CNS leptin and insulin action in the control of energy homeostasis

Authors

  • Bengt F. Belgardt,

    1. Department of Mouse Genetics and Metabolism, Institute for Genetics, Center for Molecular Medicine, Cologne Excellence Cluster on Cellular Stress Responses in Aging-Associated Diseases, Second Department for Internal Medicine University of Cologne, and Max Planck Institute for the Biology of Ageing, Cologne, Germany
    Search for more papers by this author
  • Jens C. Brüning

    1. Department of Mouse Genetics and Metabolism, Institute for Genetics, Center for Molecular Medicine, Cologne Excellence Cluster on Cellular Stress Responses in Aging-Associated Diseases, Second Department for Internal Medicine University of Cologne, and Max Planck Institute for the Biology of Ageing, Cologne, Germany
    Search for more papers by this author

Address for correspondence: Jens C. Brüning, M.D., Institute for Genetics and Center for Molecular Medicine (CMMC), Department of Mouse Genetics and Metabolism, Zülpicher Str. 47a, 50674 Cologne, Germany. jens.bruening@uni-koeln.de

Abstract

The obesity and diabetes pandemics have made it an urgent necessity to define the central nervous system (CNS) pathways controlling body weight, energy expenditure, and fuel metabolism. The pancreatic hormone insulin and the adipose tissue–derived leptin are known to act on diverse neuronal circuits in the CNS to maintain body weight and metabolism in a variety of species, including humans. Because these homeostatic circuits are disrupted during the development of obesity, the pathomechanisms leading to CNS leptin and insulin resistance are a focal point of research. In this review, we summarize the recent findings concerning the mechanisms and novel neuronal mediators of both insulin and leptin action in the CNS.

Leptin and insulin as messengers of peripheral energy levels to the CNS

The circulating levels of leptin and insulin are positively correlated with adiposity and body weight,1 and are now broadly accepted to deliver information on peripheral energy stores to the central nervous system (CNS) by acting on diverse neuron populations. In line with this notion, intracerebroventricular (i.c.v.) injection of insulin or intranasal application of insulin, which selectively mirrors CNS insulin concentration, decreases food intake and body weight in mice,2 rats,3 baboons,4 and men,5 although a recent report failed to detect an effect on food intake and body weight in rats.6 Comparably stronger than insulin's effect, leptin's ability to reduce food intake and decrease body weight is well established.7–9 It has been ascertained that insulin and leptin action in the CNS, and here especially on neurons, is essential for decreasing the food intake (anorexigenic) and eventually weight reducing effects of these two hormones,10, 11 although notably, the role of other cell types present in the CNS such as astrocytes or microglia, which do express insulin receptors (IR) and leptin receptors (LEPR),12, 13 is still poorly understood in this regard. In addition to insulin's critical roles in glucose and lipid metabolism in the periphery, leptin has also direct effects on peripheral tissues, and the interested reader is directed to a recent review on this topic.14

Hypothalamic mediators of insulin and leptin action

After the discovery that leptin and insulin mediate their effects on body weight and fuel metabolism by acting on neurons, the specific neuronal population impacted on by both hormones had to be established. It was already known that in rodents, lesions in the hypothalamus could impact body weight, either inducing weight gain or weight loss, depending on the specific region of the hypothalamus ablated. In line with this, both leptin and IR are strongly and broadly expressed in the hypothalamus, insulin action in the hypothalamus has been demonstrated to induce anorexia and weight loss,15 whereas inhibition of insulin signaling has the opposite effect.16 Similarly, hypothalamic signaling is necessary for leptin's effects on body weight.17, 18

The first breakthrough in defining primary target neurons of leptin and insulin occured in 2001, when it was demonstrated that proopiomelanocortin (POMC)-expressing neurons are depolarized by leptin treatment, which leads to an increase in neuronal activity.19 In the hypothalamic melanocortin system, POMC-expressing neurons release the POMC cleavage product α-melanocyte stimulating hormone (α-MSH), which acts on downstream target neurons (some of which located in the paraventricular hypothalamus, Figure 1) to reduce food intake, increase energy expenditure, and regulate glucose metabolism.20–22 In fact, null mutations in leptin, LEPR, POMC, its cleavage enzymes (prohormone convertases), or downstream receptors (such as Melanocortin receptor 4, MC4R) relate to hyperphagia and obesity both in rodents and humans.23 In contrast to POMC neurons, activation of arcuate nucleus (ARC) neurons coexpressing agouti-related protein (AGRP), and neuropeptide Y (NPY) induces feeding, reduce energy expenditure, and impacts on activity.24–26 AGRP acts as an inverse agonist on the MC4R, whereas NPY regulates neuronal activity by acting on several NPY receptors subtypes, ultimately blocking α-MSH-mediated anorexia. Of note, AGRP/NPY neurons also directly inhibit POMC neurons through synaptic release of inhibitory γ-aminobutric acid (GABA) on POMC soma.27 The critical role of these two types of neurons has been demonstrated by acute ablation models, which lead to hyperphagia (POMC neuron ablation) or hypophagia (AGRP neuron ablation), and following these initial experiments, it was revealed that AGRP neuron-released GABA in the hindbrain parabrachial nucleus is necessary for feeding.25, 26, 28

Figure 1.

