Brain Adipocytokine Action and Metabolic Regulation

Two types of adipose tissue, namely brown adipose tissue and white adipose tissue (WAT), are recognized (1). Brown adipose tissue is involved in heat production through nonshivering thermogenesis, a process mediated by uncoupling protein 1, located in the inner mitochondrial membrane (1). The predominant type of adipose tissue, WAT, is composed of unilocular adipocytes filled mainly with triacylglycerol and embedded in a loose connective tissue meshwork containing adipocyte precursors, fibroblasts, and immune and various cells. WAT has an abundant vascular and nervous supply and is located mainly in the subcutaneous region and around the viscera. The stored triacylglycerols in WAT provide long-term fuel reserve for the organism as a whole (2). An increase in the levels of nutrients and insulin stimulates triacylglycerol synthesis in liver and storage in WAT (2). Conversely, insulin falls during fasting, and epinephrine, glucocorticoids, and growth hormone increase, resulting in lipolysis and release of fatty acids that undergo partial oxidation in muscle and liver (2). Ketones generated from this process serve as alternate fuels for use by the brain and peripheral organs.

The worldwide increase in incidence of obesity has focused attention on the biology of WAT (3). Obesity is characterized not only by excessive WAT mass, but also by an increase in fatty acid flux and deposition of triacylglycerol and lipid metabolites in liver, muscle, pancreatic islets, and other ectopic sites (4). This condition known as “steatosis” has been linked to insulin resistance, diabetes, and organ dysfunction in obesity and aging (4). WAT in obese individuals also manifests histological and biochemical changes characteristic of inflammation (5–7). Activated macrophages in obese WAT produce cytokines, e.g., tumor necrosis factor-α and interleukin-6 (5). C-reactive protein, intracellular adhesion molecule 1, platelet-endothelial cell adhesion molecule 1, monocyte chemoattractant protein 1, and coagulation factors (e.g., plasminogen activator inhibitor 1) secreted by obese WAT have been linked to cardiovascular diseases (5–7) (Table 1).

WAT stromal and adipocytes produce enzymes that control the biosynthesis and activities of steroid hormones (8–10) (Table 1). WAT-derived aromatase catalyzes the interconversion of androstenedione to estrone and testosterone to estradiol (8). 17β Hydroxysteroid dehydrogenase converts weak sex steroids to their more potent counterparts, i.e., androstenedione to testosterone and estrone to estradiol (8). The ratio of 17β hydroxysteroid dehydrogenase to aromatase increases in obesity and has been associated with insulin resistance and hyperlipidemia in menopausal women (9). The oxidoreductase, 11β hydroxysteroid dehydrogenase type 1, mediates the conversion of cortisone to cortisol in humans and 11-dehydrocorticosterone to corticosterone in mice (10). Excess local production of active glucocorticoids has been implicated in central obesity, elevated glucose, and lipid levels and cardiovascular morbidity (10–15).

The existence of a factor secreted in proportion to energy stores in WAT, which acts in the brain to control feeding, weight and WAT mass, was first proposed by Kennedy (16) and is supported by the discovery of monogenic mutations resulting in obesity, as well as classic cross-circulation (parabiosis) experiments in rodents (17–20). The list of adipocytokines known to affect metabolism keeps growing (Table 1). This review will focus on the role of leptin as an adipocytokine primarily involved in energy homeostasis. We will also review the role of adiponectin, the most abundant adipocytokine that regulates lipid and glucose metabolism. Finally, we will discuss the biology of resistin.

The “obese” locus was first described 6 decades ago and was later shown by positional cloning to be the lep gene that encodes a secreted protein “leptin” (21). Mice and humans homozygous for leptin gene mutation (Lep^ob/ob) develop a ravenous appetite, early-onset obesity, severe insulin resistance, steatosis, hypothalamic hypogonadism, deficits of the thyroid and growth hormone axes, and immunosuppression (21–24). Leptin is expressed mainly by WAT adipocytes, though low levels are produced in the stomach, mammary gland, placenta, and skeletal muscle (3). Leptin has a relative weight of 16 kDa and circulates as free and bound forms, the former representing the bioavailable hormone. The concentrations of leptin in WAT and plasma correlate positively with WAT mass, adipocyte size, and triacylglycerol content, but the precise signals mediating the regulation of leptin synthesis and secretion are unknown. Leptin is higher in obesity and in females than males even after adjusting for body mass. This sexual dimorphism is due in part to higher production by subcutaneous WAT in females, inhibition by androgens, and stimulation by estrogens (3). Insulin, glucocorticoids, and cytokines, e.g., tumor necrosis factor-α and interleukin-6, increase leptin, whereas cold exposure and adrenergic stimulation decrease leptin (25).

