Metabolism-Independent Sugar Sensing in Central Orexin Neurons

Animal procedures were performed according to the Animals (Scientific Procedures) Act, 1986 (U.K.). Transgenic “orexin-eGFP” mice were used in all experiments to identify orexin neurons, as in our previous studies (11,25). These mice expressed enhanced green fluorescent protein (eGFP) under the control of the prepro-orexin promoter, resulting in very specific targeting of eGFP only to orexin neurons (11). Mice were maintained on a 12-h light:dark cycle (lights on at 08:00 h) and had free access to food and water. Animals (14–21 days postnatal) were killed during the light phase. Coronal slices (250–300 μm thick) containing the lateral hypothalamus were cut with a Campden Vibroslice in ice-cold extracellular solution, incubated at 35°C in carbogen-gassed extracellular solution for 30 min, and then kept at room temperature until recording, as previously described (26). Typically, between 5 and 20 cells were evaluated from each animal used.

The extracellular solution was gassed with 95% O₂ and 5% CO₂ and contained 125 mmol/l NaCl, 2.5 mmol/l KCl, 2 mmol/l MgCl₂, 2 mmol/l CaCl₂, 1.2 mmol/l NaH₂PO₄, 21 mmol/l NaHCO₃, and 1 mmol/l glucose, except where indicated otherwise. We used a “K-chloride” intracellular (pipette) solution throughout to reduce junction potential errors in measurements of current-voltage relationships (27). This solution contained 130 mmol/l KCl, 0.1 mmol/l EGTA, 10 mmol/l HEPES, 5 mmol/l K₂ATP, 1 mmol/l NaCl, and 2 mmol/l MgCl₂, pH 7.25, with KOH. Where not indicated, the full chemical names of sugars and metabolites we used were d-(+)-mannose, l-lactate, d-(−)-fructose, d-(+)-galactose, d-(+)-glucose, l-(−)-glucose, and methyl-α-d-glucopyranoside. All chemicals were from Sigma (Poole, Dorset, U.K.). Chemicals were applied extracellularly by bath perfusion, unless indicated otherwise.

Living orexin-eGFP neurons were visualized in brain slices using an Olympus BX50WI upright microscope equipped with epifluorescence as previously described (11). Whole-cell recordings, including switching to voltage-clamp mode (for recording membrane currents) and current-clamp mode (for recording membrane potential), were performed using an EPC-10 amplifier (Heka, Lambrecht, Germany) at 36°C (bath temperature controlled with Badcontroller V; Luigs & Neumann). Patch pipettes were pulled from borosilicate glass and had tip resistances of 3–6 mol/lΩ. Series resistances were in the range of 5–10 mol/lΩ and were not compensated. Data were sampled and filtered using Patchmaster software (Heka) and analyzed with custom-written Python software (http://www.python.org), Matplotlib (http://matplotlib.sourceforge.net), and Origin software (Microcal, Northampton, MA).

To monitor membrane resistance together with changes in membrane potential (Figs. 1A and 8A–C), cells were injected every 20–40 s with a fixed-amplitude (10–40 pA, 1-s duration) square pulse of hyperpolarizing current (26). The amplitude of the resulting downward deflections in membrane potential is proportional to membrane resistance (Ohm's law).

Current-voltage relationships (Figs. 2A and B, 3B, 4B, 5B and C, and 7B and C) were obtained by performing voltage-clamp ramps from 0 to −140 mV at a rate of 0.1 mV/ms ramp, which is sufficiently slow to allow leak-like K⁺ currents to reach steady state at each potential (28). In Figs. 2B and 7C, the net current-voltage relationship was fitted with the Goldman-Hodgkin-Katz (GHK) current equation (29) in the following form:

Here, I is current (in A), V is membrane potential (in V), z is the charge of a potassium ion (+1), [K⁺]_i is the pipette K⁺ concentration (in mmol/l), [K⁺]_o is the extracellular K⁺ concentration (in mmol/l), T is temperature (in Kelvins), and R and F have their usual meanings (29). P_K (a constant reflecting the K⁺ permeability of the membrane) was the only free parameter during fits. The values of P_K used to obtain the fits shown in Figs. 2B and 7C were, respectively, 4.616 × 10⁻¹⁷ and 1.554 × 10⁻¹⁷ cm/s. Values are presented as means ± SE.

