The Regulation of Glucose-Excited Neurons in the Hypothalamic Arcuate Nucleus by Glucose and Feeding-Relevant Peptides

Male 14- to 21-day-old Sprague-Dawley rats were housed with their dams at 22–23°C on a 12-h light/dark cycle and given chow (Purina Rat Chow 5001) and water ad libitum. On the day of experiment, rats were anesthetized with ketamine/xylazine and transcardially perfused as described previously (13). Brains were rapidly removed and placed in ice-cold (slushy) oxygenated (95% O₂:5% C0₂) perfusion solution. Sections (350 μm) through the hypothalamus were maintained at 34°C in oxygenated high-Mg²⁺ low-Ca²⁺ artificial cerebrospinal fluid (ACSF; in mmol/l: 127 NaCl, 1.9 KCl, 1.2 KH₂PO₄, 26 NaHCO₃, 2.5 glucose, 9 MgCl₂, and 0.3 CaCl_2, osmolarity adjusted to ∼300 mOsm with sucrose, pH 7.4) for 30 min and allowed to come to room temperature until use. Slices were then transferred to normal oxygenated ACSF (2.4 mmol/l CaCl₂ and 1.3 mmol/l MgCl₂) for the remainder of the day. High-Mg²⁺ ACSF solutions (3.1 mmol/l MgCl₂:0.3 CaCl₂ or 11 mmol/l MgCl₂:2.4 CaCl₂) were used to block presynaptic input as described previously (13).

Immediately before use, brain slices were placed in a chamber (Warner Instruments) with the temperature controlled at 33–34°C. During recording, brain slices were perfused at 6 ml/min with normal oxygenated ACSF. Viable neurons were visualized and studied under infrared differential interference contrast microscopy as described previously (13). Current- and voltage-clamp recordings (standard and perforated patch whole-cell recording configuration) from neurons in the ARC were made using an Axopatch 1D amplifier (Axon Instruments), low pass–filtered at 1 kHz, and monitored using Axoscope software (Axon). Data were simultaneously digitized at 5 kHz (Digidata 1320A; Axon Instruments) and analyzed using pClamp9 software (Axon). The junction potential between the patch pipette and the bath solutions was nulled before the gigaohm seal formation. For standard whole-cell recording, borosilicate pipettes (1.5–3.5 MΩ; Sutter Instruments) were filled with an intracellular solution containing the following (in mmol/l): 128 K-gluconate, 10 KCl, 10 KOH, 10 HEPES, 4 MgCl₂, 0.05 CaCl₂, 0.5 EGTA, 2 Na₂ATP, and 2 lucifer yellow, pH7.2, with osmolarity adjusted to 290–300 mOsm with sucrose. Only neurons with access resistance <15 MΩ were used. For perforated patch whole-cell recording, amphotericin was included in the patch pipette (final concentration 240 μg/ml, stock 60 mg/ml DMSO). Membrane potential and APF were allowed to stabilize for 10–15 min after the formation of a gigaohm seal. Neurons whose access resistance exceeded 40 MΩ were rejected. After recording, the membrane was ruptured to allow lucifer yellow access into the cell for 5 min for future immunocytochemical labeling. No differences in glucose sensitivity were observed between standard and perforated patch recording configurations. Extracellular glucose and other chemicals (Sigma) were added to the ACSF perfusion solution as described in the figures. Input resistance was calculated from the change in membrane potential in response to small 500-ms hyperpolarizing pulses (−10 or −20 pA) given every 3 s as described previously. The reversal potential was calculated from the change in membrane current in response to voltage steps from −120 mV to +60 mV from a holding potential of −60 mV. Steady-state currents were determined by measuring data points within the last 5 ms of the 200-ms voltage pulses.

