Central Nervous System Regulation of Food Intake

Some 50 years ago, Gordon Kennedy introduced the hypothesis that body fat stores are subject to homeostatic regulation through a process in which afferent signals generated in proportion to body fat mass provide negative feedback to brain areas involved in the control of food intake (1). He further proposed that the central nervous system response to this input is fundamentally catabolic in nature, being characterized by decreased food intake, increased energy expenditure, and weight loss. Accordingly, weight gain caused by a period of excessive food consumption is hypothesized to increase the delivery of "adiposity negative feedback" signals to the brain. This response in turn induces a state of negative energy balance (i.e., energy intake less than energy expenditure) that persists until excess body fat is dissipated, at which point both food intake and body weight return to their original, preintervention levels. Conversely, the effect of energy restriction to reduce body fat stores is predicted to reduce adiposity-related negative feedback to the brain. This response triggers a compensatory increase in the drive to eat that, combined with decreased metabolic rate, favors the recovery of depleted fat stores when food becomes available. The importance of this model lies not only in the light that it sheds on the mechanisms underlying energy homeostasis in normal weight individuals but in the framework it provides for studying obesity as a disorder of a regulated system rather than as the consequence of a lack of restraint or will power. Another attractive feature of this model is that it lends itself to critical hypothesis testing. For example, it predicts that genetic or acquired defects in neuronal sensing or responsiveness to input from adiposity-related signals should be interpreted by the brain as a deficit of body energy stored in the form of fat. In response, hyperphagia, reduced metabolic rate, and pathological expansion of body fat mass should occur. Evidence in support of this prediction first emerged from studies using a technique known as "parabiosis." In this paradigm, two experimental animals are surgically joined to one another, allowing a shared circulation to develop. With this strategy, Coleman showed that food intake and body weight of genetically obese mice (ob/ob) decrease when they were parabiosed to lean controls and that this weight loss was even more pronounced when they were parabiosed to mice with a different monogenic form of obesity (db/db) (2). From these observations, he inferred that ob/ob mice are obese because they lack a key adiposity negative feedback signal, whereas db/db make this signal (and hence reduce the body weight of parabiosed ob/ob partners) but are obese because they cannot respond to it. Both predictions were realized some 20 years later with the positional cloning of the ob gene locus in 1994 (3) and the demonstration that it encodes the adipocyte hormone, leptin. In this landmark paper, ob/ob mice were also shown to be homozygous for a point mutation that results in a biologically inactive leptin molecule. Thus, severe hyperphagia and obesity in ob/ob mice was hypothesized to arise from leptin deficiency, and indeed, leptin administration was subsequently shown to reverse this obesity phenotype (4, 5). Two years later, the leptin receptor was cloned (6), and the obese phenotype of db/db mice was shown to arise from a mutation in the leptin receptor gene. Together, these findings showed that genetic deficiency of either leptin or its receptor is sufficient to induce a severe obesity phenotype in mice. Combined with evidence that leptin circulates at levels proportionate to body fat stores, that it enters the brain from the circulation, that leptin receptors are expressed in brain areas associated with control of food intake and autonomic function, and that leptin administration directly into the brain potently reduces food intake and body weight (7), Kennedy's model of energy homeostasis, once viewed with understandable skepticism, was rapidly and widely accepted. The intense interest in the biology of energy homeostasis sparked by these and many subsequent findings has yielded an increasingly sophisticated understanding of its underlying mechanisms. Additional hormonal and nutrient-related signals have been identified that convey important afferent information to the brain, and the specific subsets of neurons on which they act are beginning to be identified. Similarly, mechanisms linking the response of these neurons to changes of feeding behavior and autonomic nervous system function, and the relevance of this basic information to clinical obesity, are beginning to be appreciated. Especially important is evidence that genetic deficiency of either leptin (8) or its receptor (9) cause a severe obesity phenotype in humans reminiscent of that observed in mice and that leptin-deficient humans (like ob/ob mice) exhibit a dramatic normalization of this phenotype in response to leptin replacement therapy (10). However, common forms of obesity in both humans and animal models are associated with increased, rather than decreased, plasma leptin levels (11). Because food intake in such individuals is typically normal or