP gene expression.

We next explored whether selective Foxi2 overexpression in AgRP neurons of male mice affects energy homeostasis. We observed that the body weight of chow-fed AgRP-Cre^Foxi2 mice was higher than that of control mice 9 weeks after AAV injection (Fig. 3A and B). MRI examination confirmed that AgRP-Cre^Foxi2 mice had more fat mass than control mice but no difference in lean mass (Fig. 3C). Furthermore, the individual fat pads of the AgRP-Cre^Foxi2 mice were markedly larger and weighed more than those of the control mice (Fig. 3D and Supplementary Fig. 1B). Consistently, histological examination (H-E staining) of three fat depots revealed an increase in the size of lipid droplets and adipocytes in the fat pads of AgRP-Cre^Foxi2 mice (Fig. 3E). However, selective overexpression of Foxi2 did not markedly affect the weight of the kidney, heart, or spleen with a chow diet (data not shown).

We next sought to identify the potential mechanisms of Foxi2 overexpression–induced obesity. We measured food intake and energy expenditure, which can affect energy homeostasis (28). As expected, food intake under an ad libitum feeding condition or after fasting was increased in AgRP-Cre^Foxi2 mice compared with control mice (Fig. 3F and G), which is consistent with the increased AgRP expression in AgRP-Cre^Foxi2 mice. In addition, energy consumption and activity levels were monitored by housing mice in metabolic cages. ANCOVA (29) indicated that Foxi2 overexpression significantly influenced energy expenditure during both the day and night cycles (Fig. 3H and I). However, Foxi2 overexpression did not change the utilization of carbohydrates and fat as energy sources, as revealed by RQ data (Fig. 3J). No difference in locomotor activity was observed between AgRP-Cre^GFP and AgRP-Cre^Foxi2 mice (Supplementary Fig. 1C). Upon exposure to acute cold, the rectal temperature was lower in AgRP-Cre^Foxi2 mice (Fig. 3K), further confirming that Foxi2 regulates systemic energy metabolism.

We therefore examined the expression of genes associated with thermogenesis and fatty acid oxidation in BAT harvested from AgRP-Cre^Foxi2 mice and controls. Consistent with the reduction in metabolic function in AgRP-Cre^Foxi2 mice, Foxi2 overexpression in AgRP neurons led to a decrease in the expression of thermogenic genes, including UCP1, Dio2, and Cidea. The expression of peroxisome proliferator–activated receptor-α (PPARα), an essential factor involved in fatty acid oxidation, and its target genes, including CPT1a and MCAD, was also decreased in the BAT of AgRP-Cre^Foxi2 mice (Fig. 3L). Consistently, the protein expression of UCP1 was reduced in AgRP-Cre^Foxi2 mice (Fig. 3M). We also examined glucose metabolism in these mice. GTTs and ITTs revealed impaired glucose tolerance and insulin resistance in Foxi2-overexpressing mice (Fig. 3N and O). Thus, our results suggest that increased food intake and reduced energy expenditure together contribute to obesity in AgRP-Cre^Foxi2 mice.

In addition to BAT, beige adipocytes in WAT also contribute to adipose tissue thermogenesis. We first examined the expression of thermogenic genes in the iWAT of AgRP-Cre^Foxi2 mice housed at room temperature and found that selective Foxi2 overexpression did not influence the expression of thermogenic genes (data not shown). Then, we housed the mice at 4°C for 2 days. Acute cold exposure markedly induced browning of WAT, leading to the appearance of smaller, multilocular beige adipocytes with intense UCP1 immunoreactivity in the iWAT of control mice; however, this effect was markedly attenuated in the iWAT of AgRP-Cre^Foxi2 mice. Correspondingly, cold exposure induced the expression of genes involved in thermogenesis in the iWAT of AgRP-Cre^Foxi2 mice, while Foxi2 overexpression in AgRP neurons abolished the stimulatory effects of cold exposure (Fig. 3P and Q). Protein levels of tyrosine hydroxylase, an indicator of sympathetic nerve excitability, were also significantly reduced in the iWAT of AgRP-Cre^Foxi2 mice following cold exposure (Fig. 3R). These findings suggest that Foxi2 overexpression in AgRP neurons impairs the browning of WAT in response to cold exposure.

Likewise, we obtained similar results in female AgRP-Cre^Foxi2 and control mice (Supplementary Fig. 2). Thus, Foxi2 regulation of energy metabolism is independent of sex differences.

