Rapid binge-like eating and body weight gain driven by zona incerta GABA neuron activation
Have you ever wondered why some foods are so difficult to stop eating? Well, you're not alone. Over 2% of Americans will be clinically diagnosed with binge eating disorder in their lifetime. The authors of this paper have pinpointed a specific part of the brain that might be responsible for this kind of overeating behavior. Using a mouse model, the scientists examined how a region of brain cells, known as zona incerta GABA neurons, affects feeding on high-fat and sweet foods. This study offers a better understanding of binge eating behaviors in humans and sheds light on why some food are just so hard to stop eating.
The neuronal substrate for binge eating, which can at times lead to obesity, is not clear. We find that optogenetic stimulation of mouse zona incerta (ZI) γ-aminobutyric acid (GABA) neurons or their axonal projections to paraventricular thalamus (PVT) excitatory neurons immediately (in 2 to 3 seconds) evoked binge-like eating. Minimal intermittent stimulation led to body weight gain; ZI GABA neuron ablation reduced weight. ZI stimulation generated 35% of normal 24-hour food intake in just 10 minutes. The ZI cells were excited by food deprivation and the gut hunger signal ghrelin. In contrast, stimulation of excitatory axons from the parasubthalamic nucleus to PVT or direct stimulation of PVT glutamate neurons reduced food intake. These data suggest an unexpected robust orexigenic potential for the ZI GABA neurons.
Patients receiving deep brain stimulation of the subthalamus, including the zona incerta (ZI), for the treatment of movement disorders can exhibit characteristics of binge eating (1–3), a common eating disorder characterized by recurrent episodes of consuming large quantities of food, particularly highly palatable food (4, 5). It is not clear why stimulation in the subthalamus would evoke eating, although sheep may release γ-aminobutyric acid (GABA) from the ZI in response to the sight or ingestion of food (6, 7).
The ZI is one of the least-studied regions of the brain, despite its robust projections throughout the brain (8, 9). To determine the role of the ZI in feeding and body weight regulation, we injected Cre recombinase–inducible adeno-associated viruses (AAV) expressing the optogenetic channelrhodopsin-like ChIEF fused with a tdTomato reporter [AAVdj-CAG-DIO-ChIEF-tdTomato (driven by the CAG promoter) (10, 11)] bilaterally into the rostral ZI of vesicular GABA transporter (VGAT)–Cre mice that express Cre recombinase in GABA neurons (Fig. 1A). ChIEF-tdTomato was selectively expressed in ZI GABA neurons but not in lateral hypothalamic neurons (fig. S1). Laser stimulation (1 to 20 Hz) evoked depolarizing currents in ZI ChIEF-tdTomato–expressing VGAT neurons tested with whole-cell recording in brain slices, displaying a high-fidelity correspondence with stimulation frequency (Fig. 1B). In VGAT-Cre mice with ChIEF expression, bilateral laser stimulation (20 Hz) in the ZI increased food intake, with mice rapidly consuming 35.4% of their 24-hour ad libitum high-fat food intake in just 10 min (Fig. 1, C to E, and movie S1). In control mice with tdTomato expression, consumption was only 4% of their 24-hour intake during the same period (Fig. 1E). When stimulation of 10 min ON followed by 30 min OFF was repeated four times, ZI-VGAT-ChIEF mice consumed 74% of their normal 24-hour food intake, whereas control mice consumed only 22% (Fig. 1E). Food deprivation lasting 24 hours increased ZI GABA neuron activity and excitatory neurotransmission to these neurons (Fig. 1, F to J). Ghrelin, a hormone that signals a reduced gut energy state (12), excited ZI GABA neurons and increased excitatory synaptic input onto these neurons (Fig. 1, K to M, and fig. S2).