We examined the effects of a variety of sugars and agents that affect energy metabolism on the membrane potential and currents of orexin neurons directly identified by selective expression of GFP using whole-cell patch-clamp recordings (11). We first tested 2-deoxyglucose, a nonmetabolizable glucose analog that competes with glucose for binding to enzymes and transporters, thereby acting as an energy-depleting “antimetabolite.” Thus, we expected this antimetabolite to have the opposite effects to glucose or no effects at all, because the cells that we recorded were provided with ATP from the pipette solution and hence presumably were energetically independent of extracellular glucose. However, 2-deoxyglucose mimicked the effects of glucose on the membrane potential and conductance of orexin cells, triggering dose-dependent and reversible membrane hyperpolarization and increase in membrane conductance (Fig. 1A and B, n = 30 cells). Analysis of membrane current responses to voltage-clamp ramps revealed that the changes in membrane potential were associated with activation of an ionic current that reversed at a membrane potential corresponding to E_K (Fig. 2A, n = 7 cells; predicted E_K ∼ −100 mV with our solutions) and followed an outward current-voltage rectification well described by the GHK equation (Fig. 2B, n = 4 cells), as previously described for glucose (11). Like glucose, 2-deoxyglucose thus hyperpolarizes the orexin cell membrane by activating a leak-like K⁺ current.

We also performed control experiments to ensure that the effects of 2-deoxyglucose on orexin neurons are a specific phenomenon rather than a general outcome of energy depletion caused by this antimetabolite). First, we applied 2-deoxyglucose to neurons that had similar membrane properties to orexin cells (depolarized potentials and spontaneous firing) but were located in the cerebral cortex, a nonglucosensing brain region (6). Unlike in orexin cells, in cortical neurons, 2-deoxyglucose induced neither hyperpolarization (Fig. 3A, n = 7 cells) nor activation of K⁺ currents (Fig. 3B, n = 7 cells). Second, we filled orexin neurons with 350 μmol/l tolbutamide and 5 mmol/l ATP by including these chemicals in the pipette solution used for whole-cell recording. This procedure is expected to block their K_ATP channels by ∼100% (30), thereby disrupting the ubiquitous coupling between intracellular ATP metabolism and membrane potential in neurons (31). Under these conditions, orexin cells still exhibited full-blown responses to 2-deoxyglucose (Fig. 4A and B, n = 6 cells) and glucose (n = 3; data not shown). In the presence of tolbutamide, hyperpolarization elicited by 2-deoxyglucose (17.0 ± 2.5 mV, n = 6) was not different from that in control cells (19.9 ± 2.5 mV, n = 9, P = 0.437). This implies that orexin cell responses to 2-deoxyglucose are a specific sensing phenomenon of orexin cells rather than a general neuronal response related to energy metabolism.

To further examine the effects of energy metabolism on the membrane potential of orexin cells, we stimulated them with lactate. Lactate is used as fuel for mitochondrial ATP production and is considered to be the preferred energy fuel for central neurons (32). Lactate (5 mmol/l) neither hyperpolarized orexin neurons nor activated their K⁺ currents (Fig. 5A and B). Instead, it triggered a slight membrane depolarization and an increase in action potential firing frequency, both of which were reversible upon removal of lactate (Fig. 5A, n = 10 cells). To examine membrane currents underlying these excitatory effects, we looked at membrane current-voltage relationships before and after lactate action, which revealed that lactate inhibited a membrane current with a reversal potential close to that of K⁺ (Fig. 5B, n = 4 cells). This observation is consistent with lactate being converted to ATP and the latter blocking the K_ATP channels. We observed that the net lactate-inhibited current (Fig. 5C, n = 4 cells) exhibited an inward current-voltage rectification typical of K_ATP channels (30) but not of glucose-activated K⁺ channels (11). These data imply that in orexin neurons, lactate-induced stimulation of mitochondrial ATP production causes membrane depolarization through closure of K_ATP channels and does not mimic the activating effects of glucose on leak-like K⁺ channels.

Our experiments with lactate imply that mitochondrial ATP generation does not induce glucosensing responses in orexin cells. However, it is possible that the exact site of ATP production is critical for glucose-sensing, for example, in glucose-excited β-cells; pyruvate does not trigger glucosensing responses despite being metabolized (33,34). Therefore, we next tested whether glucosensing depends on glucose metabolism in the glycolytic pathway. To block glycolytic metabolism of glucose, we pharmacologically inhibited glucokinase with high concentrations of three selective inhibitors, alloxan, d-glucosamine, and N-acetyl-d-glucosamine (35). However, in orexin neurons, glucokinase inhibition did not change the effects of glucose on the membrane potential (Fig. 6A–C, n = 14 cells) or K⁺ currents (n > 3 cells for each inhibitor; data not shown). In the presence of alloxan, d-glucosamine, and N-acetyl-d-glucosamine, respectively, a switch from 1 to 5 mmol/l glucose caused hyperpolarization of 21.7 ± 3.7 (n = 5), 22.5 ± 3.1 (n = 5), and 28.0 ± 2.4 mV (n = 4), which is indistinguishable from control value (25.1 ± 3.7 mV, n = 6; P = 0.533, 0.616, and 0.582, respectively).