Leptin has a diurnal rhythm, peaking at night in humans and morning in rodents (25). A pulsatile leptin rhythm occurs in humans and primates, but the underlying mechanisms and functional significance are unclear (25). Fasting decreases leptin levels within hours in parallel with glucose and insulin (25,26) (Fig. 1A and B). Conversely, leptin increases several hours after feeding (25,26). In contrast, adiponectin is increased by fasting (Fig. 1B and C). The nutritional regulation of leptin is likely to involve insulin and not glucose, as revealed by an increase in leptin under hyperinsulinemic clamp conditions (Fig. 1D–F; 27,28). Adiponectin, on the other hand, is reduced but not significantly by high insulin or glucose levels (Fig. 1G).

The leptin receptor belongs to the class 1 cytokine receptor family (33). At least five leptin receptor isoforms, LRa–LRe, derived from alternate splicing of lepr mRNA have been described (33). LRa, the major “short leptin receptor,” lacks the cytoplasmic domain required for JAK-STAT signaling (3). LRa is abundant in brain capillary endothelium, neurons, and peripheral tissues and has been proposed to be involved in leptin transport (3). The “long leptin receptor,” LRb, which mediates intracellular leptin signaling, is enriched in neurons in the hypothalamus and brainstem and controls feeding, metabolism and neuroendocrine function (3). Leptin enters the brain through a saturable transport system, binds to LRb, which then associates with JAK2, resulting in autophosphorylation of JAK2, phosphorylation of tyrosine residues 985 and 1138 on LRb, and activation of STAT3 (34) (Fig. 2). This cascade of events leads to translocation of STAT3 into the nucleus and transcription regulation of neuropeptides and various leptin target genes (34) (Fig. 2). Leptin terminates its own action through phosphorylation of Tyr985 and induction of suppressor of cytokine signaling 3 (SOCS3) (30) (Fig. 2). Protein-tyrosine phosphatase 1B, a well-known inhibitor of insulin action, also terminates leptin signaling through inactivation of JAK2 (35). Leptin acting through LRb has also been demonstrated to regulate insulin receptor substrate 1 and 2, mitogen-activated protein kinase, extracellular-regulated kinase, Akt, and phosphatidylinositol 3-kinase in the hypothalamus, raising the possibility of cross-talk between leptin and insulin (36).

Neuropeptide targets of leptin are classified into “orexigenic peptides,” which promote feeding and weight gain (i.e., “orexigenic”), e.g., neuropeptide Y (NPY), agouti-related peptide (AgRP), melanin concentrating hormone (MCH), and orexins, and “anorexigenic peptides,” which decrease feeding and weight, e.g., proopiomelanocortin (POMC), cocaine- and amphetamine-regulated transcript (CART), corticotropin-releasing hormone, and thyrotropin-releasing hormone (37) (Fig. 3). In the arcuate nucleus, NPY and AgRP and POMC and CART are expressed in distinct neuronal populations that project to the paraventricular nucleus and lateral hypothalamus and perifornical areas to control feeding, energy expenditure, glucose and lipid metabolism, and hormonal secretion (Fig. 3). α-Melanocyte stimulating hormone (derived from POMC) inhibits feeding and stimulates thermogenesis through activation of melanocortin 4 (MC4) receptor (Fig. 3). AgRP, which is expressed in the same arcuate neurons as NPY, is an antagonist of α-melanocyte stimulating hormone (Fig. 3). Leptin reduces feeding and weight by directly suppressing NPY and AgRP and increasing α-melanocyte stimulating hormone and CART (Fig. 3). MCH and orexins are indirectly suppressed by leptin, whereas corticotropin-releasing hormone and thyrotropin-releasing hormone are increased (Fig. 3). As predicted, lesions of the arcuate nucleus and lack of LRb and STAT3 in neurons result in obesity (38–40). The significance of LRb in POMC neurons has also been demonstrated in mice that became obese when LRb was deleted from POMC neurons (41). Furthermore, the loss of NPY and MCH attenuates obesity in leptin-deficient mice (42,43). In contrast, leptin sensitivity is enhanced in SOCS3 haploinsufficiency and neuron-specific SOCS3 ablation, leading to reduction in food intake, resistance to obesity, and decreased glucose and lipid levels (44,45).

