Special Viewpoints

Diabetes, Obesity, and the Brain

See allHide authors and affiliations

Science  21 Jan 2005:
Vol. 307, Issue 5708, pp. 375-379
DOI: 10.1126/science.1104344


Recent evidence suggests a key role for the brain in the control of both body fat content and glucose metabolism. Neuronal systems that regulate energy intake, energy expenditure, and endogenous glucose production sense and respond to input from hormonal and nutrient-related signals that convey information regarding both body energy stores and current energy availability. In response to this input, adaptive changes occur that promote energy homeostasis and the maintenance of blood glucose levels in the normal range. Defects in this control system are implicated in the link between obesity and type 2 diabetes.

More than a century ago, the renowned physiologist Claude Bernard observed that diabetes could be induced in animals by puncture of the floor of the fourth cerebral ventricle (“piqure diabetique”) (1). Although this striking finding suggested a key role for the brain in glucose homeostasis, its importance was largely neglected after the discovery of insulin in 1923. However, new findings have revived interest in the role played by the brain in both glucose homeostasis and the mechanism linking obesity to type 2 diabetes. As Bernard might have predicted, this new information suggests that a full understanding of the pathogenesis of these disorders must incorporate a role for the brain in metabolic regulation.

Evidence now indicates that the brain processes information from “adiposity signals” such as the hormones insulin and leptin, which circulate in proportion to body fat mass, and integrates this input with signals from nutrients such as free fatty acids (FFAs). In response, feeding behavior, autonomic outflow, and substrate metabolism are adjusted in ways that promote homeostasis of both energy stores and fuel metabolism. The overarching hypothesis is that in times of plenty (ample fat stores and food availability), input to key brain areas from these afferent signals leads to inhibition of both energy intake and endogenous glucose production, while simultaneously increasing energy expenditure and mobilizing fat stores (Fig. 1) (2, 3). The net effect is that when the brain senses that body energy content and nutrient availability are sufficient, further increases of stored energy (in the form of fat) and circulating nutrients (such as glucose) are resisted.

Fig. 1.

Model depicting the control of energy homeostasis and hepatic glucose metabolism by adiposity- and nutrient-related signals. Neuronal systems sense and respond to input from hormones such as insulin and leptin that are secreted in proportion to body energy stores and from the metabolism of circulating nutrients (such as glucose and FFAs). In response to this input, adaptive changes occur in energy intake, energy expenditure, and hepatic glucose production.

Conversely, a decrease in neuronal input from one or more of these afferent signals is proposed to alert the brain to a current or pending deficiency of stored energy or nutrient availability. In turn, the brain activates responses that promote positive energy balance (increased food intake and decreased energy expenditure) and raise circulating nutrient levels (increased hepatic glucose production). As body fat content and plasma glucose levels begin to increase, circulating concentrations of leptin, insulin, and FFAs increase as well. The latter are sensed in the brain, favoring the return of food intake and glucose production to their original values. The central nervous system (CNS) response to these signals is therefore catabolic in nature and is in direct opposition to the anabolic actions of insulin and FFAs on fuel storage and metabolism in peripheral tissues. Should defects arise in the CNS response to these signals, the resulting imbalance in this homeostatic system will result in elevated levels of both body fat content and hepatic glucose production. Accordingly, reduced secretion of, sensing of, or responsiveness to afferent hormonal or nutrient-related signals can be predicted to cause weight gain and insulin resistance: cardinal features that link obesity with type 2 diabetes.

The Brain as an Insulin-Sensitive Tissue

In contrast to its prominent action in liver, muscle, and fat, insulin is not a major regulator of glucose use by the brain (4). This observation, combined with the widespread belief that a peptide the size of insulin would be unable to cross the blood-brain barrier, has led to the perception of the brain as an insulin-insensitive tissue. Recent observations have revealed this presumption to be erroneous and demonstrate that even though the brain is insulin-independent (with respect to glucose use), it clearly is not insulin-insensitive.