Anterograde AAV-ChIEF-tdTomato labeling of ZI GABA cells in VGAT-Cre mice showed strong axonal projections into the paraventricular thalamus (PVT) (Fig. 2A), consistent with previous observations that some ZI cells project to the PVT (13, 14), a brain area that may contribute to energy homeostasis (15). Cre recombinase–dependent rabies virus–mediated monosynaptic retrograde pathway tracing in vGluT2–Cre recombinase mice confirmed that PVT glutamate neurons receive strong and direct innervation from ZI neurons (Fig. 2, B and C, and fig. S3). Food deprivation lasting 24 hours increased inhibitory synaptic neurotransmission to PVT glutamate neurons (fig. S4). We asked whether the PVT may be a critical target for ZI regulation of food intake. We crossed VGAT-Cre mice with vGlut2-GFP mice in which neurons expressing vesicular glutamate transporter (vGlut2) were labeled with green fluorescent protein (GFP) to study whether ZI GABA neurons release synaptic GABA to inhibit PVT glutamate neurons (16, 17). One month after AAV-ChIEF-tdTomato was injected into the ZI of these mice (Fig. 2A), photostimulation of ZI VGAT-ChIEF-tdTomato terminals in the PVT evoked GABA-mediated inhibitory currents in PVT vGlut2-GFP neurons (Fig. 2D). In vivo stimulation (20 Hz) of axon terminals from ZI GABA neurons to PVT glutamate neurons (VGATZI-PVT) evoked food-foraging behavior (movie S2). Continuous stimulation (20 Hz) for 10 min increased the intake of high-fat, sweet, and regular foods (Fig.2E) in mice with ZI-VGAT-ChIEF-tdTomato expression. No effect of laser stimulation on high-fat food intake was detected in control mice with AAV-tdTomato in the ZI (Fig. 2F). The total feeding time for ZI VGAT-ChIEF-tdTomato mice was 7.1 ± 0.5 min compared with 0.3 ± 0.1 min for controls (Fig. 2G). Photostimulation of ZI-PVT inhibitory axons evoked gnawing, but not eating, of nonnutritional wood sticks (fig. S5, A and B); photostimulation leading to food intake eliminated subsequent evoked stick gnawing. A priori wood gnawing had no effect on photostimulation-evoked food intake (fig. S5, C and D). We then measured food intake of the same mice during three successive trials of 10-min laser stimulation with a 5-min interval without photostimulation between the trials. The food intake for the first trial was 4.95 ± 0.80 kcal. The amount for the second trial was reduced substantially to 0.72 ± 0.29 kcal and 0.49 ± 0.25 kcal, respectively (Fig. 2H). Satiety feedback signals can thus attenuate ZI-induced feeding.
ZI-stimulated mice showed a preference for high-fat and sweet foods over normal food (Fig. 2I). Although mice prefer sweet and high-fat foods when stimulation is off, laser stimulation increased the relative preference for high-fat food (Fig. 2I). When normal, sweet, and high-fat foods were all available, mice consistently chose high-fat food during laser stimulation of ZI axons in the PVT (movie S3). ZI GABA neurons project to multiple brain regions, including the hypothalamus and midline thalamus (fig. S6). We therefore measured the relative contribution of stimulation of ZI somata with selective stimulation of ZI axons targeting the PVT. Stimulation of ZI VGAT cell bodies or VGATZI-PVT terminals in the PVT evoked similar levels of feeding (Fig. 2J). To further confirm the importance of the VGATZI-PVT projection in mediating ZI GABA neuron control of food intake, the type A GABA (GABAA) receptor antagonist bicuculline (Bic) was microinjected into the PVT 10 min before photostimulation of VGATZI-PVT axon terminals. Bic attenuated photostimulation-evoked feeding (Fig. 2K). That Bic did not completely block photostimulation-evoked food intake could be a diffusion limitation of Bic after application, or ZI VGAT-Cre neurons may coexpress other neurotransmitters responsible for the remaining action. These results are consistent with an early report that lesions in the area of the ZI can alter food intake (18).
