mTORC1-dependent increase in oxidative metabolism in POMC neurons regulates food intake and action of leptin.

OBJECTIVE: Nutrient availability modulates reactive oxygen species (ROS) production in the hypothalamus. In turn, ROS regulate hypothalamic neuronal activity and feeding behavior. The mechanistic target of rapamycin complex 1 (mTORC1) pathway is an important cellular integrator of the action of nutrients and hormones. Here we tested the hypothesis that modulation of mTORC1 activity, particularly in Proopiomelanocortin (POMC)-expressing neurons, mediates the cellular and behavioral effects of ROS.
METHODS: C57BL/6J mice or controls and their knockout (KO) littermates deficient either for the mTORC1 downstream target 70-kDa ribosomal protein S6 kinase 1 (S6K1) or for the mTORC1 component Rptor specifically in POMC neurons (POMC-rptor-KO) were treated with an intracerebroventricular (icv) injection of the ROS hydrogen peroxide (H2O2) or the ROS scavenger honokiol, alone or, respectively, in combination with the mTORC1 inhibitor rapamycin or the mTORC1 activator leptin. Oxidant-related signal in POMC neurons was assessed using dihydroethidium (DHE) fluorescence.
RESULTS: Icv administration of H2O2 decreased food intake, while co-administration of rapamycin, whole-body deletion of S6K1, or deletion of rptor in POMC neurons impeded the anorectic action of H2O2. H2O2 also increased oxidant levels in POMC neurons, an effect that hinged on functional mTORC1 in these neurons. Finally, scavenging ROS prevented the hypophagic action of leptin, which in turn required mTORC1 to increase oxidant levels in POMC neurons and to inhibit food intake.
CONCLUSIONS: Our results demonstrate that ROS and leptin require mTORC1 pathway activity in POMC neurons to increase oxidant levels in POMC neurons and consequently decrease food intake.

Introduction

Cells use different type of signals to convey information about energy availability and coordinate cellular functions. In mammalian cells, adenosine-triphosphate (ATP) production encompasses mitochondrial oxygen consumption, which leads to the production of reactive oxygen species (ROS). ROS, including hydrogen peroxide (H2O2) and the superoxide anion (O2), are oxidant radical species derived from oxygen acting as signal molecules able to modulate numerous intracellular pathways [1], [2].

Hypothalamic neurons can sense, transmit, and convert ROS signals into appropriate intracellular responses [2], [3]. Among these, increased ROS production has been involved in determining hypothalamic glucose and lipid sensing [4], [5], [6] and in mediating the central appetite-suppressant action of insulin [7]. Remarkably, nutrient availability directly impacts ROS levels in Neuropeptide Y/Agouti-related protein (NPY/AgRP) and proopiomelanocortin (POMC) expressing neurons of the hypothalamic arcuate nucleus (ARC) [8], [9]. In turn, ROS modify the activity of these two neuronal populations [8], [9], which control feeding behavior and peripheral metabolism by integrating information related to energy availability through neuronal inputs [10] and circulating metabolic cues [11], [12], [13]. Specifically, during fasting, ROS levels are not increased in NPY/AgRP neurons despite increased neuronal firing, thanks to a feed-forward ROS buffering mechanism involving mitochondrial uncoupling protein 2 (UCP2) [8]. However, if ROS generation is uncontrolled, firing of NPY/AgRP cells is impaired [8]. In contrast, while in fasting, ROS levels are low in hypothalamic POMC neurons, they are high upon refeeding in the same neuronal population [8], [9]. Moreover, POMC neurons are depolarized and fire more in response to H2O2, while application of the H2O2 scavenging enzyme catalase during patch-clamp recordings hyperpolarizes and inhibits them [9], [14]. Thus, a transient increase in ROS levels in POMC neurons favors satiety whereas suppression of ROS through intracerebroventricular (icv) administration of the ROS scavenger honokiol diminishes POMC neurons activity, as shown by the reduced expression of the neuronal activity marker c-fos, and promotes food intake [9]. Remarkably, ROS levels in POMC neurons of lean, chow-fed mice correlate with circulating leptin levels [9] and leptin induces ROS formation in hypothalamic cells in vitro [15]. However, whether this happens in vivo and if it is relevant for the effect of the hormone on food intake remains to be established.