Brain slices were prepared as described above, and sections (400 μm) were immediately transferred to zero-glucose/zero-pyruvate Hibernate media (Brain Bits) supplemented with 2% (vol/vol) B-27, 2.5 mmol/l glucose, 1 mmol/l l-lactate, 0.23 mmol/l pyruvate, 0.5 mmol/l l-glutamine, and 100 units/ml penicillin/streptomycin (Hibernate). The VMH was dissected and then digested in papain-containing media. Tissue was incubated for 30 min in a 30°C water bath with a platform rotating at 100 rpm and then rinsed and triturated. The cell suspension was centrifuged 4 min at room temperature at 1,000 rpm, and the pellet was resuspended with 1–1.5 ml zero-glucose/zero-pyruvate Neurobasal (Invitrogen) with 2% B27, 2.5 mmol/l glucose, 1 mmol/l l-lactate, 0.23 mmol/l pyruvate, 0.5 mmol/l l-glutamine, 100 units/ml penicillin/streptomycin, and 5 ng/ml bFGF2 (Neurobasal) (16). Cells were plated in Neurobasal on culture dishes and used within 4 days. Single ATP-sensitive K⁺ channel (K_ATP channel) currents were recorded and analyzed as described previously (17). Single-channel activity was quantified as the mean current (I) passed through the K_ATP channel for 60-s segments of data, as follows: I = NP_oi, where N is the number of channels, P_o is the open probability, and i is the single channel current amplitude.

For immunocytochemistry, all solutions were prepared in 0.1 mol/l PBS, pH 7.4, with 0.2% triton X and used at room temperature, except where otherwise noted. After recording, slices were fixed in 4% paraformaldehyde (without triton X) then sunk in 4°C 30% sucrose (also without triton X). Tissue was frozen, subsectioned at 20 μm and mounted on subbed slides. The slides were treated with 1% H₂O₂ and then blocked in 10% normal donkey serum (Jackson Immunoresearch). After an overnight treatment in 10% donkey serum with 0.02% rabbit anti-POMC precursor (Phoenix Pharmaceuticals), slides were placed in 0.2% biotinylated goat anti-rabbit IgG (Vector Laboratories). Sections were then processed using an ABC kit (Vector Laboratories) followed (as previously described) (18) by a tyramide signal amplification system (Perkin Elmer) using CY3 visualization. To amplify the lucifer yellow label, slides were retreated with 1% H₂O₂, followed by an avidin and biotin blocking kit (Vector Laboratories). They were next placed in 10% normal horse serum (Jackson Immunoresearch) and then held overnight in 1:400 biotinylated rabbit anti-lucifer yellow (Molecular Probes). The ABC procedure, followed by tyramide signal amplification system, was performed as above, this time using fluorescein visualization. After final washing, sections were coverslipped with Vectashield (Vector Laboratories) and analyzed using fluorescence microscopy. Subsections were visualized using each laser channel sequentially as well as simultaneously to avoid bleed-through between channels.

Electrophysiological and immunocytochemical techniques were used to characterize 421 ARC neurons with regard to their sensitivity to glucose, leptin, insulin, NPY, and α-MSH. In recordings made throughout the entire ARC, two types of intrinsically glucose-sensing neurons were found that were similar to the glucose-excited and glucose-inhibited neurons, which we have described previously in the VMN (13). These two subtypes of glucosensing neurons were unequally distributed between the medial and lateral ARC. When recordings were made predominantly in the medial ARC, 14% (12 of 83) of the neurons were of the glucose-inhibited subtype, whereas only 4% (3 of 83) were of the glucose-excited subtype. In contrast, when recordings were made in the lateral ARC, 13% (45 of 338) were of the glucose-excited subtype, and only 1% (3 of 338) were of the glucose-inhibited subtype (Fig. 1). The densest population of glucose-excited neurons was actually located on the lateral border of the ARC, near the cell-poor region between the ARC and the VMN. This study focuses predominantly on the ARC glucose-excited neurons.

The general characteristics of ARC glucose-excited neurons are shown in Fig. 2. Glucose-excited neurons hyperpolarize their membrane potential and decrease their APF as extracellular glucose levels decrease. This is accompanied by a decrease in input resistance (Fig. 2A, top trace). The latency of the glucose response was 5.6 ± 0.3 min (n = 38). We believe that this is caused by the time required for diffusion of glucose through the brain slice, as well as the time required for glucose metabolism to raise intracellular ATP, alter K_ATP channel activity, and finally decrease APF. The glucose response of glucose-excited neurons is intrinsic, as indicated by the persistence of this response in high-Mg²⁺ ACSF, which abolishes presynaptic input (n = 12) (Fig. 2A, bottom trace). Current-voltage relations show that the glucose-induced current reverses at −79 ± 3 mV (n = 27), which is close to the theoretical K⁺ equilibrium potential of −99 mV in our solutions (Fig. 2B). This, together with the observation that the inhibitory effect of decreased glucose is reversed by the sulfonylurea tolbutamide (100 μmol/l) (not shown), indicates that the K_ATP channel is responsible for the glucose-induced current (n = 8). Finally, decreasing extracellular glucose from 2.5 to 0.5 mmol/l or the presence of the glucokinase inhibitor alloxan (4 mmol/l) in 2.5 mmol/l glucose reversibly increases the K_ATP channel current by 222 ± 115% (n = 5; P = 0.05, Students paired t test) in cultured VMH neurons (Fig. 3). The single-channel current that was induced by both alloxan and 0.5 mmol/l glucose was blocked by tolbutamide and had a conductance of ∼46 pS, as we have shown before for the K_ATP channel in hypothalamic neurons (17).