Stimulation of anorexigenic proopiomelanocortin (POMC) cells in the hypothalamic arcuate nucleus leads to a reduction in feeding slowly over the succeeding 24 hours, whereas stimulation of orexigenic hypothalamic neurons expressing agouti-related peptide (AgRP) leads to what has previously been considered to be a rapid increase in feeding with mean latency to eat of 6.1 min (range: 1.9 to 13.8 min) (19). To test the time course and efficiency of optogenetic activation of VGATZI-PVT inhibitory inputs to evoke feeding, we used a laser stimulation protocol of 10 s ON (20 Hz) followed by 30 s OFF for more than 20 min to study ZI axon stimulation in PVT brain slices and feeding behavior. Stimulation of ZI axons with this protocol hyperpolarized and inhibited PVT glutamatergic neurons each time the light was activated (Fig. 3A). Mice immediately started feeding for each of the 30 successive trials of ZI axon laser stimulation (Fig. 3B and movie S4). The mean latency to initiate feeding was 2.4 ± 0.6 s when we used laser stimulation of 20 Hz (Fig. 3C). This is almost 100 times faster than that reported for optogenetic stimulation of the AgRP neuron soma and 500 times faster than stimulation of AgRP-PVT axon terminals (19, 20). As soon as the laser was turned off, the mice stopped eating. To test further whether photostimulation of VGATZI-PVT terminals evokes compulsive eating, food intake was measured when food was put in a brightly illuminated chamber in a two-chamber light-or-dark conflict test. Mice spent only 20% of their time in the brightly lit chamber with high-fat food when the laser was off, suggesting an aversion to the light (Fig. 3D). In spite of the light aversion, photostimulation of VGATZI-PVTterminals significantly increased the time mice spent on the illuminated side to 61% when high-fat food was available (Fig. 3D). Photostimulation increased high-fat food intake in bright light (Fig. 3E).
Binge eating has been linked to a reward-system disorder (21, 22). To test the hypothesis that the VGATZI-PVT pathway is involved in a reward state, we explored the motivational valence of VGATZI-PVT in mice by using a two-chamber place preference test. In the absence of available food, optogenetic activation of the VGATZI-PVT pathway evoked a significant preference for the chamber associated with laser stimulation compared with the control chamber (Fig. 3, F and G).
To test whether activation of the VGATZI-PVT inhibitory pathway leads to body weight gain, we selectively photostimulated this pathway for only 5 min every 3 hours over a period of 2 weeks. Photostimulation increased food intake and body weight of mice with ChIEF-tdTomato expression in ZI GABA neurons (Fig. 3, H and I). After the days of photostimulation were completed, mice showed a significantly reduced food intake compared with that of controls (Fig. 3H). The body weight of mice that showed an increase with ZI GABA neuron photostimulation gradually returned to the prestimulation body weight level of controls (Fig. 3I), consistent with the perspective that the mice return to a normal body weight set point (23) in the absence of continuing ZI activation. To test whether ZI GABA neurons exert long-term effects on energy homeostasis, we microinjected AAV-flex-taCasp3-TEVp, which expresses caspase-3 (24), into the ZI of VGAT-Cre mice to selectively ablate ZI GABA neurons (fig. S7). Ablation of ZI GABA neurons decreased long-term food intake and reduced body weight gain by 53% over 8 weeks (Fig. 3, J and K).
To explore the neuronal pathway postsynaptic to the VGATZI-PVT axon terminals, we injected Cre-inducible AAV-ChIEF–tdTomato selectively into the PVT of vGlut2-Cre mice (Fig. 4A and fig. S8A). In brain slices, laser stimulation excited PVT ChIEF-tdTomato–expressing glutamatergic neurons (Fig. 4C). Laser stimulation (20 Hz) above the PVT of ChIEF-tdTomato mice significantly inhibited normal, sweet, and high-fat food intake during 1-hour tests (Fig. 4D and fig. S8B). The mean latency for mice to stop eating was 6.1 ± 2.0 s after the laser (20 Hz) was turned on (Fig. 4E). After mice were partially fasted with only 60% of the normal food available during the preceding night, laser stimulation (20 Hz, 10 min ON followed by 10 min OFF, two times) of ChIEF-expressing PVT vGluT2 neurons reduced food intake (Fig. 4, F to H).
