Functional interaction between Ghrelin and GLP-1 regulates feeding through the vagal afferent system

The gastrointestinal tract transmits feeding-regulatory signals to the brain via neuronal and hormonal pathways. Here we studied the interaction between the orexigenic gastric peptide, ghrelin, and the anorectic intestinal peptide, glucagon-like peptide 1 (GLP-1), in terms of feeding regulation via the vagal afferents. GLP-1 preadministration 30 min before ghrelin administration to rats and mice abolished ghrelin-induced food intake, while ghrelin preadministration abolished the anorectic effect of GLP-1. Ghrelin preadministration suppressed GLP-1-induced Fos expression in the nodose ganglia (NG). Electrophysiological assessment confirmed that the initially administered peptide abolished the vagal afferent electrical alteration induced by the subsequently administered peptide. Both the growth hormone secretagogue receptor (GHSR) and the GLP-1 receptor (GLP-1R) are co-localised in a major proportion of NG neurons that innervate the stomach. In these Ghsr+Glp1r+ neurons, ghrelin preadministration abolished the GLP-1-induced calcium response. Ghrelin generated a hyperpolarising current and GLP-1 generated a depolarising current in isolated NG neurons in a patch-clamp experiment. Ghrelin and GLP-1 potently influenced each other in terms of vagally mediated feeding regulation. This peptidergic interaction allows for fine control of the electrophysiological properties of NG neurons.

Gastrointestinal signals such as peptides and neural signals modulate feeding and energy homeostasis [1][2][3][4] . The vagus nerve, one of the cranial nerves, is composed of sensory afferents and motor efferents. The primary sensory neurons of the vagal afferents are located in the bilateral nodose ganglia (NG) that relay peripheral nerve signals from the gastrointestinal tract to the nucleus tractus solitarius (NTS) in the medulla oblongata. The vagal afferents are implicated in the regulation of acute feeding behavior and long-term energy balance 5,6 . The vagal afferents are anatomically heterogeneous, and their peripheral axons in the gastrointestinal tract form characteristic sensory endings that are specialized for detection of chemical (mucosal endings) or mechanical (primarily intraganglionic laminar endings (IGLEs) and intramuscular arrays) stimuli 7,8 . The latter two serve as mechanoreceptors to detect gastrointestinal distension and luminal stroking 7 . Enteroendocrine cells distributed throughout the gastrointestinal tract produce a variety of peptides, including glucagon-like peptide 1 (GLP-1), ghrelin, cholecystokinin, and peptide YY. These peptides regulate energy homeostasis by acting directly on target tissues as hormones or by activating intrinsic and extrinsic neurons in a paracrine manner 9,10 . Gastrointestinal peptides bind to their receptors which are synthesized in the cell body of the NG neurons located in the neck and are transported to vagal afferent terminals in the gastrointestinal tract 11,12 .
At the beginning of food intake, the activity of orexigenic signals is elevated while that of anorectic signals is suppressed. With ongoing food intake, anorectic signals overcome the orexigenic tone to terminate feeding. The anorectic peptide GLP-1, which is primarily released from enteroendocrine L cells in response to meals, functions as a satiety signal 13 . Plasma GLP-1 in humans reaches its peak level within 30 min after a meal 13 . GLP-1 delays gastric emptying in human and rodents 13,14 . Ghrelin is an orexigenic peptide that is secreted from gastric endocrine cells [15][16][17] . Plasma ghrelin levels in humans rise 1-2 h before the onset of a meal and fall to trough levels 30 min after the end of the meal 18,19 . Ghrelin increases gastric motility to enhance gastric emptying 20 .
GLP-1 receptor (GLP-1R) agonists have been introduced for the treatment of type 2 diabetes mellitus 3 . Bai et al. performed single-cell RNA sequencing of the NG neurons, morphological characterization, and behavioural analysis in mice to generate a map of the NG neurons that innervate the stomach and intestine 7 . GLP-1R-expressing neurons innervating the stomach react to gastric distension and transmit signals to the hypothalamus to suppress food intake 7,21 . The growth hormone secretagogue receptor (GHSR), a cognate ghrelin receptor expressed in the NG neurons, is transported to the stomach via the vagal afferents and stimulated by ghrelin [22][23][24] . Two different routes have been proposed to convey the gastric-derived ghrelin signals to the brain: the vagal afferent nerve and blood circulation. Ghrelin transmits hunger signals to the NTS via the vagal afferents 9,25,26 , however some studies denied this transmission. A study using subdiaphragmatic vagal deafferentation in rats demonstrated that vagal afferents are not necessary for the orexigenic effect of intraperitoneal ghrelin 27 . Another researcher group reported that GHSR restoration in the hindbrain and NG in Ghsr −/− mice did not restore ghrelin-induced feeding 28 .