We next addressed the question of whether ARC glucose-excited neurons are only sensitive to profound decreases in brain glucose analogous to those that would initiate the counterregulatory response to hypoglycemia in vivo (e.g., 0.1 mmol/l) (12) or whether they are capable of sensing the small physiological changes in glucose concentration associated with daily glucose homeostasis (e.g., from 2.5 to 1 or 5 mmol/l) (10,12). As indicated in Fig. 4, APF (Fig. 4A), input resistance (B), and K_ATP channel currents (C) show dose-dependent responses to changes in extracellular glucose within the physiological range. APF and input resistance were well fit by a rectangular hyperbole, with the greatest glucose sensitivity (steepest slope) occurring <2 mmol/l (19). Thus, ARC glucose-excited neurons may be capable of sensing meal-induced fluctuations in extracellular glucose levels (10,12). In addition to sensing glucose, we hypothesized that ARC glucose-excited neurons would also be sensitive to leptin and insulin. However, ARC glucose-excited neurons are not sensitive to 10 nmol/l leptin in the presence of either 2.5 (n = 7) (Fig. 5A) or 0.1 mmol/l (n = 5) (not shown) glucose. This is in sharp contrast to VMN glucose-excited neurons, which are inhibited by 10 nmol/l leptin in the presence of 2.5 mmol/l glucose (n = 4) (Fig. 5B). In agreement with Spanswick et al. (8), the effect of leptin on VMN neurons is blocked by tolbutamide, indicating that leptin inhibits VMN glucose-excited neurons by opening the K_ATP channel (n = 1) (Fig. 5B, bottom trace). Insulin modulates ARC glucose-excited neurons in a glucose-dependent manner. We previously showed that insulin opened the K_ATP channel and inhibited ARC glucose-excited neurons in 10 mmol/l glucose (9). However, in this study, seven of eight ARC glucose-excited neurons were not affected by insulin (100 or 300 nmol/l) in the presence of 2.5 mmol/l glucose, whereas the eighth neuron actually depolarized and increased its APF by 180%. On the other hand, four of five ARC glucose-excited neurons that were exposed to the same level of insulin in the presence of 0.1 mmol/l glucose transiently depolarized by 3 ± 1 mV and increased their APF by 215 ± 115% (P = 0.05, Students paired t test) (Fig. 6). Three of these latter four neurons showed no response to insulin in the presence of 2.5 mmol/l glucose (the fourth was not tested in 2.5 mmol/l glucose).

POMC neurons in the lateral ARC (where we find glucose-excited neurons) have both NPY receptors and melanocortin autoreceptors. Thus, we postulated that ARC glucose-excited neurons might respond to NPY and α-MSH, and that they in fact belong to the POMC neuronal population (4,20). Indeed, NPY (100 and 200 nmol/l) (20,21) hyperpolarizes glucose-excited neurons by 4 ± 1 mV and completely prevents them from firing action potentials (n = 6) (Fig. 7A, top). This response persists in high-Mg²⁺ ACSF, indicating that NPY acts directly on ARC glucose-excited neurons (n = 3) (Fig. 7A, bottom). On the other hand, α-MSH (600 nmol/l) (21,22) caused a 3 ± 1 mV depolarization and increased APF by 37 ± 9% in ARC glucose-excited neurons (n = 6) (Fig. 7B, top trace). The response to α-MSH persisted in the presence of high Mg²⁺ in two of these neurons (Fig. 7, bottom trace), whereas the response was abolished in two others (not shown). This indicates that α-MSH can modulate the activity of ARC glucose-excited neurons both directly and presynaptically. Finally, we investigated whether ARC glucose-excited neurons express the POMC peptide, using double-labeled immunocytochemistry. None of the eight glucose-excited neurons that we evaluated showed POMC immunoreactivity, suggesting that at least these glucose-excited neurons were not POMC neurons (Fig. 8).