Circuitry of leptin and insulin responsive neurons. Neuron populations of the arcuate nucleus (ARC), ventromedial hypothalamus (VMH), paraventricular hypothalamus (PVH), lateral hypothalamus (LH), ventral tegmental area (VTA), raphe nucleus (RAPHE), and nucleus tractus solitarius (NTS)—which are responsive to leptin, insulin, or both—are depicted, as well as relevant synaptic contact between these neuron populations. POMC and AGRP neurons in the ARC are first-order neurons of both leptin and insulin. Whereas γ-aminobutric acid (GABA) released from AGRP neurons inhibits POMC neuron activation directly, POMC and AGRP neurons release α-MSH and AGRP respectively onto PVH neurons, which express thyroid-releasing hormone (TRH) or corticotrophin-releasing hormone (CRH), which are involved in thermogenesis and feeding, respectively. In the VMH, steroidogenic factor 1 (SF1) expressing neurons are bona fide leptin-sensitive neurons, and predominantly regulate food intake. In the lateral hypothalamus, leptin acts on GABAergic neurons, which are in synaptic contact with dopaminergic neurons in the VTA. Both leptin and insulin also regulate activity of orexin neurons in the LH through hyperpolarization (leptin) or transcriptional repression of orexin (insulin). Dopaminergic neurons in the VTA, which release dopamine in the striatum to control activity and reward, are also directly silenced by leptin signaling, whereas dopamine reuptake is under control of insulin through transcriptional control of the dopamine transporter. Although serotonergic neurons in the raphe nucleus release serotonin (5-HT) onto POMC neurons in the ARC (and striatal neurons), leptin acts directly onto 5-HT neurons as well. Finally, nucleus tractus solitarius (NTS) neurons relay and compute signals from the gastrointestinal (GI) tract and exchange information with the parabrachial nucleus (PBN). Additionally, GABA release in the PBN from AGRP neurons is essential for feeding. Note that some synaptic connections and target regions are not depicted for optimal clarity. Region descriptions are in bold and underlined, neuron descriptions are in bold, and neurotransmitters/neuropeptides are written in normal formatting. Leptin (L) or insulin (I) responsiveness is depicted in bold and italic.

With the advent of the Cre/loxP system, neuronal population-specific ablation of the IR or LEPR or their signaling components had been made possible, and mice with POMC or AGRP/NPY-specific ablation of these two receptors have been analyzed. Ablation of the leptin receptor in POMC, AGRP/NPY or both neurons led to elevated body weight and mild obesity, again underlining the role of the melanocortin pathway.29, 30 On the other hand, it was deducted from these experiments that there are other neuronal populations (and possibly extra-hypothalamic sites) that are critically targeted by leptin to inhibit feeding.

The ablation of IR from POMC and AGRP neurons surprisingly did not lead to hyperphagia or obesity, although IR signaling in AGRP neurons has an important role in systemic glucose metabolism (see below).31 This was unexpected, because insulin's anorexigenic action is at least partially dependent on an intact melanocortin system,32 and insulin clearly regulates electrical activity and transcriptional events in POMC neurons, such as localization of forkhead transcription factors.33, 34 Several recent reports have further shed light on this notion. Both insulin and leptin were known to influence membrane polarization and firing only in a subset of neurons, in the range of 50% of all POMC neurons tested, and in an opposite manner; whereas leptin depolarized POMC neurons,19 insulin clearly hyperpolarized POMC neurons.33, 35 This conundrum was elegantly resolved by the finding that these two hormones impact on different subpopulations of POMC neurons, that is there are neurons sensitive to leptin, whereas other neurons respond to insulin.36 Moreover, combined leptin and insulin action in POMC neurons has synergistic effects on glucose homeostasis (and fertility), because mice lacking both IR and LEPR on POMC neurons show drastic impairment of steady state glucose metabolism.37 Finally, it appears that in addition to the existence of subpopulations of POMC neurons with regard to receptor expression and hormone sensitivity, there are neurons which only shortly during development express POMC, some of which go on to express NPY during adulthood.38 Nonetheless, in these neurons loxP-flanked genes of interest have already been recombined due to the short-term activation of the POMC-Cre transgene, which may compromise the analysis of the resulting phenotype. Taken together, the efforts to unanimously identify the hypothalamic neurons necessary for insulin-induced anorexia have proven fruitless so far,31 with the primary leptin targets being better understood.

If hypothalamic insulin action is necessary and of consequence for food intake and body weight defense, of what nature are those neurons? First, there are several subhypothalamic nuclei, in which analysis of IR ablation has not been reported yet, for example the ventromedial and the paraventricular nuclei, which are insulin responsive and whose function is absolutely required for normal energy homeostasis.39–41 Intriguingly, leptin signaling in ventromedial hypothalamic (VMH) or lateral hypothalamic (LH) neurons is of equal importance with regards to body weight control as is leptin signaling in POMC neurons.42, 43 The neuropeptide orexin is expressed in the LH, and it is crucial for wakefulness, because mutations in orexin or its receptor are implicated in narcolepsy in mice and men.44 In the LH, leptin hyperpolarizes orexin-containing neurons, which is in line with the notion that satiety allows for periods of rest and sleep.45 Similarly, a report found that insulin, through nuclear export of the transcription factor FOXA, decreases orexin expression,46 whereas orexin application silences POMC neurons.47 Besides these targets, there are likely more hypothalamic neuron populations important for weight regulation, but there are no marker genes available yet to study them in detail.