A chemo-genetic designer receptor exclusively activated by designer drugs (DREADD) was used to test the hypothesis that silencing the cells postsynaptic to ZI GABA axons, the PVT glutamate neurons, would enhance food intake. We injected Cre-inducible AAV5-hSyn-HA-hM4D(Gi)-IRES-mCherry coding for the clozapine-N-oxide (CNO) receptor into the PVT of vGlut2-Cre mice (25, 26) (fig. S9, A and B). CNO inhibited PVT neurons with hM3D(Gi)-mCherry receptor expression (fig. S9C). Intraperitoneal CNO produced an increase in food intake during a 3-hour trial (fig. S9D). To test whether PVT neuronal activity affects body weight gain, we microinjected AAV-flex-taCasp3-TEVp into the PVT of vGlut2-Cre mice to induce Cre-dependent caspase expression and selectively ablate PVT glutamatergic neurons. To confirm that PVT vGlut2 neurons were killed by the virus-generated caspase-3, we injected the Cre-dependent reporter construct AAV-tdTomato simultaneously with AAV-flex-taCasp3-TEVp to corroborate that reporter-expressing neurons were absent after selective caspase expression. With coinjection, little tdTomato expression was detected, whereas many cells were detected with injections of AAV-tdTomato by itself, consistent with the elimination of vGluT2 neurons in the PVT (fig. S10, A to D). Ablation of PVT vGluT2 neurons substantially increased both food intake and body weight gain for an extended period (16-week study) (fig. S10, G and H).
In our monosynaptic retrograde tracing with Cre-dependent rabies virus, although less robust than the projection from the ZI, we found a substantial projection to PVT glutamate neurons from the parasubthalamic nucleus (PSTh) (Fig. 4I and fig. S11) (27, 28). That the PSTh may be involved in feeding is suggested by increased c-fos expression in the PSTh during anorexia induced by amino acid deficiency (29). To test whether the PSTh maintains an excitatory input to PVT glutamatergic neurons that could serve to antagonize binge-like eating evoked by VGATZI-PVT inhibitory pathway activation, we injected Cre-inducible AAV-ChIEF–tdTomato bilaterally into the PSTh of vGlut2-Cre mice (Fig. 4J). Restricted expression of ChIEF–tdTomato was observed in the PSTh cell bodies (Fig. 4J) and in large numbers of PSTh axon terminals in the PVT (Fig. 4J). Brain slice electrophysiology confirmed that optogenetic activation of PSTh glutamatergic neuron terminals in the PVT evoked strong glutamate-mediated postsynaptic excitatory currents in PVT vGlut2-GFP neurons, suggesting a functional role for PSTh glutamate neurons in the synaptic excitation of PVT glutamate neurons (Fig. 4K). Stimulation of PSTh glutamatergic neuron terminals in the PVT inhibited food intake (Fig. 4L). Furthermore, optogenetic activation of the vGlut2PSTh-PVT excitatory pathway in a two-chamber place-preference test generated a significant aversion associated with the laser stimulation–paired chamber (fig. S12).
Together, our data demonstrate a powerful inhibitory projection from the ZI to the PVT that can reliably generate rapid and substantial eating. That the ZI GABA cells may participate in energy homeostasis is suggested by electrophysiological data showing increased activity of these cells after food deprivation and in the presence of the empty gut–signaling peptide ghrelin. Based on retrograde rabies virus and anterograde AAV tracing, ZI axonal projections to the excitatory neurons of the PVT appear more robust than those from other known regions of the brain involved in food intake, suggesting the ZI is not a minor component; furthermore, optogenetic stimulation of the ZI generated a more robust feeding response than stimulation of the much-studied lateral hypothalamus, further suggesting that the ZI can play a substantive role in enhancing food consumption. Our study provides a potential explanation for why clinical deep brain stimulation in the ventral thalamus near the ZI can increase binge eating.
