Convergence of Pre- and Postsynaptic Influences on Glucosensing Neurons in the Ventromedial Hypothalamic Nucleus

Male 14- to 21-day-old Sprague-Dawley rats selectively bred for the traits of developing DIO or being diet-resistant (DR) when placed on a diet moderately high in fat and calories (13) were obtained from colonies at the VA Medical Center in East Orange, New Jersey. At this age, the DIO trait is not expressed, so they are referred to here as DIO-prone. They were housed with their dams on a 12:12-h light:dark cycle at 22–23°C and given low-fat diet (Purina Rat Chow 5001) and water ad libitum. On the day of the experiment, rats were anesthetized with ketamine/xylazine (80:10 mg/kg i.p.) and transcardially perfused with ice-cold oxygenated (95% O₂/5% CO₂) perfusion solution composed of the following (in mmol/l): 2.5 KCl, 7 MgCl₂, 1.25 NaH₂PO₄, 28 NaHCO₂, 0.5 CaCl₂, 7 glucose, 1 ascorbate, and 3 pyruvate (osmolarity adjusted to ∼300 mOsm with sucrose, pH 7.4). Brains were rapidly removed and placed in ice-cold (slushy) oxygenated perfusion solution. Sections (350 μm) through the hypothalamus were made on a vibratome (Vibroslice; Camden Instruments). The brain slices were maintained at 34°C in oxygenated high-Mg²⁺ low-Ca²⁺ artificial cerebrospinal fluid (ACSF; [in mmol/l]: 126 NaCl, 1.9 KCl, 1.2 KH₂PO₄, 26 NaHCO₃, 2.5 glucose, 9 MgCl₂, and 0.3 CaCl₂; osmolarity adjusted to ∼300 mOsm with sucrose; pH 7.4) with 0.2 mmol/l 2,3-butanedione monoxime for 30 min and allowed to come to room temperature. 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.

Viable neurons were visualized and studied under infrared differential-interference contrast microscopy using a Leica DMLS microscope equipped with a 40× long working–distance water-immersion objective. Current-clamp recordings (standard whole-cell recording configuration) from neurons in the VMN were made using an Axopatch 1D amplifier (Axon Instruments, Foster City, CA). Data were stored on a digital-analog tape recorder (Biologic, Claix, France) for further analyses. During recording, brain slices were perfused at 10 ml/min with normal oxygenated ACSF. Borosilicate pipettes (1–3 MΩ; Sutter Instruments, Novato, CA) were filled with an intracellular solution containing (in mmol/l): 128 K-gluconate, 10 KCl, 4 KOH, 10 HEPES, 4 MgCl₂, 0.5 CaCl₂, 5 EGTA, and 2 Na₂ATP; pH 7.2. Osmolarity was adjusted to 290–300 mOsm with sucrose. Input resistance was calculated from the change in membrane potential in response to small 500-msec hyperpolarizing pulses (−10 to −20 pA) given every 3 s. The reversal potentials for changes in membrane conductance in response to glucose were derived from the voltage response to hyperpolarizing current steps as described previously (7). Briefly, hyperpolarizing current pulses varying from −10 to −120 pA were applied at 10- or 20-pA increments. At each increment, four pulses were applied. The duration of each pulse was 500 ms, and pulses were applied every 3 s. The membrane potential response during the last 5 ms of each pulse (when dV/dt = 0) was measured, and the average of the four pulses at each amplitude was calculated. The membrane potential response was measured only after the membrane response to altered extracellular glucose had stabilized, and this value was compared with controls that were measured immediately before changing extracellular glucose. Extracellular glucose levels were altered and chemicals added as described in the figures. All chemicals were obtained from Sigma Chemical (St. Louis, MO) unless otherwise noted. Data are expressed as means ± SE.

Using physiological conditions to define GR and GS neurons, it became apparent that there were several subtypes of neurons in the VMN that fit the original description of such cells (Table 1). Two subtypes of VMN glucosensing neurons were directly responsive to alterations in extracellular glucose levels. Another three subtypes were not inherently glucose-sensing themselves, but rather are presynaptically modulated by changes in extracellular glucose.

