The neural basis of sugar preference

Abstract

When it comes to food, one tempting substance is sugar. Although sweetness is detected by the tongue, the desire to consume sugar arises from the gut. Even when sweet taste is impaired, animals can distinguish sugars from non-nutritive sweeteners guided by sensory cues arising from the gut epithelium. Here, we review the molecular receptors, cells, circuits and behavioural consequences associated with sugar sensing in the gut. Recent work demonstrates that some duodenal cells, termed neuropod cells, can detect glucose using sodium glucose co-transporter 1, and release glutamate onto vagal afferent neurons. Based on these and other data, we propose a model in which specific populations of vagal neurons relay these sensory cues to distinct sets of neurons in the brain, including neurons in the caudal nucleus of the solitary tract, dopaminergic reward circuits in the basal ganglia, and homeostatic feeding circuits in the hypothalamus that alter current and future consumption. This emerging model highlights the critical role of the gut in sensing the chemical properties of ingested nutrients to guide appetitive decisions.

“Part of the secret of success in life is to eat what you like and let the food fight it out inside.” Mark Twain c.1905

Introduction

We seek nourishment guided by our senses, and when it comes to food, they determine which are nutritious, safe, satisfying, palatable and pleasurable. Although pursuing a particular diet is a conscious process, daily food intake is largely a product of subconscious preferences and food availability. Following the agricultural revolution, food availability increased exponentially, freeing a system once used mainly for survival to also allow for the pursuit of pleasure^1,2.

Pleasurable foods, like sugar extracted from sugar cane or beets, became common crops that could be processed, preserved, and promoted in the food chain. Sugar, which was once consumed as part of natural foods such as fruit, could now be distilled into its pure form. Industrial food processing further dissociated it from sensory cues. It was now possible to produce sweeteners that lack nutritive value. But our bodies could not adapt rapidly enough to interpret such artificial stimuli leading to skewed nutritive choices followed by a steep rise in obesity and metabolic disease³. Efforts to target the taste of foods alone through non-nutritive artificial sweeteners[G] have failed to curb disorders of food intake, revealing a pressing need to understand how animals detect, distinguish and choose to consume sugar^4,5.

Our diet is broadly monitored by two categories of senses: the pre-absoprtive and the post-absorptive. The pre-absorptive senses, often referred to as exteroceptive[G], monitor food before it crosses an epithelial surface of the body. In this category are the cephalic senses of olfaction and gustation as well as the more hidden sense of the gastrointestinal epithelium⁶. The post-absorptive senses, or interoceptive[G], monitor sensory cues from absorbed nutrients reaching the liver, pancreas, and other long-term energy reporters in adipose and muscular tissues. Both pre-absorptive and post-absorptive senses include conscious and subconscious components. It is the synergy between pre-absorptive and post-absorptive sensory cues that drives behaviour by allowing the brain to create a representation of foods from their appearance to their metabolic value.

This Review details the neural basis of our preference for sugar (Box 1) — from its pre-absorptive detection to its ability to drive behavioural preferences. The functions of gustation in recognizing sugar are well reviewed^7–9. Instead, this Review focuses on the role of the gut in recognizing glucose to drive behavioural preference. Novel tools first developed for use in the CNS are now being adapted for use in the periphery, which has accelerated our understanding of gut–brain signaling. Emerging evidence suggests that the brain receives rapid sensory cues from the gut independently of the oral cavity, and that these subconscious signals may subvert conscious processes in the preference for sugars.

Box 1 |. Sugar.

The English word ‘sugar’ traces its origin to the Indian subcontinent. The sanskrit word ‘sharkara’, which characterized the ‘gravel-like’ texture of sugar, travelled with Alexander the Great through the Arabic word “sukkar”, and finally landed in Western European words like “sucre” (France) and “azucar” (Spain)¹⁶⁵. The earliest recorded use of sugar products arises even earlier, at the time of the agricultural revolution in ~8000 BCE. In fact, sugarcane is purported to be one of the first domesticated crops, and Papua New Guinean nomads were known to chew on raw sugarcane¹⁶⁶. The appeal of sugar has evolutionary roots in the fruit-based diet of great apes. The natural sugar in fruit likely served as an anticipatory cue for caloric density^167,168.