These results indicate that preventing intracellular glucose from being converted to its metabolizable form by glucokinase-catalyzed phosphorylation does not affect the ability of orexin cells to sense glucose. In fact, intracellular glucose in general is unlikely to play a key role in orexin cell glucosensing because artificial infusion of glucose into orexin cells does not trigger their inhibitory glucose responses (11). Here, we repeated the latter experiment with higher concentrations of intracellular glucose and lower concentrations of extracellular glucose than in this original study (Fig. 6D). We found that internally dialyzing orexin cells with 10 mmol/l glucose does not evoke glucosensing responses but, subsequently, that switching from 1 to 2.5 mmol/l glucose outside the same cell elicits normal responses (Fig. 6D, n = 4 cells). Thus, it is unlikely that glucose-derived energy metabolites or intracellular glucose plays a role in glucose responses of orexin neurons.

Despite the importance of orexin neurons for behavioral coordination, and the potent inhibitory effects of glucose on these cells, their sensitivity to other sugars has not been examined. Determining the sugar selectivity of glucose effects on orexin neurons can provide valuable information about the nature of the glucose-detection device in these cells, because sugar binding sites on known glucose-binding proteins exhibit specific sugar-selectivity signatures (see discussion). In addition to 2-deoxyglucose (Fig. 1), therefore, we examined responses to five other sugars, mannose, l-glucose, α-methyl-d-glucoside, fructose, and galactose. Of these, only mannose mimicked the effects of glucose on the membrane potential of orexin cells (Fig. 7A, n = 8 cells). The responses of orexin cells to mannose were also identical to glucose responses in terms of activation of membrane currents: Like glucose, mannose triggered a K-selective current (Fig. 7B, n = 6 cells) whose current-voltage relationship was well approximated by the GHK current equation (Fig. 7C). In contrast, at 5 mmol/l, none of the other sugars that we tried affected the membrane potential (Fig. 8, n > 4 cells for each sugar tested) or membrane currents (n > 3 cells for each sugar; data not shown) of orexin cells. To validate glucose-sensing viability of each cell we examined, we stimulated it with 5 mmol/l glucose at the end of the experiment, which invariably triggered a full-blown hyperpolarizing response (Fig. 8).

In this study, we document several previously unknown features of sugar-induced block of orexin neurons, cells critical for normal cognitive arousal and feeding behavior. First, 2-deoxyglucose, a nonmetabolizable glucose analog causing energy depletion, selectively mimics the inhibitory effects of glucose (Figs. 1–3). Second, stimulating ATP generation with lactate does not reproduce the effects of glucose (Fig. 5). Third, orexin cell glucosensing is not affected by glucokinase inhibitors, which block glucose entry into glycolysis (Fig. 6). Fourth, glucose-induced inhibition of orexin neurons exhibits a unique sugar-selectivity signature: It is caused by d-glucose, mannose (Fig. 7), and 2-deoxyglucose (Fig. 1) but not by l-glucose, galactose, α-methyl-d-glucoside, or fructose (Fig. 8). Together, these data indicate that behaviorally vital neurocircuits of the lateral hypothalamus contain pharmacologically novel glucose detectors that can operate independently of glucose metabolism.

To the best of our knowledge, this study is the first direct demonstration of metabolism-independent glucose sensing in glucose-inhibited neurons. Our evidence for this comes from direct electrophysiological measurements of glucosensing in identified orexin cells and is based on three complementary pharmacological observations: 1) specific and reversible stimulation of glucosensing responses by a nonmetabolizable glucose analog 2-deoxyglucose (Figs. 1–3); 2) lack of stimulation of glucosensing responses by neuronal energy fuel lactate (Fig. 5); and 3) Insensitivity of glucosensing responses to suppression of intracellular glucose metabolism to ATP with glucokinase blockers (Fig. 6). The latter experiment is an important functional complement to previous work that suggested that orexin neurons do not express glucokinase mRNA (35), because molecular determination of glucokinase expression is notoriously difficult.

