Negative feedback control of hunger circuits by the taste of food

Abstract

The rewarding taste of food is critical for motivating animals to eat, but whether taste has a parallel function in promoting meal termination is not well understood. Here we show that hunger-promoting AgRP neurons are rapidly inhibited during each bout of ingestion by a signal linked to the taste of food. Blocking these transient dips in activity via closed-loop optogenetic stimulation increases food intake by selectively delaying the onset of satiety. We show that upstream leptin receptor-expressing neurons in the dorsomedial hypothalamus (DMH^LepR) are tuned to respond to sweet or fatty tastes and exhibit time-locked activation during feeding that is the mirror image of downstream AgRP cells. These findings reveal an unexpected role for taste in the negative feedback control of ingestion. They also reveal a mechanism by which AgRP neurons, which are the primary cells that drive hunger, are able to influence the moment-by-moment dynamics of food consumption.

Keywords: AgRP; DMH; arcuate nucleus; feeding; hunger; hypothalamus; leptin receptor; taste.

Figure 1 |. AgRP neurons track the taste of food.

A, Schematic of mice equipped for fiber photometry of AgRP neurons while licking a bottle attached to a lickometer. B, Example trace of AgRP activity aligned to licks. C, Z-scored AgRP activity in fasted or fed mice drinking ensure. D, PSTH of AgRP activity and minimum z-score peri the first lick. E, Same as d but peri the last lick. F, PSTH of AgRP activity the first or last 10 minutes of drinking ensure. G, Example whole-trial and zoomed-in traces of AgRP activity during licking. H-J, PSTH of AgRP activity peri the first lick. K, Minimum z-score during licking bouts of the solutions in h-j. L, Correlation coefficient between AgRP activity and true or shuffled lick data. M,N, Z-scored AgRP activity during consumption of sweet (m), or fatty (n) solutions. O, Mean z-score over ten minutes of consumption. P, Proposed model of AgRP activity integrating different levels of signals to achieve satiation. MDG = alpha-methyl-d-glucopyranoside. N.S. p>0.05, *p<0.05, **p<0.01 for direct comparisons; compared to water (k). #p<0.05, ##p<0.01, ###p<0.001 relative to 0 (k). #p<0.05 relative to empty (o). Dots represent individual mice. Data is presented as mean ± SEM. See also Figure S1.

Figure 2 |. Ingestion-triggered dips in AgRP activity control meal duration.

A, Schematic of mice equipped with an optic ferrule above the arcuate nucleus (ARC), where AgRP neurons endogenously express ChR2. Table and schematic of stimulation protocols used. B, Schematic of experimental design involved randomly interleaving no laser trials with closed-loop trials. Average licks per 2-min trial for ChR2 and control mice. C, CDF of no laser trials preceded by either a closed-loop or no laser trial. D, Total licks for three different protocols. E, Total bouts for the interleaved closed-loop protocol shown in B (left) and tonic stimulation. F, Same as E but for bout length. G, Example feeding behavior during a closed-loop session, collapsed by trial type. H, Total bouts across all mice, separated by trial type as a distribution (left). Total bouts for ChR2 mice split by trial type and time in session (right). I, Schematic of conditioned flavor preference protocol. J, Training data for paired and unpaired flavors. K, Preference for flavor (licks for paired flavor divided by total licks) before and after training. N.S. p>0.05, *p<0.05, **p<0.01. Dots represent individual mice. Data is presented as mean ± SEM.

Figure 3 |. DMH^LepR neurons are activated time-locked to ingestion.

A, Inhibitory circuit schematic from DMH to ARC. B, Schematic and example of lens placement above GCaMP-expressing DMH^LepR neurons. Example field of view color-coded to responses during consumption. C, Schematic of single-cell calcium imaging during consumption. D, Heatmap of DMH^LepR responses (N=4–5) during consumption while fasted. E, Averaged traces of categories in D. F, Mean z-score of individual neurons over first ten minutes. G, Percentage of each category per mouse. H,I, Example T1 averaged trace during licking (grey) ensure (H) or water (I). J,K, PSTH of T1 activity peri the first or last lick of ensure (J) or water (K). L, Peak z-score during licking of individual neurons. M, Correlation coefficient for neural activity against licks. N, Same as D but for chow. O, Same as E but for chow. P,Q, Same as L,M but for chow and object. R, Generation of pseudo-photometry trace during chow consumption. DMH = dorsomedial hypothalamus. ARC = arcuate nucleus. T1 = Type 1, T2 = Type 2, T3 = Type 3, NR = no response. N.S. p>0.05, *p<0.05, **p<0.01, ****p<0.0001. Dots represent individual mice unless otherwise noted. Data is presented as mean ± SEM. See also Figures S2 and S3.

Figure 4 |. DMH^LepR neurons are activated by the taste of food.

A, Diagram of Davis Rig used for brief access taste tests during single-cell imaging of DMH^LepR neurons. B, Example traces of neurons preferring sucrose, sucralose, or both, and corresponding quantification. C, Correlation plot of activated neurons comparing the mean z-score for sucrose against either sucralose or water. Each point represents a single neuron. P-value indicates significance relative to a slope of 0. D, Example responses to sucrose concentrations. E, Mean traces of neurons activated by second-to-maximal concentration across all concentrations of sucrose. F, Peak z-score across sucrose concentrations. G, Licks per presentation. H, non-linear regression between peak z-score and licks. P-value indicates significance of fit relative to a linear regression. I-L, Same as D-H but for sucralose. N, Schematic of setup for taste panel experiment. O, K-means clustering of ensure-activated neurons across all solutions, presented as one heatmap per cluster. P, Average traces for each cluster in O, aligned to solution access. Q, Noise-to-signal ratio R, entropy, and S, noise-to-signal ratio versus entropy plot of ensure-activated neurons, colored by their preferred non-caloric taste. Data is reported as the mean, with error bars as ±SEM.

Figure 5 |. Nutrients potentiate DMH^LepR neuron responses to gustatory signals.

A, Schematic and example of intragastric (IG) infusion setup during single-cell calcium imaging. B, Heatmap of DMH^LepR neurons (N=5 mice) receiving an IG infusion of ensure. C, Same as in B but with an osmolarity-matched control. D, Quantification of neural response types. E, Top, averaged trace of activated neurons. Bottom, mean z-score. F, Example overlay generated using CellReg to cross-register neurons. G, Heatmaps of aligned neurons, divided based on response patterns: activated by both, activated only by IG ensure or licking ensure, or other. H, Averaged traces from the categories in G, and corresponding percentage out of the number activated by at least one stimulus. I, Heatmaps during sucrose or sucralose consumption (N=7). J, Percentage of type 1 or 2 neurons per mouse. K, Averaged trace of type 1 neurons during consumption. L, Total licks per time bin. M, PSTH around licking bouts using a local baseline (N=4). N,O, Quantification of M without (N) and with (O) normalization to licks. P, Summary model of nutrients potentiating gustatory responses over time. Dots represent individual type 1 neurons. N.S. p>0.05, *p<0.05, **p<0.01, ****p<0.0001 Data is presented as mean ± SEM. See also Figure S4.