It has been proven that single NG neurons express multiple receptors for anorectic substances, for example GLP-1R, cholecystokinin A receptor, and leptin receptor 7,29 , but it remains unclear whether orexigenic GHSR coexists with anorectic receptors. Several lines of study have verified individual roles of GLP-1 and ghrelin in feeding regulation 14,17,[30][31][32][33][34] ; however, it remains largely unknown how these two counteracting peptides affect the vagal sensory system. Here, we investigated the effect of interaction between ghrelin and GLP-1 on feeding regulation governed by the vagal afferents. We showed that GLP-1R and GHSR were co-expressed in a major subpopulation of NG neurons that innervate the stomach in mice. We also studied the effects of the two peptides on intracellular calcium mobilization and the membrane potentials of isolated vagal afferent neurons.

Preadministration of either ghrelin or GLP-1 abolishes feeding effect of GLP-1 or ghrelin.
To test whether the interaction between ghrelin and GLP-1 impacts on feeding regulation, we first designed feeding experiments in which these two peptides were administered in a 30-min interval. Food was given after the second peptide administration. In rats fed ad libitum, intravenous (i.v.) administration of ghrelin significantly enhanced 1-h food intake compared with saline injection (Fig. 1A). When food was given after GLP-1 administration, ghrelin kept its orexigenic effect (Fig. 1A). GLP-1 administration to rats in the dark phase suppressed feeding, while preadministration of ghrelin 30 min before GLP-1 also stimulated food intake (Fig. 1B). In 8-h fasted rats, GLP-1 significantly reduced food intake compared with saline injection, and ghrelin did not enhance feeding when it was administered 30 min after GLP-1 administration (Fig. 1C). We also studied food intake experiments in 8-h fasted mice. Ghrelin preadministration inhibited the anorectic effect of GLP-1 under this condition (Fig. 1D). In Ghsr −/− mice fasted 12 h, ghrelin preadministration did not affect GLP-1-induced anorexia (Fig. 1E), whereas ghrelin preadministration to their wild-type mice abolished GLP-1-induced anorexia (Fig. 1F). When GLP-1 was administered to Glp1r −/− mice fed ad libitum 30 min before ghrelin, GLP-1 did not affect ghrelin-induced feeding (Fig. 1G), whereas GLP-1 preadministration to their wild-type mice abolished ghrelin-induced feeding (Fig. 1H). To determine how long ghrelin and GLP-1 interact on feeding, we next administered the two peptides to mice in a 60-min interval. The results were different from those in studies in which they were administered in a 30-min interval. The peptide administered 60 min after the first peptide administration exhibited its proper effect in both cases with ghrelin and GLP-1 ( Supplementary Fig. S1).
The first peptide disappears from the plasma 30 min after its administration. We determined plasma profiles of ghrelin and GLP-1 administered to rats. When GLP-1 was administered first, its plasma concentration reached the peak at 5 min and returned to the base at 30 min when ghrelin was administered (Supplementary Fig. S2A). When ghrein was administered first, its plasma concentration also reached the peak at 5 min and returned to the base at 30 min when GLP-1 was administered ( Supplementary Fig. S2B).

GHSR and GLP-1R are expressed in NG neurons and transported within the vagal afferents.
To determine whether a distinct subset of NG neurons received signals from the gastric corpus, we performed a retrograde tracer study in mice. Right and left NG neurons were examined by fluorescence microscopy after injections of Alexa Fluor 488-conjugated cholera toxin B (CTB) into both ventral and dorsal sides of the gastric corpus ( Fig. 2A, B). The number of NG neurons labeled with Alexa 488-CTB retrogradely transported from the stomach were nearly equal (18.2% of 566 left NG neurons and 16.7% of 479 right NG neurons) (Fig. 2C).