Similarly to ROS, the mammalian or mechanistic target of rapamycin (mTOR) pathway is an important cellular integrator of the actions of nutrients and hormones on food intake [16]. mTOR is an evolutionary conserved serine/threonine kinase that forms two distinct complexes in cells (mTOR complex 1 or mTORC1, and mTOR complex 2 or mTORC2), controlling cellular proliferation and metabolism in response to nutrients, growth factors, mitogens, hormones, and cellular stress [16], [17]. We and others have shown that mTORC1 signaling localizes in NPY/AgRP and POMC neurons of the ARC and that the hypothalamic mTORC1 pathway, which includes the 70-kDa ribosomal protein S6 kinase 1 (S6K1) as one of the downstream effectors, participates in the regulation of energy balance [18], [19], [20], [21], [22], [23]. Increased hypothalamic mTORC1 activity is required for leptin-induced anorexia [18], [19], [20], [21]. However, the specific neuronal substrates in which leptin-induced mTORC1 activity translates into decreased food intake are currently unknown.

Taking into account that there is evidence linking ROS and mTORC1 signaling, mostly in aging and cancer-related studies [24], [25], here we have hypothesized that the mTORC1 pathway mediates ROS-dependent responses, specifically in POMC neurons, leading to consequent changes in food intake, and that ROS and mTORC1 signaling are intertwined and determine the effect of leptin on food intake.

By using genetic models and pharmacology, our study reveals that ROS require a functional mTORC1 pathway in POMC neurons to decrease food intake and that increased ROS and mTORC1 activity in POMC neurons are needed in order to observe the appetite-suppressant action of leptin.

Material and methods

Animals

The experiments were conducted in strict compliance with European Union Directives (2010/63/EU) and were approved by the ethical committee of the University of Bordeaux (authorization number DIR1325). All procedures involving animals were performed in accordance with the ARRIVE guidelines [26]. Animal welfare was monitored daily for the length of the study.

Two-to 3- month-old male C57BL/6J mice (Janvier, France), S6K1-KO, and their wild-type (WT) littermates were used. The S6K1 mouse strain [27], kind gift of G. Thomas and S. Kozma (IDIBELL, Barcelona, Spain) was out-crossed at least twice to C57BL/6J mice upon arrival at the Neurocentre Magendie and was maintained on heterozygous breeding. POMCCre+/+::Rptorflox/flox mice (lacking the expression of Rptor, Regulatory-associated protein of mTOR, in POMC cells and thereafter called POMC-rptor-KO) and their POMCCre−/-::Rptorflox/flox control littermates were generated by crossing POMC-Cre mice [Tg(Pomc1-cre)16Lowl/J, JAX Stock #005965, The Jackson Laboratory, USA] with Rptor-Flox mice (B6.Cg-Rptortm1.1Dmsa/J, JAX Stock #013188, The Jackson Laboratory). This conditional mouse line was generated following a 3-step backcrossing method [28].

Mice were genotyped by PCR [28] using specific primers (see Supplementary Table 1). Effective Cre-mediated deletion of Rptor in POMC neurons was further assessed by immunohistochemistry.

Mice were housed individually with chow (Standard Rodent Diet A03, 3.2 kcal/g; SAFE, France) and water ad libitum, unless otherwise specified, under a 12h light/dark cycle (lights on at 01:00 am), at 22 ± 2 °C. At the end of the behavioral experiments, mice were killed by cervical dislocation or deeply anesthetized to undergo intracardiac perfusion for tissue collection and further analysis. The number of animals used for each experiment is detailed in the figure legends.

Surgery

Using a stereotaxic apparatus (David Kopf Instruments, USA), anesthetized mice were implanted with a cannula into the lateral cerebral ventricle [28]. Correct cannula placement was confirmed by administering 5 μg of the orexigenic peptide NPY (Phoenix Pharmaceuticals Inc, France) in 1 μl of 0.9% physiological saline. The cannula was judged correctly placed if the mice ate a minimum of 0.5 g of chow within 2h after the administration of NPY, during the light phase [29]. Mice that failed to respond correctly to the NPY test were removed from the study.