Sugar generally refers to the chemical compound sucrose, which is produced by plants through photosynthesis¹⁶⁹. Sucrose is a disaccharide formed by one glucose and one fructose molecule linked via an ether bond. Cleavage of the bond via the enzyme sucrase-isomaltase is required for absorption. Expression of the enzyme has been detected in the oral cavity in saliva, on taste receptor cells, and more broadly along the brush border of the duodenum and jejunum of the small intestine. Once cleaved, glucose and fructose are absorbed via different transporters on the apical membrane of the intestine. Glucose is subsequently transported into the bloodstream, where pancreatic hormones tightly regulate its concentration. Importantly, glucose is the primary energy source for the CNS and is stored as glycogen in much lower quantities than fats. Thus, the body must continuously replenish glucose stores through gluconeogenesis or dietary intake¹⁷⁰. Fructose is then transported to the liver via the bloodstream and undergoes fructolysis before entering gluconeogenesis or lipogenesis.

Behavioural regulation of sugar intake

In 1947, Edward Adolph fed rats diets diluted with indigestible agents such as cellulose. He found that while the total amount of food consumed by the rats increased with dilute diets, the total amount of calories consumed remained the same¹⁰. This observation first associated the quantity of food consumed to its nutritional content; in other words, animals eat to meet caloric requirements. However, in the 1950s, Curt Richter observed that rats selectively increased intake of carbohydrates when challenged with the removal of the adrenal glands or the pancreas¹¹. These studies demonstrated that animals can also adjust their desire to seek out specific foods in response to the properties of ingested nutrients.

In 1952, rats with stomach fistulas were used to postulate that a solution “injected directly into the stomach serves as a reward to produce learning”¹². This role of gastrointestinal signalling in the behavioural regulation of food intake was substantiated in 1969 when Garvin Holman published a classic study in which they gave rats flavoured saccharin solutions paired with intragastric infusions of nutrient-rich eggnog or water¹³. When the rats were later given a choice between the two flavoured solutions in the absence of an intragastric infusion, they preferred the solution originally paired with nutrients. In the ensuing decades, other studies refined the techniques to further substantiate the hypothesis that the brain receives rapid signals from food stimuli detected by specific receptors in the gastrointestinal tract¹⁴.

A primary level of behavioural regulation of food intake requires animals to discriminate between different nutrients, assign value to nutrients depending on their utility, and decide whether or not to continue their consumption. Thus, nutrients entering the gut should be able to both positively and negatively reinforce consumption. The ability of nutritive substances to reduce consumption has been well documented through studies on satiation[G] and satiety[G]^15–17. However, more recent work has focused on positive reinforcement[G], which has been termed ‘appetition[G]’¹⁸. These processes are difficult to probe even in animal models, and most experimental paradigms require a training phase in which mice learn to respond for the infusion of intraluminal stimuli. For sugars, trained mice increase their licking responses towards an empty spout, which triggers intragastric self-infusion of glucose over water^19,20. Thus, mice will perform operant[G] tasks such as licking a dry spout to self-administer intra-intestinal infusions of glucose^21,22. Notably, not all carbohydrates can sustain operant behaviour to the same extent: the monosaccharide fructose is much weaker at doing so than glucose or sucrose¹⁹.

A secondary level of food intake regulation requires learning nutrient value for future decision-making. The animal must develop a preference, recall that preference, and act on it. This function of positive reinforcement of foods in associative learning has been extensively characterized using flavour–nutrient conditioning[G] paradigms, in which mice are trained to associate flavoured solutions (conditioned stimulus) with intragastric infusions of various nutrients (unconditioned stimulus). Then, in a separate session, mice select between the flavoured solutions. Using this experimental approach, it was found that the set of carbohydrates that can condition flavour preferences mirror those that can induce operant behaviour. For instance, glucose as well maltose and sucrose can strongly condition a preference. Maltose is made of two glucose molecules whereas sucrose is a mix of glucose and fructose. Contrary to glucose, intragastric infusion of fructose is less or completely ineffective at conditioning a flavour preference^23–26. The conditioned preference learning is both rapid and persistent. With strong conditioned stimuli, rats can acquire robust conditioned flavour preferences after a single conditioning day. In one study adopting such an approach, after training, a single 6 ml intragastric glucose infusion produced a long-lasting preference memory, lasting for at least 13 days post-conditioning²⁷.

The timescales of these primary and secondary levels of food intake regulation appear to be different: whereas acceptance of a solution occurs within seconds to minutes of a meal, learning probably takes place over hours across meals. Although conditioned learning has not been assessed for intervals shorter than 24 hours, the gastrointestinal tract can distinguish between specific carbohydrates within secondsREF⁹⁹. Over the last 75 years since Adolph began feeding cellulose to rats, it has become clear that the gastrointestinal tract can distinguish between nutrient stimuli and guide appetitive behaviors.