elevated, common forms of obesity seem to be associated with leptin resistance. Indeed, leptin therapy is relatively ineffective in the treatment of common forms of human obesity (12), and many behavioral, autonomic, and biochemical responses to leptin are attenuated in rodent models of "diet-induced" obesity (13), in which genetically normal animals are rendered obese by consumption of a highly palatable, high-fat diet. These observations suggest that the pathogenesis of common forms of obesity involves a complex interaction between hedonic or "reward" qualities of food, environmental factors such as the type and quantity of food available and the amount of work required to attain it, and neuroendocrine systems designed to preserve homeostasis of body fat content in the face of these variables. To be considered an adiposity signal, candidate molecules should meet several criteria (7). Such molecules should circulate at levels proportionate to body fat mass, have access to target sites in the brain, and act in the brain to promote a state of negative energy balance. Furthermore, such molecules should activate an identifiable signal transduction mechanism that can explain their effects on food intake and body weight. Last, deficient neuronal signaling by such molecules should trigger an increase of food intake and body weight. To date, leptin and the pancreatic hormone insulin are the only molecules that meet each of these criteria. Thus, even though insulin (but not leptin) exerts potent anabolic effects in peripheral tissues, both hormones promote negative energy balance through their effects in the brain. Substantial progress has been made toward the identification of neuronal subsets that mediate the actions of these two hormones, the relevant intracellular signal transduction mechanisms, and the neurocircuitry linking these neuronal responses to changes of feeding behavior and energy metabolism. Perhaps the best-defined neuronal systems involved in the response to insulin and leptin are found in the hypothalamic arcuate nucleus (ARC).1 Among these are neurons that contain the melanocortin precursor polypeptide, proopiomelanocortin (POMC), and are stimulated by both leptin (14, 15) and insulin (16). On their release from nerve terminals, melanocortin peptides such as α-melanocyte stimulating hormone bind to and activate neuronal melanocortin-4 receptors (Mc4r) present on downstream neuronal targets (17). In response to activation of Mc4r, food intake decreases, whereas metabolic rate is raised, a combination of events that promote weight loss. Conversely, pharmacological or genetic disruption of Mc4r signaling causes hyperphagia and obesity (17, 18) and prevents or attenuates the weight-reducing actions of centrally administered insulin (16) or leptin (19). These observations collectively suggest that tonic melanocortin signaling is required both for an intact response to adiposity signals and to prevent pathological overeating and weight gain. The leptin→POMC→Mc4r pathway has, thus, emerged as a fundamental component of the energy homeostasis system. Adjacent to POMC neurons in the ARC are a unique subset of cells that coexpress two orexigenic molecules, neuropeptide-Y (NPY) and agouti-related peptide (AgRP) (20). Indeed, the ARC is the only place in the body in which NPY/AgRP neurons are found. Unlike POMC neurons, NPY/AgRP neurons are inhibited by both insulin (21, 22) and leptin (23), an effect that is also implicated in the ability of these two adiposity-related hormones to promote negative energy balance. Conversely, the effect of weight loss to lower plasma insulin and leptin levels, and thereby reduce neuronal input from these two hormones, is proposed to activate NPY/AgRP neurons (24, 25) and thereby stimulate food intake (7, 25). Hyperphagic feeding induced by NPY/AgRP neurons involves not only the synaptic release of NPY, which activates feeding-related NPY receptors (Y1 and Y5 receptor subtypes), but also the release of AgRP, an endogenous antagonist of Mc4r (26), which acts to inhibit neuronal melanocortin signaling. Moreover, NPY/AgRP neurons also produce the inhibitory neurotransmitter gamma-amino butyric acid and seem to constrain the activity of adjacent POMC neurons through tonic gamma-amino butyric acid release (15). Activation of NPY/AgRP neurons, therefore, promotes positive energy balance both by activating neuronal systems that promote hyperphagia and by inhibiting neuronal systems that cause anorexia. A fundamental concept that has emerged from these observations is that the homeostatic response to weight loss involves both activation of NPY/AgRP neurons and inhibition of POMC neurons (7). Using an anthropomorphic analogy, it is as if caloric restriction causes you to "step on the accelerator" where food intake is concerned, while at the same time "taking your foot off the brake." It is this combination of responses to reduced delivery of adiposity-related signals that is hypothesized to drive hyperphagia and reduced energy expenditure that ensure the recovery of lost weight once food becomes available. In addition to input from insulin and leptin, the ARC also senses changes in energy balance conveyed by the gastric hormone ghrelin (27) and the intestinal hormone, peptide YY3–36 (PYY3–36) (28). By activating its receptor on NPY/AgRP neurons, ghrelin stimulates food intake and is the only known circulating hormone to exert this effect (29). Because circulating ghrelin originates primarily from endocrine cells in the stomach lining (27) and ghrelin levels rise before meals and drop thereafter (30), this hormone has been implicated as a physiological mediator of meal initiation. The extent to which normal meal initiation depends on an increase of ghrelin signaling, however, remains uncertain, and additional studies are warranted to address this question. PYY3–36 is cleaved from full-length PYY1–36, a member of the NPY peptide family. Unlike ghrelin, which is released from the stomach when it is empty, PYY3–36 is secreted from the hindgut on food ingestion and, depending on the type of food consumed, relatively high plasma levels of this hormone are maintained during the intermeal interval (31). As the next meal approaches, declining levels of PYY3–36 are proposed to promote meal initiation, based on evidence that the circulating form of this peptide can reduce food intake. Because PYY3–36 is closely related to NPY and is a high-affinity ligand for NPY receptors, the notion that this hormone reduces food intake seems paradoxical. To explain this paradox, circulating PYY3–36 is proposed to selectively activate Y2R, an autoinhibitory receptor expressed on NPY/AgRP neurons (28). Based on the assumption that circulating PYY3–36 gains access preferentially to the ARC (where Y2R, but not feeding-related Y1 or Y5 receptors are concentrated), this hormone seems to reduce food intake by inhibiting orexigenic NPY/AgRP neurons (28). Because NPY/AgRP neurons act, in part, by inhibiting POMC neurons, the effect of PYY3–36 to reduce food intake is further proposed to involve activation of POMC neurons (28). Whereas several investigators report a reliable food intake—inhibitory effect of PYY3–36, the reproducibility of this response has been questioned (32). Furthermore, one recent report that did show PYY3–36-induced food intake reduction in normal mice also found a similar response in mice lacking Mc4r (33), suggesting that increased melanocortin signaling is not required for its anorexic effects (33), as was originally suggested. While important concerns exist about the role played by PYY3–36 in energy homeostasis and the mechanisms mediating its effects, clinical trials using this peptide are moving forward, and early data suggest that it reduces spontaneous food intake after intravenous administration to both normal weight and obese human subjects (28, 34). Several hypothalamic cell groups have been identified that contain neurons that are innervated by ARC neurons and are implicated downstream participants in the control of energy balance. The hypothalamic paraventricular nucleus (PVN), for example, contains several distinct neuronal subsets that can affect food intake and autonomic function and are contacted by axons projecting from both NPY/AgRP and POMC neurons (35). Among these are neurons that produce either oxytocin, corticotrophin-releasing hormone, or thyrotropin-releasing hormone. Each of these peptides acts to decrease food intake, increase metabolic rate, or both, and each cell type is reported to be activated by leptin (36, 37, 38) through either an indirect mechanism involving effects in the ARC or a direct action in the PVN (36). Combined with evidence that bilateral lesions of the PVN potently induce hyperphagia and obesity (39), this brain area seems to house neurons that promote negative energy balance and play an important role in energy homeostasis, at least in part, by conveying input from the ARC to other key brain areas. Unlike the PVN, the lateral hypothalamic area and adjacent perifornical area contain neurons that stimulate food intake and promote weight gain, and bilateral lesions of this area cause anorexia and weight loss (40). Among the key lateral hypothalamic area neurons involved in body weight regulation are those that express either orexin (41) or melanin-concentrating hormone (MCH) (42). MCH seems to play an especially important role, because central administration (42) or transgenic overexpression (43) of this peptide increase food intake, whereas targeted deletion of MCH (44) or its receptor (45) causes a weight-reduced, lean, hypermetabolic phenotype. Like the PVN, the lateral hypothalamic area also receives extensive innervation from ARC neurons and is established as a critical downstream component of energy homeostasis neurocircuitry. As in peripheral tissues (46), neuronal activation of the insulin receptor, a tyrosine kinase and member of the growth factor receptor family, induces the recruitment and tyrosine phosphorylation of insulin-receptor substrate (IRS) proteins (47). Among various members of the IRS protein family, IRS-2 is concentrated in the ARC and is strongly implicated in hypothalamic control of food intake (48). Tyrosine phosphorylated IRS proteins bind to the regulatory domain of a key enzyme known as phosphatidylinositol-3-OH-kinase (PI3K) (49). Activated PI3K, in turn, phosphorylates phosphatidylinositol-bisphosphate to generate phosphatidylinositol-trisphosphate, which leads indirectly to the activation of protein kinase B (also known as Akt) and