Report

Regulation of feeding by somatostatin neurons in the tuberal nucleus

See allHide authors and affiliations

Science  06 Jul 2018:
Vol. 361, Issue 6397, pp. 76-81
DOI: 10.1126/science.aar4983

Neurons that regulate feeding

The tuberal nucleus, an area of the hypothalamus, has not been studied in great detail. Luo et al. found that GABAergic somatostatin neurons in the tuberal nucleus are functionally involved in the regulation of feeding in mice (GABA, γ-aminobutyric acid) (see the Perspective by Diano). These neurons were activated by food deprivation or hunger hormone. Loss- and gain-of-function experiments indicated that these cells are necessary and sufficient to control systemic metabolic balance. This newly described regulatory center is extensively connected with other feeding control circuits via projections to other hypothalamic nuclei.

Science, this issue p. 76; see also p. 29

Abstract

The tuberal nucleus (TN) is a surprisingly understudied brain region. We found that somatostatin (SST) neurons in the TN, which is known to exhibit pathological or cytological changes in human neurodegenerative diseases, play a crucial role in regulating feeding in mice. GABAergic tuberal SST (TNSST) neurons were activated by hunger and by the hunger hormone, ghrelin. Activation of TNSST neurons promoted feeding, whereas inhibition reduced it via projections to the paraventricular nucleus and bed nucleus of the stria terminalis. Ablation of TNSST neurons reduced body weight gain and food intake. These findings reveal a previously unknown mechanism of feeding regulation that operates through orexigenic TNSST neurons, providing a new perspective for understanding appetite changes.

The nucleus tuberalis lateralis (NTL; lateral tuberal nucleus) in the hypothalamus is revealed by the “lateral eminences on the ventral surface of the tuber cinereum” in humans and primates (1), but the homologous structure in other animals is less well defined, and knowledge about the NTL is scarce (2). Pathological or cytological changes in the NTL have been observed in several neurodegenerative and neurological diseases (28). Deep hypothalamic lesions that probably went into the NTL in human patients have led to hypophagia, and NTL pathology has been correlated with cachexia (4, 911), suggesting that the NTL plays a role in the regulation of feeding, but this notion has never been tested. The human NTL neighbors the tuberomammillary nucleus (TMN) and is marked by dense staining with somatostatin (SST) antibody (12). We therefore searched for SST-positive neurons in the mouse tuberal nucleus (TN) and studied their potential role in feeding regulation.

We bred SST-Cre mice with Ai14 reporter mice (13), and SST neurons thus were labeled with the red fluorescent protein tdTomato. In the mouse hypothalamus, SST neurons appeared in a scattered distribution in the medial-dorsal hypothalamus and in enriched clusters in the ventral hypothalamus. Specifically, SST neurons were prevalent in the arcuate area (we refer to these as ArcSST neurons) and the TN (TNSST neurons). The arcuate cluster extended slightly anteriorly, and the TN cluster extended slightly posteriorly (Fig. 1A). The arcuate nucleus plays a critical role in feeding regulation and contains well-studied orexigenic agouti-related peptide (AgRP) neurons, tyrosine hydroxylase (TH)–positive neurons, and anorexic proopiomelanocortin (POMC) neurons (1416). Immunohistochemistry showed that the ArcSST neurons were distinct from POMC- and TH-positive neurons but exhibited minor overlap (9.04 ± 2.16% of ArcSST neurons; n = 3 mice) with AgRP neurons (Fig. 1B) (17). The TNSST neurons were distinct and spatially segregated from lateral hypothalamic orexin- and melanin-concentrating hormone (MCH)–positive neurons (Fig. 1C), both of which have been implicated in feeding and metabolic regulation (18, 19). The TNSST did not show obvious overlap with histaminergic neurons [revealed by staining for histidine decarboxylase (HDC)], the key neuronal subtype in the nearby TMN region (20) (Fig. 1C and fig. S1). Therefore, as in humans, the TN in mouse is marked by a dense cluster of SST-positive neurons, which constitute a hypothalamic neuronal subtype that is distinct from the neurons currently known to support feeding and metabolic regulation. Either overnight fasting or intraperitoneal (i.p.) injection of the hunger hormone, ghrelin—the levels of which are increased by the hunger state (21)—induced robust expression of the immediate early gene c-Fos in TNSST neurons (Fig. 1, D to G, and fig. S2), but not in ArcSST neurons (fig. S3). Further analysis showed no obvious difference in the percentage of c-Fos–positive TNSST neurons at different positions, indicating a homogeneous activation of TNSST neurons (fig. S2). In acute slices of SST-Cre::Ai14 mice, bath application of ghrelin excited 7 out of 17 TNSST neurons, as determined by whole-cell recording (Fig. 1, H and I). Under the presence of synaptic blockers, 6 out of 18 TNSST neurons were excited by ghrelin under current-clamp conditions without the injection of depolarizing current, and 7 out of 16 TNSST neurons were excited by ghrelin with current injection for increasing baseline firing (fig. S4, A to F). We further characterized the TNSST neurons by staining mRNA for ghrelin receptor (Ghsr), using fluorescence in situ hybridization (FISH); we found that 17.6% (120/682) of TNSST neurons were Ghsr-positive, with the most caudal TN containing fewer Ghsr-positive SST neurons (fig. S4, G and H). Ghsr-postive and Ghsr-negative TNSST neurons were similarly activated by overnight fasting (fig. S4, I and J), indicating that both populations were involved in feeding regulation.

