Aversive learning hijacks a brain sugar sensor to consolidate memory | Nature

Download PDF

Subjects

- Feeding behaviour

- Long-term memory

- Neuroendocrinology

Abstract

The formation of food-related memories involves post-ingestion nutrient sensing signals1,2,3,4,5. Whether nutrient sensors act beyond feeding-relevant behaviour is less well understood. Here we show that an internal sugar sensor in the Drosophila brain6 is involved in memory consolidation, both in fasted flies subjected to an appetitive learning task involving a sucrose reward and in flies fed ad libitum subjected to an aversive learning task independent of food cues7,8. In the latter, spaced repetition of learning sessions, a prerequisite to induce long-term memory, lures brain fructose-sensing neurons into a fasted state through a disinhibition mechanism that transiently restores their sensing ability despite satiation9. Post-learning sugar ingestion activates disinhibited fructose-sensing neurons, which triggers memory consolidation through the release of the glycoprotein hormone thyrostimulin10,11, as in appetitive learning. The reset of fructose-sensing neurons by spaced training also results in a fasted state-like feeding behaviour, manifesting in a strong increase in sucrose preference and intake. By revealing a mechanism of non-homeostatic hunger and its critical relevance for memory consolidation, our results provide a neural circuit basis, and a cognitive value, to a behaviour akin to emotional eating.

Main

Sensing the content of ingested food is critical for the body to evaluate energy availability and accordingly set an adequate metabolic state. In addition to peripheral taste receptors, animals and humans detect the nutritional value of food through post-ingestion mechanisms involving internal nutrient sensors in the digestive tract and the brain. Internal nutrient-sensing systems are major regulators of appetite and feeding behaviour12. As such, they participate in memory processes that enable the assignment of value to food-related sensory cues13. However, the extent to which the brain’s cognitive efficiency relies on its nutrient sensors, in particular beyond food-related processes, is unclear.

Laboratory experiments using genetically tractable species allow the study of targeted neural circuits in precisely defined behavioural tasks. In the fruit fly Drosophila melanogaster, we investigated the cognitive role of a major brain sugar sensor using a Pavlovian aversive olfactory learning task comprising the association of an odorant with mild electric shocks. After such conditioning, flies develop a learned avoidance towards this odorant7. After a single learning session, this memory decays within hours. Multiple sessions spaced in time (spaced training) induce the formation of long-term memory (LTM) that lasts for days8, is protein-synthesis dependent8 and is critically linked to glucose-based metabolism in neurons of the mushroom body (MB) brain area14,15,16. The same number of immediately consecutive sessions (massed training) induces another form of consolidated memory that lasts for around 1 day, but this memory does not correspond to LTM8,16,17: in addition to being less persistent8, it involves distinct neuronal circuits within the MB17 and relies on lipid-based rather than glucose-based neuronal energy metabolism18. The differential impact on consolidation efficiency between massed and spaced learning paradigms is an experimental manifestation in flies of a well-documented cognitive phenomenon named the spacing effect8,19,20,21. Experiments in the aversive learning paradigm are typically performed on fully satiated flies with ad libitum access to food before and after conditioning.

In the fly central brain, a small set of fructose-sensing neurons (four neurons per brain hemisphere, expressing the fructose-responsive gustatory receptor Gr43a, hereafter Gr43a neurons) respond to sugar ingestion and promote feeding in a satiation-dependent manner6,9,22. Dietary sucrose contains both glucose and fructose, but fructose is also produced from glucose metabolism through the polyol pathway, so fructose sensors respond to carbohydrate intake in general6. In hungry flies, Gr43a neurons detect an increase in fructose levels after sugar intake, and their activity promotes feeding6. The nature of the molecular signalling from Gr43a neurons and their downstream targets are currently unclear. When flies are satiated, the sensitivity of Gr43a neurons to fructose is disabled by the inhibitory action of the neuropeptide tachykinin (Tk), released by upstream dFB neurons, which are localized in the dorsal layer of the fan-shaped body region (dFB)9 (Extended Data Fig. 1a).

Our experiments reveal hunger-independent sugar-sensing plasticity underlying LTM, whereby spaced training restores the sensitivity of brain fructose-sensing neurons despite satiety, enabling their activation by post-training feeding. This endows food ingestion with a cognitive value, notably as a memory consolidation signal that extends b