Furthermore, we studied energy metabolism of these mice under an HFD condition. We found that overexpression of Foxi2 in AgRP neurons aggravates HFD-induced metabolic dysfunction (Supplementary Fig. 3). Compared with control mice, HFD-fed AgRP-Cre^Foxi2 mice gained more weight because of increased fat mass (Supplementary Fig. 3A–F). Likewise, AgRP-Cre^Foxi2 mice showed an increase in food intake (Supplementary Fig. 3G and H) and decreased energy consumption (Supplementary Fig. 3I–L). Molecular mechanism studies indicated that Foxi2 overexpression decreased the expression of genes involved in thermogenesis in BAT and reduced the expression of key components of lipolytic pathways in WAT (Supplementary Fig. 3M–O). Furthermore, Foxi2 overexpression aggravated systemic glucose intolerance and impaired insulin sensitivity of metabolic tissues, including liver, WAT, and muscle (Supplementary Fig. 3P–T). Finally, compared with controls, AgRP-Cre^Foxi2 mice showed an aggravated fatty liver phenotype (Supplementary Fig. 3U–X).

We next determined whether Foxi2 overexpression influences the activity of AgRP neurons. We measured the activity of AgRP neurons in AgRP-Cre^GFP and AgRP-Cre^Foxi2 mice by performing whole-cell patch-clamp recordings on EGFP-labeled AgRP neurons (Fig. 4A).

AgRP neurons from 24-h fasted AgRP-Cre^Foxi2 mice exhibited a depolarized resting membrane potential (Fig. 4B). In addition, the AP firing rate of AgRP neurons evoked by current injection was higher in AgRP-Cre^Foxi2 mice than in control mice (Fig. 4C and D). Moreover, Foxi2 overexpression in AgRP neurons decreased the AP threshold (Fig. 4E). Furthermore, we analyzed whether selective Foxi2 overexpression affects the shape of APs. After-hyperpolarization (AHP), a component of the AP waveform, is one principal feedback mechanism in controlling the frequency and pattern of neuronal firing. The AHP in AgRP neurons was reduced in AgRP-Cre^Foxi2 mice, while Foxi2 overexpression in AgRP neurons had no significant effect on the AP amplitude (Fig. 4F–H). Changes in synaptic inputs could affect the spontaneous firing of AgRP neurons. Here, we also examined synaptic transmission at excitatory synapses of AgRP neurons in both groups of mice and found that overexpression of Foxi2 significantly increased the frequency but not the amplitude of sEPSCs in AgRP neurons (Fig. 4I–K). These results suggest that overexpression of Foxi2 increases the excitability of AgRP neurons.

To further examine the physiological function of Foxi2 in energy metabolism, we used global Foxi2 knockout (KO) mice. Western blot analysis failed to detect Foxi2 protein in the hypothalamus of KO mice (Fig. 5A). Compared with WT mice, Foxi2 KO mice showed a reduced mRNA level of AgRP, but the expression levels of NPY, POMC, and CART were not significantly changed (Fig. 5B). Consistently, the protein level of hypothalamic AgRP was reduced in Foxi2 KO mice, as revealed by immunofluorescence staining (Fig. 5C). Indeed, Foxi2 KO mice showed decreased food intake (Fig. 5D and E). The body weight of Foxi2 KO mice was lower than that of control mice starting from 18 weeks of age (Fig. 5F and G). MRI confirmed that Foxi2-deficient mice had less fat mass than control mice, while lean mass remained unchanged (Fig. 5H and I). The individual fat pads of Foxi2-deficient mice, including BAT, iWAT, and eWAT, weighed less than those of control mice (Fig. 5J and Supplementary Fig. 4A). H-E staining revealed that Foxi2 deficiency decreased the size of brown and white adipocytes and lipid droplets in adipocytes (Fig. 5K). Metabolic cage experiments showed that Foxi2 KO mice had increased energy consumption calculated by ANCOVA, whereas there were few differences in RQ or locomotor activity (Fig. 5L–N and Supplementary Fig. 4B). Consistent with the enhanced metabolic function of Foxi2-deficient mice, Foxi2 deficiency led to an increase in the expression of genes involved in thermogenesis and fatty acid oxidation (Fig. 5O and P).

We also examined glucose metabolism in these mice. GTTs and ITTs revealed that Foxi2 KO mice had enhanced glucose tolerance and improved insulin sensitivity (Fig. 5Q and R). These results suggest that depletion of Foxi2 decreases food intake and increases energy expenditure, subsequently improving metabolic phenotypes.