Molecular sugar detectors

Detection of sucrose and its byproducts by epithelial cells along the contiguous sensory surface from tongue to intestine occurs via three primary pathways: taste receptors (comprising taste receptor type 1 member 2 (T1R2) and T1R3 subunits), glucose transporters (for example, sodium/glucose cotransporter 1 (SGLT1)) and ATP-gated potassium (K_ATP) channels. These three mechanisms are shared across various tissues in both the pre-absorptive and post-absorptive sensory systems. Outside of the oral cavity, taste receptors are also expressed in the pancreas^28,29, adipocytes^30,31 and the brain^32,33. Similarly, SGLTs are not restricted to the intestine, and are also expressed in the tongue³⁴, pancreas³⁵, brain³⁶ and hepatic portal system³⁷. Within the gastrointestinal tract, T1R2–T1R3 receptors and SGLT1 occur along the apical surface facing the lumen of the gastrointestinal tract^38,39, whereas K_ATP channels detect the end products of glucose metabolism within the cell (Fig. 1)⁴⁰.

Fig. 1 |. Mechanisms of sensing glucose in gut epithelial cells.

In intestinal epithelial cells known as neuropod cells, glucose can lead to cellular depolarization and subsequent signalling molecule release in three ways. Sodium glucose co-transporter 1 (SGLT1)-dependent inflow of Na⁺ drives glucose absorption and leads to depolarization. L-type voltage-gated calcium channels (L-VGCCs) are then hypothesized to be activated, which triggers a rise in intracellular Ca²⁺ levels and the release of glutamate (1). Activation of T1R3-subunit containing taste receptors by glucose triggers a secondary signalling cascade involving the following: α-gustducin (Gα), phospholipase Cβ2 (PLCβ2), phosphatidylinositol 4,5-bisphosphate (PIP2), inositol 1,4,5-trisphosphate (IP3), IP3 receptors (IP3Rs) on the endoplasmic reticulum (ER), and TRPM5 (2). Depolarization is hypothesized to induce voltage-gated CAHLM1–CAHLM3 channels to release ATP non-vesicularly (not shown). VGCCs can also be activated that lead to increased intracellular Ca²⁺ levels and subsequent vesicular release of ATP. ATP-gated potassium (K_ATP) channels close in response to the end-products of glucose metabolism increasing the ATP:ADP ratio (3). VGCCs can then open that trigger the release of hormones such as cholecystokinin (CCK). CCKAR, CCK A receptor.

Sweet taste receptors

Sweet[G] taste receptor subunits were first identified by three independent groups and have been primarily studied in the context of taste receptor cells of the tongue^41–43. As indicated above, sweet taste is conveyed by a heterodimer of class C G-protein coupled receptors (GPCRs), namely T1R2 and T1R3^43,44. These GPCRs contain four major domains: an extracellular venus-flytrap domain, a cysteine-rich linker domain, a transmembrane seven-helix domain and an intracellular domain that interacts with G proteins. Based on homology to other receptors and modelling, it is predicted that sugar binding to the venus-flytrap domain triggers a conformational shift that closes the two flytrap lobes and activates the intracellular domain to bind its G protein^45,46. Artificial sweeteners have also been shown to activate components of T1R2–T1R3 receptors to induce sweet taste^43,47. Although the venus-flytrap domains of both T1R2 and T1R3 are the predominant sites of binding for sucrose, some artificial sweeteners such as cyclamate can act at other sites in the transmembrane domain⁴⁸. Moreover, some sensory cells, including pancreatic beta cells, can also form T1R3 homodimers that are predicted to be activated at the same domains⁴⁹. Unravelling such distinctions are necessary to understand the contributions of different organs to the perception of distinctive stimuli.

Conformation change of the sweet taste receptor leads to the activation of a secondary signalling cascade. It begins with the activation of a G protein containing the taste-specific α-subunit, known as α-gustducin. This G protein then activates phospholipase-Cβ2, which uses membrane-anchored phosphatidylinositol 4,5-bisphosphate to catalyze the formation of diacylglycerol and inositol 1,4,5-trisphosphate (IP3). IP3 subsequently activates its receptor to stimulate Ca²⁺ release from intracellular stores. Increased intracellular Ca²⁺ then triggers the opening of TRPM4 and TRPM5 channels, calcium-gated non-selective cation channels that allow the intracellular flow of Na⁺ ions^50,51. This depolarization leads to the opening of the voltage-gated non-vesicular ATP CALHM1–CALHM3 channel⁵², driving the release of ATP from the cell⁵³. Genetic knock out of T1R2–T1R3^54,55, α-gustducin⁵⁶, TRPM5⁵⁷ or CALHM1–CALHM3⁵² in mice ablates the sweet taste response in the tongue, confirming the importance of these proteins in sweet taste sensation.

