Leptin Directly Depolarizes Preproglucagon Neurons in the Nucleus Tractus Solitarius: Electrical Properties of Glucagon-Like Peptide 1 Neurons

Transgenic mice were used that expressed a modified yellow fluorescent protein (YFP-Venus) under the control of the preproglucagon promoter (15). Two founder strains, mGLU-V23-124 and mGLU-V50-144, created using mouse bacterial artificial chromosomes (BACs), were used interchangeably, since we observed no difference in the pattern of YFP expression in the brainstem (15). Animals were bred as heterozygotes on a C57/Bl6 background and were genotyped as described previously (15) before experimental use. All experiments were carried out in accordance with the U.K. Animals (Scientific Procedures) Act, 1986, with appropriate ethical approval.

Mice were anesthetized with halothane and transcardially perfused with ice-cold heparinized saline followed by ice-cold 4% paraformaldehyde in phosphate buffer. The brain was removed and postfixed with 4% paraformaldehyde at 4°C overnight. Brains were cryoprotected with 30% sucrose in PBS, and 30-μm coronal brainstem sections were cut using a cryostat.

For immunofluorescence visualization of cholinergic cells, sections were incubated for 30 min in block solution (10% rabbit serum, 1% BSA, 0.1% triton X-100 in PBS), followed by overnight incubation with a polyclonal goat antibody against choline acetyltransferase (anti-ChAT, 1:1500; Chemicon) in block solution at 4°C. After five washes in PBS, the sections were incubated in the dark for 2 h at room temperature in a rabbit anti-goat IgG CY-3 conjugate (1:500; Sigma). Sections were mounted and cover-slipped with Vectashield. An equivalent protocol was used to visualize catecholaminergic cells with a polyclonal rabbit antibody against tyrosine-hydroxylase (anti-TH, 1:500; Santa Cruz). Block solution contained 10% sheep serum, and the secondary antibody was a sheep anti-rabbit IgG CY-3 conjugate (1:500; Sigma). Mounted sections were analyzed under a Nikon upright microscope with epifluorescence to visualize fluorescence of YFP and CY-3.

Samples for RT-PCR and single-cell RT-PCR were harvested and amplified as described previously (16), using primers listed in Table 1. As negative controls for single-cell RT-PCR, reactions were routinely performed with H₂O instead of DNA sample, with solution from pipettes that were inserted into the slice without recording from a cell and with samples that were not reverse-transcribed. For positive controls, PCRs were performed on a 1:100–1:1,000 dilution of brainstem cDNA. PCR products were verified by direct sequencing (MWG Biotech, London, U.K.).

Coronal (200 μm) or horizontal (250 μm) brainstem slices were obtained from adult (>8 weeks) transgenic mice of either sex after halothane anesthesia and dissection in ice-cold low Na⁺ solution containing (in mmol/l): 200 sucrose, 2.5 KCl, 28 NaHCO₃, 1.25 NaH₂PO₄, 3 pyruvate, 7 MgCl₂, 0.5 CaCl₂, and 7 glucose (pH 7.4). After recovery at 34°C for 30 min in a solution containing (in mmol/l) 118 NaCl, 3 KCl, 25 NaHCO₃, 1.2 NaH₂PO₄, 7 MgCl₂, 0.5 CaCl₂, and 2.5 glucose (pH 7.4), slices were kept at 34°C in artificial cerebrospinal fluid (ACSF) of the following composition (in mmol/l): 118 NaCl, 3 KCl, 25 NaHCO₃, 1 MgCl₂, 2 CaCl₂, and 10 glucose (pH 7.4). Patch pipettes were pulled from thin-walled borosilicate capillaries (3–6 MΩ; Clark Electromedical Instruments, Pangbourne, U.K.) with a horizontal puller (Zeitz, Germany). Electrodes were filled with (in mmol/l) 120 K-gluconate, 5 HEPES, 5 BAPTA, 1 NaCl, 1 MgCl₂, 1 CaCl₂, and 2 K₂ATP, pH 7.2. For perforated-patch whole-cell recording, solubilized amphotericin B (Sigma) was added to the pipette solution (final concentration ∼137.5 μg/ml).