Drug administration and food intake studies

Drugs were administered through an intracerebroventricular (icv) cannula or by intraperitoneal (ip) injection. H2O2 (5 μM in 2 μL saline, icv, as in [9]) or its vehicle was administered just before the onset of the dark phase, after a 24h fast. To study the effect of a central mTORC1 inhibition on the appetite suppressant action of H2O2, mice received an acute icv injection of rapamycin [VWR International, France; 18 μg in 1 μL dimethyl sulfoxide (DMSO)], an inhibitor of mTORC1 [17], or its vehicle just before the dark phase, 40 min after an acute icv injection of H2O2 in 24h fasted C57BL/6J mice. In preliminary experiments (see Supplementary Fig. 1), we tested the dose of rapamycin (18 or 25 μg in 1 μL DMSO, icv) to combine with H2O2. Mouse recombinant leptin (obtained from Dr A. F. Parlow, National Hormone and Pituitary Program, Torrance, CA) [2.5 μg/μL in 2 μL of phosphate buffered saline (PBS), icv] or its vehicle was administered in free-fed animals 4h before the onset of the dark phase [30]. When combined with the ROS scavenger honokiol, leptin (2.5 mg/kg) was given to free-fed animals by ip injection, while honokiol was administered icv (376 μM in in 2 μL DMSO).

Immunohistochemistry (IHC)

Mice were deeply anesthetized using pentobarbital (ip) and then perfused transcardially with ice-cold PBS, pH 7.4, followed by 4% paraformaldehyde (PFA, Sigma–Aldrich, France) in PBS with 0.2% picric acid. Brains were postfixed in 4% PFA overnight at 4 °C, then cryoprotected with 30% sucrose in PBS at 4 °C. Coronal sections (30 μm) were cut with a cryostat (CM1950, Leica, Germany), collected in PBS and stored in antifreeze solution (30% ethylene glycol, 30% glycerol in KPBS) at −20 °C until further used. Brain sections from POMCCre−/-::Rptorflox/flox control and their POMC-rptor-KO littermates were processed for the co-localization of Rptor and POMC. Sections were incubated with 10% normal goat serum (Dako) and then with rabbit anti-Rptor antibody (1:500, Abcam, UK) overnight at 4 °C. The next day, sections were washed in PBS and incubated for 1h with A647-conjugated secondary goat anti-rabbit antibody (1:500, Cell Signaling). Sections were washed in PBS and incubated with Goat Fab Fragment Anti-Rabbit IgG (1:100, Jackson ImmunoResearch Laboratories, USA) to avoid cross-reactivity. Sections were washed in PBS and blocked with 10% normal goat serum (Dako) and then incubated with rabbit anti-POMC antibody (1:2000; Phoenix Pharmaceuticals, France) overnight at 4 °C. Sections were washed in PBS and incubated for 1h with A488-conjugated secondary goat anti-rabbit antibody (1:500, Cell Signaling). Fluorescent images were acquired with a confocal microscope (SP8-STED, Leica, Germany), and all sections containing the ARC (from Bregma −0.94 mm to Bregma −2.70 mm [31]) were rostro-caudally analyzed using ImageJ (https://imagej.nih.gov/ij/).

Analysis of oxidant levels in POMC neurons

Dihydroethidium (DHE, Sigma–Aldrich, France) was used, as in [8], [9], [32], to assess oxidant levels [33] in POMC neurons. Mice received first DHE (1 mg/mL in anhydrous DMSO, ip) and after 30 min, icv H2O2, leptin, or vehicle. Twenty min later, mice were deeply anesthetized with pentobarbital and perfused intracardially with 4% PFA in PBS. After post-fixation in 4% PFA overnight and cryoprotection with 30% sucrose in PBS, brains were processed and coronal brain sections (30 μm) containing the hypothalamus were obtained using a cryostat (CM1950, Leica). Sections were always protected from the light. After 30 min in PBS, they were incubated in blocking solution (10% normal goat serum in PBS containing 0.3% Triton X-100) for 1h. Primary anti-rabbit POMC antibody (1:2000, Phoenix Pharmaceuticals) was then applied overnight at 4 °C. The next day, sections were washed in PBS and incubated with the secondary antibody (goat anti-rabbit IgG alexa fluor 488, 1:1000, Cell Signaling) for 1h. Finally, the sections were mounted and confocal images were taken using a Spinning Disk microscope (Leica DMI 6000, Leica) equipped with a HQ2 camera (Photometrics, Tucson, USA). The diode lasers used were 491 and 561 mm. The objective used was a HCX PL APO CS 63× oil 1.32 NA. The z stacks were done with a piezo P721.LLQ (Physik Instrumente, Germany). Semi-quantitative analysis of DHE-associated fluorescence within POMC neurons was performed in ImageJ using a macro in which number and volume of the fluorescent spots within the cell were taken into account to determine the quantification of the fluorescent signal. The same area, designed within the ARC, was used to quantify the fluorescence within several POMC cells contained in the area of interest across different sections and animals. At least 20 POMC cells per animal were analyzed in the different experimental conditions.