Although first characterized in taste receptor cells of the tongue, T1R2–T1R3 is also expressed in cells of other mucosal surfaces, including the intestines³⁸, bladder⁵⁸ and lung⁵⁹. Indeed, taste receptor expression in the intestine has been found on electrically excitable enteroendocrine cells[G]³⁸. Enteroendocrine T1R2–T1R3 activation has been linked to release of the hormones glucagon-like peptide 1 (GLP1) and gastric inhibitory peptide (GIP), which assist in regulating SGLT1 expression⁶⁰.

Glucose transporters

A second class of apical sensors for sugar are glucose transporters. These include the SGLTs and the GLUT family of glucose transporters.

SGLT1 serves as the primary apical transporter for the monosaccharides glucose and galactose, and along with the basal membrane transporter GLUT2, is required for all subsequent post-absorptive glucose signalling⁶¹. The protein comprises 14 transmembrane helices and contains a pocket that transports one glucose molecule following two Na⁺ molecules into the cell^62,63. In sensory cells of the gut, this Na⁺ influx is sufficient to induce depolarization and subsequent opening of voltage-gated L-type calcium channels⁶⁴. Increased intracellular Ca²⁺ then triggers the release of peptide hormones such as GLP1 and GIP. Inhibition of SGLT1 pharmacologically or through a genetic knockout showed that this glucose transporter is required for the release of GLP1 and GIP^39,65. Recently, it has been demonstrated that intestinal epithelial cells labeled by the cholecystokinin (CCK) promoter release the neurotransmitter glutamate in response to glucose⁶⁶. This process depends on the activity of SGLT1. Blocking the SGLT1 transporter pharmacologically halts the release of glutamate⁶⁶. SGLT1 signalling is selective for glucose and is not activated by fructose or artificial sweeteners. Instead, fructose is absorbed by GLUT5⁶⁷. SGLT3 can similarly bind glucose and has been proposed to be a glucose sensor as well⁶⁸.

Of importance, taste receptor knockout mice show some persistence in sweet taste sensation. This phenomenon suggests the existence of taste receptor-independent mechanisms for detecting sugar in the oral cavity^54,55. Recent work has demonstrated that taste receptor cells also express SGLT1^69,70. Moreover, some of the oral afferent nerve responses to sugar depend on the activity of SGLTs⁷¹. Together, these findings suggest that the oral cavity can also utilize SGLT-dependent pathways to sense glucose. Indeed, taste-independent sugar sensing by SGLTs is preserved across species, even in Drosophila^72,73.

ATP-gated potassium channels

The majority of the knowledge on K_ATP channels has been acquired in the context of glucose-dependent insulin secretion in pancreatic beta cells⁷⁴. K_ATP channels are also found in oral taste receptor cells and intestinal enteroendocrine cells^34,40. Once glucose is transported into a beta cell, it undergoes glycolytic metabolism and culminates in the phosphorylation of ADP to ATP. As the ratio of ATP:ADP increases, K_ATP channels enter a closed conformational state that depolarizes the cell. This depolarization triggers analogous mechanisms to SGLT1-mediated Ca²⁺ release, leading to the opening of voltage-dependent L-type calcium channels and subsequent release of vesicles containing peptides or transmitters. This metabolic mechanism is a response to the end products of sugar absorption rather than the sugar molecule itself. As expected, the onset of intracellular metabolic effects of sugar come after the rapid sensing mediated by SGLT1. The existence of both rapid sensing and metabolic responses are in line with the fact that the application of sugar to the intestinal lumen leads to both a millisecond glutamatergic signal and a delayed (~2 min later) hormonal CCK signal⁶⁶. Activation of the K_ATP channel is also linked to the release of hormonal peptides such as GLP1 that contribute to regulation of glucose storage and uptake^40,64,75.