Fig. 1 The ventral hypothalamic SST neuron clusters.

(A) Serial coronal sections of SST-Cre::Ai14 mice showing distinct SST neuron clusters in both the arcuate area and the tuberal nucleus of the hypothalamus. Scale bars, 500 μm. (B) Coronal sections of SST-Cre::Ai14 mice stained for POMC, AgRP, and TH (the white arrow indicates a SST neuron that is positive for AgRP). Scale bars, 100 μm. IHC, immunohistochemistry. (C) Coronal sections of SST-Cre::Ai14 mice stained for MCH, Orx (orexin), and HDC. Scale bars, 200 μm. (D) Coronal sections of SST-Cre::Ai14 mice that were (fasted) or were not (fed) subjected to overnight fasting and immunostained for c-Fos. White arrows indicate SST neurons that are c-Fos–positive. (E) Coronal sections of SST-Cre::Ai14 mice after i.p. injection of saline or ghrelin, immunostained for c-Fos. White arrows indicate SST neurons that are c-Fos–positive. Scale bars in (D) and (E), 100 μm. (F) Percentage of c-Fos–positive SST neurons under fed or fasted conditions (mean ± SEM, n = 3, unpaired t test, *P < 0.02). (G) Percentage of c-Fos–positive SST neurons after i.p. injection of saline or ghrelin (mean ± SEM, n = 3, unpaired t test, *P < 0.02). (H) Whole-cell patch recording trace showing the change in firing frequency after bath application of ghrelin (100 nM). (I) Firing rate before and after ghrelin application (n = 17, paired t test, *P < 0.05).