Recordings were carried out in ACSF at 28–32°C. Experimental solutions were constantly bubbled with 95% O₂/5% CO₂. Most drugs were directly added to the ACSF. 6,7-Dinitroquinoxaline-2,3-dione (DNQX) was prepared as a 20 mmol/l stock solution in DMSO. Tetrodotoxin was prepared as a 1 mmol/l stock solution in Na-citrate buffer. The recording chamber (volume 2 ml) was perfused with ACSF at a rate of 4–5 ml/min. Drugs were applied either within the ACSF perfusing the bath (4–5 ml/min) or locally via pressure ejection from a glass pipette (opening diameter 5–10 μm) facing the recorded cell positioned at a distance of ∼100 μm. Exendin-4 and PYY_3–36 were obtained from Tocris Bioscience (Bristol, U.K.). Ghrelin and GLP-1_7–37 were obtained from Sigma (Gillingham, U.K.). Leptin was a kind gift from Novo Nordisk AS (Copenhagen, Denmark).

Recordings were performed in both voltage-clamp and current-clamp mode using an EPC-9 amplifier and Pulse/Pulsefit software (Heka Elektronik, Lambrecht, Germany). Currents or membrane potentials were filtered at 1 kHz and digitized at 4 kHz. Membrane resistance was monitored with 200 ms current or voltage pulses every 20 s. Current-voltage relationships were obtained either by analyzing hyperpolarizing and depolarizing 200-ms voltage or current steps, or by holding the cell at −20 mV for a minimum of 30 s and then applying a voltage ramp from −20 to −120 mV over 700 ms.

For analysis of firing frequency, membrane potential, input resistance, and spontaneous excitatory postsynaptic current frequency, the mean value over 3 min before addition of the drug and the mean over the last 3 min before drug removal were compared. For leptin and glutamate, which were washed off quickly to ascertain reversibility, peak responses were compared to the mean before drug application.

A concentric bipolar electrode (150 μm diameter) was placed in the solitary tract of horizontal brainstem slices ∼2–4 mm away from the fluorescent cell population. Either single shocks (every 10 s) or bursts of five shocks (50 Hz; every 20 s) were delivered from a stimulus isolator (DS2A; Digitimer) driven by the EPC-9 amplifier. Each stimulus had a duration of 20 μs and an amplitude of 40–70 V. Amplitude was increased using single stimuli during individual recordings until a further increase in stimulus amplitude had no further effect on excitatory post synaptic current (EPSC) amplitude. Latency was determined as the delay between the onset of the stimulus artifact and the onset of the EPSC, as described by Doyle and Andresen (17). Synaptic jitter was determined as the standard deviation of the latency for >20 successive single shocks at 0.1 Hz.

Data are given as mean ± 1 SEM. Statistical significance was tested using one-way ANOVA followed by the post hoc Tukey's test, unless stated otherwise. P values of <0.05 and <0.01 were taken to indicate that the data were significantly different.

PPG neurons in acute brainstem slices were readily identifiable by their Venus fluorescence. The majority of Venus-positive cells in these slices were found within the caudal NTS, although additional cell bodies were located further ventrally, e.g., in the reticular area. Venus-positive cells had no clear distinguishing morphological features, and their distribution overlapped with various NTS neuron populations. They were distinct both from ChAT-positive cells in the dorsal vagal motor nucleus (DMNX; Fig. 1,A) and from tyrosine hydroxylase (TH)-positive cells in the NTS—a population that includes adrenergic, noradrenergic, and dopaminergic neurons (Fig. 1 B).

Electrophysiological recordings were established from fluorescent cells in the NTS under differential interference contrast (DIC) optics (Fig. 1 C), and cytoplasm was harvested at the end of the recordings and subjected to single-cell RT-PCR for preproglucagon. Eight out of eight fluorescent cells were positive for preproglucagon and nine out of nine control cells were negative, verifying both the correct targeting of the transgene and that preproglucagon is still transcribed in PPG cells expressing Venus.