Leptin also affects neurotransmission, neuropeptide secretion, and neuronal plasticity. Leptin inhibits NPY secretion by the hypothalamus, depolarizes POMC neurons by decreasing the inhibitory tone of γ-amino butyric acid released from NPY terminals in the arcuate nucleus, and hyperpolarizes and inactivates NPY neurons (46,47). The rapid fall in leptin during fasting depolarizes NPY and AgRP neurons similar to congenital leptin deficiency, and this may underlie hyperphagia (48). We have previously reported that congenital leptin deficiency decreases brain weight, impairs myelination, and reduces several neuronal and glial proteins (49). These deficits are partially reversible in adult Lep^ob/ob mice by leptin (49). Similarly, daily subcutaneous injections of recombinant methionyl human leptin reversed deficits in gray matter in the anterior cingulate gyrus, the inferior parietal lobule, and the cerebellum in patients with congenital leptin deficiency within 6 months, and these changes persisted over 18 months (50). Leptin enhances the development of axonal projections from the arcuate nucleus to paraventricular nucleus in neonatal mice (51). Furthermore, the anorectic action of leptin is related to increases in inhibitory synapses and diminution of excitatory synapses in the hypothalamus (52). The signaling mechanisms underlying these diverse leptin actions are unknown.

Leptin was initially proposed as a hormone whose primary role was to prevent obesity by inhibiting appetite (3,21). This idea was logical, since rodents and patients lacking leptin or functional leptin receptors develop hyperphagia and obesity (21–23). However, leptin is elevated in the vast majority of obese animals and humans with no obvious leptin receptor abnormalities, yet these individuals fail to respond to high endogenous leptin levels (3). As will be discussed later, “leptin resistance” in obesity involves deficits in leptin signal transduction, associated with increased lipid build-up in muscle, liver, and various tissues.

Based on similarities between leptin-deficient (Lep^ob/ob) and fasted mice (such as hyperphagia; reduction in energy expenditure; thyroid, reproductive, and growth hormones; and immunosuppression), we hypothesized that leptin functioned primarily as a “starvation hormone” (26,53). This idea was first tested in rodents, in which leptin replacement prevented the fasting-induced changes in neuroendocrine, metabolic, and immune function (24,26). Subsequent studies confirmed that congenital leptin deficiency, lipodystrophy, and caloric restriction in humans resulted in hypogonadism and reduction in thyroid hormone, reversible by leptin replacement (22,23,54,55). Furthermore, leptin replacement prevents the fall in energy expenditure in patients subjected to chronic weight reduction, and reverses steatosis, insulin resistance, diabetes, hyperlipidemia, and hypothalamic hypogonadism in lipodystrophy, supporting a major role of low leptin level in metabolic regulation (56–58). Leptin deficiency is associated with elevation of NPY, AgRP, MCH, and orexins in the hypothalamus and reduced levels of POMC and CART (3). Thyrotropin-releasing hormone and corticotropin-releasing hormone expression is decreased in the paraventricular nucleus (3). These changes are reversed by peripheral and especially direct central nervous system injection of leptin (3).

It is possible that leptin’s role as a starvation signal conferred survival advantage during famine by limiting thyroid-mediated thermogenesis and the high energy cost of reproduction and promoting feeding and energy storage. This idea is consistent with the increase in adiposity in heterozygous patients and mice with partial leptin deficiency (59–61). An increase in energy efficiency mediated by low leptin prolongs longevity in Lep^ob/+ mice (60). Studies have suggested that low leptin may precede adiposity in primates and some indigenous human populations, but these results have not been confirmed by others (62–66).