We also challenged Foxi2-deficient mice with an HFD and studied the metabolic phenotype of these mice. We fed 6-week-old male WT and Foxi2 KO mice an HFD for 16 weeks and obtained similar results as those reported above. Foxi2 KO mice showed a decrease in food intake (Fig. 6A and B). After 16 weeks of HFD feeding, Foxi2 KO mice displayed reduced body weight, fat mass, and BAT, iWAT, and eWAT weight (Fig. 6C–F and Supplementary Fig. 5A). Likewise, H-E staining suggested that the size of adipocytes in three different fat depots in Foxi2 KO mice was also reduced (Fig. 6G). HFD-fed Foxi2 KO mice exhibited increased energy consumption and decreased RQ but no change in locomotor activity (Fig. 6H–J and Supplementary Fig. 5B). Furthermore, Foxi2 KO mice had an increased rectal temperature, confirming the enhancement of energy expenditure (Fig. 6K). Likewise, the expression levels of thermogenic genes in BAT were increased in Foxi2 KO mice (Fig. 6L). Additionally, Foxi2 deficiency increased the protein levels of p-HSL and ATGL, the rate-limiting enzymes involved in lipolysis, in iWAT (Fig. 6M). These results suggest that Foxi2 deficiency can protect against HFD-induced obesity.

Biochemical analysis revealed that serum glucose and insulin levels were significantly reduced in Foxi2 KO mice fed an HFD (Fig. 6N and O). GTTs and ITTs confirmed that Foxi2 deficiency improved glucose intolerance and insulin resistance induced by HFD feeding (Fig. 6P and Q). Consistent with the ITT data, insulin-stimulated AKT activation was enhanced in Foxi2 KO mice (Fig. 6R). Moreover, Foxi2 KO mice showed reduced liver weight and hepatic lipid accumulation, as revealed by H-E and oil red O staining (Fig. 6S and T and Supplementary Fig. 5C). Hepatic and serum TG levels, as well as serum cholesterol levels, were reduced in HFD-fed Foxi2 KO mice (Fig. 6U). Likewise, Foxi2 deficiency decreased the mRNA levels of lipogenic genes in the livers of mice, while the expression of CPT1a, PPARα, and peroxisome proliferator–activated receptor γ coactivator 1α (PGC-1α) was not markedly changed (Fig. 6V). Collectively, these data indicate that Foxi2 deficiency alleviates hepatic steatosis and improves insulin resistance in HFD-fed mice.

AgRP neurons are activated by hormonal and neural signals of heightened energy demand and stimulate feeding, at least in part, by antagonizing postsynaptic MC3R/MC4Rs through the release of AgRP (30). In the current study, we show that the transcription factor Foxi2, which was highly expressed in the hypothalamus, is a critical regulator of AgRP expression in response to nutritional status. We found that nutritional status, including fasting and HFD feeding, regulates Foxi2 expression. Fasting is well known to induce AgRP gene expression and enhance neuronal activity (31). How fasting induces Foxi2 expression and whether Foxi2 mediates the effects of fasting on AgRP neuron activity deserve further study.

In addition to Foxi2, other factors, including ATF4 and Foxo1, have been shown to regulate AgRP gene expression and neuronal activity (32,33). Of note, ATF4 has been demonstrated to regulate AgRP expression through Foxo1. Additionally, a consensus site for Stat3 exists in the promoter region of the AgRP gene, and overexpression of Stat3 causes inhibition of AgRP transcription (33,34). Thus, multiple factors form a complicated transcriptional network regulating AgRP expression in response to hormonal and nutritional cues. It is logical to speculate that Foxi2, as a transcription factor, has multiple downstream target genes and other physiological roles in the hypothalamus. ChIP sequencing in combination with RNA sequencing is required to identify the global target genes of Foxi2 in the hypothalamus. In addition to the ARC, our data suggest that Foxi2 is also expressed in the paraventricular nucleus, ventromedial hypothalamus, cerebellum, and brain stem. Further studies are required to determine the physiological functions of Foxi2 in these brain areas.

FOX proteins constitute an evolutionarily conserved family of transcription factors with an important role in development and metabolism. Based on sequence conservation, FOX genes have been divided into 19 subfamilies from Foxa to Foxs (35). Among the best studied of these factors are Foxo1 and Foxa2, both of which have been shown to regulate systemic glucose and lipid metabolism (36,37). The activity of Foxo1 is regulated by posttranslational modification, including phosphorylation and acetylation. These modifications affect Foxo1 location from the nucleus to the cytosol (36). Similar results were also observed for Foxa2, which is phosphorylated by insulin/phosphatidylinositol 3-kinase/AKT signaling and acetylated by p300 (38,39). Whether and how Foxi2 proteins are regulated by these posttranslational modifications in response to hormonal and nutritional cues remains unclear. Further studies are required to clarify these questions.