Statistics

Values are expressed as mean ± SEM. Data were analyzed using unpaired t-test, two-way ANOVA, one and two-way repeated-measures ANOVA, using genotype and treatment as independent variables, followed by Fisher-LSD post-hoc analysis, with Statistica Version 9 (Statsoft, Maisons-Alfort, France). P < 0.05 denoted statistical significance.

Results

Central administration of H2O2 requires mTORC1 signaling to decrease food intake

To investigate the possible interaction between ROS and mTORC1 signaling in the regulation of food intake, we administered icv H2O2 to C57BL/6J mice, in combination with icv delivery of the mTORC1 inhibitor rapamycin at a dose that did not alter food intake (Supplementary Fig. 1). In agreement with previous studies [9], H2O2 significantly decreased 24h food intake and body weight (Figure 1 A, B), while co-administration of rapamycin blunted these effects (Figure 1 A, B). To further assess whether mTORC1 signaling was necessary to observe the behavioral changes induced by H2O2, we then tested the effects of this ROS in WT and S6K1-KO littermates. H2O2 given icv caused hypophagia and weight loss in WT, but not in S6K1-KO mice (Figure 1 C, D). These studies therefore suggest that H2O2 engages mTORC1 signaling to modulate food intake. However, they do not allow pinpointing the exact site(s) where this action is exerted.

Figure 1

Central administration of the ROS H2O2 engages mTORC1 signaling to decrease food intake. (A) Acute icv co-administration of the mTORC1 inhibitor rapamycin with H2O2 blunts the effect of H2O2 on food intake (Repeated measures ANOVA: treatment effect F(2,15) = 6.39, P < 0.01; time effect F(3,45) = 141.33, P < 0.0001; treatment × time interaction F(6,45) = 4.59, P < 0.01, n = 5–7 mice per group) and (B) 24h body weight (BW) change (One-way ANOVA: treatment effect F(2,15) = 9.37, P < 0.005, n = 5–7 mice per group) in C57BL/6J mice. (C) Effect of an acute icv administration of H2O2 on food intake (Repeated measures ANOVA: treatment effect F(1,40) = 7.94, P < 0.01; genotype effect F(1,40) = 11.24, P < 0.005; treatment × genotype interaction F(1,40) = 6.18, P < 0.05; time effect F(3,120) = 892.17, P < 0.0001; treatment x genotype × time interaction F(3,120) = 7.58, P < 0.0005, n = 10–12 mice per group) and (D) 24h BW change (two-way ANOVA: treatment effect F(1,40) = 5.71, P < 0.05; genotype effect F(1,40) = 4.37, P < 0.05; treatment × genotype interaction F(1,40) = 9.81, P < 0.005, n = 10–12 mice per group) in WT and S6K1-KO littermates. Data are mean ± SEM. *P < 0.05, **P < 0.005, ***P < 0.0005.

mTORC1 activity in POMC neurons determines the behavioral and molecular effects of H2O2