Leptin has been proposed to play a major role in liporegulation in normal healthy individuals (4) (Fig. 4A and B). When energy intake is equal to expenditure, WAT mass remains constant and the lean tissues contain little or no fat (Fig. 4A) (4). Leptin acts directly on muscle and liver as well as indirectly through the sympathetic nervous system to increase the phosphorylation and activity of a critical energy sensor, AMP-activated protein kinase (AMPK) (67) (Fig. 4A). Activated AMPK phosphorylates acetyl-CoA carboxylase (ACC) and malonyl-CoA decarboxylase, resulting in inhibition of ACC and activation of malonyl-CoA decarboxylase. Normally, ACC catalyzes the formation of malonyl-CoA, which is the first committed step in fatty acid synthesis. AMPK reduces malonyl-CoA and thus limits lipogenesis. Malonyl-CoA inhibits carnitine palmityl transferase 1 (CPT-1), which mediates the transport of fatty acids into mitochondria to undergo oxidation. By inhibiting ACC and reducing malonyl-CoA, AMPK increases carnitine palmityl transferase 1 activity and fatty acid oxidation. Obesity is associated with high leptin level, which induces leptin resistance partly through SOCS3 induction. SOCS3 inhibits leptin signaling in the brain as well as peripheral tissues (3,4). Leptin resistance decreases AMPK activity and stimulates lipogenic enzymes—most notably ACC, fatty acid synthase, and stearoyl-CoA desaturase 1 (Fig. 4B). The latter catalyzes the synthesis of monounsaturated fatty acids (mainly oleate and palmitoleate) (68). Malonyl-CoA inhibits carnitine palmityl transferase 1 activity, reducing fatty acid oxidation. The net effect is increased fatty acid influx, steatosis, and formation of ceramide and various metabolites that impair the functions of skeletal and cardiac muscle, liver, and pancreatic islets (4) (Fig. 4B). Leptin exerts its anti-obesity and insulin-sensitizing effects partly through inhibition of stearoyl-CoA desaturase 1, which acts upstream of AMPK (68).

Other factors implicated in leptin resistance in the brain include reduction in leptin transport across the blood-brain barrier, induction of protein tyrosine phosphatase 1B activity, and dysregulation of neuropeptides (3,35). Collectively, these abnormalities increase appetite and weight, albeit to a lesser degree than congenital leptin deficiency (4).

There is increased interest in leptin’s role in glucose homeostasis (69). Leptin decreases glucose before weight loss in Lep^ob/ob mice (49; Fig. 5A). In this model, intracerebroventricular leptin administration suppresses hepatic glucose production (HGP) within 6 h (Fig. 5B). Leptin infusion for 48 h suppresses feeding and decreases weight and glucose in Lep^ob/ob mice (Fig. 5A). The reduction in glucose in pair-fed mice is due to reduction in HGP and an increase in the glucose disappearance rate (R_d) (Fig. 5B). Leptin treatment results in greater HGP suppression and increase in R_d compared with pair-feeding (Fig. 5B), confirming independent effects of central leptin treatment on weight and glucose.

We have examined the effect of intracerebroventricular leptin on glucose fluxes in wild-type C57Bl/6J mice (Fig. 6). Infusion of a dose of leptin (4 ng/h for 24 h) that did not decrease body weight increased the glucose infusion rate and suppressed HGP by 50%, but R_d was unchanged (Fig. 6). This result supports an early action of leptin on hepatic glucose metabolism. In rat, intracerebroventricular leptin infusion stimulates gluconeogenesis but does not affect glucose production, as a result of a compensatory decrease in glycogenolysis (70). Pharmacological blockade of melanocortin prevents leptin’s ability to stimulate gluconeogenesis; however, inhibition of glucose production and glycogenolysis is independent of melanocortin signaling (70). Short-term voluntary overfeeding induces resistance to the effects of systemic insulin and leptin on liver glucose metabolism (71). Leptin administered intracerebroventricularly restores insulin sensitivity by inhibiting glucose production mainly by decreasing glycogenolysis (70). Together, these studies establish critical roles of leptin and MC4 receptor in glucose regulation that could be harnessed for treatment of diabetes.