Intestinal sugar detection and SGLT1

In 2008, de Araujo et al. tested the ability of mice lacking TRPM5 to detect sucrose⁷⁶. While the mice were unable to distinguish the sweetener sucralose from water, Trpm5 knockout mice retained the ability to distinguish sucrose over water^76,77. This finding suggested that sugar preference is independent of sweet taste receptors. Moreover, in intragastric conditioning paradigms, the sweet taste receptor agonist sucralose is insufficient for conditioning a flavour preference⁷⁸ and both pharmacological and genetic silencing of T1R3 do not alter the reinforcing properties of glucose⁷⁸.

It became plausible that the ability to recognize sugar depends on a set of receptors and perhaps cells located outside the tongue. Some reports suggested that this receptor could be SGLT1. Its agonist α-methylglucopyranoside can condition a preference, and both pharmacological and genetic silencing of SGLT1 prevent mice from being conditioned to develop a preference for glucose^25,79. These results suggest that SGLT1 is the sensor for glucose appetition in the gastrointestinal epithelium.

Gut sensory transduction

Despite progress in understanding the molecular sensors for sugars, the identity of the gut epithelial cells responsible for the discrimination and subsequent reinforcing properties of sugar remained mysterious for decades^4,80,81. Although the connections from gut-to-brain include both humoral and neural pathways⁸², the prevailing model of nutrient sensation had been that the pre-ingestive senses utilized epithelial transducers synaptically connected to nerves, while post-ingestive senses utilized endocrine cells that secreted hormones^83–85.

Studies of sugar sensing in the gut thus emphasized circulating peptides⁸⁵. These hormonal pathways have mostly been linked to satiety^86,87, but some early studies postulated that the desire and preference to consume sugar could be due to the hormone CCK^88,89. However, early experiments established that pharmacological inhibition of the CCK A receptor via intraperitoneal injection does not affect preference conditioning for intragastric infusion of the glucose polymer polycose⁹⁰. In addition, when the hormone peptide YY(3–36) was infused intravenously at low concentrations, food intake decreased and at high concentrations conditioned robust taste aversion⁹¹. Hepatic portal vein infusions of GLP1 only reduced food consumption⁹². The hormone ghrelin is known to stimulate feeding, but pharmacological inhibition and genetic ablation of the ghrelin receptor did not impact sugar conditioning⁹³. Thus, the existing evidence suggests that consumption and preference of sugar is unlikely to rely on gut hormonal signals.

Flavour–nutrient conditioning paradigms showed that the preference and desire to consume sugar depended on sensory cues emanating from the very proximal small intestine. Such an ability of sugar to condition a flavor preference disappears in surgical preparations where the proximal small intestine is bypassed^94–96. These studies showed that sugar detection occurred in the duodenum and jejunum, but not in the distal ileum or the stomach^94–96. In 2015, it was shown that enteroendocrine cells throughout the intestine and colon of the mouse synapse with nerves⁹⁷. The discovery of synaptic neurotransmission in these nutrient sensing cells immediately opened the possibility that appetitive decisions and other visceral behaviors could be guided by rapid sensory cues emanating from the gastrointestinal tract⁹⁸. The endocrine term was limiting for a cell engaging in neurotransmission, so such cells were termed neuropod cells[G]⁶⁶ (Box 2).

Box 2 |. Neurotransmission from gut to brain.

In 2011, an anatomical peculiarity was observed in a subset of enteroendocrine cells that opened a new mechanism for intestinal communication. Some of these sensory cells had a cytoplasmic protrusion at the base of the cells which resembled a neuronal axon¹⁷¹. Subsequent work confirmed that some enteroendocrine cells, both with and without a basal extension, could from synapses with enteric, vagal and splanchnic nerves^{66,97,172,173}. To account for the neuron-like function in these cells, the term neuropod cell was extended to those epithelial cells engaging in neurotransmission⁶⁶. This gut neuroepithelial sensory circuit could serve to account for the chemical, mechanical and thermal properties of ingested nutrients. The similarity between neuroepithelial circuits in the gastrointestinal tract and other sensory systems like those of taste, olfaction and vision raised the hypothesis that there exists continuity between pre-ingestive and post-ingestive sensory mechanisms.