Materials and Methods
Figs. S1 to S12
Movies S1 to S4
References and Notes
1. L. B. Zahodne et al., J. Neuropsychiatry Clin. Neurosci. 23, 56–62 (2011).
2. P. Amami et al., J. Neurol. Neurosurg. Psychiatry 86, 562–564 (2015).
3. L. Novakova et al., Neuroendocrinol. Lett. 32, 437–441 (2011).
4. J. I. Hudson, E. Hiripi, H. G. Pope Jr., R. C. Kessler, Biol. Psychiatry 61, 348–358 (2007).
5. S. A. Swanson, S. J. Crow, D. Le Grange, J. Swendsen, K. R. Merikangas, Arch. Gen. Psychiatry 68, 714–723 (2011).
6. D. L. Benson, P. J. Isackson, C. M. Gall, E. G. Jones, Neuroscience 46, 825–849 (1992).
7. K. M. Kendrick, M. R. Hinton, B. A. Baldwin, Brain Res. 550, 165–168 (1991).
8. N. Urbain, M. Deschênes, Neuron 56, 714–725 (2007).
9. J. Chen, A. R. Kriegstein, Science 350, 554–558 (2015).
10. J. Y. Lin, M. Z. Lin, P. Steinbach, R. Y. Tsien, Biophys. J. 96, 1803–1814 (2009).
11. X. Zhang, A. N. van den Pol, Nat. Neurosci. 19, 1341–1347 (2016).
12. M. Tschöp, D. L. Smiley, M. L. Heiman, Nature 407, 908–913 (2000).
13. J. S. Lee, E. Y. Lee, H. S. Lee, Brain Res. 1598, 97–113 (2015).
14. G. J. Kirouac, M. P. Parsons, S. Li, J. Comp. Neurol. 497, 155–165 (2006).
15. T. R. Stratford, D. Wirtshafter, Brain Res. 1490, 128–133 (2013).
16. E. E. Hur, L. Zaborszky, J. Comp. Neurol. 483, 351–373 (2005).
17. H. Huang, P. Ghosh, A. N. van den Pol, J. Neurophysiol. 95, 1656–1668 (2006).
18. B. A. Stamoutsos, R. G. Carpenter, L. Grossman, S. P. Grossman, Physiol. Behav. 23, 771–776 (1979).
19. Y. Aponte, D. Atasoy, S. M. Sternson, Nat. Neurosci. 14, 351–355 (2011).
20. J. N. Betley, Z. F. Cao, K. D. Ritola, S. M. Sternson, Cell 155, 1337–1350 (2013).
21. E. M. Schulte, C. M. Grilo, A. N. Gearhardt, Clin. Psychol. Rev. 44, 125–139 (2016).
22. R. M. Kessler, P. H. Hutson, B. K. Herman, M. N. Potenza, Neurosci. Biobehav. Rev. 63, 223–238 (2016).
23. R. E. Keesey, T. L. Powley, Annu. Rev. Psychol. 37, 109–133 (1986).
24. C. F. Yang et al., Cell 153, 896–909 (2013).
25. B. N. Armbruster, X. Li, M. H. Pausch, S. Herlitze, B. L. Roth, Proc. Natl. Acad. Sci. U.S.A. 104, 5163–5168 (2007).
26. G. M. Alexander et al., Neuron 63, 27–39 (2009).
27. M. Goto, L. W. Swanson, J. Comp. Neurol. 469, 581–607 (2004).
28. S. Li, G. J. Kirouac, Brain Struct. Funct. 217, 257–273 (2012).
29. X. Zhu et al., Am. J. Physiol. Endocrinol. Metab. 303, E1446–E1458 (2012).
30. Acknowledgments: We thank J. N. Davis and J. Paglino for manuscript suggestions and Y. Yang for technical assistance. Support provided by NIH DK084052 and NS48476.