ROS may inhibit food intake by activating hypothalamic POMC neurons, which mostly rely on oxidative metabolism [9]. Hence, we evaluated whether the ability of H2O2 to modify food intake and body weight required mTORC1 in POMC neurons. In order to do so, we generated mice lacking Rptor (a protein necessary for the assembling of mTORC1 [34], [35]) and therefore characterized by defective mTORC1 activity. Rptor was expressed in 83.87% of POMC neurons of POMCCre−/-::Rptorflox/flox controls, while the protein was present in only 20.98% of POMC cells of POMC-rptor-KO littermates (Supplementary Fig. 2A). POMC-rptor-KO mice did not show any evident alteration in food intake or body weight under normal housing conditions as compared to controls (Supplementary Fig. 2 B, C). However, similarly to what previously observed using S6K1 mice, while POMCCre−/-::Rptorflox/flox controls decreased their food intake and body weight in response to icv delivery of H2O2, POMC-rptor-KO mice did not (Figure 2 A, B). Under these experimental conditions, in which animals were fasted for 24h before the administration of H2O2, we observed that POMC-rptor-KO, independently of the pharmacological treatment, ate less than control littermates (Figure 2 A, B). Accumulation of oxidants, as assessed by DHE-associated fluorescence, was evident in hypothalamic POMC neurons from controls receiving H2O2, but it was lacking in POMC-rptor-KO littermates (Figure 2 C, D). A trend (P = 0.1) toward a genotype effect was also observed, with decreased DHE-associated fluorescence in POMC-rptor-KO mice as compared to POMCCre−/-::Rptorflox/flox controls (Figure 2 C, D).

Figure 2

H2O2 requires activity of mTORC1 in POMC neurons to modulate food intake and intracellular oxidant levels. (A) Effect of an acute icv administration of H2O2 on food intake (Repeated measures ANOVA: treatment effect F(1,25) = 3.99, P = 0.05; genotype effect F(1,25) = 1.25, P = 0.31; treatment × genotype interaction F(1,25) = 4.73, P < 0.05; time effect F(3,75) = 624.43, P < 0.0001; treatment x genotype × time interaction F(3,75) = 6.63, P < 0.0005, n = 7–8 mice per group) and (B) 24h BW change (Two-way ANOVA: treatment effect F(1,25) = 1.04, P = 0.31; genotype effect F(1,25) = 0.33, P = 0.56; treatment × genotype interaction F(1,25) = 6.07, P < 0.05, n = 7–8 mice per group) in POMCCre−/-::Rptorflox/flox control and POMC-rptor-KO littermates. (C) Representative images of DHE-associated fluorescence in POMC neurons of control and POMC-rptor-KO littermates after icv administration of H2O2 and (D) related quantification (Two-way ANOVA: treatment effect F(1,8) = 1.70, P = 0.22; genotype effect F(1,8) = 2.44, P = 0.1; treatment × genotype interaction F(1,8) = 5.9, P < 0.05, n = 3 mice per group). Data are mean ± SEM. *P < 0.05, **P < 0.005. Scale bar: 10 μm.

Thus, these data imply that mTORC1 activity in POMC neurons is required to observe both behavioral and molecular changes caused by exogenous ROS increase.

The hypothalamic mTORC1 pathway is involved in determining the actions of leptin on food intake [18], [21]. In turn, leptin increases formation of ROS in vitro [15] and acts onto POMC neurons to modulate food intake [36], [37]. Therefore, we finally tested whether the ability of leptin to modulate food intake requires ROS and mTORC1 in POMC neurons.

Leptin requires ROS formation to modulate food intake

First, we demonstrated that an acute icv administration of leptin rapidly increased oxidant levels in POMC neurons (Figure 3 A, B). Consequently, we assessed whether ROS scavenging could blunt the hypophagic action of leptin. To this purpose, an icv administration of the ROS scavenger honokiol at a dose that did not alter food intake (Figure 3C) was combined with a peripheral injection of leptin. As expected, leptin significantly reduced food intake, while co-administration of honokiol completely prevented leptin-induced hypophagia (Figure 3C).