In contrast, the role of hypothalamic insulin signaling in control of peripheral glucose metabolism is well established,48, 49 where electrical inhibition of neurons expressing AGRP has been identified as a major component of hepatic glucose production regulation likely through vagal innervations.31 Leptin's well-known ability to improve systemic glucose metabolism has also been shown to depend on hypothalamic circuits.50, 51 Indeed, there is now conclusive evidence that leptin signaling in POMC neurons is predominantly necessary for regulation of systemic glucose homeostasis. This notion was deducted from the finding that reexpression of the LEPR only in POMC neurons, whereas all other cells lack the LEPR, was sufficient to mostly normalize glucose homeostasis (but not weight), whereas db/db mice (lacking LEPR globally) suffer from early onset uncontrolled diabetes.52

Notably, it has not yet been proven that CNS insulin and leptin action play the same role in humans, because patients undergoing liver transplantation and therefore hepatic deenervation show only modest changes in glucose production and metabolism, although these findings are hard to qualify due to the pathologies leading to the liver transplantation itself, immunosuppression therapy, and other endocrine abnormalities detected after the transplantation.53 Moreover, one group demonstrated that in dogs, CNS insulin action appears to have only a small effect on glucose metabolism, underlying the need for further studies in nonrodent species.54

In addition to glucose metabolism, both CNS insulin and leptin action has been demonstrated to regulate lipid uptake and/or metabolism in the white adipose tissue, highlighting the broad potency of these two hormones to coordinately orchestrate fuel partitioning.55–57

Leptin and insulin intracellular signaling cascades

Because leptin and insulin have profound effects on transcriptional and electrical events in neurons, the signaling events evoked by these hormones are of high interest (Figure 2). Leptin activates intracellular signaling cascades through the recruitment of the Janus kinase (JAK) to the LEPR, where it phosphorylates several key residues on the LEPR. Signal transducer and activator of transcription (STAT) 3 proteins binds to the phosphorylated LEPR, and are themselves phosphorylated by JAKs.58 This allows for dimerisation, and subsequent translocalization into the nucleus, where the STAT3 proteins bind to and regulate transcription of target genes.24 The role of this pathway especially in POMC transcription is well defined because leptin treatment increases POMC transcription.59 On the other hand, leptin is able to stimulate the phosphatidylinositol-3-kinase (PI3K), which is also insulin's main intracellular signaling cascade.60 Here, insulin will stimulate binding of the regulatory subunit of the PI3K to phosphorylated insulin receptor substrates (IRS), which allows for activation of the catalytic subunit of the PI3K. PI3K catalyzes the phosphorylation of the membrane lipid phosphatidylinositol-4,5-bisphosphate (PIP2) and thus generates phosphatidylinositol-3,4,5-trisphosphate (PIP3). PIP3 can bind to and activate ion channels, but is also recognized by phosphatidylinositol-dependent kinase 1, which phosphorylates several proteins such as the kinase AKT to elicit downstream signaling events.61 Although it is widely accepted that leptin activates PI3K signaling at least in specific neurons, the signaling cascade linking leptin stimulation to PI3K activation has not yet been fully resolved. Leptin action may also reduce the degradation of PIP3 to PIP2 by phosphorylation and thus deactivation of the PIP3 dephosphatase phosphatase and tensin homologue (PTEN).62 On the other hand, the adapter protein SH2B1 has been shown to recruit JAK and IRS proteins in a supercomplex, thus allowing crosstalk between these two pathways.63 Importantly, SH2B1 has also been linked to human obesity.64 Leptin-induced PI3K signalling in the hypothalamus has been linked to peripheral glucose homeostasis and food intake,60, 65 although the neuron populations mediating both effects are not completely elucidated.

Figure 2.

Intracellular signaling cascades activated by insulin and leptin. Insulin and leptin are able to stimulate activity of the phosphatidylinositol-3-kinase (PI3K), which subsequently results in phosphorylation and nuclear exclusion of the forkhead transcription factors FOXO1 and FOXA2. FOXO1 is a negative regulator of carboxypeptidase E (CPE) expression in POMC neurons, whereas it stimulates transcription of AGRP in AGRP neurons. Leptin-activated STAT3 binds to POMC and AGRP promoters, where it stimulates (POMC) or inhibits (AGRP) expression via recruitment of histone modifying enzymes. FOXA2 was demonstrated to bind to the orexin and MCH promoters, where it stimulates expression of these neuropeptides. Insulin stimulation is known to increase expression of the dopamine transporter (DAT) gene in dopaminergic neurons. Both leptin and insulin are able to hyperpolarize neurons by PI3K-mediated opening of ATP-dependent potassium channels (KATP channels) and subsequent potassium outflow. On the other hand, leptin is able to depolarize neurons and thus increase the firing rate by opening a nonspecific cation channel by a PI3K and JAK-dependent pathway, and pharmacological manipulation has led to the conclusion that this channel may be the transient receptor potential (TRP) channel. Note that for clarity, several other pathways have been omitted (see text).

