Orexin A Preferentially Excites Glucose-Sensitive Neurons in the Lateral Hypothalamus of the Rat In Vitro

Wistar rats aged 2–3 weeks were anesthetized with sodium pentobarbital (40 mg/kg intraperitoneally) and decapitated in accordance with U.K. Home Office procedures. For optimization of tissue viability (18), the brain was immediately immersed in chilled sucrose-based artificial cerebrospinal fluid (aCSF) and bubbled with 95% O₂/5% CO₂, the composition of the medium being (in mmol/l) 2.5 KCl, 1.2 NaH₂PO₄, 26 NaHCO₃, 0.1 CaCl₂, 1.3 MgCl₂, 10 glucose, and 189 sucrose. Transverse hypothalamic slices (200–400 μm) at the level of the median eminence were cut using a tissue slicer (Vibroslice 752/M; Campden Instruments, Cambridge, U.K.) and then incubated for 1 h in the same medium. One slice was then immersed in a recording chamber and continually perfused (6–8 ml/min) at room temperature with gassed aCSF that had the following composition (in mmol/l): 120 NaCl₂, 2.5 KCl, 1.2 NaH₂PO₄, 26 NaHCO₃, 2.4 CaCl₂, 1.3 MgCl₂, and 10 glucose. The slice was allowed to equilibrate in the recording chamber for an additional 1 h before study.

Intracellular recording microelectrodes were made from borosilicate glass on a Flaming/Brown model P-87 micropipette puller (Sutter Instruments, Novato, CA). The electrodes had resistances of 120–180 MΩ when filled with 2 mol/l KCH₃SO₄. Neurobiotin (Vector Laboratories, Burlingame, CA) 2% wt/vol was included in the electrode filling solution. Standard current clamp recordings were made using an Axoprobe-1A amplifier (Axon Instruments, Burlingame, CA). Current and voltage signals were digitized (sample rate ∼5 kHz, filtered at 3 kHz) and analyzed using SPIKE2 software (Cambridge Electronic Design, Cambridge, U.K.). Input resistance of the neuron was monitored by measuring the amplitude of the voltage transient in response to a 0.1- to 0.3-nA hyperpolarizing current step (500 ms duration). The change in number of evoked action potentials was tested using a 0.1- to 0.5-nA depolarizing current pulse (500–1,000 ms duration).

Bath glucose concentration was altered to 3 or 15 mmol/l from basal 10 mmol/l, for 3-min periods, to identify glucose-sensing neurons as described below. After re-equilibration in 10 mmol/l glucose for 30 min, orexin A (GlaxoSmithKline, Harlow, U.K.; 1 μmol/l in aCSF) was applied to the slices for 2–4 min by switching between reservoirs containing standard aCSF and the orexin solution. Switching artifacts were rarely apparent.

Changes in membrane potential and intrinsic neuronal properties in response to changes in glucose concentration or addition of orexin A were expressed as mean ± SE and compared versus baseline and among the three cell types, using the paired or unpaired Student’s t test as appropriate. A change in input resistance of >20% was regarded as significant. Changes in firing frequency in response to addition of orexin A were measured during 100-s intervals and expressed as percentage increase above baseline. These data were analyzed initially using two-way analysis of variance with cell type and time as independent variables, followed by the unpaired t test to examine differences between cell types at individual time points.

For visualization of neurobiotin-filled neurons, slices were fixed overnight in 4% paraformaldehyde and washed in four changes of 0.1 mol/l phosphate-buffered saline (PBS). Slices were then incubated in 1% Triton X-100 in PBS for 2 h and in 1:100 streptavidin-dichlorotriazinylaminofluorescein (Jackson, West Grove, PA) in 1% Triton for another 2 h before being washed four times in PBS. Orexin A immunoreactivity was identified using standard indirect immunofluorescence techniques. The slice was incubated with rabbit polyclonal antisera raised against native orexin A (SmithKline Beecham; 1:250 to 1:500 dilution with 1% Triton X-100) for 24 h at room temperature. Slices were washed four times in PBS, then incubated for 4–6 h at room temperature with a conjugated Cy3-tagged secondary antibody (Jackson) 1:100 in PBS containing 1% Triton X-100. After four additional washes in PBS, slices were floated onto gelatin-coated slides and mounted with Vectashield mounting medium (Vector Laboratories).

