Glucokinase Is the Likely Mediator of Glucosensing in Both Glucose-Excited and Glucose-Inhibited Central Neurons

All work was approved by the institutional animal care and use committee of the East Orange VA Medical Center. Animals were housed individually on a 12:12 h light:dark schedule at 22–23°C with food (Purina rat chow #5001) and water available ad libitum, except where otherwise specified. For all but the Ca²⁺ imaging studies, outbred adult male Sprague-Dawley rats were either purchased (Charles River) or bred from an in-house colony of rats selectively bred from the outbred Sprague-Dawley stock for their propensity to gain weight on a high-energy diet (diet-induced obesity [DIO]) or resist such weight gain (diet resistance [DR]) (14). Animals were enrolled in studies between the ages of 2 and 5 months. For the Ca²⁺ imaging studies, outbred male pups between the ages of 12 and 20 days were used.

Under ketamine/xylazine anesthesia, right internal carotid artery PE10 catheters were implanted, facing cephalad (15). Rats were acclimated to the infusion procedure by daily carotid infusions of 0.9% saline for 10 min at 40 μl/min. After 3–4 days of saline infusion, rats were infused for 1 h with either glucose (4 mg · kg⁻¹ body wt · min⁻¹, n = 22) or equiosmolar mannitol (4 mg · kg⁻¹ · min⁻¹, n = 13) to perfuse the forebrain without altering plasma glucose or insulin levels (15). At 3 h after infusion onset, pentobarbital-anesthetized rats were perfused with 4% paraformaldehyde. Brains were removed, sunk in sucrose, and sectioned at 40 μm for c-fos immunocytochemistry with rabbit anti-fos antibody and diamino benzidine, with nickel intensification as chromagen (15).

GK mRNA in situ hybridization was performed in three rat models with known defects in brain glucosensing:

• Rats selectively bred to express the DIO trait have defective central glucosensing (14,16). Seven ad libitum–fed, selectively bred DIO male rats were compared with seven selectively bred DR rats with normal glucosensing. Their brains were removed for assessment of GK mRNA between 0800 and 1000 h.

• The outbred population of Sprague-Dawley rats contains obesity-prone (DIO-prone) and obesity-resistant (DR-prone) rats that differ in intake, body weight, and fat only when fed a diet of moderate fat and caloric density (17). DIO-prone rats have defective glucosensing (18–20) and can be identified prospectively by their high 24-h urinary norepinephrine excretion, as compared with DR-prone rats (17). Of 38 chow-fed, 3-month-old male rats, the 14 with the highest norepinephrine levels, as assessed by high-performance liquid chromatography (17), were designated as DIO-prone, and the 14 with the lowest levels were designated as DR-prone. Half of each of these groups were maintained on ad libitum chow and half were fasted for 48 h. Animals were decapitated between 0800 and 1000 h.

• A single hypoglycemic bout impairs central glucosensing because it significantly diminishes the counterregulatory response to subsequent hypoglycemia (21). Here, two groups of outbred rats were prospectively defined as DIO-prone (n = 11) versus DR-prone (n = 12) by their 24-h urine norepinephrine excretion. All rats were food deprived overnight, and then six DIO-prone and six DR-prone rats were injected by tail vein with insulin, 5 units/kg. The remaining five DIO-prone and six DR-prone rats were injected with 0.9% saline. Tail blood glucose levels fell to nadirs of ∼1.7 mmol/l over 30–60 min in insulin-injected rats. At 70 min, 1 g/kg glucose was given intraperitoneally, and food was returned. Rats were killed 48 h after the injections.

Rats were decapitated and their brains were removed, frozen on dry ice, and maintained at −80°C until sectioned at 15 μm. Sense and antisense riboprobes were generated from plasmids containing inserts for mRNA of rat GK (GenBank no. M25807, bases 1–1422), neuropeptide Y (NPY) (no. M20373, bases 1–511), Kir6.2 (no. X97041, bases 107–599), GAD65 (no. M72422, bases 1–1966), pro-opiomelanocortin (POMC) (no. X03176.1, POMC exon 3, bases 46–200; and J00759.1, POMC exon 2, bases 88–742), and orexin (no. AF041241, bases 187–473). Riboprobes were labeled with ³⁵S (GK), fluorescein (GK), or digoxigenin (NPY, POMC, Kir6.2, GAD65, and orexin). Single and double in situ hybridizations were performed as previously described (12). In other conditions, double-label in situ hybridization was carried out using anti-fluorescein antibodies with fluorescent detection and anti-digoxigenin antibodies with NBT/BCIP (nitroblue tetrazolium chloride/5-brom-4-chloro-3-indolyl phosphate) visualization (9).

