Hypothalamic Reactive Oxygen Species Are Required for Insulin-Induced Food Intake Inhibition: An NADPH Oxidase–Dependent Mechanism

All animal experiments were carried out in strict accordance with European Communities Council Directive 86/609/EEC and the Guide for the Care and Use of Laboratory Animals, edited by the National Research Council. For this study, we used 7- to 8-week-old male C57BL6/J mice (Harlan Laboratories). The animals were kept in a temperature-controlled room (22 ± 1°C) on a 12-h light/dark cycle with free access to food and water. Depending on the experiment being conducted, mice were fed ad libitum with standard diet (22.4% fat, 60.9% carbohydrate, and 16.7% protein, 2.9 kcal/g; R04T25; SAFE, Augy, France) or high-fat diet (42.5% fat, 42.5% carbohydrate, and 15% protein, 4.4 kcal/g; customized; SAFE). For the fasting group, food was removed just before dark onset (1900 h). For tissue collection, mice were killed by cervical dislocation.

Under anesthesia (0.4% isoflurane, 100% O₂), mice underwent stereotaxic surgery to implant a chronic stainless steel cannula (Charles River). The third cerebral ventricle was targeted using the following coordinates from bregma: anterior-posterior, –0.825 mm; dorsal-ventral, –5 mm; and medial-lateral, 0 mm. The cannula was fixed to the skull using dental cement. Mice were finally housed individually and were allowed 1 week for recovery before the experiment.

Injections into the third cerebral ventricle were performed in awake mice. The injections (1 μl) consisted of either 0.4 μU insulin (100 units/ml Actrapid recombinant human insulin; Novo Nordisk), 0.1 mmol/l Trolox (Calbiochem), 0.1 mmol/l diphenyleneiodonium (DPI; Sigma-Aldrich), or vehicle. Insulin, Trolox, and DPI were prepared as stock solutions in diluting medium for soluble insulin injection (Novo Nordisk), 1% ethanol, and 1% DMSO, respectively. For injections, all drugs and their respective vehicle were diluted in 0.9% NaCl. Injections were performed over 1 min. After the infusion, the guide cannula was kept in place for an additional 30 s to allow the drugs to diffuse away from the cannula tip.

Food was removed 1 h before intracerebroventricular injection. Insulin was injected 4 h before dark onset (1500). Intracerebroventricular injections of Trolox or DPI were performed 30 min prior to the 0.4 μU insulin injection. Just before the beginning of the dark period (1900), food (10 g) was presented to the mice. Then, 12 and 24 h later, food intake and body weight were measured.

Blood samples were collected from the tail 2 h after 4 μU insulin intracerebroventricular injection and used for measurements of blood glucose. It was directly measured on a glucose analyzer (Accu-Chek Active meter).

The dye 2′,7′-dichlorofluorescein diacetate (H₂DCFDA; Molecular Probes) was used to monitor intracellular change in H₂O₂ levels. After intracerebroventricular injections, hypothalami were dissected and immediately frozen in liquid nitrogen. After rapid thawing, tissues were homogenized (20 strokes) with a Dounce homogenizer and a B-type pestle in 250 μl of a ROS buffer (150 mmol/l KCl, 20 mmol/l Tris, 0.5 mmol/l EDTA, 1 mmol/l MgCl₂, 5 mmol/l glucose, and 0.5 mmol/l octanoïc acid, pH 7.4). Homogenates were exposed to 16 μmol/l H₂DCFDA and were incubated at 37°C for 30 min under agitation. Reaction was stopped with 125 μl of 70% ethanol plus 125 μl of 0.1 N HCl. Homogenates were then centrifuged at 3,000g for 15 min at 4°C. Supernatants were collected, neutralized with 175 μl of 1 mol/l NaHCO₃, and centrifuged at 6,000g for 15 min at 4°C. As previously described (9,10), ROS level was evaluated by measuring fluorescence intensity with a microplate reader (Victor; Wallac). Intensity of fluorescence was expressed as arbitrary units per microgram of proteins.

Dissected hypothalami were immediately frozen in liquid nitrogen. Tissues were homogenized (20 strokes) with a Dounce homogenizer and a B-type pestle in a lysis saline solution (150 mmol/l KCl, and 3 mmol/l EDTA, pH 7.4). Homogenates (50 μl) were mixed with 450 μl of 5% metaphosphoric acid. Samples were then centrifuged at 1,500g for 10 min at 4°C. Supernatants were collected and used to detect reduced glutathione (GSH). Glutathione assay was performed by reverse-phase high-performance liquid chromatography as previously described (21). The results are expressed as ficomoles per microgram tissular protein.