We next injected Cre recombinase–inducible adeno-associated viruses (AAVs) expressing chemogenetic or optogenetic effectors into the TN of SST-Cre mice (Fig. 2) and analyzed their eating behavior (fig. S5). After the mice habituated in behavioral chambers for 4 days (fig. S6A), chemogenetic activation of TNSST neurons by expressing the excitatory DREADD (designer receptor exclusively activated by designer drug) hM3D and injection of CNO in the late morning (11 a.m.) dramatically promoted food consumption within the subsequent 3 hours and significantly enhanced eating time and frequency (Fig. 2, A to E). Similar results were found when CNO was injected at 5 p.m., when mice were more spontaneously active (fig. S6, B and C). No effect of CNO on eating behavior was observed in control mice with green fluorescent protein (GFP) expressed in TNSST neurons (Fig. 2, B to E, black traces). After expressing Cre-dependent channelrhodopsin (ChR2) in TNSST neurons, we characterized the optogenetic response of TNSST neurons in acute slices (fig. S7, A to F) and used a light intensity that can reliably excite TNSST neurons for in vivo experiments (22). In vivo unilateral optogenetic stimulation reliably and progressively promoted eating as stimulation frequency was increased (fig. S7, G and H). Moreover, repeated optogenetic stimulation (20 Hz for 1 hour) reliably enhanced eating frequency, which required the presence of chow (fig. S8A); in a control experiment, chemogenetic activation of TNSST neurons did not elicit gnawing of a wood stick (fig. S8B). As reported recently (17), we also found that chemogenetic activation of ArcSST promoted eating (fig. S6D). We further examined the necessity of TNSST neurons in homeostatic eating by expressing the inhibitory DREADD κ opioid receptor (KORD) in SST-Cre mice and injecting salvinorin B (SalB) to inhibit the activity of TNSST neurons (Fig. 2F) (23). Injecting SalB at 10 p.m., at the start of the peak of homeostatic eating, significantly reduced cumulative eating time and eating frequency, but not eating bout duration (Fig. 2, G to I), which did not correlate with eating frequency in homeostatic eating (fig. S5). SalB injection in GFP control animals did not produce a significant difference from vehicle-injected animals (Fig. 2, G to I, black traces). Inhibition of TNSST neurons by SalB also significantly reduced food consumption in refeeding after overnight fasting (Fig. 2J). Because injecting SalB at the peak of the animals’ dark cycle may cause stress and interfere with homeostatic eating, we expressed Cre-dependent archaerhodopsin (ArchT) in the TN of SST-Cre mice and optogenetically inhibited these neurons by using light from 10 p.m. to 12 a.m., without disturbing the mice’s activity (Fig. 2K). Again, optogenetic inhibition of TNSST neurons significantly reduced cumulative eating time and eating frequency, but not eating bout duration (Fig. 2, L to N). The same optogenetic protocol did not cause changes in eating behavior in GFP control mice (Fig. 2, L to N, black traces). To test whether TNSST neurons have long-term effects on energy homeostasis, we injected AAV-flex-taCasp3-TEVp to express caspase-3 and ablate TNSST neurons in SST-Cre mice (24) (fig. S9). Ablation of TNSST neurons reduced body weight gain by 56% over 10 weeks (Fig. 2O). We further studied the impact of ablating TNSST neurons by using another set of mice in metabolic chambers, 4 weeks after injecting viruses, when the TNSST-ablated mice did not show significant body weight differences from the control mice but gained body weight significantly more slowly (fig. S10). TNSST ablation resulted in mild but significantly reduced daily and dark-cycle food intake, reduced prandial drinking, mildly reduced respiratory exchange ratio, similar total energy expenditure, and mild changes in activity (fig. S10). Because both positive and negative valence mechanisms appear to play a role in controlling feeding (2528), we explored the motivational valence of TNSST in mice. In the absence of food, chemogenetic activation of the TNSST neurons, paired with CNO injection, elicited a significant preference for the chamber (Fig. 2P).

Fig. 2 The TNSST neurons regulate eating.

(A) AAV-DIO-hM3D-mCherry was bilaterally injected into SST-Cre mice (DIO, double-floxed inverse orientation). (B) Food consumption, (C) cumulative eating time, (D) mean eating bout duration, and (E) eating frequency during the 3 hours after CNO injection at 11 a.m. (n = 7, **P < 0.01, paired two-tailed t test). SST-Cre mice were injected with AAV-FLEX-GFP as a control (n = 6, P > 0.5, paired two-tailed t test). n.s., not significant. (F) AAV-DIO-KORD-mCitrine was bilaterally injected into SST-Cre mice. (G) Cumulative eating time, (H) eating frequency, and (I) eating bout duration during the 3 hours after SalB injection (n = 6, *P < 0.05, paired two-tailed t test). SST-Cre mice were injected with AAV-FLEX-GFP as a control (n = 8, P > 0.5, paired two-tailed t test). (J) SST-Cre mice were injected with AAV-DIO-KORD-mCitrine, and food consumption after overnight fasting was measured [n = 7, *P < 0.05, one-way analysis of variance (ANOVA)]. (K) AAV-FLEX-ArchT-GFP was bilaterally injected into SST-Cre mice. (L) Cumulative eating time, (M) eating frequency, and (N) eating bout duration during 2 hours of yellow light illumination (n = 6, *P < 0.05, **P < 0.01, paired two-tailed t test). SST-Cre mice were injected with AAV-FLEX-GFP as a control (n = 7, P > 0.5, paired two-tailed t test). (O) Body weight gain on a normal chow diet for SST-Cre mice bilaterally injected in the TN with either AAV-FLEX-GFP as a control (n = 4) or AAV-flex-taCasp3-TEVp (n = 5) for ablating TNSST neurons (ablation versus control: repeated measures two-way ANOVA, F = 24.21, P < 0.001). (P) Experimental design of conditioned place preference with chemogenetic activation of TNSST neurons. A heat map shows the percent occupancy time before and after conditioning sessions. Red bar, chemogenetic activation side. The bar graph shows the change in occupancy time after conditioning sessions for SST-Cre mice injected bilaterally with AAV-DIO-hM3D-mCherry (n = 4) or AAV-FLEX-GFP (n = 3) (**P < 0.01, one-way ANOVA). Bold traces, means ± SEM; faint traces, individual mice. 3v, third ventricle. LED, light-emitting diode. Scale bars in (A), (F), and (K), 500 μm.