Evidence of insulin action in the brain emerged 25 years ago with the demonstration in a primate model that food intake decreases when a low dose of insulin is delivered directly to the brain by continuous intracerebroventricular (icv) infusion (5). When this fact was combined with evidence that insulin circulates at levels proportionate to body fat mass, that circulating insulin is transported into the brain, and that insulin receptors are concentrated in brain areas involved in the control of food intake and autonomic function (6), insulin emerged as a candidate “adiposity negative feedback” signal in the central control of energy homeostasis. When it was later shown that icv insulin administration also reduces hepatic glucose production (by increasing liver sensitivity to insulin) (7), insulin was hypothesized to be a central regulator of both energy homeostasis and glucose metabolism.

The hypothesis that brain insulin action is required for intact glucose homeostasis is supported by a series of recent findings. Studying mice genetically deficient in the insulin receptor, Accili and colleagues showed that selective expression of insulin receptors in only two tissues (liver and pancreas) substantially reduces diabetes severity and prolongs life-span (8). Even more surprising is that rescue of insulin receptor expression in the brain, in addition to the liver and pancreas, confers near-complete protection against hyperglycemia. In parallel studies undertaken in normal rats, chronic blockade of hypothalamic insulin receptor signaling was shown to cause hepatic insulin resistance and to increase hepatic glucose production (7, 9). Combined with evidence that mice with neuron-specific insulin receptor deletion are overweight, insulin-resistant, and glucose-intolerant (10), these data demonstrate that neuronal insulin signaling is required for intact control of both body fat mass and glucose homeostasis.

A potentially confounding aspect of these studies is that reduced neuronal insulin action increases food intake and body weight, and these effects may also cause insulin resistance. Nevertheless, available data suggest that central insulin action can regulate energy homeostasis and glucose metabolism via neuronal systems that are, at least in part, independent of one another (9).

In peripheral tissues, insulin signal transduction involves the insulin receptor substrate–phosphatidylinositol 3-OH kinase (IRS-PI3K) pathway (11). The idea that this cellular pathway is a key mediator of neuronal insulin action is supported by evidence that the actions of icv insulin on both food intake and hepatic glucose production are blocked by icv pretreatment with either of two PI3K inhibitors (7, 12). Because defective IRS-PI3K signaling has been implicated in the pathogenesis of insulin resistance in peripheral tissues (11, 13), these observations raise the unanswered question of whether neuronal insulin resistance might also occur by this mechanism. This possibility takes on heightened significance in view of evidence that IRS-PI3K signal transduction in hypothalamic neurons is also activated by the adipocyte hormone leptin (14), and that deletion of IRS2 from hypothalamic neurons results in obesity and insulin resistance (15, 16).

Leptin and Glucose Homeostasis

The neuronal response to leptin receptor activation involves the Janus kinase–signal transducer and activator of transcription (Jak-STAT) pathway (17). Among the proteins induced by leptin-mediated STAT signaling is suppressor of cytokine signaling-3 (SOCS3), which inhibits leptin activation of the Jak-STAT pathway (17). Interestingly, SOCS3 also potently inhibits signaling by insulin receptors (18), and sensitivity to both insulin and leptin is augmented in mice with reduced neuronal expression of SOCS3 (19, 20). That these mice are protected against diet-induced obesity suggests further that SOCS3-mediated attenuation of the neuronal response to adiposity signals is required for weight gain induced by consumption of a highly palatable, energy-rich diet. Combined with evidence that both leptin- and insulin-induced signaling involves the IRS-PI3K pathway, the cellular actions of these two adiposity-related hormones appear to overlap at multiple levels within neuronal systems that are important to both energy homeostasis and glucose metabolism (21).

Genetic leptin deficiency in ob/ob mice (22) is associated not only with pronounced hyperphagia and obesity but with insulin resistance and mild-to-moderate diabetes. Although impaired glucose metabolism in these mice is clearly driven by their severe obesity, leptin deficiency per se appears to make an independent contribution, because the glucose-lowering effect of leptin occurs at doses below those needed to reverse obesity (23) and cannot be reproduced by simple caloric restriction (24). In addition, the hyperglycemic consequences of impaired leptin signaling are dependent on coexistent defects in insulin secretion that are, at least in part, genetically determined. Background genes that influence endocrine pancreatic function are therefore important determinants of the predisposition to diabetes in genetic models of deficient leptin signaling. The lack of such genes may explain why children with genetic leptin deficiency are not reported to have diabetes, although they are severely obese (25).