Foxi2 overexpression in AgRP neurons induced AgRP gene expression, thereby increasing food intake and decreasing energy metabolism, and subsequently caused glucose intolerance and impaired insulin sensitivity and obesity. Of note, to explore the effects of Foxi2 overexpression in AgRP neurons, we performed stereotactic injection of AAV-GFP (as a control) and AAV-Foxi2 into the ARC of AgRP-IRES-Cre mice. Each group consisted of seven mice. To verify the anatomical accuracy and effectiveness of viral injection, we examined the Foxi2 expression levels in ARC and other brain areas of all these sacrificed mice by quantitative PCR and immunofluorescence assay. We found that Foxi2 was specifically induced in the ARC of six AAV-Foxi2–injected AgRP-IRES-Cre mice, while its expression was not altered in the ARC of one mouse. Thus, this mouse was excluded from further study.

In contrast, global Foxi2-deficient mice gained less weight upon aging and were resistant to HFD-induced obesity. In the current study, we used male mice in most cases, except we used chow-fed female mice with selective overexpression of Foxi2 in AgRP neurons. This is a weakness of the study. In the future, female Foxi2 KO mice are required to further confirm our results. Additionally, we performed GTTs and ITTs using the traditional methodology, which has several weaknesses. First, in a typical GTT experiment, mice are fasted overnight. However, a shorter fast (5–6 h) is more physiological for mice. Second, the traditional approach for GTTs or ITTs performed in mice is to base the dose of glucose or insulin on the weight of the mouse. Lean mass (muscle, brain, and liver) is the principal site of glucose disposal. However, modulating Foxi2 expression in mice changed the fat mass but not the lean mass. Thus, it is more appropriate to base the dose of glucose and insulin on lean mass rather than on body weight (40). Although ANCOVA showed that modulating Foxi2 expression in mice similarly influences energy consumption and body weight as age, we point out that these mice were maintained at room temperature (∼23°C) rather than at thermoneutrality (30°C). The 23°C house condition is a chronic thermal stress to mice, which may have influenced the outcome of the metabolic studies (41,42). It is more appropriate to study the effects of Foxi2 on energy metabolism of mice at thermoneutrality to avoid the thermal stress.

We show that global Foxi2 KO mice have decreased food intake and increased energy metabolism. Notably, Foxi2 is also expressed in other brain areas and BAT. Thus, Foxi2 deficiency in these tissues may also contribute to these phenotypes of Foxi2 KO mice. Moreover, Foxi2 may be involved in brain development (15,43), which might indirectly affect systemic energy metabolism. Thus, AgRP-specific Foxi2 KO mice were required to confirm its function in AgRP neurons in the future.

Previous studies have indicated that AgRP expression levels are positively related to AgRP neuron activity and food intake. For example, fasting induces AgRP mRNA levels and activates AgRP neuron activity (31). Leptin and insulin act in the ARC of the hypothalamus to reduce expression of the AgRP gene and inhibit AgRP neuron activity (30,33,44). Furthermore, Foxo1 activates AgRP gene transcription and influences food intake of mice in response to leptin treatment (33). Finally, PGC-1α regulates AgRP expression in response to fasting, and PGC-1α deficiency in AgRP neurons led to a reduction of food intake (45). Consistent with these data, our data indicate that Foxi2 overexpression stimulates AgRP expression and enhances AgRP neuronal activity (whole-cell patch-clamp experiments). Thus, it appears that an increase in AgRP expression levels drives the activity of AgRP neurons on the basis of these data. On the other hand, insulin and leptin have been shown to inhibit AgRP neuronal activity through the phosphatidylinositol 3-kinase/phosphatidylinositol 4-phosphate 3 axis. Phosphatidylinositol 4-phosphate 3 opens K_ATP channels, thereby producing an outward flow of K⁺ ions and neuron hyperpolarization (46). Whether and how Foxi2 influences K⁺ ions and AgRP neuron hyperpolarization remain unknown. The exact mechanism of Foxi2 activation of AgRP neuron activity deserves further studies.

In summary, our work identifies Foxi2 as a critical regulator of AgRP neuron activity. Targeting Foxi2 in the hypothalamus might be a therapeutic strategy for treating obesity and metabolic diseases.

Funding. This work was supported by the National Natural Science Foundation of China (grants 81825004 and 81730024), the National Key Research and Development Program of China (grant 2018YFA0800601), and the Scientific and Technological Research Project of Xinjiang Production and Construction Corps (grants 2018AB018 and 2021AB028).

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

Author Contributions. Y.F., S.S., and C.G. conducted the experiments. Y.F., S.S., C.G., W.Q., Y.J., L.T., L.Z., and X.D. acquired data. Y.F., S.S., C.G., W.Q., Y.J., Y.G., L.Z., J.Z., and X.D. analyzed data. Y.F., Y.L., H.S., and Y.C. designed the research studies. Y.F. and Y.C. wrote the manuscript. L.T., Y.G., J.Z., X.L., H.S., Y.L, and Y.C. provided reagents. Y.C. 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.