To explore the circuit mechanisms of feeding regulation by TNSST neurons, we examined the projection pattern of TNSST neurons by injecting AAV-FLEX-GFP into the TN of SST-Cre mice (Fig. 3A). Similarly to AgRP neurons (29), TNSST neurons projected to various brain regions, including the paraventricular nucleus (PVN), the bed nucleus of the stria terminalis (BNST), the central amygdala (CeA), and the periaqueductal gray (PAG), with sparse projection to the paraventricular nucleus of the thalamus (PVT) and the parabrachial nuclei (PBN) (Fig. 3A). By injecting the retrograde tracers CTB488 and CTB647 into the BNST and PVN, respectively, in SST-Cre:Ai14 mice, we found that the majority (87.7%) of PVN-projecting TNSST neurons projected to the BNST, and the majority (67.8%) of BNST-projecting TNSST neurons projected to the PVN (Fig. 3B). Using a similar method, we found that 36.1% of CeA-projecting TNSST neurons projected to the BNST, and 45.7% of BNST-projecting TNSST neurons projected to the CeA; 7.4% of BNST-projecting TNSST neurons projected to the PAG, and 30.5% of PAG-projecting TNSST neurons projected to the BNST; and, lastly, 13.8% of CeA-projecting TNSST neurons projected to the PAG, and 29.6% of PAG-projecting TNSST neurons projected to the CeA (Fig. 3, C to E, and fig. S11). Therefore, in contrast to AgRP neurons, different subpopulations of which project to different brain regions in a one-to-one pattern (29), the TNSST neurons showed a one-to-many projection pattern.

Fig. 3 The TNSST neurons regulate feeding through projections to the PVN and BNST.

(A) Top left, diagram of a sagittal brain section summarizing the axonal projections of TNSST neurons of SST-Cre mice injected with AAV-FLEX-GFP. Other panels show GFP-expressing axon terminals in different brain regions. Scale bars, 200 μm. (B to E) Left, diagrams illustrating the injection of retrograde tracers in different brain regions of SST-Cre::Ai14 mice. Center, red TNSST neurons took up CTB647 or CTB488 injected in different projection sites. Yellow arrows indicate the TNSST neurons positive for both CTB647 and CTB488. Green arrows indicate the TNSST neurons positive only for CTB488. Blue arrows indicate the TNSST neurons positive only for CTB647. Scale bar, 50 μm. Right, Venn diagrams illustrate the quantifications from three mice. (F to I) Top, diagrams showing optogenetic stimulation of different axon terminals of TNSST neurons injected with AAV-DIO-ChR2-mCherry. Bottom, bar graphs showing the change in eating frequency after optogenetic stimulation (bars, means ± SEM; black lines, individual mice; *P < 0.05, **P < 0.01, paired two-tailed t test).