Mayer and Thomas’s (23) glucostatic hypothesis proposes that reduced glucose availability is sensed by the brain as a trigger for meal initiation. Campfield and Smith (24) later demonstrated that small dips in plasma glucose preceded meal initiation. Although decreased glucose availability may serve as a signal for meal initiation, it is more likely to be a critical signal for integrating autonomic activity and whole-body metabolism (25). The present studies demonstrate that ARC glucose-excited neurons respond in a dose-dependent fashion to physiological changes in extracellular glucose. This supports the hypothesis that the brain is capable of sensing small changes in glucose associated with daily (meal-to-meal) fluctuations in extracellular glucose. Interestingly, ARC glucose-excited neurons are very sensitive to decreases in extracellular glucose levels between 1.5 and 0.5 mmol/l. This raises the possibility that decreased activity of ARC glucose-excited neurons signals glucose deficit and possibly contributes to meal initiation. This is consistent with the decrease in VMH glucose levels from 1.4 to 0.7 mmol/l seen in the fed versus fasted state (12). ARC glucose-excited neurons show a slight response to increased extracellular glucose levels from 5 to 10 mmol/l. Under physiological conditions, brain glucose levels generally are ∼20–30% of plasma levels (10,12). Thus, except in extreme hyperglycemic diabetic conditions, it is unlikely that extracellular brain glucose levels ever exceed 5 mmol/l. However, neuronal processes in the medial ARC do extend into the median eminence, where the blood-brain barrier is permeable (26). Thus, it is possible that ARC glucose-excited neurons are exposed to higher glucose levels than the rest of the brain. Our data showing that they respond to increased glucose levels >5 mmol/l are consistent with this theory. However, it is important to note that this response is small compared with that <2.5 mmol/l glucose.

The mechanism by which glucose-excited neurons detect small changes in extracellular glucose is similar to that in pancreatic β-cells. We showed previously that a subpopulation of isolated VMN glucose-excited neurons, identified by increasing calcium oscillations as extracellular glucose levels increase from 0.5 to 2.5 mmol/l, express glucokinase mRNA (14). The glucokinase inhibitor alloxan (much like low glucose) decreases these calcium oscillations (14). In the present study, we directly evaluated single K_ATP channel currents using the cell-attached patch clamp configuration in cultured VMH glucose-excited neurons. Here, alloxan and 0.5 mmol/l glucose increased K_ATP single-channel currents to a similar degree. This effect was completely reversible, indicating that it was not caused by any possible neurotoxic effect of alloxan. In isolated neurons, similar results were obtained with three other glucokinase inhibitors that have no neurotoxicity (14). This supports the hypothesis that inhibition of glucokinase in glucose-excited neurons reduces glycolytic flux, and consequently ATP production, causing the K_ATP channel to open (14,15).

Leptin, in 10 mmol/l glucose, opens the K_ATP channel and inhibits VMH glucose-excited neurons (8). However, in the present study of ARC glucose-excited neurons, leptin did not alter APF in the presence of 2.5 or 0.1 mmol/l glucose. This may be caused by differences in extracellular glucose levels used in these studies (i.e., 10 vs. 2.5 or 0.1 mmol/l). However, we and others (8) found that the same concentration of leptin (10 nmol/l) used here opened the K_ATP channel and inhibited VMN glucose-excited neurons in 2.5 mmol/l glucose. It is more likely that this population of ARC glucose-excited neurons, which are located at the dorso-lateral border of the POMC field, do not possess leptin receptors. We have previously shown that only 30% of VMN glucose-excited neurons express mRNA for the long form of the leptin receptor (ObRb) (15). It is possible that the ObRb is heterogeneously distributed among glucose-excited neurons in the ARC as well.