Protein concentration of samples was determined using the DC protein assay kit (Bio-Rad) according to the manufacturer's instructions.

Data are the means ± SE. Depending on the experiment being conducted, data were compared by unpaired Student's t test, the Mann-Whitney U test, or one-way ANOVA. Differences among groups were considered significant at P < 0.05.

To study the role of hypothalamic insulin signaling in food intake regulation, we first aimed to select the best paradigm of acute insulin injection (Fig. 1,A). We chose a third-ventricle injection because acute insulin administration into the lateral ventricle was without effect on food intake in mice (22). In comparison to vehicle-injected animals, insulin significantly reduced food intake 12 and 24 h after food presentation (Fig. 1,B). This represented a 40% decrease in daily food intake. Concurrently to this food intake inhibition, insulin also significantly reduced body weight at 12 and 24 h (Fig. 1,C). To ensure that intracerebroventricularly injected insulin did not indirectly affect food intake through diffusion out of the brain, we injected a higher dose of insulin and measured glycemia. No change in plasma glucose was observed in this condition, showing that the insulin effect on food intake only depends on its effect on the brain (Fig. 1 D).

To determine whether insulin could induce ROS increase in the hypothalamus, insulin was injected into the third ventricle at a dose that inhibits food intake. Mice were killed 5, 15, and 30 min later, and the ROS level was evaluated using the fluorescent redox-sensitive dye H₂DCFDA. Figure 2 A shows that insulin induced a significant increase (+36%) in the ROS level 15 min postinjection. This insulin-induced ROS increase was transient because no modification of fluorescence intensity was observed at either 5 or 30 min postinjection.

We then designed experiments to examine the physiological relevance of this insulin-induced hypothalamic ROS elevation. For this purpose, we assessed food intake in response to acute intracerebroventricular insulin injection in combination with an intracerebroventricular pretreatment of Trolox, a ROS scavenger. We first verified that Trolox alone did not induce significant change in basal ROS levels and food intake, and then we investigated the ability of Trolox to inhibit insulin-induced ROS production. As illustrated in Fig. 2,B, Trolox delivery 30 min prior to insulin completely prevented insulin-stimulated ROS increase. We then measured the effect of Trolox pretreatment on food intake response. As for the ROS level, intracerebroventricular Trolox injection completely suppressed insulin-induced food intake inhibition without significantly modifying basal food intake (Fig. 2 C). Therefore, these results indicate that the transient insulin-induced ROS elevation in the hypothalamus is required for food intake inhibition.

In peripheral tissues, ROS produced by membrane-bound NADPH oxidase are involved in the insulin signaling cascade. We questioned whether, as in the periphery, NADPH oxidase was involved in hypothalamic insulin ROS signaling to regulate food intake. For this purpose, we assessed food intake in response to acute intracerebroventricular insulin injection in combination with prior intracerebroventricular administration of DPI, a NADPH oxidase inhibitor. To test the ability of DPI to inhibit the ROS level increase, we first measured the hypothalamic ROS level 15 min after insulin injection with DPI pretreatment. DPI delivery was effective in fully preventing insulin-stimulated ROS elevation (Fig. 3,A). In this condition, DPI completely restored basal food intake in insulin-injected mice (Fig. 3 B). DPI treatment alone did not significantly modify basal ROS levels as well as food intake. Therefore, these results indicate that NADPH oxidase–dependent ROS increase within the hypothalamus is required for the central effect of insulin on food intake inhibition.

We then tested whether a modification of insulin effect on food intake was systematically associated with a change of ROS level within the hypothalamus. For this purpose, we chose two physiological situations that produce central insulin unresponsiveness: 18-h fasting (energy deficiency) (20) and 3 days' high-fat diet (energy excess) (7). Figure 4,A shows that these two conditions effectively suppressed insulin-induced food intake inhibition. According to our hypothesis, insulin did not promote elevation of ROS level within the hypothalamus 15 min after intracerebroventricular injection (Fig. 4 B).