Another molecular target of both leptin and insulin is the AMP-dependent kinase (AMPK). Low cellular energy levels will increase the AMP/ATP ratio, which is sensed by AMPK and converted into a cellular response to induce ATP generation and reduce ATP consumption. In the hypothalamus, AMPK activation increases food intake, and both leptin and insulin have been shown to decrease the phosphorylation and thus activation of AMPK in the hypothalamus,66 although it is unresolved how exactly activation of AMPK is blocked by both hormones. AMPK action in POMC and AGRP neurons plays a role in the neuron's response to ambient levels of glucose, whereas leptin's and insulin's effect on neuron firing is not affected.67 Besides these pathways, leptin (and insulin) are able to induce mitogen-activated protein kinases (MAPK) such as extracellular signal-regulated kinase (ERK). Although ERK signaling mediates some of the effects of leptin on food intake in the hypothalamus,68 it is unknown if insulin uses this pathway to maintain body weight.

Transcription, membrane potential, and synaptology

As true regulators of neuronal activity, leptin and insulin change membrane potential of target neurons to control firing rate and thus neuropeptide and neurotransmitter release. Both leptin's and insulin's effect on POMC neuron firing has been extensively studied, with leptin depolarizing and insulin hyperpolarizing a subset of POMC neurons. With regards to leptin, the cation-channel opened by leptin stimulation has been elusive for some years, with recent reports implicating both leptin-stimulated JAK and especially PI3K signaling in opening of transient receptor potential channels in POMC neurons, and it will be highly informative to see if this holds true also for other leptin-stimulated neurons, such as VMH neurons.69, 70 On the other hand, both insulin and leptin have been shown to be able to activate ATP-dependent potassium (KATP) channels, which leads to potassium outflow, hyperpolarization and a reduction of the firing rate.71, 72 This has been well demonstrated for AGRP neurons.31, 73, 74 Mechanistically, it has been proposed that PI3K activation leads to local accumulation of PIP3, which binds to KATP channels, increasing the probability for an “open” channel, and reducing the affinity for ATP.75 On the other hand, PIP3 generation alone may not be sufficient, because actin filament stabilization prevents insulin-stimulated KATP channel opening, whereas the mechanisms induced by insulin (and leptin) to control actin filament dynamics are poorly understood.74 As the name suggests, KATP channels are sensitive to cellular levels of ATP, that is, they are closed by intracellular rises in ATP. Glucose-sensitive neurons (such as POMC neurons) are able to sense a rise in ambient glucose concentrations, because the increase in cytoplasmic ATP in response to glycolysis closes the KATP channels, which eventually results in depolarization and an increase in firing rate.76 PI3K activation and insulin stimulation have also been reported to depolarize AGRP neurons, although the channels involved have not been identified, underscoring the diversity of findings regarding electrophysiological responses to insulin and leptin, as recently discussed.67, 76, 77 Taken together, depending on the neuronal population, acute insulin and leptin application may depolarize or hyperpolarize target neurons, effects which may be accounted for by differential expression of their receptors, that of ion channels or intracellular signaling intermediates determining the net result.

Both leptin and insulin directly control transcription of target genes, including neuropeptides. As mentioned before, leptin-activated STAT3 signaling controls POMC transcription in POMC neurons.59 In the same vein, leptin and insulin stimulation leads to phosphorylation and nuclear exclusion of the forkhead transcription factor FOXO1, allowing for STAT3 binding to the promoter and transcription of POMC.33, 34, 78, 79 FOXO1 is also a negative regulator of carboxypeptidase E (CPE) expression, which is important for a distal step in processing POMC into its cleavage products, i.e., α-MSH.80

Regarding expression of orexigenic neuropeptides, FOXO1 and STAT3 again compete for binding to the promoters of AGRP and NPY, with FOXO1 being an activator of transcription of these orexigenic neuropetides, and STAT3 being inhibitory.78, 79 Indeed, leptin's ability to reduce AGRP/NPY expression depends on PI3K signaling,81 and insulin stimulation excludes FOXO1 from the nucleus of AGRP neurons.34 Because mice with constitutive STAT3 activation only in AGRP neurons are hyperactive and lean, STAT3 signaling in these neurons surprisingly regulates locomotor activity, although the downstream neurons mediating this phenotype are unknown.24 Interestingly, mice with constitutive STAT3 signaling in POMC neurons are hyperphagic and mildly obese, due to suppressor of cytokine signaling (SOCS) 3 overexpression.82 SOCS3 inhibits activation of the leptin signaling cascade at the level of the receptor, and SOCS3 expression is under control of STAT3 signaling, thus constituting a negative feedback loop. Because leptin levels are chronically elevated in obesity, SOCS3 levels are increased in the hypothalamus of obese mice,83 and ablation of SOCS3 in the brain or POMC neurons ameliorates high fat diet- (HFD)induced obesity.84, 85 Nonetheless, hyperleptinemia alone does not induce SOCS3-mediated leptin resistance and consequently obesity, because leptin-transgenic mice remain leptin sensitive.86 Interestingly though, hyperleptinemia predisposed leptin-transgenic mice to obesity when challenged with a HFD. SOCS3 is also an inhibitor of insulin signaling through degradation of IRS proteins, thus (at least in POMC neurons) hyperleptinemia or STAT3 overactivation will concurrently lead to cellular insulin resistance.82, 87

Leptin also controls expression of multiple neuropeptides in neurons downstream of POMC and AGRP/NPY neurons, for example in the paraventricular neurons (PVN). Here, leptin has been shown to increase expression of thyroid releasing hormone (TRH), which is a positive regulator of energy expenditure.88 Moreover, leptin stimulation affects transcription of many more genes, nonetheless mice with leptin receptor deficiency only in the PVN have not been generated yet, thus the direct and indirect targets (through neuropeptide/neurotransmitters released by upstream neurons) are currently indistinguishable.89