Double-label immunocytochemistry and in situ hybridization used techniques similar to those above (9). Slides were hybridized with ³⁵S-labeled GK riboprobe and carried through the final wash with 0.1× sodium chloride–sodium citrate at 60°C (see above). After blocking in 1% BSA, rabbit anti–tyrosine hydroxylase (Chemicon), rabbit anti–neuron-specific nuclear protein (Chemicon), or rabbit anti–glial fibrillary acidic protein (GFAP) (Dako) were applied at a 1:1,000 dilution. Biotinylated goat anti-rabbit antibody (Vector) was then applied at a 1:250 dilution, followed by Vectorstain peroxidase ABC kit (Vector). Peroxidase was visualized using diaminobenzidine or Vector SG reagent as chromagens. The sections were fixed in gluteraldehyde and dipped for autoradiography. Double-labeled slides were observed under bright-field illumination using differential focusing to separate silver grains and chromagen reaction. Photography at 60–100× magnification was performed using light-field illumination to visualize the chromagen reaction, with a separate exposure using dark-field illumination in a different focal plane to visualize autoradiographic grains. The two images were then separately adjusted for contrast and combined digitally.

Slices were taken through the hypothalamus at the level of the paraventricular nucleus rostrally, the infundibular stalk caudally, and the optic tract laterally to include the paraventricular, arcuate (ARC), dorsomedial, and ventromedial hypothalamic (VMN) nuclei. A second sample was dissected from the caudal hindbrain at the level of the middle to caudal end of the nucleus tractus solitarius (NTS) to include the nuclear tract and the area postrema. Samples were weighed, frozen on dry ice, and homogenized. RNA was extracted using a Purescript RNA isolation kit (Gentra Systems). Samples were diluted to yield RNA extracted from 0.1 mg total tissue. They were reverse-transcribed using a Superscript first-strand synthesis kit for RT-PCR (Gibco) and then treated with DNase (Promega). PCR was performed (12) using primers for rat GK (forward GTGGTGCTTTTGAGACCCGTT, reverse TTCGATGAAGGTGATTTCGCA; corresponding to GenBank no. M25807, bases 965–985 and 1305–1285, respectively) or cyclophilin (forward GGGGAGAAAGGATTTGGCTA, reverse ACATGCTTGCCATCCAGCC; corresponding to no. M19533, bases 166–185 and 422–404). The magnesium concentration was 15 mmol/l.

Single VMN neurons were prepared from the brains of 12- to 20-day-old outbred Sprague-Dawley rats with slight modifications of the procedures of Routh et al. (7). After decapitation, brains were rapidly removed and placed in ice-cold oxygenated (95% O₂/5% CO₂) perfusion buffer (in mmol/l: 2.5 KCl, 1.25 NaH₂PO₄, 28.0 NaHCO₃, 7.0 MgCl₂, 0.5 CaCl₂, 7.0 glucose, 1 ascorbate, and 3 pyruvate). Coronal hypothalamic sections (350 μm) containing the VMN were cut on a vibratome and then immediately transferred to Hibernate-A medium supplemented with 0.1% (vol/vol) B-27. After shaking for 10 min at 34°C, slices were transferred to Hibernate-A medium containing papain (2 mg/ml), incubated for 30 min at 30°C, and allowed to recover for 10 min in enzyme-free solution. The VMN was punched with a 500-μm blunt needle, placed into Hibernate-A medium, and gently triturated using a fire-polished Pasteur pipette. Triturated pieces were allowed to settle for 2 min. The supernatant was removed, and the sediment was resuspended and subjected to two further triturations. The combined cells from the three triturations were resuspended, plated on poly-d-lysine (5 μg/cm²)–coated glass coverslips, and maintained in a HEPES-buffered balanced salt solution (in mmol/l: 135 NaCl, 5 KCl, 1 CaCl₂, 1 MgCl₂, 10 HEPES, pH 7.4, and 2.5 glucose). Fluorescent imaging measurement of intracellular Ca²⁺ ([Ca²⁺]_i) was carried out using fura-2, essentially as previously described (22). VMN neurons attached to coverslips were loaded with 2.5 μmol/l fura-2 acetoxymethyl ester (Molecular Probes) for 20 min in Hanks’ balanced salt solution (HBSS) containing 0.25% (wt/vol) fatty acid–free BSA. Cells were washed twice with HBSS to remove free fura-2 and transferred to a thermostatically regulated microscope chamber (37°C). Fura-2 fluorescence images, acquired by alternating excitation at 340 and 380 nm and emissions (420–600 nm), were collected using a cooled, charge-coupled device camera. Values for [Ca²⁺]_i were calculated from the 340/380-nm fluorescence ratio after correction for autofluorescence. Autofluorescence values were obtained at the end of each experiment by permeabilizing the neurons with digitonin (20 μg/ml) in an intracellular-like medium containing a Ca²⁺-EGTA buffer and 2 mmol/l ATP/Mg. The fura-2 calibration parameters were determined in vitro using a half-maximal binding (K_d) value of 224 nmol/l (23).