Because cellular ROS levels result from the balance between their production and removal by various antioxidants, we quantified hypothalamic glutathione, which is considered as a major cellular ROS scavenging system (23). Figure 4 C shows that hypothalamic GSH levels were increased by 18-h fasting (+20%) and 3 days' high-fat diet (+36%, P < 0.05). Therefore, impairment of insulin-induced inhibition of food intake was associated with a lack of ROS increase in the hypothalamus, likely because of higher GSH level. These results strengthen the crucial role of hypothalamic ROS signaling in the anorexigenic effect of insulin.

In the current study, we describe the involvement of an NADPH oxidase–dependent ROS signaling pathway in the anorexigenic insulin effect. We demonstrate for the first time the physiological relevance of this concept in conscious mice.

Insulin was administrated into the third ventricle at a dose (2.4 nmol/l) that is consistent with brain insulin receptor affinity (24,25) and that produces maximal food intake inhibition in mice (2) and unpublished results). At this physiological concentration, insulin induced a transient ROS level elevation within the hypothalamus (Fig. 2). The current data demonstrate for the first time in vivo that insulin can effectively induce a change in ROS levels in the central nervous system. Despite a relatively low magnitude (+36%), this ROS level increase is similar to that observed in the hypothalamus after glucose, lipid, or ghrelin stimulation (9,–11). These subtle ROS changes play a key role in the control of the autonomous nervous system and food intake. We show that this insulin-induced ROS increase is required for food intake inhibition, thus demonstrating its physiological relevance. First, the suppression of insulin-stimulated ROS production using pharmacological tools prevented insulin inhibition of food intake. Second, in two nutritional conditions of central insulin unresponsiveness (i.e., 18-h fasting or 3 days' high-fat diet), impairment of the insulin effect on food intake was associated with a lack of insulin-stimulated ROS elevation within the hypothalamus. These data strongly support the notion that cellular ROS constitute a pivotal second messenger in hypothalamic signaling pathways involved in energy homeostasis regulation, especially by controlling nutrient sensing (9,10) and hormonal signaling (11).

The involvement of an oxidant signal in the action of insulin has been suggested for decades, mainly with the observation that oxidants, including hydrogen peroxide, mimic insulin's effects on glucose transport and lipogenesis in adipose cells (26). A few years later, insulin itself was shown to elicit the generation of hydrogen peroxide in adipocytes (27). More recently, several elegant studies carried out by Mahadev and colleagues (13,,–16) clearly demonstrate that a physiological transient insulin-induced ROS elevation mediates the increase of glucose uptake through GLUT4 translocation to the plasma membrane in adipocytes. Finally, creating mildly and transiently oxidative intracellular conditions has been reported to enhance insulin receptor activation, suggesting that optimal insulin responsiveness involved a process of “redox priming” of the β-subunit (17,28). Our results reinforce these previous data. Altogether, they strongly support the idea that a ROS increase induced by insulin 1) forms an integral part of insulin signaling pathway in the central nervous system as in the periphery and 2) is relevant in whole-body energy homeostasis.

The observed insulin-induced ROS increase peaks and dissipates in a few minutes. Nevertheless, and in agreement with the study by Brown et al. (2), feeding behavior is significantly inhibited for several hours after acute intracerebroventricular insulin injection (Fig. 1). In adipocytes, the early insulin-stimulated production of ROS is required for insulin to elicit its late cellular responses, especially by activating the insulin receptor substrate (IRS)/phosphoinositide (PI)-3 kinase signaling pathway (13,,–16,18,29). Within the hypothalamus, both insulin-induced ROS and IRS/PI 3-kinase signaling pathways are essential for the long-term inhibitory insulin effect on food intake (Fig. 2) (30). Reciprocally, diet-induced impairment of insulin effect on food intake is associated with an alteration of these two insulin-activated signaling pathways (Fig. 4 A and B) (7,31). The time course of insulin-induced ROS level increase is similar to that observed for insulin-induced PI 3-kinase activation in the hypothalamus, with activation a few minutes after stimulation, persisting for 10–15 min, and gradually declining over 30–60 min (32). Altogether, these data strongly suggest that the early insulin-induced ROS increase that we measured within the hypothalamus is a pivotal component of insulin-induced IRS/PI 3-kinase signaling pathway activation, leading to a long-term decrease in food intake, likely through modulation of the expression of genes such as neuropeptide Y (33). Actually, hypothalamic neuropeptide Y and pro-opio melanocortin C are well known to play a key role in food intake regulation and to be down- or upregulated by hormones such as leptin, ghrelin, or insulin (2,11,33,34). Because insulin-induced PI 3-kinase activation occurs in these neuronal subpopulations (32), one can speculate that the insulin-induced ROS increase occurs at least in one of these, as has been recently demonstrated for ghrelin (11).