Dual fluorescence studies were performed using a Zeiss LSM510 confocal microscope. Dichlorotriazinylaminofluorescein (neurobiotin) was excited using the 488-nm light from an argon ion laser, and emitted light was collected off a 545-nm dichroic mirror through a 530- to 550-nm band-pass filter. Cy3 fluorescence (orexin A) elicited by the 543-nm line of a helium-neon laser was collected through a 545-nm dichroic mirror and a 560-nm long-pass filter. Spillover between fluorochromes was minimized by sequential scanning with the alternate excitation light sources and photomultiplier tubes were protected by appropriate filters. With the use of the supplied Zeiss software, overlapping confocal slices were combined to create three-dimensional renders of the stack of images.

Axons and dendrites showed classical features (19) and were discriminated as follows. Dendrites arose from the neuron soma and tapered gradually; they branched typically into two, and these second-order dendrites were reduced in diameter. By contrast, the axon arose from a cone-shaped hillock on the soma or a principal dendrite, as a small-caliber process that maintained the same diameter as it ran from the soma. At branch points, the axonal branches had the same diameter as the parent axon. Axons of orexin neurons typically carried varicosites and terminated in boutons.

GSNs comprised 13 of the 29 (45%) LHA neurons characterized in detail. GSNs were identified by showing an increased spontaneous firing rate (>100% for >2 min) when the bath glucose concentration was lowered from 10 to 3 mmol/l (Fig. 1A). At 10 mmol/l glucose, resting membrane potential (Vm) of the GSNs was 64.7 ± 1.7 mV, spontaneous firing rate was 0.13 ± 0.01 Hz, and input resistance was 212 ± 25 MΩ. Exposure to 3 mmol/l glucose caused prompt depolarization of 5.2 ± 1.0 mV and a 150% increase in firing rate (0.29 ± 0.02 Hz; P < 0.01). Conversely, 15 mmol/l glucose inhibited spontaneous firing (0.09 ± 0.03 Hz; P < 0.01 vs. 10 mmol/l glucose). These cells were then subjected to one or more additional tests, described below.

Addition of orexin A (1 μmol/l) caused rapid, marked, and prolonged excitation of GSNs. When tested at 10 mmol/l glucose, orexin A caused prompt (within 40 s) depolarization of 9.9 ± 2.2 mV (n = 13) and an immediate 500% increase in spontaneous firing rate (P < 0.01) that declined gradually over several minutes (Figs. 1B and 2). Further addition of orexin A within 60 min elicited smaller depolarizations of 5–8 mV, indicating desensitization. Orexin A also altered the intrinsic neuronal properties of GSNs, consistent with increased excitability. When the GSNs were clamped at their baseline Vm, the input resistance was significantly increased by 34% (284 ± 18 vs. 212 ± 25 MΩ, n = 8; P < 0.05). In addition, the depolarizing threshold current needed to trigger an action potential was reduced by 37% (0.07 ± 0.01 vs. 0.11 ± 0.2 nA, n = 8; P < 0.01), whereas the number of spikes induced by a stepped depolarizing current was doubled (4.2 ± 0.7 vs. 2.2 ± 0.4, n = 7; P < 0.01) (Fig. 1C).

The magnitude and time course of depolarization suggested that orexin A acted directly on the impaled GSNs. For verification of this, GSNs (n = 3) were rechallenged with orexin A in the presence of tetrodotoxin (1 μmol/l). No significant reduction in the depolarization induced by orexin A was seen (data not shown), confirming a direct postsynaptic action. Orexin A also antagonized the inhibitory effects of hyperglycemia on GSNs, restoring the spontaneous firing rate at a glucose concentration of 15 mmol/l (0.39 ± 0.15 vs. 0.09 ± 0.03 Hz, n = 3; P < 0.01).

