International Journal of Obesity (2001) 25, Suppl 5, S68–S72 ß 2001 Nature Publishing Group All rights reserved 0307–0565/01 $15.00 www.nature.com/ijo

PAPER Glucosensing neurons do more than just sense glucose BE Levin1* 1

Neurology Service, VA Medical Center, E Orange, and Department of Neurosciences, New Jersey Medical School, Newark, New Jersey, USA The brain regulates energy homeostasis by balancing energy intake, expenditure and storage. To accomplish this, it has evolved specialized neurons that receive and integrate afferent neural and metabolic signals conveying information about the energy status of the body. These sensor – integrator – effector neurons are located in brain areas involved in homeostatic functions such as the hypothalamus, locus coeruleus, basal ganglia, limbic system and nucleus tractus solitarius. The ability to sense and regulate glucose metabolism is critical because of glucose’s primacy as a metabolic substrate for neural function. Most neurons use glucose as an energy substrate, but glucosensing neurons also use glucose as a signaling molecule to regulate neuronal firing and transmitter release. There are two types of glucosensing neurons that either increase (glucose responsive, GR) or decrease (glucose sensitive, GS) their firing rate as brain glucose levels rise. Little is known about the mechanism by which GS neurons sense glucose. However, GR neurons appear to function much like the pancreatic b-cell where glycolysis regulates the activity of an ATP-sensitive Kþ (KATP) channel. The KATP channel is composed of four pore-forming units (Kir6.2) and four sulfonylurea binding sites (SUR). Glucokinase (GK) appears to modulate KATP channel activity via its gatekeeper role in the glycolytic production of ATP. Thus, GK may serve as a marker for GR neurons. Neuropeptide Y (NPY) and pro-opiomelanocortin (POMC) neurons in the hypothalamic arcuate nucleus are critical components of the energy homeostasis pathways in the brain. Both express Kir6.2 and GK, as well as leptin receptors. They also receive visceral neural and intrinsic neuropeptide and transmitter inputs. Such metabolism-related signals can summate upon KATP channel activity which then alters membrane potential, neuronal firing rate and peptide=transmitter release. The outputs of these neurons are integral components of effector systems which regulate energy homeostasis. Thus, arcuate NPY and POMC neurons are probably prototypes of this important class of sensor – integrator – effector neurons. International Journal of Obesity (2001) 25, Suppl 5, S68 – S72 Keywords: glucose responsive; glucose sensitive; neuropeptide Y; POMC; glucokinase; SUR; KATP

Through evolution it has become necessary for organisms to sense the state of their energy homeostasis in order to maintain a proper balance during times of famine and plenty. Although many sites within the body can respond to differences in energy balance, the brain stands out as the primary site for sensing, integrating and effecting signals essential for maintenance of homeostasis. Since prolonged energy deficit is the most life-threatening, the body has evolved numerous overlapping and redundant systems to be sure that the organism is driven to seek food when bodily

*Correspondence: BE Levin, Neurology Service (127C), VA Medical Center, 385 Tremont Ave, E Orange, NJ 07018-1095, USA. E-mail: [email protected]

energy stores become depleted. There are systems that signal both short- and long-term changes in homeostatic balance. Glucose is a prototypical short-term signal since its stores in the body are limited and the brain and other organs have a critical dependence upon it as a primary energy source. Thus, significant deviations from a fairly narrow range of plasma glucose levels lead to significant counter-regulatory responses designed to restore glucose availability. While it is clear that the brain can both sense and respond to severe hyper- or hypoglycemia,1 – 5 it is less certain that minor ultradian variations in glucose can serve as a signal to modulate physiologic functions such as food intake.6,7 In fact, glucosensing neurons in the brain do seem capable of responding to relatively small changes in ambient glucose concentrations by altering their membrane potential and firing rate. This defines a ‘glucosensing neuron’, ie a