All experiments began with neurons held at 2.5 mmol/l glucose (3). Neurons that showed spontaneous [Ca²⁺]_i fluctuations in 2.5 mmol/l glucose and cessation of these fluctuations in 0.5 mmol/l glucose were defined as GE. Those without spontaneous [Ca²⁺]_i fluctuations in 2.5 mmol/l glucose, but with spontaneous fluctuations at 0.5 mmol/l glucose, were defined as GI (3). Those neurons that exhibited no change in their spontaneous fluctuations of [Ca²⁺] _i with alterations of extracellular glucose were designated as nonglucosensing (NG). Then, various doses of four different selective inhibitors of GK were applied. These included freshly prepared alloxan (0.1–10 mmol/l) (24,25), mannoheptulose (1–5 mmol/l) (4,5,24,26), glucosamine (5–10 mmol/l) (27), and N-acetylglucosamine (1–4 mmol/l) (28). In addition, catalase (4,000 units/ml) or dimethylthiourea (4 mmol/l) was added to GE and GI neurons in the presence of alloxan (4 mmol/l) to exclude the possibility that the inhibitory action of alloxan on GE neurons might be caused by the formation of oxygen free radicals (29). Finally, neurons were tested with either 50–200 μmol/l tolbutamide or 20 nmol/l glutamate for responsiveness.

In situ hybridization autoradiographic films were read by an experimentally blinded observer. Measurements were made of the areas (mm²) over which the hybridized, radiolabeled GK probe exposed the film in the paraventricular nucleus, ARC, VMN, and dorsomedial hypothalamic and medial amygdalar nuclei (30). For each brain, the three largest areas in each of these nuclei were selected. The product of area and optical densities of these areas was then averaged as arbitrary units to provide a semiquantitative estimate of the amount of GK mRNA expressed (21). Treatment effects were then analyzed by one-, two-, or three-way ANOVA with post hoc analysis by Scheffe’s F test.

In situ hybridization demonstrated that GK mRNA was present only in cells with neuronal phenotypes (neuron-specific nuclear protein, Kir6.2, NPY, POMC, GAD65, tyrosine hydroxylase) but not in astrocytes (GFAP) (Fig. 1). However, GK expression by oligodendroglia or microglia cannot be ruled out because these cell types were not assessed. Furthermore, lateral hypothalamic orexin neurons did not express GK mRNA (Fig. 1), even though GK mRNA is present in the lateral hypothalamus, as assessed by RT-PCR (12). Similarly, although we previously found no GK mRNA by in situ hybridization in the area postrema–NTS (12), GK mRNA was demonstrable there by RT-PCR (Fig. 2).

Bilateral forebrain perfusion by intracarotid glucose infusions increased c-fos expression in a number of forebrain areas (Fig. 3B, E, and H), as compared with equiosmolar mannitol control infusions (Fig. 3A, D, and G). This suggests direct or indirect activation of GE neurons. This glucose-induced c-fos expression paralleled GK expression in several brain areas. In the paraventricular nucleus, both GK (Fig. 3C) and c-fos (Fig. 3B) expression were predominantly seen in the ventral tier and the dorsal cap (not shown) of cells where autonomic efferent neurons reside (31). In the medial amygdalar nucleus, GK mRNA expression was concentrated in the postero-dorsal and postero-ventral portions (Fig. 3I), where the greatest number of c-fos–expressing cells was seen (Fig. 3H). There was some mismatch between the relative quantity and/or subnuclear localization of GK mRNA (Fig. 3F) versus c-fos (Fig. 3E) expression in the ARC, VMN, and dorsomedial nucleus. In the ARC, GK mRNA expression was relatively uniform, whereas c-fos–expressing cells were relatively sparse and more laterally placed in areas where POMC neurons are located (32). In the VMN, GK mRNA was concentrated at the ventrolateral and dorsomedial poles, whereas c-fos–expressing cells were relatively sparse and mostly located in the dorsomedial subdivision. In the dorsomedial nucleus, GK expression was relatively diffuse, whereas only a few scattered cells expressed c-fos. These mismatches may be due to the fact that both GE and GI neurons are intermixed in these areas. Although both GE and GI neurons appear to express GK (see below), only GE neurons would be activated and, therefore, express c-fos when ambient glucose levels were raised.