We then optogenetically stimulated TNSST axon terminals in different brain regions and examined their roles in feeding regulation. Stimulating the TNSST terminals in the PVN and BNST strongly promoted feeding, but stimulating the terminals in the CeA and PAG had no effect on feeding (Fig. 3, F to I, and fig. S12). To better understand the neural transmission between TNSST neurons and their downstream neurons, we performed FISH for SST and Gad1 or Gad2, the two major enzymes for producing γ-aminobutyric acid (GABA) (30). We found that 91.4% of TNSST neurons expressed Gad1, and 87.8% expressed Gad2 (Fig. 4, A and B). We further confirmed by immunohistochemistry that the majority of TNSST neurons produced GABA (Fig. 4C). Moreover, no TNSST neurons were positive for VGlut2 (Fig. 4D), the major glutamatergic neuron marker in the hypothalamus (25). To directly examine the contribution of GABA signaling on the effect of activating TNSST neurons, we locally injected CNO with or without the GABAA receptor antagonist bicuculline into the PVN of SST-Cre mice that had been injected with AAV-DIO-hM3D in the TN. Local injection of CNO into the PVN stimulated feeding similarly to i.p. injection of CNO (1.43 ± 0.14 g versus 1.20 ± 0.11 g chow; P = 0.27), but co-injecting bicuculline into the PVN dramatically reduced this effect (1.20 ± 0.11 g versus 0.43 ± 0.05 g; P < 0.01) (Fig. 4E). Therefore, TNSST neurons stimulated feeding through the inhibition of downstream neurons, a mechanism similar to that of AgRP neurons (31, 32). Despite this similar effect in promoting feeding, ablating AgRP neurons leads to severe starvation (33), but we found that ablating TNSST neurons resulted in reduced body weight gain—an effect that is comparable to that of ablating lateral hypothalamic or zona incerta GABAergic neurons (28, 34)—suggesting a functional differentiation between AgRP and other hypothalamic orexigenic neurons.

Fig. 4 The TNSST neurons are GABAergic.

FISH for (A) SST and Gad1 or (B) SST and Gad2 in C57BL/6J mice. Scale bars, 200 μm. (C) Immunohistochemical staining for GABA in SST-Cre::Ai14 mice. Scale bars, 200 μm. (D) FISH for SST and VGlut2 in C57BL/6J mice. Scale bar, 200 μm. (E) SST-Cre mice were injected bilaterally with AAV-DIO-hM3D-mCherry in the TN, and CNO was injected either i.p. or by infusion cannula implanted over the PVN with or without bicuculline (Bic). The same animal was subjected to all three different injections on different days, and the amount of food consumption 3 hours after drug injection was analyzed (n = 6, **P < 0.01, paired t test). Gray bars, means ± SEM; black lines, individual mice. (F) TNSST neurons in the mouse TN can be activated by hunger signals, including ghrelin. Activation of TNSST neurons stimulates feeding through inhibiting downstream neurons in the PVN and BNST, a mechanism similar to the activation of AgRP neurons in the arcuate area.

Our results show that, as in humans (12), the TN in mice is marked by SST-positive neurons. This fact provided us with genetic access to study the function of this previously enigmatic brain region. We found that TNSST neurons are required for maintaining normal body weight gain, and TNSST neurons project to various brain regions and promote feeding by inhibiting downstream neurons. Taken together, our results reveal an important role of the TN in energy homeostasis and point to a previously unknown circuit mechanism of feeding regulation that operates through orexigenic TNSST neurons (Fig. 4F). These findings promise a better understanding of the appetite and body weight changes.

Supplementary Materials

www.sciencemag.org/content/361/6397/76/suppl/DC1

Materials and Methods

Figs. S1 to S12

References (3537)

References and Notes

Acknowledgments: We thank M. Stryker for critical reading of the manuscript. Funding: This work was supported by an A*STAR Investigatorship provided by the Biomedical Research Counsel (BMRC) of A*STAR (1530700142 to Y.F.), intramural funding from A*STAR BMRC (to W.H.), the National Natural and Science Foundation of China (81771215 to J.H.), and National Medical Research Council Collaborative Research Grants (NMRC/CBRG/0099/2015 and NMRC-OFIRG-0037-2017 to K.K. and S.S.). Author contributions: Y.F. conceived of the project and wrote the manuscript. S.X.L., J.H., Q.L., and Y.F. designed and implemented the study. H.M. performed retrograde tracing and local CNO injection. C.-Y.L., K.K., and S.S. contributed electrophysiological recordings. A.M.-Y.K., Y.L.T., and M.H. performed immunohistochemical staining. J.Y.L., and H.L. performed the metabolic chamber study. J.S. helped with in vivo optogenetic stimulation. L.Y.Y. and J.G. performed magnetic resonance imaging. All authors discussed the results and commented on the manuscript. Competing interests: The authors declare no competing interests. Data and materials availability: All data deeded to evaluate the conclusions of the paper are present in the paper and the supplementary materials.
View Abstract

Subjects

Navigate This Article