The most common subtype of inherently glucosensing VMN neuron will be referred to herein as glucose-excited (GE) because its action potential frequency changed in parallel with changes in extracellular glucose. Of the VMN neurons, 14% (14 of 100) fit this criterion. Under control conditions (2.5 mmol/l glucose), their resting membrane potential (RMP) was −43 ± 2 mV (n = 14), and their input resistance was 876 ± 99 MΩ (n = 10). GE neurons were reversibly hyperpolarized by 6.0 ± 1 mV, and their action potential frequency was decreased when extracellular glucose levels were decreased from 2.5 to 0.1 mmol/l (n = 11) (Fig. 1A) or from 5 to 0.1 mmol/l (n = 3). Input resistance was reduced by 29.5 ± 3.2% when glucose levels were decreased from 2.5 to 0.1 mmol/l (n = 8), indicating that conductance was increased. This conductance increase reversed at −95 ± 3 mV (theoretical K_eq = −99 mV), suggesting that a K⁺ channel was activated (n = 4) (Fig. 1B). GE neurons showed no further excitation as glucose levels were increased to >2.5 mmol/l (n = 4). Moreover, a reduction in extracellular glucose from 2.5 to 1.0 mmol/l was not sufficient to inhibit GE neurons. However, these same neurons were inhibited when the glucose level was subsequently reduced to 0.5 mmol/l (n = 4). The sulfonylurea drug tolbutamide reversed the inhibitory effects of 0.1 mmol/l extracellular glucose (n = 4) (Fig. 1A). Finally, the inhibitory effect of 0.1 mmol/l glucose persisted under conditions of high Mg²⁺ (3.1 mmol/l) and low Ca²⁺ (0.3 mmol/l) (n = 2), which remove presynaptic transmission (Fig. 1C) (14,15). Thus, the inhibitory effect of decreasing extracellular glucose within the physiological range appears to be mediated by a K_ATP channel on the cell body of VMN GE neurons.

The second, and less common, subtype of inherently glucosensing VMN neuron will be referred to as the glucose-inhibited (GI) neuron because its action potential frequency varied inversely with extracellular glucose levels (1). Only 3 of 100 VMN neurons were found to be GI. Their RMP in 2.5 mmol/l glucose was −48 ± 2 mV, and their input resistance was 590 ± 125 MΩ. These GI neurons reversibly depolarized by 4 ± 3 mV and increased their action potential frequency as extracellular glucose decreased from 2.5 to 0.1 mmol/l (Fig. 2A, top trace). This was associated with a 32 ± 11% increase in input resistance (n = 3), indicating a decrease in conductance. This response persisted under conditions of high Mg²⁺ and low Ca²⁺ in all three neurons, suggesting that they are inherently glucosensing neurons (Fig. 2A, bottom trace). Unlike the GE neurons, which reversed at approximately −90 mV, the conductance decrease of these GI neurons reversed at −50 ± 5 mV (n = 3; theoretical Cl⁻ equilibrium potential = −57 mV) in both normal- and high-Mg²⁺/low-Ca²⁺ ACSF. It would be desirable to show that alteration of the Cl⁻ gradient caused an appropriate shift in the reversal potential for the response to glucose in these GI neurons. However, the paucity of GI neurons in the VMN (3 of 100) makes this extremely difficult and would significantly hinder timely presentation of this information. Nevertheless, these data are consistent with the hypothesis that decreasing extracellular glucose levels increased the action potential frequency and decreased conductance in GI neurons by inactivation of a Cl⁻ channel (Fig. 2B).