Besides direct transcriptional control and membrane potential regulation, leptin (and purportedly insulin) signaling is able to change the synaptic input onto neurons. For example, mice lacking leptin show decreased numbers of glutamatergic (=excitatory) synapses on POMC neurons and increased glutamatergic input on NPY neurons, both of which is rapidly, that is in hours, normalized upon leptin treatment.90 Other hormones, for example estrogen, a known anorexigenic hormone, or ghrelin, which stimulates food intake, also regulate synaptic input,90, 91 thus synaptic rewiring may be an important level of regulation used by many different hormones involved in energy homeostasis. Nonetheless, it is still unresolved by which mechanism and which attractant the synapses are recruited or repelled, and, if this is due to signaling on the presynaptic or postsynaptic neurons. Although a role for insulin in synaptic plasticity in neurons directly linked to energy homeostasis has not been reported, postsynaptic insulin action in hippocampal neurons is known to recruit GABA receptors and thus sensitize cells for this inhibitory neurotransmitter,92 which at least opens up the question if insulin plays a similar role in hypothalamic neurons.

Extrahypothalamic neurons targeted by insulin and leptin

Clear evidence that hypothalamic neurons do not contribute all of the weight-regulating effects of both leptin and insulin has opened the search for other nuclei expressing LEPR and IR, and analysis of their importance with regards to energy homeostasis. All of these nuclei and neuropeptide circuits had been previously shown to control body weight regulation, and it is now obvious that insulin and leptin act on almost all levels of feeding, including food recognition, food liking, and meal initiation.

Dopamine

The generation of mice lacking dopamine in the brain led to the discovery that these mice show reduced activity, less food intake, and would die if not treated with l-dopa, which is metabolized to dopamine.93 Notably, if these mice are crossed to mice lacking leptin, the resulting animals are also hypophagic and die.94 The role of the dopaminergic circuit concerning addiction to drugs such as alcohol, amphetamines or cocaine but also to the rewarding aspect of food is well established,95 with dopaminergic neurons located in the substantia nigra (SN) and ventral tegmental area (VTA) projecting to many brain nuclei involved in activity, decision-making and activity, such as the frontal cortex, hippocampus or the striatum. Intriguingly, there is now growing evidence that obesity is also linked to dysfunction of the dopaminergic system, as striatal dopamine D2 receptor binding is reduced in obese patients as measured by positron emission tomography.96 Most importantly, leptin and insulin impact on midbrain dopaminergic neurons to regulate food-finding behavior and eventually body weight.95 Thus, LEPR are expressed on VTA dopaminergic neurons, which are hyperpolarized after stimulation with leptin ex vivo.97, 98 Moreover, leptin microinjection into the VTA reduces food intake, whereas ablation of the receptor only in the VTA increased the sensitivity of these mice to the rewarding aspect of highly palatable food, such as sucrose.98 The crucial role for VTA dopamine neurons in the regulation of energy homeostasis is underlined by the finding that direct leptin action on LH neurons also signals to VTA dopamine neurons by synaptic contact, decreasing food intake and thus body weight.43 Besides LEPR, the IR is also expressed on VTA (and SN) dopaminergic neurons.99 Intracerebroventricular insulin treatment has been demonstrated to increase expression the dopamine transporter (DAT) in dopaminergic neurons.100 Dopamine from the synaptic cleft is taken up by DAT back into the presynaptic neuron, and thus stops to stimulate postsynaptic neurons. Hence, insulin may act through this cascade to decrease the rewarding effect of food, which is in line with the findings from multiple experimental paradigms.95

Serotonin

In the CNS, serotonin (5-Hydroxytryptamin) is a neuropeptide expressed only in the raphe nucleus located in the midbrain. Besides being involved in mood regulation, serotonin clearly plays an important role in the control of weight. Thus, molecules increasing the bioavailability of serotonin (by inhibiting reuptake into the presynaptic neuron) such as fenfluramin were very effective in reducing body weight in patients, although side effects partially due to serotonin receptor activation in the heart led to its withdrawal from the market. Interestingly, serotonin acts on the melanocortin system to reduce hunger. Thus, serotonin nerve terminals are found on POMC neurons, serotonin treatment increases excitation and thus the firing rate of POMC neurons, and mice with reexpression of serotonin receptor type 2 only on POMC neurons on a background of serotonin type 2c receptor knockout greatly normalizes body weight compared to the obese HTR2c receptor knockout mice.101, 102 Although previous reports had already detected significant LEPR expression in the (dorsal) raphe nucleus,103 recent data indicates that leptin acts directly on serotonergic neurons to control food intake and body weight, because mice lacking LEPR only on serotonergic cells showed similar obesity compared to mice lacking all LEPR (i.e., db/db mice).104 The authors could furthermore show that leptin treatment decreased the number of action potentials of serotonin neurons.104 Because serotonin release onto POMC neurons is thought to mediate its anorexigenic effects at least in part, it is not immediately obvious how this inhibitory effect of leptin would lead to an outcome of reduced feeding and body weight.