Although leptin had no effect on our ARC glucose-excited neurons, insulin did modulate their activity in a glucose-dependent fashion. We previously showed that insulin in the presence of 10 mmol/l glucose opened the K_ATP channel and inhibited VMH glucose-excited neurons (9). We report here that there is no effect of insulin on ARC glucose-excited neurons in 2.5 mmol/l glucose; however, in the presence of 0.1 mmol/l glucose, insulin actually increases their APF. We hypothesize that there are multiple mechanisms (i.e., excitatory and inhibitory) by which insulin modulates the activity of ARC glucose-excited neurons. In high glucose, the inhibitory effect on the K_ATP channel may predominate. However, as glucose levels decrease, the K_ATP channel moves progressively toward the open state. As this occurs, the effect of opening the K_ATP channel has less of an overall effect on neuronal activity. In 2.5 mmol/l glucose, these opposing effects may be balanced, with the excitatory effect becoming dominant as glucose levels decrease to 0.1 mmol/l. The excitatory effect of insulin may be attributable to increased glucose transport into glucose-excited neurons because we recently found a subpopulation of VMN glucose-excited neurons expressing mRNA coding for both the insulin receptor and the insulin-dependent glucose transporter GLUT4 (15). An interaction between glucose and insulin is not surprising because the brain is never exposed to either insulin or glucose in isolation.

NPY and α-MSH also modulate the activity of ARC glucose-excited neurons. The observation that NPY inhibits and α-MSH stimulates ARC glucose-excited neurons is consistent with the opposing roles of these peptides in the regulation of food intake and energy balance (1,2,4,27) as well as in neuronal excitability (28–30). NPY is a potent orexigen whose expression is upregulated by fasting (1,31). On the other hand α-MSH produces satiety and increases energy expenditure (4). These data lead to the overall hypothesis that increased activity in glucose-excited neurons reflects adequate energy stores. Thus, in both steady-state glucose levels and in the presence of α-MSH, these neurons are almost maximally active. However, when glucose levels fall or NPY is present, their activity ceases. The excitatory effect of insulin in 0.1 mmol/l glucose is consistent with this hypothesis. That is, in the presence of insulin, a transient drop in glucose is not as serious an indicator of reduced glucose availability, and therefore glucose-excited neurons are not totally inhibited. It is important to note, though, that the excitatory effect of insulin in low glucose was transient. Thus, even though insulin attenuated the inhibitory effect of low glucose, it did not abolish it, emphasizing that low glucose is a very powerful signal to the brain.

Finally, ARC glucose-excited neurons were not POMC immunoreactive, despite their location in the lateral ARC (4). Most glucose-excited neurons were found intermixed with POMC neurons in the lateral-most aspect of the POMC field. However, because glucose-excited neurons were not distributed through the entire POMC field, it is unlikely that all POMC neurons are glucose excited. Furthermore, POMC neurons respond to leptin, whereas ARC glucose-excited neurons did not (20). One technical issue that might have led to a negative result is dialysis of POMC-containing vesicles during standard whole-cell recording. This could prevent us from detecting labeling. However, this is unlikely because of the size of peptide-containing vesicles. Intracellular compartmentalization would also act to prevent dialysis. However, to rule out this possibility, we used perforated patch recording and ruptured the membrane 5 min before concluding the experiment. This allowed lucifer yellow to mark the cell, but it presumably would not allow time for significant dialysis of the large POMC vesicles. Using this technique did not reveal POMC immunoreactivity. Therefore, we conclude that ARC glucose-excited neurons are not of the POMC phenotype, but rather downstream from the POMC and NPY neurons.

In conclusion, ARC glucose-excited neurons dose-dependently increase APF and decrease K_ATP channel currents as glucose levels increase throughout the entire physiological range; however, they are particularly sensitive to decreases in extracellular glucose. Their activity is increased under conditions when glucose levels are adequate (e.g., steady-state to high glucose, α-MSH, and insulin) and decreased in conditions of glucose deficit (e.g., low glucose and NPY). The fact that they are not POMC immunoreactive, but instead are sensitive to α-MSH and NPY, suggests that they are downstream from these important neuronal populations. Overall, the data presented herein place ARC glucose-excited neurons in a pivotal position to serve as putative integrators of information regarding peripheral and central energy homeostasis.

This study was supported, in part, by National Institutes of Health Grants DK55619 and DK53181 and the Research Service of the Veterans Administration.

We would like to thank Clint Rosenfeld and Dr. Lester Sultatos for their valuable assistance in fitting and interpreting our glucose dose-response data as well as Drs. Hreday Sapru and Vineet Chitravanshi for assistance with subsectioning of the brain slices.