To investigate the expression of Ghsr and Glp1r, NG neurons labeled with Alexa 488-CTB were collected and studied by single-neuron mRNA analysis. Both Ghsr-and Glp1r-expressing neurons account for 70.8%, Ghsr alone 20.8%, Glp1r alone 6.3%, neither of them 2.1% (Fig. 2D). To examine the transportation of GHSR and GLP-1R within the vagal afferents, we studied bindings of 125 I-ghrelin and 125 I-GLP-1 in the vagal segments. The transportation of these receptors from the NG neurons towards the afferent endings were blocked by ligation of the vagal segments. Radioactivities of both peptides were detected at the proximal to ligature (Fig. 2E). We applied immunocytochemistry to study colocalization of GHSR and GLP-1R in isolated NG neurons. Dispersed NG neurons of mice were double stained with anti-GLP-1R and anti-GHSR antibodies. Both receptors were  www.nature.com/scientificreports/ the gastric vagal afferent of rats administered ghrelin and GLP-1. Ghrelin attenuated the vagal afferent activity (Fig. 3A), whereas GLP-1 enhanced it (Fig. 3B). When GLP-1 was administered 30 min after ghrelin, GLP-1-enhanced activity was not observed (Fig. 3A). Conversely, ghrelin administration 30 min after GLP-1 did not attenuate GLP-1-enhanced afferent activity (Fig. 3B). We next studied Fos induction as a proxy for neuronal activation in mice. GLP-1 induced Fos in NG neurons (left 70.5% and right 71.0%) (Fig. 3C,D). This Fos expression was reduced by ghrelin administration before 30 min (left NG 21.7% and right NG 20.4%) (Fig. 3C,D). By contrast, ghrelin administration before 60 min did not affect GLP-1-induced Fos (left NG 68.5% and right NG 69.8%) ( Supplementary Fig. S4). This finding supports the result in the feeding study operated in a 60-min interval in mice ( Supplementary Fig. S1). GLP-1 also induced Fos in the NTS (Fig. 3E). Ghrelin preadministration abolished GLP-1-induced Fos in the NTS (Fig. 3E). Ghrelin induced Fos in the hypothalamic arcuate nucleus (ARC) (Fig. 3F). GLP-1 preadministration abolished ghrelin-induced Fos in the ARC (Fig. 3F).
Preadministration of ghrelin or GLP-1 affects calcium response modulated by the second peptide administration and the two peptides have opposite effects on current-voltage. To study neuronal activation of GLP-1 and ghrelin on the vagal afferents, we measured calcium response in single neurons isolated from mouse NG. GLP-1 evoked calcium response in Glp1r + Ghsr + neurons and ghrelin 10 min after GLP-1 did not affect the response (Fig. 4A,C). GLP-1 10 min after ghrelin failed to evoke calcium response in Glp1r + Ghsr + neurons (Fig. 4B,D). GLP-1 evoked calcium response in Glp1r + Ghsr − neurons ( Supplementary  Fig. S5A), but not in Glp1r − Ghsr − neurons ( Supplementary Fig. S5B). Ghrelin did not evoke calcium response in Glp1r − Ghsr + neurons ( Supplementary Fig. S5E, F). After calcium measurement, all neurons were individually collected to study Glp1r and Ghsr expression by qRT-PCR (Fig. 4E).
We next performed whole-cell voltage-clamp study to examine the electrical properties of the NG neurons in response to ghrelin or GLP-1 administration. Current responses of the NG neurons to voltage clamp (− 80 to 40 mV) were used to generate the I/V relationships of ghrelin-or GLP-1-induced responses. Ghrelin generated hyperpolarizing current (Fig. 4F,H) and GLP-1 generated depolarizing current (Fig. 4G,I) in Ghsr-and Glp1rexpressing NG neurons, respectively.

Discussion
This study demonstrated how vagal afferents modulate the functional interaction between GLP-1 and ghrelin in NG neurons to regulate feeding. The administration of ghrelin and GLP-1 at a 30-min interval abolished the feeding effect of the peptide administered second; however, when they were administered at a 60-min interval, this abrogation did not occur. This finding was consistent with the fact that ghrelin preadministration reduced GLP-1-induced Fos expression in the NTS when administered at a 30-min interval but not at a 60-min interval. Here, when the two peptides were administered at a 30-min interval, the plasma concentration of the first peptide returned to baseline upon administration of the second peptide. This suggests that the first peptide did not affect the second peptide as a circulating hormone.