Nucleus tractus solitarius (NTS)

The NTS of the brainstem is a relay between afferent input from the gut, with synaptic connections to the hypothalamus, the parabrachialis nucleus, but also higher brain regions such as the forebrain.105 Gut distention due to food intake in combination with release of gut hormones such as glucagon-like peptide (GLP) signals to the brainstem via the vagus nerve to induce satiety, and this input is then forwarded and computed in the aforementioned nuclei. Leptin signaling in the brainstem is clearly relevant for meal size regulation, because LEPR expression is found in the brainstem, injection of leptin into the brainstem parenchyma at very small doses induces phosphorylation of STAT3, and is able to reduce meal size and thus food intake.106 Importantly, a subpopulation of NTS neurons exists that both integrates leptin action (through the LEPR) and gastrointestinal input (through vagal innervation).105 Moreover, leptin potentiates gastrointestinal (GI) signals; thus when gastric distention and leptin injection into the 4th ventricle (in close contact with the NTS), at levels that do not reduce food intake, are combined, a significant reduction in food intake is achieved.105 The NTS also harbors a population of POMC neurons, which are activated by the anorexigenic gut hormone cholecystokinine, and signal to MC4R in the brainstem to subsequently reduce food intake.107 Although it is not clear of insulin has similar properties with respect to NTS-mediated food intake suppression, the synergy between leptin (and possibly insulin) and GI signals is of high interest, because GI hormones such as GLP (and its derivates) or oxyntomodulin positively affect glucose homeostasis and/or body weight, and importantly, obesity may not induce resistance against gut hormones to the same extent as against leptin and insulin.108

The mechanisms of CNS insulin and leptin resistance

Frustratingly, leptin and insulin's ability to control energy homeostasis is abrogated both in obese animal models and individuals suffering from obesity. Because lifestyle interventions alone are not sufficient to normalize body weight in most individuals, therapeutic interventions appear to be necessary for treatment of obese patients. Thus, to understand the mechanisms by which the weight-reducing effects of leptin and insulin are blunted is of utmost importance (Figure 3).

Figure 3.

Mechanisms implicated in CNS leptin and insulin resistance. Both leptin and insulin must pass the blood–brain barrier (BBB) by a saturable and active transport mechanism. Binding of C-reactive protein (CRP), whose expression is elevated in obesity to leptin inhibits interaction of leptin with its receptors, and has been speculated to impair BBB transport of leptin. Hyperlipidemia, hyperglycemia, and obesity by itself have also been shown to impair insulin transport across the BBB. Elevated circulating levels of saturated fatty acids, such as palmitate, may activate toll-like receptor TLR / Myd88-dependent signaling, which results in inhibitor of NFKB kinase activation, which is able to both directly impair leptin signaling, but may also upregulate expression of suppressor of cytokine signaling (SOCS) 3, which then acts on the leptin receptor. IKK and SOCS3 are also involved in the induction of CNS insulin resistance. Inflammatory signals, for example tumor necrosis factor-(TNF)α, can activate inflammatory signalling cascades, such as IKK but also c-Jun N-terminal kinase (JNK)1. JNK1 has been implicated in blocking insulin signaling directly at insulin receptor substrates through inhibitory serine phosphorylation. Both IKK and JNK1 can be activated by high-fat diet (HFD) induced endoplasmatic reticulum (ER) stress, induced by accumulation of misfolded proteins in the ER, which leads to the unfolded protein response (UPR). Palmitate also regulates localization and activation of protein kinase C theta (PKCΘ), which may induce insulin resistance through serine phosphorylation of IRS proteins. Low-grade inflammation will induce expression of protein phosphatase 1B (PTP1B), which dephosphorylates and thereby deactivates IR, IRS, LEPR, and JAK proteins. Finally, HFD induces apoptosis of hypothalamic neurons in rats, and it is unknown if this holds true also for other model organisms.

The first obstruction that both leptin and insulin must surpass is the blood–brain barrier (BBB), at least in brain regions where the BBB is tight (in the ARC, the BBB is only weakly developed). Both leptin and insulin are transported across the BBB via saturable mechanisms, which are impaired by obesity and associated pathologies such as hypertriglyceridemia.109–112 A more direct impact on leptin signaling has been linked to the low-level chronic inflammation found in obesity, that is C-reactive protein (CRP), secreted by the liver and increased in obese patients, binds to leptin and inhibits interaction with LEPR, thus leading to leptin resistance.113 Notably, leptin itself induces hepatic CRP expression, and as such hyperleptinemia itself has been shown to be a negative regulator of leptin sensitivity. Thus, chronic hyperleptinemia in combination with additional factors such as HFD will lead to overexpression of SOCS3, which induces both a blockade of leptin signaling at the level of the leptin receptor,114 and insulin resistance, because SOCS3 targets IRS proteins for ubiquitin-mediated degradation.87 Further to this, SOCS3 ablation from all neurons strongly protects from diet-induced obesity.84

Besides SOCS3, other negative regulators of leptin and insulin signaling have been shown to be involved in the induction of resistance to these two hormones, such as protein phosphatase 1b (PTP1B).115 Hypothalamic PTP1B is increased upon obesity (and interestingly aging), it may dephosphorylate JAKs, STATs, the IR as well as IRS proteins.116, 117 Both SOCS3 and PTP1B are involved in inflammatory processes. The notion that obesity induces low-level inflammation in the periphery, especially the adipose tissue, has been known since the late 1980s, when it was demonstrated that adipose tissue of both murine obesity models and that of obese patients expresses higher levels of tumor-necrosis factor-α (TNF-α),118 which is released into circulation, and induces insulin resistance in target tissues such as skeletal muscle and adipose tissue itself.119 Indeed, in that setting, macrophages invade the adipose tissue, likely recruited to take up overloaded and thus dying adipocytes,120 and release proinflammatory cytokines which in a positive feedback loop will attract more macrophages.121