In perforated patch recordings, it was evident soon after sealing the electrode onto an individual cell that the intrinsic electrical activity of PPG neurons could be divided into three types: regular firing, silent, and burst firing (Fig. 1,D). These phenotypes persisted under stable perforated-patch conditions. Of 135 recorded neurons, the majority (n = 99) were spontaneously active with a mean firing frequency of 1.9 ± 0.2 Hz. These neurons exhibited a resting membrane potential of −51 ± 1 mV (Fig. 1,D and E), with an average input resistance of 1.05 ± 0.03 GΩ and capacitance of 28 ± 1 pF. Another 11 cells were electrically silent, with a resting potential of −55 ± 2 mV, an input resistance of 0.84 ± 0.09 GΩ, and a capacitance of 27 ± 3 pF (Fig. 1,D and F). Action potentials could be elicited in these cells by current injections of 14 ± 1 pA (n = 9). The remaining 25 neurons showed burst-firing activity, with spontaneous fluctuations of the membrane potential by almost 20 mV (Fig. 1,D and E) that persisted in 0.5 μmol/l tetrodotoxin (n = 5). These cells had a whole-cell capacitance of 21 ± 1 pF. In all three cell types, injection of hyperpolarizing current pulses (200 ms) caused a delay of ∼1 s before the return of spontaneous activity (Fig. 1 F), indicating that PPG neurons express A-type K⁺ currents. All further experiments for this study were performed on the spontaneously active cell population.

The location of PPG neurons in the NTS, a region with an intact blood-brain barrier, suggests that these cells would be predominantly modulated by synaptic inputs. To investigate excitatory synaptic inputs onto PPG neurons, cells were held at −70 mV in standard whole-cell recordings, under which conditions EPSCs were observed as inward currents. Spontaneous EPSCs (sEPSCs) occurred at a frequency of 3.0 ± 0.5 Hz with an amplitude of 23 ± 2 pA (n = 21) and were largely glutamatergic, as demonstrated by their inhibition by kynurenic acid (1 mmol/l) or DNQX (10 μmol/l; Fig. 2,A and B). In support of the idea that PPG neurons receive direct glutamatergic input, glutamate (0.1 mmol/l) triggered membrane depolarization and a reduction in the membrane resistance. Both actions were reversed by DNQX (20 μmol/l), indicating the involvement of non-NMDA type glutamate receptors (Fig. 2 C and D).

One potential source of excitatory glutamatergic inputs are vagal sensory neurons. Previous studies in horizontal brainstem slices have shown that catecholaminergic NTS neurons receive direct monosynaptic input from afferent vagal fibers in the solitary tract (18,–20). We therefore used this recording configuration to examine whether PPG neurons are also responsive to solitary tract stimulation. Tract stimulation reliably evoked EPSCs in PPG neurons with a mean latency of 3.2 ± 0.3 ms and a mean peak amplitude of 126 ± 17 pA (n = 14) (Fig. 3 A and B).

Further assessment of whether the observed responses were monosynaptic or polysynaptic was performed by examining their failure rate, latency, and jitter (the variability of the latency). Of the cells, 11 of 14 had a failure rate of <5%, and the majority of cells displayed a jitter of <300 μs, suggesting that the responses are predominantly, but not entirely, explained by monosynaptic connections. DNQX (10 μmol/l) completely blocked EPSCs evoked by tract stimulation in all cells tested (n = 7), and kynurenic acid (1 mmol/l) reduced their amplitude by 58% (n = 3; Fig. 3,C and D). In 10 of 14 cells, the EPSC amplitude became progressively smaller in response to a burst of five shocks at 50 Hz (Fig. 3 A, Ci, and Cii). These results suggest that the majority of PPG cells receive direct synaptic input from the solitary tract (and are thus second-order neurons) and that both the direct and indirect excitatory inputs are glutamatergic.

GLP-1_7–37 (100–200 nmol/l) had no effect on the resting membrane potential, membrane resistance, or firing frequency of PPG cells (n = 6; Fig. 4,A and B) and did not affect the current response to voltage ramps between −20 and −120 mV. Similarly, the stable GLP-1 analog exendin-4 (100 nmol/l) failed to alter the membrane potential, membrane resistance, or firing frequency of these cells (n = 5). These findings suggested that NTS PPG neurons do not express functional GLP-1 receptors. This conclusion was further supported by single-cell RT-PCR analysis of 12 cells from which cytoplasm was harvested at the end of the recordings. PPG expression was detected in all 12 cells, thus verifying their phenotype, but no cell tested positive for GLP-1 receptor mRNA. By contrast, GLP-1 receptor mRNA was detected successfully in a 1:1,000 dilution of mouse brainstem cDNA (Fig. 4 C).

Although these findings exclude a direct effect of GLP-1 on PPG cells, we also examined whether presynaptic GLP-1 receptors might modulate synaptic inputs onto PPG neurons. However, local application of 10 nmol/l exendin-4 had no significant effect on either the amplitude or frequency of sEPSCs (n = 5; Fig. 4 D and E).