other intracellular signal transduction cascades. In peripheral tissues such as liver, fat and muscle, activation of the IRS-PI3K pathway is crucial for insulin stimulation of glucose uptake and other metabolic responses (46, 49). Moreover, impaired signal transduction through the IRS-PI3K mechanism is implicated in the pathogenesis of insulin resistance in common metabolic disorders such as obesity and type 2 diabetes (49, 50). The hypothesis that neuronal insulin action requires signal transduction through the IRS-PI3K pathway received support initially from an electrophysiological study showing that the effect of insulin to hyperpolarize hypothalamic "glucose-responsive" neurons (neurons whose firing rate is increased by high glucose concentrations) was blocked by inhibitors of PI3K (51). A subsequent in vivo study performed by our group showed that intracerebroventricular infusion of either of two PI3K inhibitors effectively prevents insulin-induced anorexia (47). Combined with our finding that hypothalamic signaling through the IRS-PI3K pathway is induced after intracerebroventricular or systemic insulin treatment (47), these observations suggest a key role for this signal transduction mechanism in neuronal insulin action. Unlike the insulin receptor, the leptin receptor is a member of the class 1 cytokine receptor superfamily (6). Based on this discovery, leptin receptor—mediated cell signaling was hypothesized to involve the "Jak-Stat" transduction cascade, a pathway known to be activated by other members of this receptor family, and clear evidence of this followed shortly thereafter. Specifically, the binding of leptin to its receptor was shown to activate a member of the Janus Kinase tyrosine kinase family (JAK-2), that is anchored to the leptin receptor in the plasma membrane (52). Activated JAK-2 phosphorylates multiple intracellular substrates, including signal transducer and activator of transcription 3 (STAT3), a transcription factor implicated in the actions of several cytokines. Bates et al. (53) sought to determine the physiological role played by leptin signaling through the JAK-STAT pathway in the neuroendocrine control of energy homeostasis. They showed that "knock-in" of a mutant leptin receptor that cannot activate STAT-3 results in mice with a hyperphagic, obese phenotype comparable with that associated with global leptin receptor deficiency. In comparison, these mice exhibit intact reproductive function, increased growth, and are less diabetic than db/db mice that lack all leptin receptor function. Together, these observations indicate that STAT-3 is required for some, but not all, actions of leptin. This conclusion is compatible with in vitro evidence that leptin receptor activation elicits cellular responses that are far too rapid in onset and offset to be explained by a STAT3-mediated transcriptional mechanism (54) and that signal transduction pathways distinct from JAK-STAT are engaged by the leptin receptor. Based on extensive similarities in the central actions of insulin and leptin, the hypothesis that leptin might also signal through IRS-PI3K was raised, and several observations now support this hypothesis (55, 56). Furthermore, the ability of leptin to reduce food intake (like that of insulin) is attenuated by prior intracerebroventricular infusion of an inhibitor of PI3K (55). Thus, the two known adiposity-related hormones (insulin and leptin) seem to reduce food intake through a mechanism that requires intact PI3K signaling. Leptin-induced reductions of intake also require STAT3. Two recent papers have extended these findings by showing that obesity, insulin resistance, leptin resistance, and ultimately, diabetes result from deletion of IRS-2 from both hypothalamus and pancreatic β cells using a Cre-lox technique (48, 57). Collectively, these observations support a critical role for neuronal signaling through IRS-PI3K in the central control of energy homeostasis. Identifying the specific neurons responsible for these effects is a key priority for future studies. Because impaired IRS-PI3K signal transduction clearly contributes to the pathogenesis of common forms of insulin resistance in peripheral tissues (49, 50), the hypothesis can be entertained that impaired neuronal signaling through this pathway also occurs in some obese individuals (56). Should this prove to be the case, an attenuated central nervous system response to the two known adiposity signals should result in the defense of an elevated level of body weight. These considerations raise the interesting but as yet untested possibility, as suggested by the phenotype of mice in which IRS2 is deleted from hypothalamic neurons (48, 57), that reduced hypothalamic signaling through IRS-PI3K contributes to the pathogenesis of common forms of obesity. Although the hypothalamus is clearly a key brain area for processing afferent input involved in energy homeostasis, the mechanisms coupling this response to brain areas that control specific aspects of feeding behavior remains to be defined. hypothesis that a that areas that input from adiposity signals to areas such as the nucleus of the that to This concept is on the that during signals such as the peptide are released into the circulation and promote by activating neurons