Control of feeding by a bottom-up midbrain-subthalamic pathway

Abstract

Investigative exploration and foraging leading to food consumption have vital importance, but are not well-understood. Since GABAergic inputs to the lateral and ventrolateral periaqueductal gray (l/vlPAG) control such behaviors, we dissected the role of vgat-expressing GABAergic l/vlPAG cells in exploration, foraging and hunting. Here, we show that in mice vgat l/vlPAG cells encode approach to food and consumption of both live prey and non-prey foods. The activity of these cells is necessary and sufficient for inducing food-seeking leading to subsequent consumption. Activation of vgat l/vlPAG cells produces exploratory foraging and compulsive eating without altering defensive behaviors. Moreover, l/vlPAG vgat cells are bidirectionally interconnected to several feeding, exploration and investigation nodes, including the zona incerta. Remarkably, the vgat l/vlPAG projection to the zona incerta bidirectionally controls approach towards food leading to consumption. These data indicate the PAG is not only a final downstream target of top-down exploration and foraging-related inputs, but that it also influences these behaviors through a bottom-up pathway.

Fig. 1. Vgat l/vlPAG cells encode approach to food and eating.

a Surgery for recording l/vlPAG vgat calcium transients (top). Example l/vlPAG vgat traces (bottom). b Order of assays. c L/vlPAG vgat activity centered at eating onset (cricket assay n = 223 neurons; walnut assay n = 204 neurons). d Representation of behaviors in principal component space (PC) during cricket hunting (see example on left). Clustering quality was measured by silhouette score, which was higher than the chance level of zero (dotted red line) (n = 4 mice; one-sample two-tailed t-test, t-statistic = 2.79, p = 0.059). e Same as d, but for walnut (n = 4 mice; one-one-sample two-tailed t-test, t-statistic = 14.37, p = 0.001). f Behaviors were decoded above chance (red dotted line) (n = 4 mice; one-sample two-tailed t-test, cricket t-statistics: approach = 3.87, p = 0.031; eat = 5.59, p = 0.011; walnut t-statistics: approach = 4.75, p = 0.018, eat = 7.53, p = 0.005). g Cells co-registered across assays. h Mahalanobis distance between points from two clusters that display higher overlap (left panel with light and dark blue points) and two clusters that are well-separated (right panel with light and dark green points). i Mahalanobis distance between approach and eating clusters across the cricket and walnut assays in n-dimensions (n = # of co-registered neurons). The distance between eating clusters across assays is smaller than the distance between approach clusters, indicating the representation of eating is more conserved than food approach (approach sample n = 527 time points, eat sample n = 1958 time points; two-tailed Wilcoxon rank-sum test, z-score = 22.88, p < 0.001). j Same as d, but for co-registered cells, showing conserved representation of behaviors across assays (n = 4 mice; one-one-sample two-tailed t-test, t-statistic = 4.22, p = 0.024). k The distance between individual points and cluster center is smaller in the eating cluster than the approach cluster, indicating the representation of eating is more conserved across assays (approach sample n = 1049 time points, eat sample n = 5492 time points; two-tailed Wilcoxon rank-sum two-tailed test, z-score = 10.35, p < 0.001). ***p < 0.001, **p < 0.01, *p < 0.05, † p = 0.059. Data are presented as mean values +/- SEM. Source data are provided as a Source Data File.

Fig. 2. L/vlPAG vgat cells are more active immediately prior to eating.