Three additional subtypes of VMN glucosensing neurons did not have inherent glucosensing properties. Rather, their action potential frequencies and membrane properties were altered by presynaptic inputs in response to changes in extracellular glucose. Of these three noninherently glucosensing subtypes of VMN neurons, the first subtype responded to decreased extracellular glucose levels, whereas the other two subtypes responded to increased extracellular glucose levels. That is, 14% (14 of 100) of the VMN neurons reversibly depolarized by 2.4 ± 0.5 mV (RMP = −44 ± 1 mV) in response to a decrease in glucose levels from 2.5 to 0.1 mmol/l and increased their action potential frequency (Fig. 3A, top trace). In this case, input resistance was reduced by 16 ± 2.7% (from −934 ± 99 MΩ in 2.5 mmol/l glucose; n = 10), indicating increased membrane conductance. However, this response was lost when presynaptic transmission was removed (n = 3) (Fig. 3A, middle trace). These neurons will be referred to as presynaptically excited by decreased glucose (PED) neurons. The current voltage relations for the increased conductance in response to decreased extracellular glucose in these PED neurons were parallel between −50 and −100 mV (n = 3) (Fig. 3B). This suggests that neither K⁺ nor Cl⁻ channels were involved. In current-clamp studies of spontaneously active neurons, action potentials obscured accurate measurement of the voltage response to positive current injection. Thus, we were unable to investigate the current-voltage relations above the RMP (approximately −50 mV) and could not determine whether their current-voltage relations remained parallel or intersected at a more positive voltage.

As mentioned above, the last two subtypes of VMN neurons altered their action potential frequency in response to increased extracellular glucose concentrations that were >2.5 mmol/l as a result of presynaptic inputs (Table 1). The first of these was reversibly inhibited as extracellular glucose levels were increased. Of the neurons, 11% (10 of 92) hyperpolarized by 6 ± 2 mV (RMP = −44 ± 2 mV) and decreased their action potential frequency when extracellular glucose was increased from 2.5 to 10 mmol/l (n = 8) (Figs. 4A and B) or from 5 to 10 mmol/l (n = 2) (Fig. 4D). Input resistance was 759 ± 80 MΩ in 2.5 mmol/l glucose and decreased by 22.2 ± 4.8% when extracellular glucose was increased to 5 or 10 mmol/l (n = 7). The γ-aminobutyric acid type A (GABA_A) receptor antagonist bicuculline reversed the inhibitory effect of 10 mmol/l glucose (n = 5) (Fig. 4A). The inhibitory effect of increased extracellular glucose was also abolished under conditions of high Mg²⁺ and low Ca²⁺ (n = 2) (Fig. 4B). Furthermore, these neurons were also inhibited by tolbutamide (n = 2) (Fig. 4C). This suggests that this subtype of VMN neuron is not inherently glucosensing. Instead, it appears to receive synaptic input from a GABAergic glucosensing neuron. Tolbutamide mimics the inhibitory effect of high glucose, suggesting the involvement of a K_ATP channel on the presynaptic locus. Thus, these neurons were presynaptically inhibited as glucose levels were raised (PIR neurons). Two PIR neurons were also inhibited when extracellular glucose was decreased from 2.5 or 5 to 0.1 mmol/l, and they were stimulated by tolbutamide (as shown in Fig. 4D). On the other hand, one PIR neuron was excited as glucose levels were reduced from 2.5 to 0.1 mmol/l. These differences are likely to reflect differences in the maintenance of individual synaptic inputs after the actual slice procedure. That is, presynaptic neurons that provide the VMN with inhibitory input as glucose levels rise may synapse on a variety of different VMN neurons, including GE and GI neurons as well as noninherently glucosensing neurons, as suggested above. The in vivo occurrence of such synapses is difficult to determine in this in vitro brain slice preparation because these inputs may not always be intact, depending on the exact slice location.