in the In the of and several other of gastric effects on neurons indirectly through activation of afferent neurons in the This response, in turn, a crucial role in meal and therefore, an important of amount of food during meals Several observations suggest that leptin administration both the and the response to administered The response to input from signals therefore, hypothesized to be regulated by input from adiposity-related but this effect involves hypothalamic input from adiposity signals or a direct action of leptin in the further to and are are important for an understanding of changes in body fat mass compensatory of energy intake on a In addition to hormonal generated in proportion to body the brain also senses and to nutrient-related signals from changes of intracellular energy content or in either the or of such as Some of these signals are generated in response to a decrease of substrate whereas arise from a of Among the are cellular responses by protein kinase a enzyme that is activated when cellular energy are depleted by an increase in the of to of cellular responses is by activation of this of which to increased substrate uptake and and thereby to raise intracellular In the activation of is hypothesized to be a signal of that leads to of food intake and body weight Moreover, both insulin and leptin seem to inhibit activity in this brain area Evidence that reductions of food intake require inhibition of the that adiposity-related signals promote negative energy in part, through inhibition of this In addition to as the brain glucose in a subset of neurons "glucose-responsive" and a signal that membrane and neuronal In neurons, the for this glucose effect the mechanism glucose stimulates insulin from pancreatic β cells glucose in these cells activates allowing and thereby the activation of by glucose in ARC neurons is attenuated by insulin and leptin through a mechanism (51). Both the underlying mechanism and extent to which these neurons to the actions of insulin and leptin in the neuroendocrine control of energy homeostasis to be of molecules is function that is implicated in the central nervous system control of energy homeostasis. These key in the intracellular and of fat are hypothesized to as a cellular of primarily when both glucose and are in Because several different that can be predicted to raise hypothalamic content intracerebroventricular infusion of either a acid or an inhibitor of to weight and inhibition of hypothalamic NPY gene an increase of levels of these molecules seems to activate responses that those of insulin and leptin. that the regulation of neuronal content and its to plasma levels and the intracellular mechanism a of content neuronal function are for a more understanding of this system in energy homeostasis. such as and are molecules that function as for either of two receptors and The hypothesis that these molecules in energy homeostasis originally from the that induced by contains the receptor is associated with increased food This effect is by administration of to experimental animals and, of receptors reduces food intake and thus, as a for obesity treatment in this area has on effects of on qualities of food, and the of this action might in energy homeostasis is studies suggest that receptor signaling leptin action in the hypothalamus and that this effect is required for the pathogenesis of forms of obesity. As in other brain both and receptors are present in the and leptin treatment hypothalamic levels, whereas that reduce leptin signaling have the effect Moreover, pharmacological or genetic deletion of receptors reduce body fat mass, increase leptin and both hyperphagia and obesity Collectively, these observations suggest that play a physiological role to leptin signaling and in this to weight gain. In this receptors can be as leptin and the defense of a reduced level of body to the specific neuronal subsets involved in inhibition of signaling and to the mechanism underlying neuronal between leptin and will to the role of this system in energy homeostasis. to the mechanisms underlying energy homeostasis have an for and studying targets for in the treatment of obesity and related metabolic Based on the of key molecules in this many of these targets are in to obesity As an example, the mechanisms underlying neuronal resistance to adiposity signals has clear that prevent or reverse this resistance can be predicted to the defense of a reduced level of body fat. one might to input from the hormone ghrelin or to PYY3–36 or other receptor of feeding-related NPY receptors (Y1 or or also have as of neuronal might also target neuronal systems that are from the including of MCH receptors, implicated as a for or treatment of obesity in rodent models These are a of the many with the to lower the level of body fat Based on the extensive in this neuroendocrine control however, it seems to that that target multiple receptor systems will be to induce and weight loss of body and that the of such will on of changes in and As the obesity an on human to the neuroendocrine for both normal and energy homeostasis are increasingly This information provides both a framework which to the pathogenesis of obesity and a for the of treatment for this This work was by and and by the and of the of