Figure 3

Leptin needs ROS formation to modulate food intake. (A) Representative images of DHE-related fluorescence in POMC neurons after icv administration of leptin in C57BL/6J mice and (B) related quantification (Unpaired t-test, one-tailed: t(9) = 1.94, P < 0.05, n = 5–6 mice per group). (C) Effect of an acute ip administration of leptin combined with icv administration of the ROS scavenger honokiol on 2h food intake in C57BL/6J mice (Two-way ANOVA: honokiol effect F(1,27) = 0.80, P = 0.38; leptin effect F(1,27) = 5.33, P < 0.05; honokiol × leptin interaction F(1,27) = 4.53, P < 0.05, n = 7–9 mice per group). Data are mean ± SEM. *P < 0.05, **P < 0.005. Scale bar: 10 μm.

mTORC1 in POMC neurons determines the ability of leptin to modulate food intake and intracellular oxidant levels

Lastly, to verify whether leptin requires mTORC1 in POMC neurons to regulate food intake and ROS formation, we investigated the central actions of the hormone in POMC-rptor mice. Icv administration of leptin significantly decreased food intake and body weight in POMCCre−/-::Rptorflox/flox, but not in POMC-rptor-KO littermates (Figure 4 A, B). The appetite suppressant action of the hormone was associated with an accumulation of oxidants in POMC neurons, which was lacking in cells of POMC-rptor-KO mice (Figure 4 C, D). A significant genotype effect was observed, with POMC-rptor-KO mice having less DHE-associated fluorescence than controls (Figure 4 C, D).

Figure 4

Leptin requires activity of mTORC1 in POMC neurons to modulate food intake and intracellular oxidant levels. (A) Effect of an acute icv administration of leptin on food intake (Repeated measures ANOVA: treatment effect F(1,27) = 11.1, P < 0.005; genotype effect F(1,27) = 0.50, P = 0.49; treatment × genotype interaction F(1,27) = 3.79, P = 0.06; time effect F(3,81) = 314.54, P < 0.0001; treatment x genotype × time interaction F(3,81) = 7.44, P < 0.0005, n = 7–9 mice per group) and (B) 24h BW change (Two-way ANOVA: treatment effect F(1,27) = 6.01, P < 0.05; genotype effect F(1,27) = 2.91, P = 0.10; treatment × genotype interaction F(1,27) = 3.2, P = 0.08, n = 7–9 mice per group) in POMCCre−/-::Rptorflox/flox control and POMC-rptor-KO littermates. (C) Representative images of DHE-related fluorescence in POMC neurons of control and POMC-rptor-KO littermates after icv administration of leptin and (D) related quantification (Two-way ANOVA: treatment effect F(1,12) = 5.04, P < 0.05; genotype effect F(1,12) = 15.79, P < 0.005; treatment × genotype interaction F(1,12) = 6.51, P < 0.05, n = 3–5 mice per group). Data are mean ± SEM. *P < 0.05, **P < 0.005, ***P < 0.0005. Scale bar: 10 μm.

Overall, these findings imply that mTORC1 activity in POMC neurons is required for the ability of leptin to induce ROS production in POMC neurons and consequently regulate food intake.

Discussion

In recent years, a series of studies have established that a transient increase in hypothalamic ROS is part of a second messenger intracellular response to an increase in energy availability, which participates to the inhibition of feeding behavior. Indeed, increased hypothalamic ROS levels contribute to the decreased feeding observed in association with experimental hypertrygliceridemia [5] and to central delivery of insulin [7], implying that ROS participate in the intracellular decoding of the information provided by both nutrients and hormones. Accordingly, exogenous ROS and ROS scavengers can directly impact food intake by modulating the neuronal activity of NPY/AgRP and POMC neurons [8], [9], [14].

Here we demonstrated that an acute icv administration of H2O2 decreased food intake and body weight in lean chow-fed C57BL/6J mice. These data agree with previous investigations showing same behavioral changes in response to central administration of H2O2 [9]. Work by Andrews et al. and Diano et al. also demonstrated that hypothalamic POMC neurons appear to have high oxidants levels, as indicated by intracellular DHE-associated signal, when these cells are activated to promote satiety [8], [9]. Increase in ROS levels would therefore result from mitochondrial activation due to full oxidation of glucose in POMC neurons during positive energy balance [8], [9].

Our work now establishes that ROS signaling requires mTORC1 activity specifically in POMC neurons to regulate food intake. This is relevant not only for the action of ROS (and specifically H2O2) on feeding, but also for the ability of leptin to increase oxidant levels in POMC neurons and in turn inhibit food intake.