GRNs were defined as cells showing >50% reduction in spontaneous firing rate when glucose was decreased from 10 to 3 mmol/l, and these accounted for 10 (34%) of the LHA neurons sampled. At 10 mmol/l glucose, GRNs had a Vm of 61.2 ± 1.9 mV and spontaneous firing rate of 0.19 ± 0.04 Hz, which fell to 0.06 ± 0.01 Hz (P < 0.01) in 3 mmol/l glucose (Fig. 3A). Five of these cells were also tested at a glucose concentration of 15 mmol/l, which increased their firing rate to 0.46 ± 0.05 Hz (P < 0.01).

Orexin A caused excitation of GRNs, but these effects were significantly less marked than on GSNs and were comparable to those elicited by orexin A in the glucose-indifferent neurons that constituted the remaining six (21%) neurons studied (Figs. 2, 3A, and 3C). At 10 mmol/l glucose, orexin A–induced depolarization of GRNs was 6.3 ± 1.4 mV (n = 10; P < 0.05 vs. GSNs) and in glucose-indifferent neurons was 4.2 ± 1.4 mV (n = 6; P > 0.05 vs. GRNs). Two-way analysis of variance showed significant effects of cell type (P = 0.024) and of time (P = 0.02) on firing rate after application of orexin A (see Fig. 2). The spontaneous firing rate of GRNs increased to a peak of only 200% above baseline (P < 0.05 vs. GSNs). The magnitude of this response to orexin A was closely similar to that of glucose-indifferent cells, and the firing rates of these two cell types did not differ significantly at any time point (Fig. 2).

Confocal microscopy of fluorescein-labeled GSNs revealed fusiform or pyramidal cell bodies with complex ramifying processes, some extending up to 800 μm from the perikaryon (Fig. 4A). Cells identified electrophysiologically as GRNs showed broadly similar structural features. In some sections, two to five contiguous cells of similar morphology were labeled (Fig. 4B), even though only a single GSN or GRN had been injected with neurobiotin.

As in previous reports (20,21,22), immunostaining for orexin A showed rounded or elongated cell bodies in the LHA and perifornical area, with abundant intensely labeled fibers, many carrying the characteristic varicosities. None of the labeled GSNs or GRNs contained orexin A or had varicose processes. However, both GSNs and GRNs were intimately related to orexin neurons (Fig. 4). Varicose orexin-cell axons were frequently entwined around and in close contact with the dendrites of GSNs (Fig. 4C). Conversely, axons and dendrites of GSNs were seen in close contact with dendrites of orexin neurons (Fig. 4A). Orexin axons also passed close to GRNs processes, while dendrites and axons of GRNs were in close proximity to dendrites of orexin neurons (Fig. 4D).

This is the first demonstration of the detailed morphology of the glucose-sensing neurons of the LHA and of their anatomic and functional relationship to another hypothalamic neuronal system implicated in autonomic regulation. Both GSNs and GRNs have extensive and complex dendritic systems and long axons that could interact with other neuronal systems in and possibly beyond the LHA. The spread of intracellularly injected label into chains of morphologically similar neurons (Fig. 4B) suggests that glucose-sensing neurons are coupled through gap junctions and thus may form a functional network.

Particularly striking are the intimate reciprocal relationships between orexin neurons and glucose-sensing cells. We did not find any glucose-sensing neurons (either GSNs or GRNs) to contain orexin A. It has been reported that intracellular labeling can mask γ-aminobutyric acid–like immunoreactivity in hippocampal neurons (23), and it is possible that orexin immunoreactivity was similarly concealed here. However, none of the fluorescein-labeled cells in our study seemed to have the varicose processes typical of orexin neurons (20,21,22). We therefore conclude that at least some GSNs and GRNs are distinct from orexin neurons, although because orexin neurons are relatively sparse, we cannot exclude the possibility that some glucose-sensing cells may express orexin. We previously reported that acute hypoglycemia induced Fos positivity in a population of LHA neurons of which only 9% were orexin-immunoreactive (17); we presume that the nonorexin-containing remainder of these hypoglycemia-activated cells includes the GSNs.