A variety of gene defects have been identified that disrupt adipogenesis, causing a disorder known as lipodystrophy that is characterized by a loss of body adipose tissue, leptin deficiency, and a unique and severe form of insulin resistance and diabetes (26). Mouse models recapitulate the key features of the human disorder and have yielded substantial new insights into both its metabolic basis and its treatment (27). Because lipodystrophic mice are lean and, by definition, have reduced or absent fat mass, their body weight phenotype contrasts sharply with the severe obesity of ob/ob mice. Yet lipodystrophic and leptin-deficient mice share key features in common, including hyperphagia, insulin resistance, diabetes, and markedly reduced leptin signaling. Because the diabetes phenotype of lipodystrophic and leptin-deficient mice is ameliorated by leptin administration (28), a role for leptin deficiency in both diabetes syndromes is suggested (29). Moreover, icv administration of leptin at a low dose reversed the metabolic disturbance of lipodystrophic mice as effectively as systemic administration of a much higher dose (30), suggesting that the antidiabetic effect of leptin in this setting involves an action in the brain.

Lipodystrophic diabetes also develops in mice that express insulin receptors in the liver and pancreas but otherwise lack the insulin receptor gene (8). Surprisingly, this lipodystrophy is prevented by the expression of insulin receptors in the brain, in addition to the liver and pancreas (8). How might insulin action in the brain affect the predisposition to lipodystrophy? One possibility is suggested by the recently discovered mutations of the BSCL2 gene. Although the function of the protein encoded by this gene (termed “seipin”) remains to be determined, these mutations are responsible for one variant of the Berardinelli-Seip congenital lipodystrophy syndrome associated with mental retardation in humans (31, 32). This gene is highly expressed in the brain but only modestly in adipocytes (31), suggesting a role for the CNS in the pathogenesis of lipodystrophy in humans.

Hypothalamic Targets of Insulin and Leptin Action

The arcuate nucleus, situated adjacent to the floor of the third ventricle in the mediobasal hypothalamus, contains neurons that exert potent effects on food intake, energy expenditure, and glucose homeostasis and are regulated by input from both hormonal and nutrient-related signals (2, 33). “Anabolic” neurons coexpress neuropeptide Y (NPY) and Agouti-related peptide (AgRP), two peptides that potently stimulate food intake and reduce energy expenditure, and thereby promote weight gain (3436). These neurons are inhibited by leptin and insulin (33); consequently, reduced neuronal input from these hormones increases hypothalamic signaling by both peptides. Central administration of NPY causes insulin resistance and glucose intolerance, even when its effects on food intake are prevented (37, 38). Under conditions of reduced hypothalamic signaling by insulin and leptin, increased NPY signaling may therefore contribute not only to the resultant hyperphagia and weight gain but to systemic insulin resistance and glucose intolerance as well (Fig. 2).

Fig. 2.

Neurocentric model depicting sites where defects in the negative feedback regulation of energy balance and glucose production predispose to weight gain and insulin resistance. Defects in the secretion of insulin or leptin (1), in the hypothalamic sensing of adiposity- or nutrient related signals (2), or in the neuronal responsiveness to these inputs (3) predispose to both positive energy balance and increased glucose production. If sustained, these will result in pathological weight gain and insulin resistance.

By comparison, the anabolic effects of AgRP arise from antagonism of neuronal melanocortin receptors (MC3r and MC4r) (36, 39) that serve to limit food intake and body weight. Like the response to NPY administration, chronic blockade of central melanocortin receptors causes weight gain and insulin resistance (40), although the extent to which AgRP affects glucose metabolism independently of its effects on body fat mass awaits further study.

Melanocortins are peptides derived from posttranslational processing of the precursor proopiomelanocortin (POMC), and POMC neurons in the arcuate nucleus innervate the same hypothalamic areas supplied by fibers from NPY/AgRP neurons. Unlike NPY/AgRP neurons, however, POMC neurons are stimulated by input from insulin and leptin (41, 42), and the binding of melanocortins to MC3r and MC4r inhibits food intake and promotes weight loss (43). When neuronal input from leptin and insulin is reduced, therefore, POMC neurons are inhibited whereas NPY/AgRP neurons are activated, responses that in turn can cause hyperphagia, insulin resistance, and glucose intolerance. Although changes in autonomic outflow to the liver and other tissues have been proposed, the efferent mechanism whereby output from arcuate nucleus neurons is linked to the control of glucose metabolism in peripheral tissues remains uncertain. Moreover, the arcuate nucleus is by no means the only area of the brain that can process input from adiposity- and nutrient-related signals, and an important priority is to clarify the contribution of other brain areas to the control of energy homeostasis and glucose metabolism.