Another possibility of the direct effects of GLP-1 and ghrelin in the brain has been proposed since both peptides cross the brain blood barrier 25,35 . Further studies are needed how peripherally administered ghrelin and GLP-1 activate the neurons in the brain; directly via their respective receptors, via the vagal afferents or both.
Luminal stroking and circular stretching of the stomach lead to feelings of hunger and satiety, respectively 3,7 . In humans, GLP-1 inhibits meal-induced gastrointestinal motility via vagal pathways 13 . GLP-1R-expressing NG neurons functions as mechanoreceptors that innervate gastric IGLEs and detect gastric distension 36 . Selective deletion of the GLP-1R in NG neurons accelerated gastric emptying in rats 14 . The GHSR is also expressed in tension-sensitive NG neurons that innervate the stomach 37 . Furthermore, ghrelin reduces the mechanosensitivity of gastric vagal afferents in mice and ferrets 38 . Together, ghrelin and GLP-1 may interact to regulate feeding by modulating gastric motility. The right and left vagal nerves enter the abdomen as two trunks that divide into distinct primary branches at the subdiaphragmatic esophageal level 39 . The ventral and dorsal surfaces of the stomach are innervated by gastric branches that originate from neurons in the right and left NG, respectively 39,40 . The small intestine is innervated by neurons from both the right and left NG 40 . In this study, a retrograde tracer experiment in which CTB was injected in equal into the ventral and dorsal sides of the stomach indicated that nearly the same number of neurons from the right (16.7%) and left (18.2%) NG innervated the stomach, supporting the branching pattern of the vagal gastric neurons. We also found four groups of stomach-projecting NG neurons on the basis of the expression of GHSR and GLP-1R (both GHSR-and GLP-1R-expressing neurons, 70.8%; GHSR alone, 20.8%; GLP-1 alone, 6.3%; neither, 2.1%). A major portion (91.6%) of stomach-projecting NG neurons expressed GHSR, supporting the previous finding of GHSR expression in 98% of tension-sensitive NG neurons that were retrogradely traced from the stomach 37 . To explore the effect of ghrelin and GLP-1 on the cellular excitability of Glp1r + Ghsr + neurons, we studied these neurons' intracellular signalling mechanisms.
Activation of GLP-1R, a G αs -coupled GPCR, increased cAMP concentrations in rat NG neurons and enhanced their excitability by suppressing potassium currents and enhancing membrane depolarization, thereby increasing intracellular calcium 41,42 . GHSR activates a G αq/11 -coupled GPCR in the pituitary to stimulate intracellular calcium mobilization from the endoplasmic reticulum 33 , whereas in NG neurons, Grabauskas et al. showed that ghrelin-induced hyperpolarization was suppressed by pertussis toxin, a G αi protein inhibitor 43 . They concluded that by activating the G αi -PI3K-Erk1/2-K ATP pathway in NG neurons, ghrelin evoked potassium currents and thereby causes hyperpolarization 43 . GPCRs can bind to distinct classes of heterotrimeric G-proteins in different cell types. We here showed that ghrelin preadministration abolished the GLP-1-induced calcium response in Glp1r + Ghsr + neurons, supporting that GHSR in NG neurons is coupled to G αi protein. Using a patch-clamp experiment combined with single-neuron mRNA analysis, we verified that Glp1r + Ghsr + neurons were depolarized www.nature.com/scientificreports/ by GLP-1 and hyperpolarized by ghrelin. These physiological effects on Glp1r + Ghsr + neurons may modulate the anorectic or orexigenic signals that they convey to the brain. In summary, we demonstrated that ghrelin and GLP-1 influence feeding regulation through their opposite effects on the cellular excitability of vagal afferent neurons. This peptidergic interaction sheds light on the fine control of feeding regulation.

Jugular vein cannulation.