The second subtype of VMN neuron to alter its action potential frequency in response to increasing the extracellular glucose concentration to >2.5 mmol/l was presynaptically excited when extracellular glucose levels were raised (PER neuron). That is, 9% (8 of 92) of the VMN neurons reversibly depolarized by 3 ± 1 mV (RMP = −48 ± 2 mV; n = 8) and increased their action potential frequency when extracellular glucose was increased from 2.5 mmol/l to 5 or 10 mmol/l (Fig. 3A, lower trace). Input resistance was 978 ± 132 MΩ in 2.5 mmol/l glucose and decreased by 14.7 ± 4% when extracellular glucose levels were raised. This effect was lost when presynaptic input was abolished under conditions of high Mg²⁺ and low Ca²⁺ (n = 2) (Fig. 3A, middle trace). Interestingly, these PER neurons were also excited when extracellular glucose levels were decreased from 2.5 to 0.1 mmol/l (n = 5) (Fig. 3A, top and bottom traces). The response to decreased glucose persisted under conditions of high Mg²⁺ and low Ca²⁺ in one of these VMN PER neurons, whereas in another, the response to decreased glucose was abolished when presynaptic transmission was inhibited. The remaining three PER neurons that were also excited as glucose decreased were not evaluated under conditions of high Mg²⁺ and low Ca²⁺. Thus, some neurons that provide excitatory presynaptic input to VMN neurons when extracellular glucose increases to >2.5 mmol/l synapse on either GI or PED neurons. As mentioned above, this is likely to be caused by differences in the maintenance of synaptic inputs in the individual slice preparations.

Finally, tolbutamide altered the action potential frequency of several different types of VMN neurons. As expected, tolbutamide (200 μmol/l) depolarized and increased the action potential frequency of GE neurons as well as the two PIR neurons that were also inhibited by 0.1 mmol/l glucose (n = 6). GI and PED neurons were depolarized and their action potential frequencies increased in the presence of tolbutamide (n = 2). In contrast, tolbutamide hyperpolarized and decreased the action potential frequencies of VMN PIR neurons that were not inhibited as glucose was decreased to 0.1 mmol/l (n = 2). Surprisingly, tolbutamide also depolarized and increased the action potential frequency of three VMN neurons that had no apparent response to changes in extracellular glucose.

In contrast to the DR rats, VMN glucosensing neurons in DIO-prone rats were both fewer in number and showed abnormal responses to glucose (Table 1). Only 6% (3 of 53) of the VMN neurons in the DIO-prone rats were inhibited by 0.1 mmol/l glucose (Figs. 5A and B, top trace). These VMN GE neurons were hyperpolarized to a similar extent as those in DR rats (6 ± 1.7 mV). However, two of three neurons were abnormal, as illustrated by their slow and blunted response (Fig. 5A) and incomplete recovery (Fig. 5B). Furthermore, only 1 of the 36 (2.7%) VMN neurons in DIO-prone rats was inhibited by increasing extracellular glucose to 10 mmol/l. In this case, the action potential frequency decreased slightly, and the neuron was hyperpolarized by only 2 mV (Fig. 5B, middle trace). This neuron was stimulated by tolbutamide in the presence of 10 mmol/l glucose (Fig. 5B, lower trace).

Only 6% (3 of 53) of the VMN neurons in the DIO-prone rats depolarized (1 ± 0.4 mV; n = 3) and increased their action potential frequency when extracellular glucose levels were decreased from 2.5 to 0.1 mmol/l. For two of these three neurons, the response was barely detectable because of a high steady-state action potential frequency (Fig. 6A). The subtype of these neurons could not be determined because this high action potential frequency precluded investigation of presynaptic effects. Finally, only 2.7% (1 of 36) of the VMN neurons in DIO-prone rats depolarized (3 mV) and increased their action potential frequency when extracellular glucose was increased to >5 mmol/l (Fig. 6B, bottom trace). This neuron was not excited by decreasing extracellular glucose levels (Fig. 6B, top trace).