In a further series of experiments, we examined the effects of two more gut-derived peptides involved in appetite control. PYY_3–36 (100 nmol/l), which is co-released with GLP-1 from enteroendocrine L-cells, was without effect on membrane properties and sEPSC frequency or amplitude in voltage-clamp recordings (n = 6; Fig. 4,D and E). In accordance with these results, no Y2 receptor message was detected with single-cell RT-PCR analysis of six PPG neurons. Ghrelin (100 nmol/l), which is released from the stomach as an orexinergic signal, also failed to elicit any significant change in membrane potential, membrane resistance, or spontaneous action potential firing rate in PPG cells (n = 5; Fig. 4 F).

To investigate whether the anorexigenic peptide leptin has direct effects on PPG neuron activity, we tested its effect in perforated patch current clamp recordings. Leptin (20 nmol/l) applied locally caused a depolarization of 9 ± 2 mV that was accompanied by an increase in firing frequency and a decrease in membrane resistance in eight of eight PPG cells (Fig. 5). This response was reversible upon washout of leptin.

Activation of PPG neurons in the NTS has been linked to a reduction of food intake in a number of studies that demonstrated c-Fos expression in these cells in response to satiety signals such as gastric distension, consumption of sweetened milk, or elevated leptin levels (6,–8,21,22). However, because of the inability to identify PPG neurons in live tissue, nothing has been known about their electrical properties, or the direct modulation of their activity. Our results describe for the first time the electrical properties of PPG neurons located in the caudal NTS and their direct activation by leptin.

Under resting conditions, the majority of PPG neurons in the NTS were spontaneously active, firing action potentials at a frequency of ∼2 Hz. In this respect, they resemble the neighboring population of paraventricular nucleus–projecting catecholaminergic neurons that were also reported to be spontaneously active (18). POMC neurons in the NTS, by contrast, receive visceral afferent input but are electrically silent unless stimulated (19). At present, we do not know how the level of spontaneous activity of PPG neurons translates to their release of GLP-1 from nerve terminals. Recent studies by Wan et al. (23,24) investigating the effects of GLP-1 and GLP-1 antagonists on dorsal vagal neurons using an equivalent preparation in rat found no evidence of tonic GLP-1 release onto pancreas-projecting vagal motoneurons. This might be taken as an indication that the resting activity of the PPG neurons is not sufficient to induce local release of GLP-1, although whether there is any tonic release of GLP-1 at the level of the hypothalamus remains unknown.

NTS PPG neurons display A-type K⁺ currents, a feature that they share with other second-order NTS neurons such as TH-positive cells (18) as well as vagal preganglionic neurons of the adjacent dorsal vagal motor nucleus (DMNX [25]). In accordance with the hypothalamic projection of PPG neurons, A-type K⁺ currents have been described previously in NTS neurons projecting to the paraventricular nucleus, but not those projecting to the caudal ventrolateral medulla (26).

Most PPG neurons in the NTS were found to be second-order neurons that receive strong glutamatergic synaptic input from the solitary tract and express non-NMDA type glutamate receptors (27). These excitatory inputs would be predominantly from visceral vagal afferents, in line with the finding that c-Fos is induced in GLP-1 neurons upon gastric distension (6).

Rises in peripheral GLP-1 have been linked to activation of neurons located in the NTS. Our electrophysiological and expression data provide direct evidence that NTS PPG neurons do not have functional GLP-1 receptors. While we cannot exclude the possibility that they might respond to activation of presynaptic GLP-1 receptors, this seems unlikely to be the case because we observed no effect of GLP-1 or exendin-4 on firing rate or spontaneous synaptic activity. Thus, even if peripheral GLP-1 was able to cross the blood-brain barrier, it seems unlikely that it would have a local effect on the activity of PPG neurons. It remains possible, however, that potential connections between GLP-1–sensitive cells in the area postrema and PPG cells in the NTS might have been lost in the slice preparation.