Glucosensing neurons BE Levin

S69 neuron that uses glucose as a signaling molecule to alter cell function and neuronal activity. This distinguishes glucosensing neurons from the majority of neurons which utilize glucose simply as a metabolic substrate to fuel increases in neuronal activity and metabolic demands. As it turns out, glucosensing neurons respond to more than just short-term alterations in glucose availability. In brain areas such as the hypothalamus, glucosensing neurons also contain receptors for insulin, leptin, monoamines and other transmitters and peptides involved in energy homeostasis.8 – 12 Thus, many or all glucosensing neurons respond to both short- and long-term signals relating to both the physical and affective components of energy homeostasis. These same neurons are generally connected to efferent pathways involved in regulating both energy intake and expenditure and so fulfill the criteria for a true metabolic sensor – integrator – effector. While such sensors may also be located in the periphery, the best characterized of these lie in brain areas such as the hypothalamus, caudal medulla, substantia nigra and locus coeruleus.812 – 16 Of these, the neuropeptide Y and proopiomelanocortin neurons of the hypothalamic arcuate nucleus appear to be prototypic examples of metabolic sensor – integrator – effector neurons.5,10,17 In general, there are two classes of glucosensing neurons. Neuronal activity in glucose sensitive (GS) neurons is the reciprocal of ambient glucose levels. In glucose responsive (GR) neurons, activity parallels changes in ambient glucose.8,12 – 16 Thus, as glucose levels rise, GR neurons fire more rapidly and GS neurons decrease their firing rate. Although first described more than 25 y ago,14 almost nothing is known about the mechanism used by GS neurons to sense glucose. On the other hand, GR neurons appear to function much like the pancreatic b-cell which uses an ATPsensitive Kþ (KATP) channel to sense glucose.10,16,18 When the intracellular ATP=ADP ratio is increased by metabolism of glucose, ATP is bound to the KATP channel causing it to become inactive (closed). This raises intracellular Kþ levels leading to membrane depolarization, opening of a voltagedependent Ca2þ channel and cell firing (Figures 1 and 2). KATP channels also reside on axon terminals of some gammaaminobutyric acid (GABA) and glutamate neurons.12,19,20 Raising ambient glucose levels around these terminals leads to transmitter release. In fact, many published studies describing neurons as ‘GS’ or ‘GR’ based on changes in their firing rates when ambient glucose levels were changed did not account for the fact that glucose can both stimulate GR neurons to fire and simultaneously release GABA or glutamate onto both these same GR neurons and onto non-GR neurons. Thus, the resultant activity of a given neuron in the face of changing glucose levels does not guarantee that it has intrinsic glucosensing properties nor that the resulting firing rate is necessarily due to an effect of glucose solely on that cell.12 Rather, the resultant firing rate of a given neuron is the sum of inhibitory and excitatory inputs produced by the direct and indirect effects of glucose acting through the KATP channel.

The KATP channel is an octomeric structure containing four inwardly rectifying pore-forming units (Kir6.2) and 4 sulfonylurea receptors (SUR).21 Channel function requires the presence of both Kir6.2 and SUR subunits. However, there are at least three SUR isoforms. SUR1 is a high-affinity receptor found in b-cells and some neurons.5,21,22 SUR2A and 2B are low affinity and are found in neurons, cardiac, skeletal and smooth muscle.21 The grouping of Kir6.2 with various combinations of SUR1 or SUR2 isoforms may play a major role in determining the characteristics of the KATP channel on a given cell or axon terminal.23 This may explain the highly pleomorphic responses of individual neurons when ambient concentrations of glucose and=or sulfonylureas are changed.15,23 While the KATP channel has a clear role in modulating membrane potential in GR neurons, it is less certain how ATP=ADP ratios at the channel are modulated within the narrow range needed to act as a regulator of channel function. In the b-cell, both the GLUT2 glucose transporter and glucokinase (GK) have been proposed as candidates for such a regulatory role.24 In the brain, glucose levels are generally about 20% of those in the plasma, ie in the low mM range.15 Under these conditions, GLUT3, which is found on most neurons, is generally saturated because of its low Km.5 Similarly, hexokinase I is present in most neurons and its Km is outside the usual physiologic range of brain glucose levels making it an unlikely regulator of the rate of neuronal glycolytic production of ATP.5 However, we17 and others16 have shown that GK is also present in the brain. Moreover, it is expressed in relatively few sites in the brain and many of those contain populations of known glucosensing neurons.17 Among the neurons which express GK are neuropeptide Y (NPY) and pro-opiomelanocortin (POMC) neurons in the arcuate nucleus17 and GABA neurons in the ventromedial nucleus, while orexin neurons in the lateral hypothalamus (where GS neurons predominate) do not contain GK (unpublished results). Similarly, SUR2 binding25 and mRNA expression (unpublished data) are highly localized to these same hypothalamic nuclei which are known to contain GR neurons with KATP channels. Both Kir6.2 and SUR1 are widely distributed in neurons throughout the brain (22, 10) while GK and SUR2 appear to have a much more limited distribution.17 Thus, one or both of these may be critical components of true GR neurons. Other neurons containing the KATP channel may utilize it as a neuroprotective device by which severe limitations of energy availability as occur during hypoxia or hypoglycemia would lower intracellular ATP, hyperpolarize the neuron by activating (opening) the KATP channel and prevent neurotoxic cell death found under these conditions.4,5 On the other hand, the requirement for GR neurons to be highly sensitive to small changes in ambient glucose may have necessitated them to abrogate the neuroprotective function of the KATP channel. This may explain why presumed GR neurons in the arcuate nucleus appear to undergo apoptotic cell death during transient bouts of non-coma-producing hypoglycemia.4 International Journal of Obesity