Our findings also indicate that orexin neurons make synaptic contact on GSNs and excite them via release of orexin A. Our imaging method cannot identify synapses directly, but the juxtaposition of orexin axons and GSNs processes (Fig. 4C) and the powerful, tetrodotoxin-resistant depolarization of GSNs induced by orexin A strongly suggest synaptic contact. Thus, GSNs seem to be target neurons for orexin in the LHA. It is intriguing that some of our images suggest that GSNs may also synapse onto orexin neurons; this raises the possibility that these two neuronal populations regulate each other in a reciprocal manner, but this must remain speculative until the neurochemical identity of GSNs has been established. Orexin A also caused excitation of GRNs; however, this was significantly less than in GSNs, being no greater than that elicited in LHA neurons indifferent to glucose and consistent with the peptide’s effects on neurons elsewhere, e.g., in the locus ceruleus (24).

The orexins have been attributed various physiological functions, and it is likely that this population of neurons is functionally heterogeneous. There is sound circumstantial evidence that both orexin neurons (11,12,13,14,15,16,17) and GSNs (1,2) could be involved in hypoglycemia-induced feeding. We now suggest that a subset of orexin neurons are responsible for regulating certain lateral hypothalamic GSNs and that this interaction is important in controlling aspects of feeding behavior. GSNs may mediate the acute appetite-stimulating effect of orexin A injected into the LHA (13). Moreover, GSNs may drive feeding during hypoglycemia, which is known to excite orexin neurons (15,16,17) and would be predicted to enhance orexin A release.

In theory, the orexin neurons could also help to terminate glucoprivic feeding by relaying inhibitory satiety signals to the GSNs. It is not known whether the projection from the NTS that conveys inhibitory signals to the GSNs terminates directly on these cells. Our recent studies show that a subset of orexin neurons are stimulated by hypoglycemia but promptly inhibited by eating and that their activation parallels that of neurons in the NTS (17). We propose that the orexin neurons are the primary point of impact of the inhibitory pathway relayed via the NTS and that they regulate the GSNs accordingly. At present, we cannot determine whether the glucose sensitivity of orexin neurons is an intrinsic property of these cells or results from inputs from glucose-sensing cells elsewhere, notably in the LHA or NTS.

Finally, it is possible that orexin neurons and GSNs also play a role in normal feeding, which is also regulated by changes in glucose availability, although much more subtle than those induced by hypoglycemia or glucose antimetabolites. It has been shown that small glucose falls (∼0.3–0.5 mmol/l) precede spontaneous feeding episodes in normal rats and that administration of exogenous glucose to abolish these dips can delay feeding (25). In vivo electrophysiological studies indicate that such small glucose dips can be detected by some “class 1” GSNs in the LHA (5). Immunoneutralization of orexin A (26) or administration of an orexin 1 receptor antagonist (27) inhibits short-term food intake in normal rats. We speculate that this is due at least in part to reduced excitation by orexin A of GSNs, thus effectively lowering the blood glucose level at which eating is triggered.

X.L. was supported by the Scottish Executive and additional funding was from GlaxoSmithKline U.K. and the Liverpool Diabetes Research Action Fund.

We are indebted to Drs. Martyn Evans, Jon Arch, and Shelagh Wilson (GlaxoSmithKline, Harlow, Essex, U.K.) for generous gifts of reagents and much useful and stimulating discussion and to Profs. Ian Silver and Maria Ereciñska and Dr. Gül Erdemli for helpful comments on the manuscript.