Neuropod cells are a subset of enteroendocrine cells that are differentiated by their neural function. Transcript and protein analyses have shown that subpopulations of enteroendocrine cells express pre-synaptic proteins such as synapsin 1, piccolo, bassoon, MUNC13B, RIMS2, latrophilin 1 and neurexin 2⁹⁷. Ultrastructural investigations have revealed that neuropods contain neurofilaments (see the figure, part a) as well as both dense core vesicles containing peptides and synaptic clear vesicles containing neurotransmitters (see the figure, part b)¹⁶⁴. In vitro, neuropod cells synapse with co-cultured neurons, and optogenetic activation of neuropod cells triggers excitatory postsynaptic potentials in connected neurons within 60 milliseconds⁶⁶. In vivo, monosynaptic tracing using rabies virus demonstrated that small intestinal neuropod cells form monosynaptic connections with vagal neurons to rapidly communicate nutrient sensory information to the brain (see the figure, part c)⁶⁶. In a series of studies utilizing organoid cultures, Bellono et al. showed that colonic enterochromaffin cells synapse with peripheral nerves to rapidly transduce the presence of irritants using serotonin¹⁷². Subsequent results showed that these cells could also form purinergic synapses¹⁷³. In vivo, luminal optogenetic manipulation of neuropod cells shows that these cells are required for sugar preference⁹⁹. Parts a and b adapted with permission from REF.¹⁶⁴. Part c adapted with permission from REF.⁶⁶.

In 2018, Kaelberer et al. showed that duodenal neuropod cells synapse with vagal neurons to convey the presence of sugar⁶⁶. These cells express SGLT1 and increase intracellular Ca²⁺ in response to glucose. Glucose perfusion onto these cells induces the release of glutamate, which can evoke excitatory postsynaptic currents in co-cultured neurons⁶⁶. Silencing duodenal neuropod cells using optogenetics or pharmacological inhibitors of ionotropic glutamatergic receptors completely abolished the rapid phase of vagal firing induced by glucose infusion. Although these cells were first identified using the CCK promoter, rapid transmission of sugar stimuli onto vagal afferents is independent of the hormone CCK. Inhibition of the CCKA receptor with the pharmacological inhibitor devazepide does not affect rapid vagal firing induced by glucose infusion in the duodenum. Thus, these findings established that the ability of the brain to receive in milliseconds sensory stimuli from sugar entering the intestine depends on glutamatergic signaling from neuropod cells.

In 2022, Buchanan et al. developed a new tool to interrogate the contribution of visceral sensory cues to behaviour. They developed a flexible fibre optic cable to drive light-sensitive opsins in specific sensory cells exposed to the gut lumen. In this way, it was now possible to interrogate the contribution of neuropod cells to sugar preferences in real time⁹⁹. When trained mice are presented with two bottles containing either sucralose or isosweet sucrose solutions, they prefer sucrose. However, silencing duodenal neuropod cells labeled by the CCK promoter prevents the animals from distinguishing sugar from artificial sweetener. Moreover, pharmacological inhibition of ionotropic glutamate receptors in the duodenum also extinguished preference for sugar over sweetener. These findings demonstrated that the animal distinguishes sugar from sweetener guided by intestinal glutamatergic signalling. In other words, neuropod cells guide the animal’s consumption and preference for sugar by rapidly detecting sugar. Whether glutamatergic neurotransmission from neuropod cells is necessary for other behaviours such as the development or recall of sugar preferences remains to be tested.

From the gut to the brain

Vagal pathways

The vagus nerve contains both sensory and motor fibres that connect the brainstem with the viscera, including the gut^100,101. The entry of nutrients into the duodenum stimulates vagal firing¹⁰², and specific vagal neurons are tuned to macronutrients¹⁰³. Both sugars⁶⁶ and artificial sweeteners⁹⁹ entering the duodenal lumen rapidly stimulate vagal firing, but the vagal responses appear to be mediated by specific subpopulations of sensory vagal nodose neurons (Fig. 2)^81,104. Vagal responses to sucrose depend on SGLT1⁸¹ and are confined to the small intestine as colonic perfusion of sucrose does not increase vagal firing⁹⁹. Although the colonic mucosa is sparsely innervated by the vagus nerve, SGLT1 expression is absent in the colonic epithelium⁹⁹. Importantly, vagal nerve terminals do not reach the lumen of the intestine^105,106 and do not express SGLT1^99,105,107. Therefore, vagal neurons must engage with intestinal neuropod cells that act as the primary sensory transducers. And at least for sugars, vagal neurons depend on sensory inputs from neuropod cells. Whether the vagus serves as a relay from the intestine to brain or has some role in integrating information from multiple primary sensors in other visceral organs remains to be tested.

