Abstract
Despite the important and exciting work done over the past few years to understand the mammalian circuitry regulating eating, there remain unexplained elements. Neurons in the hypothalamic arcuate nucleus (ARC) expressing Agouti-related peptide (AgRP - they also express the more abundant Neuropeptide Y [NPY] and GABA) are thought to signal hunger, are activated by fasting and reliably cause voracious eating when stimulated {1,2}. They also are considered to send an aversive signal that motivates animals to eat {3,4}, despite the fact that they suppress an aversive signal by releasing GABA in the parabrachial nucleus {5}. However, the immediate reduction seen in AgRP neuron firing when the animal is presented with a sensory food cues occurs despite the animals continuing to eat {3,4}. The mechanism by which this occurs is explored in this elegant study. ARC AgRP neurons receive innervation from a population of GABAergic neurons in the dorsomedial hypothalamus (DMH) that express leptin receptors (LEPR) and prodynorphin (PDYN) and that respond to food-related sensory input {5}. The authors examined firing activity in these DMHLEPR neurons as mice learned a 2-choice task involving a light cue and a liquid treat - the neurons responded to the light cue. This response became more rapid and accurate with practice, consistent with learning, and did not happen in naïve mice. These cells also responded to caloric (but not noncaloric) infusions to the gut, but more slowly and weakly; chemogenetic inhibition of these neurons did not affect satiation or satiety. The authors then used modified rabies viruses and channelrhodopsin to identify upstream activators of the DMHLEPR neurons. This uncovered a single site containing glutamatergic neurons, in the lateral hypothalamus (LH) consistent with the recognized role of the LH in sensory cue detection. Optogenetic excitation of the glutamatergic terminals in DMH of these LH neurons during the dark cycle (when AgRP cells are active and drive feeding) resulted in decreased food intake, consistent with inhibition of the AgRP neurons. Stimulation of these LH-DMH terminals also caused food-restricted animals to prefer the stimulation side in a place preference test, consistent with a positive, not aversive, effect. Using an elegant approach in which the DMHLEPR neurons expressed GcaMP6 and LH glutamatergic cells projecting to DMH expressed an inhibitory DREADD, the activity of LH cells was shown to be important for the DMHLEPR response to food cues, demonstrating that the LH projections mediate the DMHLEPR neurons output to the AgRP cells. Conversely, inhibition of the LH neurons resulted in strong reductions in food cue-induced inhibition of the AgRP cells. The behavioral evidence in the 2-choice task indicated that the mice learned from the repeated presentation of a stimulus and a food cue. Acutely inhibiting DMHLEPR neurons (chemogenetically) in highly-trained mice did not affect their performance in the 2-choice task. Blocking the DMH-ARC circuit by causing the DMHLEPR neurons to express tetanus toxin (thereby crippling activity-dependent GABA release from their terminals) greatly increased the amount of time it took naïve mice to learn the task. This effect was restricted to the PDYN-expressing DMH neurons, as expressing tetanus toxin on this population had the same effect as expressing it in the DMHLEPR neurons. Using tetanus toxin expression in the LH neurons projecting to DMH similarly reduced the rate of learning in this task. These manipulations had no significant effect on body weight when restricted to the DMHPDYN neurons or their upstream excitatory inputs. Finally, they examined whether this response generalized to other ingestive targets. In water-restricted animals, inhibition of GABA release by DMH-AgRP neurons did not affect the rate of acquisition of the task by naïve mice. Overall, this technically stellar paper defines an important role for the LH-DMHPDYN+LEPR neurons in learning cues that initiate food intake, and indicates the importance of suppressing the aversive AgRP neuron input in helping the animal learn. However, I still remain unclear as to how the disinhibition of the PBN neurons by suppression of the AgRP/NPY/GABA cells results in net food intake, while the strong activation of these same cells result in robust food intake {1,4,6-8}. There is enough evidence that some of the projection pathways from AgRP cells are direct and unbranched, and that the population is not uniform from that perspective. It would be interesting to learn if the suppression of food intake and the stimulation of food intake involves different neurons, or different firing rates of the same neurons {7,8}. Clearly there is more yet to learn!