a Order of assays. b L/vlPAG vgat cells are more active prior to eating (blue rectangle). c Left: Representative cell showing activation prior to eating cricket. Right: pre-post eating activity of all cells (n = 223 cells; two-tailed Wilcoxon signed rank test, z-value = 4.36, p < 0.001). d Same as c, but for walnut (n = 204 cells; two-tailed Wilcoxon sign rank test, z-value = 3.78, p < 0.001). e Left: More cells were negatively modulated by eating cricket (n = 4 mice; two-one-sample two-tailed t-test, t-statistic=3.15, p = 0.026). Right: Mean traces (± 1 s.e.m.) of all cells modulated by eating. (eat- n = 86 cells, eat+ n = 37 cells). f Same as e, but for walnut (left: n = 4; two-one-sample two-tailed t-test, t-statistic=5.88, p = 0.009; right: eat - n = 110 cells, eat + n = 36 cells). g Cells co-registered in cricket 1 and cricket 2. h General linear model (GLM) weights for representative cell in crickets 1 and 2. i Correlation of eating GLM weight across assays (coregistered cricket n = 78 cells, coregistered walnut n = 74 cells; Spearman correlation, left: p < 0.001, right: p = 0.005). j Example cell co-registered in cricket 1 and 2 shows decreased activity after eating onset in both assays. k (left pair of bars) Eat- walnut cells show decreased activity during eating of both walnut (blue) and cricket 1 (grey) (n = 53 cells; two-tailed Wilcoxon signed rank test, walnut z-value = 2.18, p = 0.029; coregistered cricket z-value = 2.72, p = 0.007) (right pair of bars). Same as the two bars on the left, but for cells co-registered in cricket 1 and 2 (n = 35 cells; two-tailed Wilcoxon signed rank test, cricket 1 z-value = 4.49, p < 0.001; coregistered cricket 2 z-value = 3.44, p = 0.001). l Left: Activity of eat- cells in cricket 1 (light blue). Activity for the same cells co-registered in cricket 2 (grey). Right: Correlation of activity centered at eating onset for eat- (light blue) and eat+ (dark blue) cells co-registered in cricket 1 and 2 (eat- n = 35 cells, eat+ n = 17 cells; Spearman correlation). m Same as l, but for cells co-registered in walnut and cricket 1 (eat - n = 53 cells, eat + n = 15 cells; Spearman correlation; for r-value comparison, Fisher r-to-z transformation, p = 0.004). ***p < 0.001, **p < 0.01, *p < 0.05. All error bars and error bands correspond to +/- SEM. Source data are provided as a Source Data File.

Fig. 3. L/vlPAG vgat cell activity bidirectionally modulates eating of cricket and walnut.

a Left: Scheme showing experimental approach for expressing inhibitory (Arch) and excitatory (ChR2) opsins in l/vlPAG vgat cells. Right: Representative Image showing fiber placement and viral expression in l/vlPAG vgat cells. Similar images were obtained in all 13 animals used. b Experimental timeline. c Optogenetic inhibition of l/vlPAG vgat cells decreases amount of walnut eaten (YFP n = 6, Arch n = 7; right: Wilcoxon rank-sum, z-value = 2.50, p = 0.006; Wilcoxon signed rank (Arch), z-value = −2.20, p = 0.028). d Inhibition of these cells also increases latency to first cricket attack, latency to successful predation and the probability of successful predation (YFP n = 7, Arch n = 8, left: Wilcoxon rank-sum, z-value = −2.03, p = 0.042; Wilcoxon signed rank (Arch), z-value = 2.38, p = 0.013; middle: Wilcoxon rank-sum, z-value = −2.72, p = 0.007; Wilcoxon signed rank (Arch), z-value = 2.52, p = 0.006; right: Wilcoxon rank-sum, z-value = 2.10, p = 0.036; Wilcoxon signed rank (Arch), z-value = −2.25, p = 0.024). e Optogenetic excitation of l/vlPAG vgat cells decreased latency to eat walnut and increased amount walnut eaten (YFP n = 5, ChR2 n = 15, left: Wilcoxon rank-sum, z-value = 1.96, p = 0.049; Wilcoxon signed rank (ChR2), z-value = −3.41, p < 0.001; right: Wilcoxon rank-sum, z-value = −3.14, p = 0.002; Wilcoxon signed rank (ChR2), z-value = 3.41, p < 0.001). f Excitation of ChR2-expressing l/vlPAG vgat cells decreased latency for the first cricket attack and the latency to successful predation. This manipulation also increased the probability of successful predation. (YFP n = 6, ChR2 n = 9, left: Wilcoxon rank-sum, z-value = 3.13, p < 0.001; Wilcoxon signed rank (ChR2), z-value = −2.67, p = 0.008; middle: Wilcoxon rank-sum, z-value = 3.15, p < 0.001; Wilcoxon signed rank (ChR2), z-value = −2.67, p = 0.008; right: Wilcoxon rank-sum, z-value = −3.21, p < 0.001; Wilcoxon signed rank (ChR2), z-value = 2.69, p = 0.007). ***p < 0.001, **p < 0.01, *p < 0.05. Data are presented as mean +/- SEM. Source data are provided as a Source Data File.