Three broad categories of VMN neurons were identified by their changes in [Ca²⁺]_i oscillations in response to altered ambient glucose concentrations. GE neurons increased [Ca²⁺]_i oscillations when glucose levels were raised to 2.5 mmol/l, and they reduced oscillations at 0.5 mmol/l (Fig. 4A, C, D, E, F, J, K, and L). GI neurons decreased [Ca²⁺]_i oscillations when glucose levels were raised to 2.5 mmol/l and increased them when glucose was reduced to 0.5 mmol/l (Fig. 4B and G). NG neurons were of two types. Most were inactive at either 0.5 mmol/l or 2.5 mmol/l glucose (Fig. 4H and I), whereas some showed spontaneous [Ca²⁺]_i oscillations at either concentration (data not shown). Neither type of NG neuron altered its [Ca²⁺]_i oscillations when glucose concentrations changed. Of 109 VMN neurons evaluated, 21 (19.3%) were GE, 15 (13.8%) were GI, and 73 (67.0%) were NG. All GE neurons responded to the sulfonylurea tolbutamide by increasing [Ca²⁺]_i oscillations. However, 10–30% of both GI and NG neurons also responded to 50–200 μmol/l tolbutamide by increasing their [Ca²⁺]_i oscillations (Fig. 4B and H) (3).

Four inhibitors of GK activity were assessed in these neurons (Table 1). A “response” in GE neurons held at 2.5 mmol/l glucose was defined as abolition of [Ca²⁺]_i oscillations in the presence of the inhibitor. A “response” in GI neurons held at 2.5 mmol/l glucose was defined as an increase in [Ca²⁺]_i oscillations in the presence of the inhibitor. The majority of the trials were carried out with alloxan, which had an half-maximal inhibitory concentration (IC₅₀) of ∼3 mmol/l for inhibition of [Ca²⁺]_i oscillations in GE neurons between 0.1 and 10 mmol/l. Overall, 71% of GE and 72% of GI neurons responded to alloxan (Table 1; Fig. 4C, D, E, F, and G). In GE neurons, the alloxan effect was reversed by glutamate (Fig. 4C), tolbutamide (Fig. 4D), or a glucose concentration approximately twofold greater than that of alloxan (data not shown). In addition, alloxan’s effect was unaltered in either GE or GI neurons by the addition of the free radical scavengers catalase (4,000 units/ml) (Fig. 4E) or dimethylthiourea (4 mmol/l) (Fig. 4F; data not shown for GI neurons). NG neurons were unaffected by alloxan and could still be stimulated by either tolbutamide (Fig. 4H) or glutamate (Fig. 4I). Glucosamine (5–10 mmol/l) (Table 1; Fig. 6J), N-acetylglucosamine (1–4 mmol/l) (Table 1; Fig. 4K), and mannoheptulose (1–5 mmol/l) (Table 1; Fig. 4L) inhibited [Ca²⁺] _i oscillations in GE neurons and increased [Ca²⁺]_i oscillations in GI neurons held at 2.5 mmol/l glucose. The four GK inhibitors reduced [Ca²⁺]_i oscillations in 29–100% of GE neurons and increased them in 72–100% of GI neurons (Table 1). Although the number of neurons examined with glucosamine, N-acetylglucosamine, and mannoheptulose was relatively small, glucosamine was the most effective inhibitor, and N-acetylglucosamine was the least effective in GE neurons. On the other hand, alloxan was least effective in GI neurons, yet it still stimulated [Ca²⁺]_i oscillations in 72% of them (Table 1).