Glucosensing neurons BE Levin

S70

Figure 1 Model of a KATP channel on a brain glucose responsive neuron which utilizes the KATP channel to sense glucose and other metabolic and neural signals. When glucose is at low levels, relatively low levels of intracellular ATP are available to act at the KATP channel. Tyrosine kinase receptors for insulin or JAK=STAT mediated receptors for leptin or cytokines activate intracellular phosphoinositide 3 kinase (PI3-K) leading to phosphorylation of phosphatidylinositol-4-phosphate (PIP) to phosphatidylinositol-4, 5-bisphosphate (PIP 2) and phosphatidylinositol-3,4,5-triphosphate (PIP 3). PIP 2 and PIP 3 stabilize the KATP channel in the activated (open) state, possibly by causing disaggregation of actin filaments associated with the channel and by lowering the sensitivity of the channel to ATP. Kþ leaves the neuron causing membrane hyperpolarization and reduced cell firing. (Adapted from Harvey et al27,28 and Baukrowitz and Fakler.29)

The exact role of such neurons in modulating physiologic functions remains in question. While small changes in plasma glucose have been shown to occur preceding meals,6 it is uncertain whether the brain or periphery has sensing mechanisms sensitive enough to respond to such small changes. Clearly, severe hypoglycemia can be sensed by the brain2 – 4 and this leads to a profound counterregulatory sympathoadrenal response which mobilizes remaining available glucose within the body. Similarly, relatively large increases in plasma glucose can selectively activate the sympathetic nervous system.1,26 While extremes of glucose availability evoke physiologic responses, it has been extremely difficult to establish a definitive connection between true physiologic changes in cell firing and homeostatic functions such as changes in food intake, energy expenditure and storage. Despite these difficulties, mounting evidence suggests that glucosensing neurons in some brain areas such as the hypothalamus are equipped to monitor more than just ambient glucose levels. Neurons containing the KATP channel are responsive to leptin, insulin, neurotransmitters and peptides.8,9,11 Several types of evidence in cultured cell lines point to a role for phosphatidyl inositol 3,4 phosphate (PIP International Journal of Obesity

2) and 3,4,5 phosphate (PIP 3) in modulating KATP channel function21,27 – 29 as a result of activation of G proteincoupled, JAK-STAT or tyrosine kinase-coupled receptors29 (Figures 1 and 2). Normally, ATP derived from GK-mediated glycolysis of glucose binds to and inactivates the KATP channel (Figure 1). Similarly, binding of sulfonylureas to SUR inactivates the channel. Both these events lead to membrane depolarization and increase neuronal firing. Generally, conditions that increase intracellular phosphorylation stabilize the channel in the activated state (open) and counteract the effects of either ATP or sulfonylurea binding.30 Similarly, conditions that decrease phosphorylation tend to inactivate the channel. Thus, agonists of G-protein-coupled receptors linked to activation of phospholipase C and=or PI 3 phosphatase reduce phosphorylation of PIP 2 and PIP 3, making the inactive state of the KATP channel more stable. This provides a theoretical mechanism for monoamines and neuropeptides which act on this pathway to promote inactivation of the channel and increased cell firing. On the other hand, activation of the JAK-STAT- or tyrosine kinase-coupled pathways by leptin, insulin or cytokines can activate phosphoinositide 3 kinase (PIP 3) kinase. This leads to phosphorylation of PIP to PIP 2 and PIP 3. Both of these can inhibit