Fig. 2 |. Heterogeneity within nodose neurons.

a | Schematic showing individual subpopulations of vagal sensory nodose neurons that respond to different stimuli. b | GCaMP3 calcium responses of nodose neurons in response to mechanical stretch and nutrient perfusion. In the stomach, most activated neurons are activated by stretch, while in the intestine, distinct populations respond to mechanical and chemical stimuli. c | GCaMP6f imaging of vagal nodose neurons. Subpopulations of neurons imaged were selective for glucose and not artificial sweetener acesulfame K (AceK). Only neurons that had a response to glucose are shown. Each row represents an individual cell, and colour represents normalized fluorescence. Part b adapted with permission from REF.¹⁰⁴. Part c adapted with permission from REF.⁸¹.

Despite its central role in gut–brain signaling, the vagus nerve’s function in sugar preference has been debated. For over a century, the prevailing idea was that vagal signalling from the small intestine to the brain is a negative feedback signal to inhibit future consumption^4,108. In 2018, Han et al. countered this theory by identifying a population of “reward neurons” in the right vagus¹⁰⁹. Using optogenetics together with viral tracing, they manipulated a specific population of vagal afferents innervating the proximal gut independently of efferents or fibres that do not innervate the gut. Activating these gut-innervating neurons was sufficient for mice to perform arbitrary operant behaviours such as nose poking. Moreover, stimulating these same neurons was sufficient to condition a preference for an arbitrary location. Such self-stimulating behaviours confirmed the existence of vagal afferents capable of inducing reward¹⁰⁹. The study went on to use these vagal neurons to modulate the rewarding effects of intragastric perfusions of fat emulsion, and future work should address whether sugars engage similar circuits.

The challenge of investigating the function of the vagus lies in a lack of tools to target specific subpopulations of vagal afferents¹¹⁰. Classic techniques such as surgical vagotomy are limited for several reasons. The vagus nerve can regrow^111,112, nerve transection does not fully differentiate between afferents and efferents, and nerve transection is often not organ specific. Such limitations produce results confounded by many unintended effects such as high mortality rate in experimental animals, failure to thrive, change in gastric motility and emptying, and other secondary effects. Pharmacological approaches such as capsaicin ablation are also incomplete. Nevertheless, these techniques have been used in numerous studies to conclude that reward learning from intra-intestinal glucose is vagal-independent^96,113–115.

A recent study used a genetic tetanus toxin strategy in mice to silence all sensory vagal neurons expressing the vesicular glutamate transporter 2 in the nodose ganglion[G]. The study found that these mice did not prefer glucose over an artificial sweetener⁸¹. Such seemingly contradictory pieces of evidence may point to redundancy in the system. When stimuli are well separated in time, other signalling pathways, including humoral and post-absorptive hepatic signalling, may be recruited to assure proper reinforcement and association^116,117. However, in complex scenarios in which multiple stimuli are consumed within one meal, gut–neural circuits may be engaged to ensure proper stimulus discrimination and association¹¹⁸.

New efforts to subclassify vagal neurons guided by transcription patterns have identified that vagal neurons innervating the mucosa of the proximal small intestine specifically express genes such as Gpr65 or Ntsr1^105,107. Cre-lines to specifically modulate these subpopulations have also been developed, and such tools will enable future experimentation to differentiate the function of specific vagal subpopulations^104,119.

Splanchnic pathways

Another set of neurons that project from the intestine to the brain are splanchnic nerves. These nerves derive from the celiac and superior mesenteric ganglia and enter the brain via the spinal cord. Limited evidence exists to determine the function of these neurons in sugar sensing. While some studies have not observed significant changes in food consumption^120,121, surgical ablation of these afferents using a complete transection of the ganglia reduces the strength of conditioned preferences for glucose¹¹⁵. More recently, splanchnic lesions were shown to reduce the responsiveness of central feeding circuits in response to glucose¹²². Similar to vagotomies, surgical transection of splanchnic nerves to interrogate their contribution to behaviour has several limitations. Future intersectional genetic approaches using viruses carrying chemical or optical actuators will help elucidate the contribution of these nerves to rapid appetitive decisions.

Central circuits in sugar preference

Neuronal signals from the viscera enter the brain through the brainstem before they are conveyed to a distributed network of overlapping circuits throughout the brain involved in controlling behaviour. Here, we describe the primary brainstem neurons that integrate signals first encountered by the vagus nerve, the current understanding of the role of upstream dopaminergic reward circuits in the basal ganglia, and homeostatic feeding circuits in the hypothalamus involved in triggering consumption and learning the value of sugar.