An alternative pathway by which peripheral GLP-1 might modulate PPG cell activity in the NTS is via GLP-1 receptors on vagal afferent neurons (14). GLP-1 receptor mRNA has been detected in the inferior ganglion of the afferent vagus (the nodose ganglion), and electrical activity in the vagus nerve was shown to respond to GLP-1 application (14). It is therefore possible that NTS PPG cells, which appear to be predominantly second-order neurons with regard to solitary tract stimulation, would relay this vagal signal, resulting in central GLP-1 release in the brainstem and hypothalamus. In support of this hypothesis, it was shown that appetite suppression and arcuate nucleus expression of c-Fos elicited by peripheral GLP-1 were attenuated by vagotomy (28). The idea that circulating GLP-1 is detected by peripheral rather than central GLP-1 receptors was also suggested by the finding that intraperitoneal but not intracerebroventricular application of exendin (9–39) antagonized the appetite suppressive effects of peripheral GLP-1 (29). However, direct evidence that NTS PPG neurons are activated via a vagal pathway when GLP-1 is released postprandially from gastrointestinal L-cells is still lacking.

GLP-1 receptors have been reported on area postrema neurons as well as in many other locations throughout the brainstem (5,30). Electrophysiological studies in brainstem slices have demonstrated that pancreas-projecting dorsal vagal neurons are both directly and indirectly modulated by GLP-1 (23,24). The lack of GLP-1 responsiveness of PPG cells in the NTS suggests that PPG neurons do not receive inputs from the same cells as pancreas-projecting dorsal vagal neurons. One possibility is that they are actually the source of the GLP-1 that affects vagal output.

PYY_3–36 is a satiety peptide (31), released from intestinal L-cells in parallel with GLP-1. Its peripheral administration has been shown to increase c-Fos expression in cell bodies in the area postrema and NTS, including populations that were negative for tyrosine hydroxylase and might therefore overlap with PPG cells (32). In brainstem slices, however, we were unable to elicit either pre- or postsynaptic responses to PYY_3–36 in NTS PPG neurons. Peripheral injections of PYY_3–36 and GLP-1 receptor agonists have synergistic effects on appetite suppression in rats and mice (10). Destruction of vagal afferent neurons using capsaicin prevented the reduction of food intake by exendin-4, but not that by PYY_3–36 (33), suggesting that unlike GLP-1, the effect of PYY does not require intact sensory neurons. Although the PYY_3–36-responsive receptor type, Y2, has been identified in a number of brainstem regions including the area postrema and NTS (34), it remains unclear whether the appetite suppressive action of PYY_3–36 depends on NTS PPG neurons.

The adipocyte-derived hormone leptin directly and rapidly increased the electrical activity of PPG neurons. The electrophysiological data suggest that this is mediated by the activation of a depolarizing conductance, in contrast to the inhibitory effect of leptin on K⁺ channels described previously in ventromedial hypothalamic neurons (35). These results provide a potential explanation for previous findings that c-Fos immunoreactivity in PPG neurons is induced by leptin exposure (7). They also support previous observations that PPG neurons in mice express leptin receptor mRNA (9) and that leptin activates STAT3 phosphorylation, a marker of direct leptin action, in PPG neurons in mice, but not rats (21). The immediate effect of leptin on PPG cell electrical activity suggests that a component of the effect of this hormone, normally considered as a longer-term regulator of food intake, is mediated by enhancing postprandial satiety signals involving NTS PPG neurons. By contrast, the orexigenic peptide ghrelin (36), which is released from the stomach in anticipation of food ingestion, failed to affect the electrical activity of PPG neurons. Thus, although ghrelin has been reported to attenuate GLP-1–induced inhibition of food intake and gastric emptying (11), our results suggest that this does not happen at the level of the NTS PPG neurons.

In conclusion, our data provide a first functional in vitro characterization of PPG neurons in the caudal NTS—a study made possible by the development of Venus-tagged GLP-1 mice (15), which reliably express a fluorescent marker in PPG cells in the brainstem. Interestingly, we find that the activity of NTS PPG neurons is rapidly and directly modulated by leptin but is unaffected by GLP-1, PYY, or ghrelin. These findings have important implications for our understanding of the role of NTS PPG neurons in relaying appetite signals between the periphery and the hypothalamus.

This study was supported by the Medical Research Council, U.K. (K.H. and S.T. Ref. G0600928), and the Wellcome Trust (FMG Ref. 088357 and FR Ref. 084210).

No potential conflicts of interest relevant to this article were reported.

K.H. researched most data, contributed to discussion, and edited the manuscript; F.R. contributed to study design and edited the manuscript; F.M.G. contributed to the study design and edited the manuscript; and S.T. designed the study, researched some data, and wrote the manuscript.