Interestingly also in the hypothalamus, increased expression of both TNF-α and Interleukin 6 have been reported upon obesity development,122 accompanied by activation of inflammation-sensitive kinases such as inhibitor of NFKB kinase (IKK) and c-Jun N-terminal kinase (JNK).122–124 Moreover, inhibition of IKK signaling improves leptin and insulin sensitivity, whereas JNK1 inhibition ameliorates hypothalamic insulin resistance123–125. Mechanistically and in line with the accepted role for SOCS3 in leptin resistance, IKK activation may positively control SOCS3 expression, especially in AGRP neurons.124 JNK1 on the other hand, may phosphorylate IRS proteins on serine residues, to inhibit activating tyrosine phosphorylation.126 JNK1 may also inhibit TRH expression and thus energy expenditure, although it is unknown if this is due to JNK1 action directly in the PVN neurons.123, 127

In this context, it is notable that TNF-α receptor knockout mice show increased energy expenditure potentially due to elevated TRH levels, which may suggest a TNF-α-> JNK1->TRH-regulatory circuit.128 On first view, this data may indicate that increased TNF-α and Interleukin 6 signaling may be underlying the activation of the proinflammatory signaling cascades found in the hypothalami of obese animals, yet closer inspection may yield other conclusions. First, mice lacking interleukin 6 develop adult-onset obesity,129 indicating that IL6 signaling is necessary for normal energy homeostasis. Second, it is unclear if in the mouse model system centrally applied TNF-α is anorexigenic or orexigenic,130, 131 which may be dependent on the concentration used. Third, TNF-α signaling has been demonstrated to protect brain tissue against multiple cellular stressors such as (glutamate) excitotoxicity, axonal injury or oxidative stress.132, 133 In this context it is important to note, that hyperglycemia, a hallmark of diabetes and obesity, induces neuronal oxidative stress,134 mitochondrial abnormalities,135 and most importantly, diet-induced obesity leads to loss of hypothalamic neurons due to apoptosis in rats.136 It has also been noted that chronic cilliary neurotrophic factor (CNTF) administration reduces body weight dependent on neurogenesis in the hypothalamus,137 whereas the acute anorectic effect of CNTF is mediated via gp130 receptor signaling in POMC neurons,138 indicating that gain or loss of neurons known to be relevant in energy homeostasis may be involved in the development and prevention of obesity. Finally, it has not yet been revealed if cytokines produced in peripheral tissues such as adipose tissue are the source of the low-level inflammation in the hypothalamus, or if microglia, astrocytes and/or neurons produce these cytokines, which then act in an autocrine/paracrine manner. It is likely, that both peripheral and locally produced cytokines are involved in the inflammatory processes in hypothalamic tissue of obese animals. Further decisive experiments regarding the diet-induced dysfunction and degeneration of neurons involved in energy homeostasis will be highly informative.

Besides cytokines, (saturated) fatty acids such as palmitate have been directly implicated in both leptin and insulin resistance by eliciting activation of inflammatory kinases such as IKKs or JNKs. It has long been acknowledged that hyperlipidemia induces peripheral insulin resistance, partially due to activation of toll-like receptor (TLR) 4 signaling.139 In the hypothalamus, i.c.v. injection of palmitate at a dose which does not affect food intake partially blocks the ability of leptin to activate STAT3 signaling and reduce food intake.140 This was not the case in mice lacking neuronal MYD88, a scaffold protein necessary for TLR2/4-mediated activation of IKK and JNK signaling.140 Similar studies in rats could show that palmitate application induced IKK activation, and subsequently hypothalamic insulin resistance.141 Notably, this study confirmed that even when total caloric intake was similar between groups, the group consuming HFD showed greater CNS insulin resistance.141 Finally, fatty acids may induce local expression of cytokines, which then may lead to leptin resistance and obesity.142 Fatty acids also activate protein kinase (PKC) theta, which might then translocate to the plasma membrane and induce inhibitory serine phosphorylation on IRS proteins.143 This report underlined the notion that unsaturated fatty acids, such as oleate, do not induce CNS insulin resistance,143 and in fact may have anorexigenic properties, by inhibition of KATP channels of POMC neurons.144 It should be noted, that as of now, there has been no conclusive evidence that saturated fatty acids such as palmitate directly bind to TLRs, and moreover, there has been the intriguing finding that under conditions of obesity, small amounts of intestinal lipopolysaccharide (LPS, the prototypical TLR activating molecule) may enter the circulation due to a leakage of the intestinal barrier function, which then might induce low-grade inflammation.145