Core body temperature of anesthetized rats was maintained at 37 °C using a heating blanket. A sterilised i.v. catheter (Braintree Scientific) containing pyrogen-free heparin saline (500 U/ ml) was inserted into the jugular vein. Rats were acclimated to the experimental conditions by daily handling during 7 days. Only rats exhibiting progressive weight gain after surgery were used in the experiments.
Culture preparation and immunocytochemistry. Bilateral NGs resected from mice were immersed in hibernateA media (BrainBits). Each ganglion was cut into five pieces, placed in HibernateA minus Ca 2+ media (BrainBits) containing 1 mg/ml collagenase/dispase (Roche Diagnostics) and incubated for 90 min at 37 °C. Neurons were dispersed by gentle titration through pipettes and they were washed three times with fresh NbAc-tive4 media (BrainBits) and plated onto poly-D-lysine/laminin (BD Bioscience)-coated coverslips, followed by incubation (5% CO 2 in air at 37 °C) for 24 h. Cultures were rinsed in 10 mM phosphate buffered saline (PBS) and fixed with 4% paraformaldehyde (PFA) in phosphate buffer (PB). Ganglion neurons were incubated with the first primary antibodies; anti-GLP-1R or anti-GHSR at 4 °C overnight, and then with the second primary antibodies; anti-GHSR or anti-Pan Neuronal Marker (Supplementary Table S1) at 4 °C overnight. They were then reacted with corresponding Alexa Fluor 488-or 594-conjugated secondary IgG (Supplementary Table S1) at RT for 1 h. Neurons were observed under a fluorescence microscope (Olympus). Retrograde tracing studies. The abdomen of mice (n = 3) was exposed after a midline laparotomy. Alexa    Table S1) at 4 °C. Images of immunofluorescence staining were observed using a Nikon C2 + confocal laser scanning microscope (Nikon, Tokyo, Japan). Fos-positive neurons in three specimens from each animal were counted using ImageJ.

Vagal branch ligation and autoradiography.
Electrophysiological recordings of the vagal nerve. Multi-unit neural discharge in gastric vagal afferent fibres was recorded extracellularly. Rats were anesthetised by an i.p. injection of urethane (1 g/kg) (Sigma-Aldrich). The electrophysiological study was performed entirely under anaesthetisation as described previously 45 . After the gastric branch of the vagal nerve was visualised, we placed filaments isolated from the peripheral cut end of the ventral branch for the recording of afferent nerve activity via a pair of silver wire electrodes connected through an alternating current-coupled differential amplifier (ER-1; Cygnus Technology) to PowerLab (AD Instruments). The nerve activities were recorded and analysed using LabChart v8 software (AD Instruments measurements in an integrated fluorescence microscope (Keyence BZ-X700), images were captured at 10-s intervals; 340-and 380-nm excitation filters were used for Fura-2-AM dual-wavelength excitation ratio imaging 46 . NG neurons were stimulated with 10 nM either ghrelin or GLP-1, and 10 min later stimulated with the other peptide. At the end of experiments, cells were exposed to 100 mM KCl (Nacalai Tesque) for 2 min. Cells exhibiting more than 200% of the 340/380 fluorescence ratio in response to KCl were used for the analysis. All data were expressed as percent changes from the average of the 340/380 fluorescence ratio in response to ghrelin or GLP-1.
Single-cell RT-PCR. All neurons collected in retrograde tracing studies and used to assess calcium response and patch-clamp recording were picked up by the UnipicK + system (NeuroInDx) immediately after each measurement. RNA was isolated from single neurons with PicoPure RNA Isolation Kit (Applied Biosystems). qRT-PCR (primer sets, Supplementary Table S2) was conducted with SYBR Premix Ex Taq (2 ×) (Takara Bio Inc.) on a Thermal Cycler Dice Real-Time System II (Takara Bio Inc.). In the retrograde tracer studies, percentage of stomach-innervating NG neurons expressing Ghsr or Glp1r were analysed. In the patch-clamp recording, only Ghsr + Glp1r + neuron cells were used for the final analysis.
Statistical analysis. Statistical analyses were performed by one-way ANOVA followed by Bonferroni's post-test for multiple comparisons, as appropriate. When two mean values were compared, analyses were performed by unpaired t-test. All data are expressed as means ± s.e.m. A value of P < 0.05 was considered to be statistically significant.