The present studies demonstrate that glucosensing by VMN neurons involves a complex convergence of pre- and postsynaptic mechanisms. These mechanisms are summarized in Fig. 7. Under physiological conditions, there are two types of neurons with inherent glucosensing properties in the VMN (Table 1). The first, which we have named GE, is analogous to the classic GR neuron (1,2), in which the inhibitory effect of decreasing extracellular glucose from 2.5 to 0.1 mmol/l is mediated by a somatic (or dendritic) K_ATP channel. Interestingly, VMN GE neurons were also inhibited when extracellular glucose decreased from 2.5 to 0.5 mmol/l but not from 2.5 to 1 mmol/l (n = 4). This suggests that GE neurons only respond to the large decreases in extracellular glucose that accompany profound systemic hypoglycemia, rather than the relatively small changes expected with the 10–15% dips in plasma glucose levels that precede some meals (16). Thus, VMN GE neurons may be more likely to play a role in the counterregulatory response to hypoglycemia (17) and glucoprivic feeding (18) than in the physiological control of meal initiation. This is consistent with recent studies of mice having deletions of the Kir6.2 pore-forming unit of the K_ATP channel (Kir6.2^−/−). These mice had no functional VMN GR neurons, and their counterregulatory response and stimulation of feeding induced by glucoprivation were attenuated. On the other hand, they maintained normal ingestive responses to the physiological regulators of ingestion, leptin, and neuropeptide Y (4). However, it is important to note that there are no data concerning actual meal-to-meal variations in extracellular brain glucose levels. Moreover, the data of Silver and Erecinska (12) show that the relation between plasma glucose levels and extracellular brain glucose levels is not linear at <2.5 mmol/l. Thus, further studies are needed before a final role in glucose homeostasis is assigned to the GE neuron.

A less common type of VMN neuron with inherent glucosensing properties is analogous to the GS neuron, which is found in low abundance in the VMN (1). We refer to these as GI neurons because they are inhibited by glucose. GI neurons were excited when extracellular glucose was decreased from 2.5 to 0.1 mmol/l. Further dose-response curves with glucose will be required to determine the functional response range of these VMN GI neurons. Regardless of the range of glucose needed to alter their firing, the present data lead us to hypothesize that their direct response to decreased glucose levels is mediated by the inactivation of a Cl⁻ channel. Because decreasing extracellular glucose levels should lower intraneuronal ATP levels, such a Cl⁻ channel should be responsive to changes in the ATP-to-ADP ratio. The cystic fibrosis transmembrane regulator (CFTR) is one such Cl⁻ conductance, and it is activated by ATP and blocked by sulfonylureas (19). Moreover, CFTR mRNA and protein are expressed in human and rat hypothalamus (20,21). There is also an ATP-activated Cl⁻ conductance in pancreatic islet cells. Decreased levels of ATP or high levels of the sulfonylurea gliburide inactivate this channel and depolarize the islet cells (22). Given these facts, it is interesting that the sulfonylurea tolbutamide stimulated GI neurons in our study. Although this may reflect the ubiquitous nature of the K_ATP channel (23), it may also indicate a lack of specificity of the sulfonylureas for the K_ATP channel. The SUR and CFTR belong to the same family of ATP-binding cassette transporters and have significant homology (6,24). Thus, tolbutamide may depolarize GI neurons by blocking the same Cl⁻ conductance that is reduced when extracellular glucose (or ATP) levels are decreased. These data also suggest that caution is warranted when using sulfonylureas as the sole indicator of the presence of GE neurons and/or K_ATP channels. For this reason, GE neurons in this study were characterized by a response to glucose that was not only blocked by sulfonylureas but also reversed at the K⁺ equilibrium potential. Finally, although the GI neurons we observed were similar, we cannot be certain that they were identical to the GS neurons previously described by Oomura (1). This is because the GI neurons in our study also received presynaptic input from neurons that were excited when extracellular glucose levels were increased to >2.5 mmol/l. Thus, it is uncertain whether these GI neurons would change their action potential frequency in response to a linear change in glucose between 0 and 10 mmol/l.