That insulin induces a rapid and transient increase in ROS level within the hypothalamus led us to investigate the involvement of NADPH oxidase. Indeed, this enzyme can be rapidly activated by insulin, leading to a rapid and transient ROS increase in peripheral cells (14,15). In our experimental conditions and similarly to Mahadev et al. (15), pretreatment with a NADPH oxidase inhibitor (DPI) completely abolished 1) the rapid and transient insulin-induced ROS increase within the hypothalamus and 2) the inhibitory insulin effect on food intake (Fig. 3). We can thus conclude that NADPH oxidase plays a critical role in the insulin signaling pathway in the brain as in the periphery. There is now compelling evidence showing the presence of NADPH oxidase homologues in rodent or human brain tissue (35,,–38). Within the hypothalamus, immunoreactivity for NADPH oxidase subunits, such as p47^phox and gp91^phox, was found in arcuate, ventromedial, and paraventricular nuclei, which are well known to be involved in energy homeostasis regulation (36). Moreover, increasing data demonstrate the importance of redox signaling derived from NADPH oxidase in normal central nervous system processes, such as long-termpotentiation and hippocampal-dependent memory or the regulation of the cardiovascular system (38,,,–42). Further in vivo studies would permit us to confirm that an acute insulin-induced NADPH oxidase activation mediates other hypothalamic insulin effects involved in energy homeostasis.

After fasting or short-term high-fat feeding, the brain became less responsive to insulin, which resulted in a loss of its effects on food intake (Fig. 4,A) (7,20). In this context, insulin-induced hypothalamic ROS elevation was suppressed (Fig. 4,B). To explain this result, we explored the involvement of antioxidant defense systems. Indeed, the rise in free fatty acid availability during fasting (because of lipolysis) and short-term high-fat feeding (because of diet) likely induced an increase in hypothalamic lipid oxidation (9,11,43), a mechanism known to produce ROS (44). In the physiological range, excess ROS is scavenged at the cellular level by stimulating the synthesis of antioxidant defense systems (44), as already demonstrated within the hypothalamus of food-deprived animals (9,45). We reported that hypothalamic GSH level is effectively higher after high-fat feeding and to a lesser measure after fasting (Fig. 4 C), with a concomitant hypothalamic normal ROS level. These data strongly suggest that 1) high-fat diet induced excessive hypothalamic ROS production, which had been well regulated by the cellular scavenging system, and 2) this increased antioxidant defense system could consequently overquench ROS produced under insulin stimulation. Such a cellular mechanism has been already proposed to explain the impaired peripheral insulin sensitivity observed during physiological conditions such as pregnancy (46). In animal models of long-term high-fat diet–induced obesity (47,48) associated with hypothalamic insulin resistance (47), oxidative stress appears in the brain (48). This cellular stress has been proposed to be a critical mechanism underlying development of insulin resistance in obesity (49,50). We thus hypothesized that 1) the early diet-induced central insulin resistance is a consequence of an adaptive mechanism to avoid cytotoxic ROS production, although 2) in the long-term, the resistance might be caused by an overwhelming of antioxidant defense systems, and thus oxidative stress (44).

In conclusion, we report that insulin injected into the third ventricle promoted a subtle and transient increase in ROS level within the hypothalamus. This mechanism is required for insulin food intake inhibition and involves NADPH oxidase (Fig. 5). We thus suggest that 1) an NADPH oxidase–dependent ROS-sensitive signaling pathway is implied in hypothalamic insulin action, and 2) diet-induced impairment of hypothalamic ROS homeostasis may contribute to cerebral insulin resistance.

This work was funded in part by Agence Nationale de la Recherche Grant ANR-05-PNRA-004 and a grant from the Institut Benjamin Delessert. T.J. received a fellowship from the Ministère de l'Enseignement Supérieur et de la Recherche and from the Fondation pour la Recherche Medicale.

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

We are indebted to M. Nibbelink for her technical help. We thank M. Rigoulet for helpful discussions and insightful comments.