Nutrient Sensing and the Central Control of Glucose Homeostasis

In addition to processing input from leptin and insulin, hypothalamic neurons involved in peripheral glucose metabolism also sense and respond to intracellular signals that reflect ongoing cellular energy status. Like other cell types, neurons possess fuel-sensing mechanisms that not only ensure that cellular energy needs are met but, in specialized cells, influence firing rate, gene expression, or other cellular functions. The enzyme adenosine monophosphate (AMP)–activated protein kinase (AMPK) is activated in response to falling intracellular adenosine triphosphate levels (which are typically accompanied by rising AMP content) and is an example of such a cellular fuel sensor (44). In the arcuate nucleus, AMPK activation induces hyperphagia and weight gain resembling the response to reduced hypothalamic input from insulin and leptin (45). Further, arcuate nucleus AMPK activity is inhibited by both insulin and leptin, and the ability of leptin to reduce food intake appears to require this inhibition of AMPK (45). The extent to which this fuel-sensing mechanism mediates effects of adiposity-related hormones in the central control of glucose metabolism is an active area of study.

In addition to mechanisms for sensing the depletion of cellular energy reserves, the hypothalamus can also respond to an increase of nutrient availability. One example is the intracellular esterification of FFAs into long-chain fatty-acyl–CoA molecules (LCFACoA, derived from both exogenous FFAs and de novo FFA synthesis) (46, 47). Accumulation of LCFACoA in key neurons is hypothesized to constitute a cellular signal of plenty that increases when both glucose and FFAs are in abundant supply. In the hypothalamus, acute increases of FACoA levels can potently inhibit food intake while also increasing hepatic insulin sensitivity (3). This area of research was originally inspired by the observation of leptin- and insulin-like effects (reductions of food intake, hypothalamic NPY gene expression, and blood glucose levels) after administration of C75, an inhibitor of fatty acid synthase (48). This drug is predicted to increase intracellular FACoA content through the accumulation of malonyl-CoA, a molecule that inhibits FACoA oxidation. The extent to which the accumulation of FACoA [or of malonyl-CoA, also proposed to have signaling properties (49)] mediates the actions of C75 is unclear, because this compound also inhibits AMPK (50) and has neurotoxic effects (51).

The work of Rossetti and colleagues strongly supports a role for FACoA signaling in the central control of both food intake and glucose homeostasis. They found that icv infusion of oleic acid (a long-chain FFA) in rats potently reduces food intake, inhibits Npy gene expression, and increases hepatic insulin sensitivity (52) (Fig. 2). Similar responses were reported after icv infusion of an inhibitor of CPT-1 (3), the mitochondrial transfer protein that controls the rate of fatty acid oxidation. This intervention increased hypothalamic FACoA levels and again triggered behavioral, metabolic, and hypothalamic responses that resemble the central actions of insulin and leptin. Furthermore, central CPT-1 inhibition activated neurons in brainstem areas that control parasympathetic outflow and increased hepatic insulin sensitivity through a mechanism that is blocked by vagotomy (53). Thus, hypothalamic lipid sensing is proposed to regulate hepatic glucose metabolism via the activation of vagal efferent fibers that supply the liver (Fig. 2). Remaining challenges include efforts to clarify whether hypothalamic FACoA content is sensitive to changes in plasma FFA levels, how fluctuations of intracellular nutrient metabolism affect neuronal function, how they are integrated with input from insulin and leptin, and how this information is transduced into behavioral, autonomic, and metabolic responses.