Fig. 4. Optogenetic excitation of l/vlPAG vgat cells produces exploratory foraging, place preference and compulsive eating.

a Experimental timeline (bottom). b L/vlPAG vgat stimulation increases rearing in an empty box (YFP n = 7, ChR2 n = 7; Wilcoxon rank-sum, z-value = 3.07, p < 0.001) and also c climbing in an environment with a ladder (YFP n = 7, ChR2 n = 7; Wilcoxon rank-sum, z-value = −3.28, p < 0.001). d L/vlPAG vgat activation increased headdips into holes in the holeboard test, and also increased distance traveled. Lower rows: Representative tracks (YFP n = 6, ChR2 n = 6; left: Wilcoxon rank-sum, z-value = −2.81, p = 0.005; right: Wilcoxon rank-sum, z-value = −2.428, p = 0.015). e Left: Behavioral heat maps showing l/vlPAG vgat stimulation induced real-time place preference (YFP n = 6, ChR2 n = 10; Wilcoxon rank-sum, z-value = −3.48, p < 0.001; Wilcoxon signed rank (ChR2), z-value = 3.09, p = 0.002). f A ball was moved on a trajectory spelling “BG” (upper left). L/vlPAG vgat activation decreased distance to ball, increased correlation of the position of mouse and ball and decreased the head angle offset between mouse and ball (YFP n = 7, ChR2 n = 5; Wilcoxon rank-sum; left: z-value = 2.76, p = 0.006; middle: z-value = −2.43, p = 0.015; right: z-value = 2.76, p = 0.006). g L/vlPAG stimulation decreased latency to climb a ladder and then bite the walnut, and increased time spent in the walnut zone (YFP n = 6, ChR2 n = 5; bottom left: Wilcoxon rank-sum, z-value = 2.71, p = 0.007; Wilcoxon signed rank (ChR2), z-value = −2.02, p = 0.043; bottom right: Wilcoxon rank-sum, z-value = −2.65, p = 0.008; Wilcoxon signed rank (ChR2), z-value = 2.02, p = 0.043). h An object and a walnut were suspended in opposite corners (upper left). L/vlPAG vgat activation increased rearing and time spent in the walnut corner, but not the object corner (YFP n = 6, ChR2 n = 5; bottom left: Wilcoxon rank-sum (walnut), z-value = −2.65, p = 0.008; bottom right: Wilcoxon rank-sum (walnut), z-value = −2.65, p = 0.008). i To measure compulsive eating, walnuts were placed on a shocking grid to evaluate if mice endured foot-shocks to access walnuts. L/vlPAG vgat stimulation (shock+laser epoch) increased consumption. (YFP n = 5, ChR2 n = 7; Wilcoxon signed rank (baseline vs. shock) YFP: z-value = 1.75, p = 0.08; ChR2: z-value = 2.37, p = 0.018; Wilcoxon rank-sum (shock+laser), z-value = 2.76, p = 0.01) ***p < 0.001, **p < 0.01, *p < 0.05, † p = 0.08. Data are presented as mean +/- SEM. Source data is in the Source Data File.