Glucose-inhibited orexin neurons thus appear to operate very differently from classical glucose-excited neurons. In the latter, glucosensing responses are not triggered by 2-deoxyglucose, mimicked by lactate, and abolished by glucokinase inhibitors (36). This clearly indicates that glucose-excited neurons need glucose metabolism for their sugar-sensory activity, and it has been recently theorized that glucose-inhibited neurons follow a similar strategy (13,35). However, for glucose-inhibited neurons, such theories have so far been examined by measuring the effects of reducing glucokinase mRNA expression on cytosolic Ca²⁺ signals in neuronal cultures (13) and by looking at effects of pharmacological glucokinase inhibitors and activators on spontaneous Ca²⁺ signals (13,35). How these data relate to the coupling between extracellular glucose and membrane currents in situ is difficult to interpret. On the other hand, direct patch-clamp measurements from glucose-inhibited neurons of the ventromedial hypothalamus suggested that these cells share at least some similarities with the orexin neurons analyzed here. For example, like orexin cells (Fig. 5), the ventromedial glucose-inhibited neurons are not inhibited by lactate (37). A comprehensive evaluation of whether energy-independent glucosensing is found widely in glucose-inhibited cells requires the development of specific markers for identifying glucose-inhibited neurons in other brain regions, combined with direct measurements of their electrical activity. Currently, our study indicates that, in glucose-inhibited orexin neurons at least, conventional glucose metabolism is not required for sugar sensing.

Although the physiological importance of orexin neurons and their ability to sense glucose have been recognized for several years, the nature of the glucose detector in orexin cells remains elusive. Our study provides the first pharmacological characterization of this sugar-sensing device by examining the effects of other sugars on orexin cells. For the orexin cell glucosensor, we found activation by d-glucose, 2-deoxyglucose (Fig. 1), and mannose (Fig. 7), but not by fructose, d-galactose, α-methyl-d-glucoside, or l-glucose (Fig. 8). Thus, the glucosensing device in glucose-inhibited orexin cells appears pharmacologically distinct from that in glucose-excited neurons, because the latter is activated by galactose but not by 2-deoxyglucose (36).

The sugar selectivity profile defined here provides important information about the orexin cell glucose detector because known glucose-binding proteins exhibit characteristic sugar-selectivity signatures. For example, the glucose transporter protein hSGLT3, which acts as an extracellular glucose sensor not requiring transport or metabolism of the sugar for its sensing role, is activated by glucose and α-methyl-d-glucoside but not by galactose, fructose, or mannitol (38). In contrast, both glucose and galactose are active substrates of its relative, SGLT1. Another family of glucose transporters, GLUTs, which can also operate as glucose receptors independently of their ability to transport or metabolize sugars (39), exhibit yet different sugar selectivity. For example, GLUT2, unlike SGLT3, is activated by fructose. The sugar selectivity of the orexin cell glucosensor thus appears distinct from the glucose-binding proteins mentioned above: Unlike hSGLT3, it is not activated by α-methyl-d-glucoside; unlike SGLT1, it is not activated by galactose; and unlike GLUT2, it is not activated by fructose. We are not aware of any glucose-binding proteins that have similar sugar selectivity to the glucose detector in orexin neurons. This provides a valuable criterion for future screening of different glucose-sensing molecules as candidate molecular correlates for the orexin cell glucosensor.

Here, we studied neurochemically defined orexin neurons, whose physiological importance in the control of feeding and wakefulness is well established (19) and whose firing can lead to behavioral alterations if it changes by as little as 5 Hz (18). Understanding the mechanisms regulating the electrical activity of orexin can thus provide direct insights into cellular bases of vital behaviors. In higher animals, complex brains produce survival advantages by predicting and then avoiding danger before it becomes disruptive to bodily functions. The dissociation between sensing and use of metabolic fuels may allow the brain to sense glucose levels and to carry out compensatory adjustments before cellular metabolism is disrupted. It is tempting to speculate that the energy-independent glucose sensing described here can occur outside the glucose concentration range where extracellular glucose affects intracellular ATP levels. This would enable metabolism-independent glucose-inhibited neurons to act as “predictive” sensors, and metabolism-dependent glucose-excited neurons to act “emergency alarms” signaling dangerously low glucose levels. Comparisons of glucose sensitivity of lateral hypothalamic neurons suggest that glucose-inhibited neurons may sense glucose in a higher concentration range than glucose-excited neurons (26). Further elucidation of relative roles of glucose-excited and glucose-inhibited neurons are a key subject for future research. In this context, our finding that glucose-inhibited orexin neurons do not sense galactose (Fig. 8D) offers a potential new way to discriminate between physiological effects of activation of glucose-inhibited versus glucose-excited neurons, because the latter do sense galactose (36). For example, a suppression of feeding by glucose but not galactose (e.g., as reported by Kurata et al. [40]) may suggest an involvement of glucose-inhibited neurons. Future in vivo studies guided by the cellular information presented here should provide new insights into physiological and pathological impact of different glucosensing neurocircuits.

This study was funded by the Biotechnology and Biological Sciences Research Council.