Glucosensing neurons BE Levin

S71

Figure 2 Inactivation of the KATP channel on a glucose responsive neuron by increased extracellular glucose and=or application of sulfonylureas. The ATP=ADP ratio is raised by glycolysis mediated by glucokinase (GK). ATP is bound to the KATP channel and sulfonylureas to the sulfonylurea receptor (SUR), both of which inactivate (close) the channel. This leads to increased intracellular Kþ leading to membrane depolarization with opening of a voltage-gated Ca2þ channel, influx of extracellular Ca2þ and neuronal firing. Neurotransmitters and peptides which act at a G protein-coupled receptor (GPCR) can lead to activation of phospholipase C (PLC) and=or phosphoinositide 3 phosphatase (PI3-P). Both of these enzymes lead to lowering of PIP 2 and PIP 3, which tends to stabilize actin filaments associated with the KATP channel enhancing the inactivation of the channel caused by ATP and sulfonylurea binding to the channel. Transmitters or peptides which inhibit PLC or PI3-P through their actions on the GPCR would tend to oppose the actions of glucose or sulfonylureas. (Adapted from Harvey et al27,28 and Baukrowitz and Fakler.29)

ATP binding which would favor the active configuration of the KATP channel.29 This process appears to involve a structural disaggregation of actin filaments associated with the channel.28 Although such data were derived from in vitro cell systems, it is quite possible that leptin and insulin could utilize a similar mechanism in neurons to hyperpolarize the cell membrane in GR neurons. Thus, the KATP channel on GR neurons might provide a central focus for the actions of multiple metabolic and neural signals from the periphery and surrounding the brain, allowing such neurons to monitor and regulate energy homeostasis. While this does not exclude other mechanisms for control of energy homeostasis, it does provide a model system to explore the concept of a central metabolic sensor – integrator – effector neuron.

References 1 Levin BE. Glucose increases rat plasma norepinephrine levels by direct action on the brain. Am J Physiol 1991; 261: R1351 – R1357. 2 Borg WP, Sherwin RS, During MJ, Borg MA, Shulman GI. Local ventromedial hypothalamic glucopenia triggers counterregulatory hormone Release. Diabetes 1995; 44: 180 – 184.

3 Borg MA, Sherwin RS, Borg WP, Tamborlane WV, Shulman GI. Local ventromedial hypothalamus glucose perfusion Blocks counterregulation during systemic hypoglycemia in awake rats. J Clin Invest 1997; 99: 361 – 365. 4 Tkacs NC, Dunn-Meynell AA, Levin BE. Apoptosis and reduced arcuate nucleus NPY and POMC MRNA in non-coma hypoglycemia. Diabetes 2000; 49: 820 – 826. 5 Levin BE, Dunn-Meynell AA, Routh VH. Brain glucose sensing and body energy homeostasis: role in obesity and diabetes. Am J Physiol 1999; 276: R1223 – R1231. 6 Campfield LA, Smith FJ. Functional coupling between transient declines in blood glucose and feeding behavior: temporal relationships. Brain Res Bull 1986; 17: 427 – 433. 7 Mayer J. Glucostatic mechanism of regulation of food intake. New Engl J Med 1953; 249: 13 – 16. 8 Kow L-M, Pfaff DW. Actions of feeding-relevant agents on hypothalamic glucose-responsive neurons in vitro. Brain Res Bull 1985; 15: 509 – 513. 9 Spanswick D, Smith MA, Groppi VE, Logan SD, Ashford ML. Leptin inhibits hypothalamic neurons by activation of ATPsensitive potassium channels. Nature 1997; 390: 521 – 525. 10 Dunn-Meynell AA, Rawson NE, Levin BE. Distribution and phenotype of neurons containing the ATP-sensitive Kþ channel in rat brain. Brain Res 1998; 814: 41 – 54. 11 Spanswick D, Smith MA, Mirshamsi S, Routh VH, Ashford ML. Insulin activates ATP-sensitive K þ channels in hypothalamic neurons of lean, but not obese rats. Nature Neurosci 2000; 3: 757 – 758.