An important finding of these studies is that there are many neurons in the VMN that have no inherent glucosensing capacity of their own. Instead, their firing rate is regulated by presynaptic inputs from other glucosensing neurons that are presumably outside of the VMN. One subtype of these neurons (PED) is excited by decreasing extracellular glucose from 2.5 to 0.1 mmol/l. Such PED neurons differ from GI neurons in several ways. First, the glucose modulation of their action potential frequency and membrane properties is presynaptic. Second, conductance is increased rather than decreased. Finally, because the current-voltage relations were parallel between −60 and −120 mV, this conductance increase appears unrelated to opening of K⁺ or Cl⁻ channels. If the current-voltage relations remain parallel at more positive potentials, it might suggest the involvement of an ATP-dependent pump such as the Na⁺/K⁺-ATPase. This has been previously suggested for GS neurons (1). Two additional subtypes of VMN neurons are presynaptically modulated by raising extracellular glucose levels. The first of these (PIR) was inhibited at extracellular glucose levels to >2.5 mmol/l. Because bicuculline reversed and tolbutamide mimicked the effect of increased extracellular glucose, we hypothesize that this response may be mediated by inactivation of a K_ATP channel on a presynaptic GABA cell body or nerve terminal. A comparable finding has been described for GABAergic inputs to the substantia nigra (25,26,27). Although the VMN itself contains GABA neurons (28), we found no VMN neurons with inherent glucosensing responses at levels >2.5 mmol/l glucose. Thus, it is likely that the presynaptic inputs to PIR neurons were from GABA neurons whose cell bodies lay outside the VMN. A likely candidate is the population of neuropeptide Y neurons in the adjacent arcuate nucleus that co-express GABA (29). These neurons are regulated by metabolic perturbations and have molecular properties that make them likely to be GR neurons (30). Also, neurons in the arcuate nucleus might be exposed directly to plasma glucose levels that exceed extracellular brain glucose levels because the arcuate nucleus is adjacent to the median eminence, which has a defective blood brain barrier (31).

Increasing extracellular glucose from 2.5 mmol/l to either 5 or 10 mmol/l presynaptically excited another subtype of VMN neuron (PER). An increase in extracellular glucose concentration to 5 mmol/l would be seen in the brain after a meal (12). In addition, these neurons were also excited when glucose levels were decreased. Thus, these PER neurons may be part of a regulatory system that becomes active when extracellular glucose levels increase above a steady-state level of ∼2.5 mmol/l. This could include the hypothalamo-pituitary and autonomic systems, which participate in the assimilation, expenditure, and storage of ingested calories.

Thus, our data support that of others suggesting that VMN glucosensing neurons are involved in the responses to both pathological and physiological stimuli. The DIO rat exhibits several abnormalities of physiological function related to defects in central glucose sensing (9,10,11). Here, we show that these defects may be related to abnormalities in VMN glucosensing neurons. Not only were VMN glucosensing neurons of all subtypes fewer in number in DIO-prone rats, but their glucose responses were also abnormal. The K_ATP channel appears to play a role in the glucose sensing of both GE and PIR neurons. Thus, abnormalities in GE and PIR neurons are consistent with the reduced low-affinity sulfonylurea binding in the VMN and arcuate of DIO-prone rats (11). These data suggest that defective VMN glucosensing neurons may play a role in the altered central glucose sensing in DIO-prone rats.

In summary, we have shown that GE and GI neurons in the VMN respond directly to physiological changes in extracellular glucose, but only when glucose falls to <2.5 mmol/l. GE neurons use the K_ATP channel to sense glucose, whereas GI neurons may use a Cl⁻ conductance. Importantly, other VMN neurons alter their firing rates in response to a variety of extracellular glucose concentrations, but none of these neurons are inherently glucosensing. Instead, the observed effects are caused by presynaptic inputs from other glucosensing neurons, whose cell bodies may reside outside the VMN. Finally, DIO is associated with defective glucosensing neurons.

This study was supported by an American Diabetes Association Career Development Award (to V.H.R.), by NIH Individual National Research Service Award Grant 1F32NS10335-01 (to V.H.R.), and by National Institute of Diabetes and Digestive and Kidney Diseases Grant RO1-53181 (to B.E.L.).