Fig. 5. L/vlPAG vgat cells receive inputs from regions that control feeding.

a In order to identify inputs to PAG vgat cells, vgat-Cre mice were injected with AAV5-hsyn-FLEX-TVA-P2A-eGFP-2A-oG in the PAG. Three weeks later, mice were injected with EnvA-dG-rabies-mCherry in the same location. b Timeline of viral injections. c Expression of viral vectors in the vlPAG and the lPAG in vgat-Cre mice. Double-labelled GFP (green) and mCherry (red)-expressing vgat PAG cells (shown in yellow in ‘Merge’ panel) show starter cells for monosynaptic retrograde tracing. d mCherry-expressing (red) cells that project to PAG vgat cells in the bed nucleus of the stria terminalis (BST), medial preoptic area (MPA), central amygdala (CeA), Zona incerta anterior and posterior regions (ZIa and ZIp, respectively), lateral hypothalamic area (LHa) and cuneiform nucleus (Cun). e Quantification of retrogradely-labelled cells across brain regions. Structures are ordered from anterior to posterior locations. Note that the Zona incerta (ZI) provides the most prominent input to PAG vgat cells (n = 6807 starter cells and 16,704 retrogradely labeled cells). Anterior-posterior coordinates in mm from Bregma (B) is displayed in the lower left corner of each image. acp anterior commissure, posterior part, aca anterior commissure, anterior part, AH anterior hypothalamus, BST bed nucleus of the stria terminalis, BLA basolateral amygdala, CeA central amygdala, Cun cuneiform nucleus, cp cerebral peduncle, DMH dorsomedial hypothalamus, DR dorsal raphe, HDB nucleus of the horizontal limb of the diagonal band, IC inferior colliculus, LHa lateral hypothalamus area, MPA medial preoptic area, dPAG lPAG and vlPAG dorsal, lateral and ventrolateral periaqueductal gray, respectively, PH posterior hypothalamus, PMd dorsal premammillary nucleus, SNl, SNc, SNr lateral, compact and reticular parts of the substantia nigra, respectively, VMH ventromedial hypothalamus, ZIa and ZIp zona incerta anterior and posterior regions, respectively. Data are presented as mean values +/- SEM. Source data are provided as a Source Data File.

Fig. 6. Bottom-up l/vlPAG vgat cells’ projections to the zona incerta and other regions that control feeding.

a In order to characterize the anterograde connectivity of l/vlPAG vgat cells, vgat-Cre mice were injected with AAV5-EF1a-DIO-YFP in the PAG. b Expression of YFP in the l/vlPAG of vgat-Cre mice. c YFP axon terminals that project to the bed nucleus of the stria terminalis (BST), d hypothalamus and e zona incerta (ZI). b–e Similar images were obtained in all 5 animals used. f Scheme showing injections used to characterize monosynaptic GABAergic input from PAG vgat cells to the zona incerta. g (Left panel) Representative traces of ex vivo slice recordings of zona incerta neurons following optogenetic excitation of ChR2-expressing axon terminals of vgat l/vlPAG cells. Traces show recordings in the absence of drugs (control), 0.5 µM tetrodotoxin (TTX), 0.5 µM TTX + 100 µM 4-Aminopyridine (4AP) and 25 µM bicuculline (BIC) (from the top to the bottom). g (Right panel) Quantitative analysis of the amplitude of light-evoked inhibitory postsynaptic currents from the zona incerta for the control (n = 12 cells from 5 mice), 0.5 µM TTX (n = 9 cells from 5 mice), 0.5 µM TTX + 100 µM 4-Aminopyridine (n = 9 cells from 5 mice) and 25 µM bicuculline conditions (n = 6 cells from 3 mice). One-way ANOVA test following by a means comparison Tukey test, NS Not Significant, aca anterior commissure, AH anterior hypothalamus, LHa lateral hypothalamus area, VMH ventromedial hypothalamus. Data are presented as mean values +/- SEM. Source data are provided as a Source Data File.