Adiponectin is abundantly secreted by WAT adipocytes (72). The primary structure of adiponectin consists of an NH₂-terminal signal sequence, a variable domain, a collagen-like tail domain, and COOH-terminal globular head domain (72). Adiponectin shares strong sequence homology with C1q and types VIII and X collagen, and the globular domain resembles tumor necrosis factor-α. Unlike leptin and other polypeptide hormones, which circulate at picograms or nanograms per milliliter, adiponectin circulates at very high levels (micrograms per milliliter). Native adiponectin exists as homotrimers that form low-molecular-weight hexamers and high-molecular-weight complexes. High-molecular-weight adiponectin is increased by thiazolidinediones and thought to mediate the biological activity of adiponectin (30,31,72).

In contrast to leptin, adiponectin is decreased in obesity, is inversely related to glucose and insulin, and increases during fasting (72) (Fig. 1B and C). Adiponectin deficiency results in insulin resistance, glucose intolerance, dyslipidemia, and increased susceptibility to vascular injury and atherosclerosis (31,73,74). Adiponectin reverses these abnormalities by increasing fatty acid oxidation, suppressing gluconeogenesis, and inhibiting monocyte adhesion, macrophage transformation, proliferation, and migration of smooth muscle cells in blood vessels (30,31,72–74). These actions of adiponectin are associated with AMPK activation and modulation of inflammatory signals, in particular nuclear factor κB (72).

Putative adiponectin receptors (AdipoR1 and R2) containing seven-transmembrane domains, but structurally and functionally distinct from G protein–coupled receptors, have been identified (72). AdipoR1 and R2 are widely expressed in the brain and peripheral tissues and are reported to bind adiponectin, activate AMP kinase, and inhibit ACC in liver, muscle, and blood vessels (72). We and others have found that AdipoR1 and R2 are highly expressed in the paraventricular nucleus, amygdala, and area postrema and are diffusely localized in the periventricular areas and cortex (R.S.A., A. Ferguson, T. Kadowaki, P. Sanna, unpublished data). Other investigators have demonstrated binding of adiponectin to T-cadherin but not AdipoR1 and R2 and have proposed that T-cadherin affects the bioavailability of adiponectin (H. Lodish, unpublished data; 75).

Peripheral adiponectin treatment decreases body fat by enhancing energy expenditure and fatty acid oxidation (76). Chronic adiponectin treatment reduces food intake, weight, glucose, and lipids in obese rats (77). Moreover, adiponectin and leptin are inversely related to seasonal changes in WAT mass and adipocyte lipid content in mammalian hibernators (78). Thus, we hypothesized that adiponectin may act centrally to regulate metabolism (29). In agreement, the full-length adiponectin, globular form, and a mutant protein unable to form hexamers increased brown adipose tissue thermogenesis, enhanced lipid oxidation, and lowered glucose after intracerebroventricular injection (29). Lep^ob/ob mice, a model in which adiponectin is reduced, were highly sensitive to central and systemic adiponectin treatment (29). Adiponectin potentiated the effect of leptin on thermogenesis and fatty acid oxidation, and both adipocyte hormones induced Fos protein immunostaining in the paraventricular nucleus and increased brown adipose tissue uncoupling protein 1 expression, suggesting activation of hypothalamic sympathetic circuits (29). Importantly, agouti (A^y/a) mice that are incapable of melanocortin signaling failed to respond to leptin or adiponectin, implying an overlap in central neuronal targets (29).

We have confirmed that adiponectin knockout mice (ADPko) bred on C57Bl/6J background develop insulin resistance, manifested by a decrease in glucose infusion rate and an increase in HGP (Fig. 7A and B). Adiponectin deficiency does not seem to affect R_d (Fig. 7C). Intracerebroventricular injection of mammalian adiponectin transiently decreases glucose, triglycerides, and nonesterified fatty acid (NEFA) and increases ketones within 4 h (Fig. 7D–G). These results support a central action of adiponectin. Because adiponectin is increased during fasting, we assessed whether ADPko mice would respond abnormally to fasting and refeeding (26). In wild-type mice, plasma glucose, insulin, triglyceride, and fatty acid levels fell and ketones rose during fasting and were restored within 48 h after refeeding (data not shown). Although basal glucose and triglycerides were slightly lower in ADPko mice, they responded appropriately to fasting and refeeding (data not shown), indicating that adiponectin is not critical to acute changes in energy balance.