GK mRNA was examined by in situ hybridization in three models of defective central glucosensing. First, selectively bred DIO rats had a selective, 44% increase in GK mRNA expression in the ARC as compared with DR rats (P = 0.001) (Fig. 5A). Second, outbred DIO-prone rats had a generalized increase in GK mRNA expression across all five areas assessed (F [4,109] = 3.26, P = 0.014) (Fig. 5B). By post hoc analysis, this increase, as compared with DR-prone rats, was significant in the ARC (124%; P = 0.001), dorsomedial hypothalamic (68%; P = 0.01), and medial amygdalar (82%; P = 0.01) nuclei. Third, rats studied at 48 h after a prior bout of hypoglycemia had a generalized increase in GK mRNA expression across all regions compared with saline-treated rats (F [1,63] = 8.83, P = 0.004) (Fig. 6A). These increases reached post hoc statistical significance in the VMN (F [1,18] = 9.39, P = 0.001) of both DR-prone (37% higher) and DIO-prone (26% higher) rats subjected to hypoglycemia and in the medial amygdalar nucleus (F [1,18] = 20.41, P = 0.0001) for both DR-prone (82% higher) and DIO-prone (60% higher) phenotypes. On the other hand, there were no differences in GK expression, regardless of DR- or DIO-prone phenotype, associated with a 48-h fast (Fig. 6B).

The current studies tested the hypothesis that GK is a critical mediator of glucosensing in neurons. We previously showed that GK mRNA is selectively expressed in brain areas containing mixed populations of both GE and GI neurons. A notable exception was the NTS, where in situ hybridization showed no signal (12). But RT-PCR here shows GK mRNA to be present in that area. In addition, we show that GK mRNA is expressed in a variety of neuropeptide and transmitter neurons involved in glucose- and metabolic-sensing brain areas. Also, the subnuclear distribution of GK mRNA within selected brain nuclei overlaps significantly with in vivo glucose-activated c-fos expression. We also show for the first time that four different GK inhibitors decrease spontaneous [Ca²⁺] _i oscillations in isolated VMN GE neurons and increase oscillations in GI neurons held at physiologically relevant glucose concentrations. Finally, we present the novel finding that brain GK mRNA expression is increased in three different in vivo models in which there is defective central glucosensing, but it was unaffected by a 48-h fast. All of these findings point to a role for GK as the likely mediator of neuronal glucosensing.

GE neurons are similar to pancreatic β-cells in many ways. Both appear to utilize GK to regulate the ATP-to-ADP ratio (4–6,12). When glucose metabolism increases this ratio, the K_ATP channel (composed of the Kir6.2 K⁺ pore-forming unit and a sulfonylurea receptor) (33) is inactivated. This depolarizes the neuron, accompanied by Ca²⁺ influx through a voltage-activated Ca²⁺ channel (33). Sulfonylurea receptor occupancy also inactivates the channel and depolarizes the cell (33). The K_ATP channel is ubiquitous in neurons throughout the brain and is thus necessary, but not sufficient, to be the sole regulator of GE neuron glucosensing (8,9). Similarly, glucose transport is probably not a key regulator of neuronal glucosensing because the predominant neuronal transporter, GLUT-3, is ubiquitous and saturated at physiological brain glucose concentrations (34). Other brain glucose transporters (GLUT-1, -2, -4, and -8) have either inappropriate anatomic or cell type localizations (35–38).

As in the pancreatic β-cell, (6), our data show that GK is the likely rate-limiting step in glucosensing in both GE and GI neurons. As compared with hexokinase I, which has a ubiquitous neuronal distribution, low Michaelis constant (K_m), and end product inhibition (39, 40), GK has the proper kinetics, lack of end product inhibition, and selective distribution required for this role (6,11,12). GK is expressed in brain areas in which GE neurons have been identified electrophysiologically (1,3–5) and where intracarotid glucose stimulates c-fos expression. GK is also expressed in areas where GI neurons reside (3, 41) and where hypoglycemia induces c-fos expression (42). Using high-stringency conditions, we previously found no GK expression by in situ hybridization in lateral hypothalamus or the area postrema–NTS (12) where glucosensing neurons reside (43–45). But GK mRNA is demonstrable in both areas by RT-PCR, suggesting that in situ hybridization is not sensitive enough to demonstrate the low levels of GK mRNA present (12).