Obesity, the Brain, and Type 2 Diabetes

Conventional wisdom links type 2 diabetes to obesity by virtue of the insulin resistance that arises from an excess of body fat. If pancreatic β cells cannot appropriately increase insulin secretion, glucose intolerance and ultimately frank hyperglycemia ensue. These same observations, however, are compatible with an alternative, neurocentric model (Fig. 3). This model is predicated on four key observations highlighted in this essay. First, the brain is not insulin-insensitive; on the contrary, it uses input from insulin, leptin, and nutrient-related signals to regulate both body fat content and hepatic insulin sensitivity (Fig. 2). Second, impaired neuronal signaling by these afferent signals causes hyperphagia, weight gain, and hepatic insulin resistance through mechanisms that are at least partly independent of one another. Third, obesity is strongly associated with biochemical resistance to both insulin and leptin. Fourth, defective insulin secretion (which presumably reduces insulin delivery to the brain as well as to other tissues) is a prerequisite for type 2 diabetes (54). Together these observations support a model in which reduced neuronal insulin and leptin signaling contributes to the link between excess body fat and disordered glucose metabolism.

Fig. 3.

Neurocentric model linking obesity to the pathogenesis of insulin resistance and type 2 diabetes. Reduced neuronal input from adiposity- or nutrient-related signals favors both positive energy balance and hepatic insulin resistance. As weight gain progresses to obesity, worsening insulin resistance increases the demand for insulin secretion. When combined with a β-cell defect (which reduces insulin action in the brain and periphery), type 2 diabetes ensues.

According to this model, obesity and impaired glucose metabolism can be predicted to arise from any of several defects that affect how the brain receives or processes input from key adiposity- or nutrient-related signals. Reduced insulin secretion can be invoked as a primary event, because reduced insulin delivery to the brain favors both weight gain and hepatic insulin resistance. As obesity progresses, a further deterioration of insulin sensitivity occurs (due to increased body fat content) that, together with impaired insulin secretion, will cause plasma glucose levels to increase. Although such increases may initially be limited by a compensatory increase of insulin secretion, the ability of the β cell to meet the demand posed by progressive weight gain and insulin resistance may ultimately reach its limit, leading to overt hyperglycemia. If obesity causes resistance to insulin in neurons as well as in peripheral tissues, a vicious cycle is created that accelerates weight gain and hepatic insulin resistance and thereby hastens diabetes onset (Fig. 3).

Defects in either the neuronal sensing of, or response to, afferent hormonal or nutrient-related signals can also set in motion a pathological cascade that progresses ultimately to obesity and diabetes. Because convergent signal transduction (for example, via the IRS-PI3K signaling pathway) and termination (for example, SOCS3) mechanisms mediate the neuronal actions of insulin and leptin, defects within a single biochemical pathway can potentially cause resistance to the central actions of both hormones (21). This in turn can be predicted to induce hyperphagia, weight gain, hepatic insulin resistance, and glucose intolerance. The feasibility of this concept is strengthened by evidence implicating impaired IRS-PI3K signal transduction in the insulin resistance of peripheral tissues in diabetic humans and animal models (13). When combined with a β-cell defect, a feed-forward mechanism is again set in motion whereby reduced insulin and leptin action in the brain and periphery initially favors weight gain and insulin resistance, progressing to glucose intolerance and ultimately diabetes. Because functional resistance to both leptin and insulin is common among the obese, this hypothesis warrants careful consideration (21) (Fig. 3).

Direct evidence in support of these predictions was recently provided from studies of mice in which IRS2 was selectively deleted from pancreatic β cells and hypothalamic neurons (15, 16). Because IRS-2 is necessary for β-cell survival, a gradual, progressive loss of β-cell mass occurs in these mice and predisposes them to diabetes. Because IRS-2 is also implicated in neuronal signaling by insulin and leptin, deletion of this protein from the hypothalamus impairs afferent input from the two known adiposity signals. Consequently, these animals develop obesity and insulin resistance that, combined with β-cell dysfunction, progress to glucose intolerance and finally to diabetes. Disruption of signaling via the IRS-PI3K pathway is therefore sufficient to cause obesity and diabetes, even when this defect is limited to only the brain and pancreas.

A neurocentric model to explain the link between obesity and diabetes also predicts that the risk of these disorders is strongly increased by environmental factors that favor weight gain (such as an abundance of highly palatable, energydense foods combined with a minimal requirement for physical activity). Accordingly, therapies that restore neuronal signaling by key afferent signals may prove beneficial for both obesity and diabetes, especially when combined with adjustments in diet and physical activity. Therefore, as Bernard anticipated in 1854, progress in understanding and treating diabetes will require an improved understanding of brain systems that control body fuel homeostasis and energy storage.

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article