Fig. 7. Vgat l/vlPAG projection to zona incerta is activated during approach to food.

a (left) Scheme for fiber photometry recording. (right) vgat-Cre mice were injected with a vector encoding cre-dependent GCaMP6s in the l/vlPAG. A fiberoptic cannula was implanted above the zona incerta (ZI) to record calcium transients in l/vlPAG vgat axon terminals in the zona incerta. b Image showing GCaMP6s-expressing l/vlPAG vgat axon terminals in the ZI. Similar images were obtained in all 8 animals used. c Experimental timeline. d Example recording showing that activity in the l/vlPAG vgat projection to zona incerta is higher during approach to walnut (green), and lower during walnut eating (blue) (top). Bottom: same as top, but for recordings from the same mouse in the cricket assay. e Left: Average activity in the l/vlPAG vgat projection to zona incerta during walnut approach and eating. Right: same as left, but for the cricket assay. The l/vlPAG vgat projection to zona incerta is activated prior to approach (f) and is inhibited following eating onset (g) in both assays (walnut n = 5, cricket n = 8; one-sample t-test, approach t-statistics: walnut=3.03, p = 0.039, cricket = 4.37, p = 0.003; eat t-statistics: walnut = −4.73, p = 0.009, cricket = −6.85, p < 0.001). h Mice were exposed to the environment depicted in the scheme. Mice were able to see and smell the walnut through the wire mesh but were not able to eat or touch it. i Heat map showing approach activity in the l/vlPAG vgat projection to the zona incerta for the assay shown in h. Note activity is highest in the entrance to the walnut corridor. j Average activity during approach onset to object or walnut. k Significantly elevated activity was observed during approach to walnut, but not object (n = 5; one-sample t-test, t-statistic (walnut) = 3.45, p = 0.026). ***p < 0.001, **p < 0.01, *p < 0.05. Data are presented as mean values +/- SEM. All error bars and error bands correspond to +/- SEM. Source data are provided as a Source Data File.

Fig. 8. Activity in the l/vlPAG vgat projection to zona incerta is necessary and sufficient for consumption of walnut and crickets.

a Left: scheme showing viral strategy to express the opsins Arch (inhibitory) or ChR2 (excitatory) in l/vlPAG vgat cells. A fiberoptic cannula was placed over the zona incerta (ZI) to inhibit or excite the l/vlPAG vgat projection to the zona incerta. Right: Image showing ChR2-YFP expressing l/vlPAG axon terminals in the ZI. Opt: optic tract. Similar images were obtained in all 21 animals used. b Experimental timeline. c Inhibition of the l/vlPAG vgat projection to the zona incerta decreased eating of walnut (YFP n = 10, Arch n = 8; left: Wilcoxon rank-sum, z-value = −1.60, p = 0.11; right: Wilcoxon rank-sum, z-value = 2.45, p = 0.014; Wilcoxon signed rank (Arch), z-value = −2.52, p = 0.006). d This manipulation also increased latency to first cricket attack and decreased the probability of killing the cricket (YFP n = 13, Arch n = 8; left: Wilcoxon rank-sum, z-value = −2.03, p = 0.042, Wilcoxon signed rank (Arch), z-value = 2.12, p = 0.034; right: Wilcoxon rank-sum, z-value = 2.06, p = 0.039). e Excitation of the l/vlPAG vgat projection to the zona incerta increased consumption of walnut (YFP n = 6, ChR2 n = 9; left: Wilcoxon signed rank (ChR2), z-value = −2.52, p = 0.006; right: Wilcoxon rank-sum, z-value = −2.53, p = 0.006; Wilcoxon signed rank (ChR2), z-value = 2.52, p = 0.006). f This manipulation also decreased latency to successful cricket predation and increased the probability of killing the cricket (YFP n = 6, ChR2 n = 9; middle: Wilcoxon rank-sum, z-value = 3.18, p < 0.001, Wilcoxon signed rank (ChR2), z-value = −2.67, p = 0.008; right: Wilcoxon rank-sum, z-value = −2.47, p = 0.014; Wilcoxon signed rank (ChR2), z-value = 2.46, p = 0.014). ***p < 0.001, **p < 0.01, *p < 0.05, † p = 0.11. Data are presented as mean values +/- SEM. Source data are provided as a Source Data File.