Ca²⁺ imaging in dissociated GE and GI VMN neurons shows that they function within the physiological range of brain glucose levels (3, 46). GE neurons decreased, and GI neurons increased, [Ca²⁺] _i fluctuations as extracellular glucose levels fell from 2.5 to 0.5 mmol/l glucose. As shown in electrophysiological studies of VMN neurons (3), all GE neurons and some GI and NG neurons responded to sulfonylureas, suggesting that they contain the K_ATP channel. The relative proportion and the glucose and sulfonylurea response properties of dissociated VMN GE and GI neurons identified by Ca²⁺ imaging here are similar to those seen in whole-cell current-clamp studies of VMN neurons in hypothalamic slices (3). This suggests that the evoked [Ca²⁺] _i oscillations are a reasonable surrogate for altered neuronal activity. Furthermore, the use of dissociated cells specifically excludes presynaptic inputs from producing changes in [Ca²⁺]_i oscillations (3). The results with four different GK inhibitors suggest that glucosensing in these dissociated neurons depends on GK regulation of ATP production. GK inhibition should lower the ATP-to-ADP ratio, activate the K_ATP channel, and inactivate GE neurons. Similarly, lowering the ATP-to-ADP ratio by GK inhibition in quiescent GI neurons at 2.5 mmol/l glucose activated them. This might be mediated by an ATP-activated K⁺ channel (47), a chloride channel (3), or the Na⁺-K⁺-ATP pump (41). GK inhibition was specific to glucosensing neurons because the inhibitors had no effect on NG neurons, which presumably utilize hexokinase I to mediate glycolysis (40). Although alloxan is probably the most selective GK inhibitor (24,25), it can also produce toxicity by generation of free radicals (48). However, the doses and exposure times used here produced no toxicity. Alloxan’s effects were washed out by drug removal or high glucose concentrations. Cells also remained responsive to activation by glutamate and sulfonylureas, and alloxan increased activity in GI neurons. Finally, alloxan’s effect was unaltered by free radical scavengers. In GE neurons, alloxan half-maximally inhibited Ca²⁺ influx at ∼3 mmol/l, whereas the IC₅₀ in β-cells is ∼2–5 μmol/l (25,49). This reduced potency suggests that glucosensing neurons lack GLUT-2, which is the preferred glucose transporter for alloxan in the β-cell (50). Finally, the heterogeneous responses to the GK inhibitors might reflect variations in GK activity, similar to what is seen in pancreatic β-cells (51).

Many of the neurons that express GK mRNA are involved in energy homeostasis and respond to a variety of metabolic signals from the periphery, such as leptin and insulin. This suggests that they are actually “metabolic sensors” (2). This includes ARC NPY (52) and POMC (53), as well as γ-aminobutyric acid (54) and locus ceruleus noradrenergic neurons (55). ARC NPY neurons appear to be GI (52) and POMC neurons GE (56), and both contain GK mRNA. This is compatible with our findings that both GI and GE neurons utilize GK for glucosensing. Metabolic perturbations might also affect neuronal glucosensing by altering the expression of GK. Although brain and pancreatic GK are homologous (5,11), pancreatic GK was reduced by fasting (57), whereas brain GK mRNA was not. However, brain GK expression was increased in animals with defective central glucosensing. The increased hypothalamic GK expression seen 48 h after hypoglycemia might be a compensatory response to hypoglycemia-induced neuronal damage and consequent defective central glucosensing produced by prior hypoglycemia (21). It might also play a pathogenic role in the development of hypoglycemia-associated autonomic failure. Similarly, the increased brain GK expression in both outbred DIO-prone and selectively bred DIO rats might be a compensatory response to their multiple defects in central glucosensing (3, 16,18–20,58).

In conclusion, GK appears to be the rate-limiting step controlling glucosensing in both GE and GI neurons. It is expressed in neurons with known glucosensing and metabolic sensing properties. Inhibition of GK activity alters the kinetics of [Ca²⁺]_i oscillations in both GE and GI VMN neurons, suggesting that GK activity is critical to their functioning as glucosensors. GK expression is also specifically upregulated in rat models that are known to have defective central glucosensing, but it is unaffected by a state of negative energy balance induced by fasting. Taken together, these data suggest that manipulation of brain GK might afford a means to alter central glucosensing.

This work was funded by the Research Service of the Veterans Administration and the National Institute of Diabetes and Digestive and Kidney Diseases (DK-53181, to A.A.D.-M., B.E.L., and L.K.) and an American Diabetes Association Career Development Award and NRSA-1F32NS10335-01 (to V.H.R.).

The authors thank Antoinette Moralishvilli, Charlie Salter, Guofeng Zhou, and Odeal Gordon for technical assistance.