Our findings in particular demonstrate that both acute pharmacological inhibition of the mTORC1 pathway with central delivery of rapamycin as well as whole-body deletion of S6K1 prevented the ability of H2O2 to decrease food intake. Finally, defective mTORC1 activity by deletion of Rptor in POMC neurons led to the inability of H2O2 to both induce accumulation of oxidants in these cells and inhibit food intake. Thus, these data suggest that mTORC1 activity is required in order for POMC cells to respond to ROS, and consequently decrease food intake. However, the exact molecular events connecting mTORC1 with ROS in POMC or other neuronal populations still need to be elucidated.

Several studies conducted in the fields of aging and cancer show that mTORC1 can affect both production and buffering of ROS [24]. mTORC1 can be purified in the mitochondrial fraction and increased mTORC1 activity increases mitochondrial oxygen consumption and oxidative metabolism, while its inhibition with rapamycin lowers mitochondrial ATP synthetic capacity while lowering mitochondrial ROS production [38], [39]. Accordingly, mTORC1 was shown to control mitochondrial oxidative function through a YY1 (yin-yang 1) - PGC-1α (peroxisome proliferator-activated receptor-γ coactivator-1α) transcriptional complex [40] and phosphorylation of the signal transducer and activator of transcription 3 (STAT3) [41]. Thus, mTORC1 activity may determine the relative balance between mitochondrial and non-mitochondrial sources of ATP [38]. Besides, our findings show that exogenous H2O2 was unable to increase oxidant levels in POMC neurons of POMC-rptor-KO mice. This piece of evidence suggests that deletion of rptor increases ROS scavenging in these cells. In support of this interpretation, several studies have shown that rapamycin decreases ROS levels in different cell types by increasing anti-ROS enzymes superoxide dismutase and catalase through stimulation of autophagy [39], [42], [43]. The latter is due to rapamycin-dependent inhibition of mTORC1 activity [42] and reduction of rptor protein [44]. Accordingly, in vivo deletion of rptor in endothelial and beta cells induces autophagy and decreases oxidative stress [45], [46].

Diano and colleagues demonstrated that ROS levels in hypothalamic POMC neurons of lean animals positively correlated with circulating levels of leptin [9], indicating that leptin may induce ROS formation in this neuronal population. We now demonstrate that exogenous leptin i) increases oxidant levels in POMC neurons, ii) exerts its appetite-suppressant action through ROS, as the ROS scavenger honokiol prevents the anorectic effect of leptin, and iii) engages mTORC1 in POMC neurons to modulate both oxidants production in POMC cells and food intake. Therefore, similarly to other signals of positive energy balance [4], [5], [7], leptin requires ROS production, a molecular response that necessitates activation of mTORC1 in POMC neurons to modulate food intake and body weight. The critical role of mTORC1 in POMC neurons in regulating leptin-dependent effects on food intake and body weight is in agreement with previous investigations showing that phosphatidylinositide 3-kinase (PI3K), which is upstream of mTORC1, is equally required in POMC neurons for the effects of leptin on energy balance [47], [48]. Of note, mice that lack S6K1 in POMC neurons are still responsive to leptin [22]. Thus, leptin would specifically need mTORC1 rather than S6K1 in POMC neurons to decrease food intake.

We have previously mentioned studies demonstrating how ROS affect both NPY/AgRP and POMC neuronal firing [8], [9], [14]. Similarly to what has been observed concerning the action of leptin on food intake, PI3K is required for the ability of leptin to increase firing of POMC neurons, while S6K1 is dispensable [22], [47]. However, POMC neurons lacking S6K1 have lower resting membrane potential and spike firing frequency than control POMC cells, implying a reduction in basal transmitter release [22]. Similarly, sustained overactivation of mTORC1 in POMC neurons, induced by genetic manipulation or old age, impairs cellular activity, leading to hyperpolarization and silencing of POMC neurons through increase in the activity of ATP-sensitive potassium channels [49]. Thus, mTORC1 signaling plays an important role in the regulation of POMC neuronal activity; however, further investigations are needed in order to understand the exact link with ROS in this context.

In conclusion, our study has identified mTORC1 as a key intracellular pathway involved in the central regulation of food intake by ROS. Considering the known interactions existing between mitochondrial function, ROS and the mTORC1 signaling in aging and cancer, it will be particularly relevant to further investigate this molecular intertwining and its role on the function of neuronal circuits regulating energy balance in obesity and type 2 diabetes.