International Journal of Obesity

Glucosensing neurons BE Levin

S72

12 Levin BE. Glucose-regulated dopamine release from substantia nigra neurons. Brain Res 2000; 874: 158 – 164. 13 Mizuno Y, Oomura Y. Glucose responding neurons in the nucleus tractus solitarius of the rat, in vitro study. Brain Res 1984; 307: 109 – 116. 14 Oomura Y, Ooyama H, Sugimori M, Nakamura T, Yamada Y. Glucose inhibition of the glucose-sensitive neurone in the rat lateral hypothalamus. Nature 1974; 247: 284 – 286. 15 Silver IA, Erecinska M. Extracellular glucose concentrations in mammalian brain: continuous monitoring of changes during increased neuronal activity and upon limitation in oxygen supply in normo-, hypo-, and hyperglycemic animals. J Neurosci 1994; 14: 5068 – 5076. 16 Yang X, Kow L-M, Funabashi T, Mobbs CV. Hypothalamic glucose sensor. Similarities to and differences from pancreatic B-cell mechanisms. Diabetes 1999; 48: 1763 – 1772. 17 Lynch RM, Tompkins LS, Brooks HL, Dunn-Meynell AA, Levin BE. Localization of glucokinase gene expression in the rat. Brain 2000; 49: 693 – 700. 18 Ashford, MLJ, Boden PR, Treherne JM. Glucose-induced excitation of hypothalamic neurones is mediated by ATP-sensitive Kþ channels. Pflugers Arch 1990; 415: 479 – 483. 19 Amoroso S, Schmid-Antomarchi H, Fosset M, Lazdunski M. Glucose, sulfonylureas, and neurotransmitter release: role of ATPsensitive Kþ channels. Science 1990; 247: 852 – 854. 20 Lee K, Dixon AK, Rowe ICM, Ashford MLJ, Richardson PJ. The high-affinity sulfonylurea receptor regulates KATP channels in nerve terminal of the rat motor cortex. J Neurochem 1996; 66: 2562 – 2571. 21 Aguilar-Bryan L, Clement JP 4th, Gonzalez G, Kunjilwar K, Babenko A, Bryan J. Toward understanding the assembly and structure of KATP channels. Physiol Rev 1998; 78: 227 – 245.

International Journal of Obesity

22 Karschin C, Ecke C, Ashcroft FM, Karschin A. Overlapping distribution of KATP channel-forming unit Kir6.2 subunit and the sulfonylurea receptor SUR1 in rodent brain. FEBS Lett 1997; 401: 59 – 64. 23 Liss B, Bruns R, Roeper J. Alternative sulfonylurea receptor expression defines metabolic sensitivity of K-ATP channels in dopaminergic midbrain neurons. EMBO J 1999; 18: 833 – 846. 24 Matschinsky FM. Glucokinase as glucosensor and Metabolic signal Generator in pancreatic B-cells and Hepatocytes. Diabetes 1990; 39: 647 – 652. 25 Dunn-Meynell AA, Routh VH, McArdle JJ, Levin BE. Low affinity sulfonylurea binding sites reside on neuronal cell Bodies in the Brain. Brain Res 1997; 745: 1 – 9. 26 Levin BE, Sullivan AC. Glucose, insulin and sympathoadrenal activation. J Auton Nerv Sys 1987; 20: 233 – 242. 27 Harvey J, McKay NG, Walker KS, Van der Kaay J, Downes CP, Ashford, ML. Essential role of phosphoinositide 3-kinase in leptin-induced K(ATP) channel activation in the rat CRI-G1 insulinoma cell line. J Biol Chem 2000; 275: 4660 – 4669. 28 Harvey J, Hardy SC, Irving AJ, Ashford MLJ. Leptin Activation of ATP-sensitive Kþ (KATP) channels in rat CRI-G1 insulinoma cells involves disruption of the actin cytoskeleton. J Physiol 2000; 527: 95 – 107. 29 Baukrowitz T, Fakler B. KATP channels: linker between phospholipid metabolism and excitability. Biochem Pharmac 2000; 60: 735 – 740. 30 Routh VH, McArdle JJ, Levin BE. Phosphorylation modulates the activity of the ATP-sensitive Kþ channel in the ventromedial hypothalamic nucleus. Brain Res 1997; 778: 107 – 119.