Whether adiponectin enters the brain is controversial (79,80). Iodinated globular adiponectin does not cross the blood-brain barrier in mice (79). Nonetheless, murine cerebral microvessels express AdipoR1 and R2, which are upregulated during fasting (79). Furthermore, globular adiponectin inhibited interleukin-6 release from brain endothelial cells, providing a potential mechanism of action (79). The significance of these findings is uncertain, since very little, if any, globular adiponectin circulates in mammals (P.E.S., H. Lodish, personal communication). Adiponectin, in particular the trimeric form, has been demonstrated in human cerebrospinal fluid using gel filtration chromatography (P.E.S., unpublished data). Moreover, adiponectin protects human neuroblastoma SH-SY5Y cells from apoptosis induced by the mitochondrial complex I inhibitor, 1-methyl-4-phenylpyridinium, suggesting a direct action on neurons (81). It is possible adiponectin enters the brain via the circumventricular organs, e.g., area postrema, median eminence, and subfornical organ, located outside the blood-brain barrier.

Resistin belongs to a family of cysteine-rich COOH-terminal domain proteins called resistin-like molecules (RELMs). Resistin is expressed and secreted by WAT adipocytes in rodents and was named for its ability to induce insulin resistance (82). Multimeric complexes of resistin and RELMβ have been identified (83). Each promoter consists of a COOH-terminal disulfide-rich β-sandwich head and an NH₂-terminal α-helical tail, and the latter associates to form three-stranded coils, linked by interchain disulfide linkages to form tail-to-tail hexamers. Resistin circulates as hexamers and trimers. There appears to be a discrepancy between resistin mRNA expression and protein levels in the circulation, such that plasma levels are increased in obesity while mRNA levels in WAT are reduced (32). As with leptin, resistin falls during fasting and increases during refeeding (32). These changes are controlled, at least in part, by insulin and glucose (32).

As predicted, systemic treatment or transgenic overexpression of resistin decreases insulin’s ability to suppress hepatic glucose output, and this is associated with induction of SOCS3 (84,85). Conversely, ablation of the retn gene or reduction in resistin protein through antisense oligonucleotide treatment improves insulin sensitivity through AMPK activation (86,87). Resistin inhibits adipogenesis, whereas the loss of resistin function increases body weight and fat and enhances insulin sensitivity (88,89). Thus, resistin has significant roles in energy and glucose homeostasis. We have found that loss of resistin in leptin-deficient mice exacerbates obesity by further decreasing energy expenditure, but insulin sensitivity is enhanced (R.S.A., unpublished data). Recent reports have described inhibition of food intake and induction of hepatic insulin resistance by intracerebroventricular resistin administration (90,91).

In contrast to rodents, resistin is secreted by mononuclear cells and activated macrophages in humans (92). Resistin single-nucleotide polymorphisms have been linked to obesity and lipid and glucose abnormalities in some studies (93). Resistin appears to be elevated in WAT and serum in obesity and insulin resistance, although other studies have failed to establish such a relationship (94). Resistin has been associated with increased risk of inflammation and atherosclerosis in humans (94,95).

This review emphasizes the roles of adipocytokines in the control of energy homeostasis and various metabolic processes. So far, most of the evidence is derived from genetic and pharmacological approaches in rodents. In the case of leptin and adiponectin, the main actions of these adipocytokines on energy balance and glucose and lipid metabolism are similar between rodents and humans. In contrast, the roles of resistin, visfatin, retinol binding protein 4, and various adipocytokines are yet to be clarified (95–97). More than a decade after being discovered, the precise mechanisms regulating secretion of leptin and adiponectin are unclear. Furthermore, their transporters and signaling pathways that mediate diverse actions in various tissues have yet to be fully ascertained. The biology of adiponectin is further complicated by complex forms. Understanding these processes will provide a framework for studying other adipocytokines and offer insight into the pathophysiology of obesity, diabetes, and related metabolic diseases.

Adiponectin has been shown to increase blood pressure without affecting heart rate following microinjection in the area postrema (AP) of rats. Cells in the AP were either hyperpolarized or depolarized by adiponectin, thus proving a possible mechanism for the central regulation of cardiovascular function (98).

This work was supported by National Institutes of Health Grants RO1-DK62348, PO1-DK49210, and P30-DK19525 (to R.S.A.), and R01-DK55758, R24-DK071030, and R03-EY014935 (to P.E.S.).