Besides both low-grade inflammation and hyperlipidemia, another crucial event in the pathology of peripheral insulin resistance is endoplasmatic reticulum (ER) stress. Protein folding in the ER is necessary for normal cellular homeostasis, as misfolding of nascent proteins can have deleterious results, ultimately compounding cell viability. In obese animals, the need for translation is increased (at least in liver and the pancreas), whereas at the same time folding capability is limited. Accumulation of unfolded proteins is then sensed by specific receptors, which will initiate an adaptive program, termed unfolded protein response (UPR). During the UPR, the expression of chaperones, proteins assisting in protein folding will be increased, whereas global translation is reduced to resolve the ER stress.146 Note that upon enduring unresolved ER stress, the cell will initiate an apoptosis program.146 In peripheral tissues, especially the liver, ER stress has been detected in diabetic and obese mouse models, and application of chemical chaperones can greatly reduce glucose intolerance and insulin resistance in these animals.147, 148 ER stress is not restricted to peripheral tissues; instead it appears to play a major role in the induction of hypothalamic leptin and insulin resistance. Neuronal ablation of XBP-1, a transcription factor involved in the UPR, leads to massive leptin resistance, and obesity.149 Hence, i.c.v. application of chemical chaperones ameliorates leptin resistance.149 Moreover, i.c.v. application of chemical reagents such as tunicamycin or thapsigargin known to induce ER stress partially inhibits the efficacy of i.c.v. leptin and insulin as well.150 In line with the notion that hyperlipidemia, inflammation and ER stress are interlinked, hyperlipidemia and inflammation may induce hypothalamic ER stress, and ER stress induces cytokine expression, which may be involved in a specific decrease of anorexigenic POMC expression, possibly due to POMC neuron apoptosis.124, 136, 142, 151 Although several reports now point to the induction of ER stress upon hyperlipidemic or inflammatory insults also in the hypothalamus, the mechanisms by which ER stress is induced, are less well defined (and vice versa, the mechanisms by which chaperone treatment improves hypothalamic leptin sensitivity). Saturated fatty acids may directly impair ER homeostasis, for example by changing the lipid composition of the ER membrane, which is followed by calcium depletion and breakdown of ER function.152–154 Notably, overabundance of saturated fatty acids such as palmitate may also induce generation of ceramides, a lipid species involved in insulin resistance and apoptosis. Ceramide generation has been linked to insulin resistance in peripheral tissues such as liver and skeletal muscle.155 Palmitate is necessary for the de novo generation of ceramides, palmitate supplement acutely induces ceramide generation, and ceramide accumulation induces insulin resistance by blocking AKT activation and dephosphorylation of activated AKT in peripheral tissues or nonneuronal cell lines.156–158

Conclusion

In this review, we attempt to give an informing overview on the current concepts of insulin and leptin targets, the intracellular cascades activated, and the pathomechanisms leading to CNS resistance against these two anorexigenic hormones. We have summarized the most urgent questions in need of clarification in Table 1. It is clear from the existing data that although hypothalamic circuits are necessary for control of energy homeostasis, several others such as the dopaminergic and serotonergic neurons critically contribute to normal body weight regulation. It is also accepted that besides leptin and insulin, there are dozens of other hormones such as ghrelin, but also metabolites such as glucose and lipids, which directly impact weight regulation and glucose metabolism. Moreover, there is the worrying fact that obesity and diabetes, especially in the pregnant mother, have negative and importantly long-term effects for the unborn child (see review in this volume by S. Ozanne, Ref. 159), which suggest that in our rapidly aging and westernized societies, the diabesity pandemics will grow. Thus, although environmental changes and lifestyle adaptation will hopefully be able to throttle the increase in obesity and diabetes with beneficial effects on other aging-related diseases, we have to try to develop potential treatment options, which may eventually include intranasal delivery of neuropeptides,160 or self-inactivating gene therapy.161

Table 1.  Urgent questions
• Are there leptin- but not insulin-responsive subpopulations in all nuclei involved in energy homeostasis? How can we identify and differentiate between these subpopulations in vivo by transgenesis to address their function?
• Does insulin influence synaptic rewiring of neurons? Is this cell-intrinsic or postsynaptically controlled? Can we determine the molecular pathways controlled by leptin and insulin to affect axon guidance and synaptic contact?
• If leptin-mediated regulation of serotonin is crucial for energy homeostasis, is insulin involved in this as well?
• Does VTA/NTS/raphe leptin and insulin resistance exist? Do the same mechanisms occur as in the hypothalamus?
• Can we identify further extrahypothalamic targets mediating leptin's and insulin's effects, such as olfactory neurons?
• How can we connect and integrate genes found by human genome–wide association studies (GWAS) into the known signaling pathways of leptin and insulin? Is there a connection to be found at all?
• Can we identify marker for the neurons implicated in computing gastrointestinal signals and insulin/leptin signaling? Are GI signals in the CNS modulated by insulin?

We would like to stress that although one may think that we have identified all major brain subregions involved in energy homeostasis, we have likely only identified a small proportion. Further progress in the analysis of extra-hypothalamic nuclei and integration of these findings including the role of clearly relevant genes discovered by human genome–wide association studies (GWAS) of human obesity (such as FTO)162, 163 into our existing concepts will eventually lead to novel therapeutical targets and better treatments for the diabesity pandemic.

Acknowledgments

We apologize to all colleagues whose important contribution could not be cited due to space limitations. We thank G. Schmall and T. Rayle for excellent secretarial assistance and all members of the Brüning lab for helpful discussion of the manuscript. This work was supported by grants from the CMMC (TV1) and the DFG (Br. 1492/7–1) to J.C.B. CECAD is funded by the DFG within the Excellence Initiative by German Federal and State Governments.

Conflicts of interest

The authors declare no conflicts of interest.