Nucleus tractus solitarius

Afferent terminals from pre-absorptive circuits terminate in the nucleus tractus solitarius (NTS). This nucleus forms a column of cell bodies within the medulla oblongota with general topographic distribution depending on the location of the input¹²³. The rostral portion of the NTS (rNTS) receives inputs mainly from the oral cavity, while the more caudal portion (cNTS) receives glutamatergic inputs from visceral organs, including the gut^106,124. NTS neurons also receive inputs from humoral circulating factors such as leptin, as well as feedback from neural circuits arising in the hypothalamus^125,126. NTS neurons project to many regions across the brain, including the dorsal motor nucleus of the vagus, hypothalamus, central nucleus of the amygdala, parabrachial nucleus, and other brainstem areas involved in adrenergic (locus coeruleus), dopaminergic (ventral tegmental area) and serotonergic signalling (dorsal raphe)^127,128. NTS neurons are primarily glutamatergic and co-release peptides such as norepinephrine, GLP1 and CCK¹²⁹. The NTS forms a heterogeneous population of neurons that can induce both satiation and appetition^130,131.

Distinct populations of cNTS neurons, as measured by cFos expression, are activated by intragastric and intestinal glucose perfusions^132,133. Tan et al. recently identified that a population of proenkephalin-expressing neurons within the cNTS is activated by duodenal glucose infusion⁸¹. These same neurons are not activated by the artificial sweetener acesulfame K. Acesulfame K binds to sweet taste receptor T1R2–T1R3 but does not serve as a substrate of the sugar receptor SGLT1. Ablating synaptic activity in these brainstem neurons using tetanus toxin does not prevent discrimination of either glucose or the artificial sweetener from water. However, their manipulation does prevent mice from discriminating between glucose and artificial sweetener, and from subsequently developing a preference for glucose over sweetener in a 48-hour test. Moreover, chemogenetically activating these neurons was sufficient to condition a flavour preference⁸¹. Monosynaptic tracing studies from neuropod cells in the duodenum have shown connections with vagal afferent neurons that project to the NTS⁶⁶. These vagal neurons most likely overlap with the peripheral reward neurons that project to NTS proenkephalin-expressing neurons, but this hypothesis remains to be tested.

Dopaminergic reward circuits

Reward and reinforcement learning are linked to dopaminergic circuits that facilitate goal-directed behaviours and promote the ‘wanting’ of food rewards^134,135. These circuits are conceptualized to mediate subconscious aspects of food reward and selection⁴. Eliminating dopaminergic signalling in mice does not prevent sugar consumption, and mice can still discriminate sugar from water¹³⁶. This deficiency did reduce the total number of consumption bouts[G], but the rate of licking and bout size were greater¹³⁶, suggesting that dopamine is not required to detect sucrose palatability[G]. By contrast, pharmacological inhibition of dopamine D1 receptors in the striatum during training sessions but not during testing sessions prevented flavour preference conditioning¹³⁷. Thus, dopamine signalling underlies the learned associations between food value and environmental cues rather than intrinsic sugar preferences.

Two independent dopaminergic circuits arise in the midbrain: the mesolimbic[G] system projecting from the ventral tegmental area (VTA) to the ventral striatum, and the nigrostriatal[G] system projecting from the substantia nigra pars compacta (SNPc) to the dorsal striatum (Fig. 3). Outputs from these two basal ganglia regions project to motor feeding circuits to control behaviour. Oral and intra-intestinal infusion of sucrose shows increased activation of neurons and dopamine release in both pathways^77,138–141. However, in sham-fed rats, in which the oral cavity is exclusively stimulated, sucrose only induces dopamine release in the ventral striatum¹³⁹. Thus, while the mesolimbic system receives signals from oral and post-ingestive sites, the nigrostriatal system only receives input from the intestine and post-absorptive sensors.

Fig. 3. |. Dopaminergic circuits in sugar sensing.

Two distinct dopaminergic pathways are activated by the experience of sugar in the mouse brain. Mesolimbic dopamine release triggered by connections between the ventral tegmental area (VTA) and the ventral striatum (Str) are produced in response to oral, intestinal and hepatic portal vein infusion of sugar. By contrast, nigrostriatal dopamine release between the substantia nigra pars compacta (SNPc) and the dorsal striatum are only triggered by hepatic portal vein and intestinal signalling in response to sugar. All peripheral inputs enter the brain through the nucleus tractus solitarius (NTS), where oral inputs enter the rostral (rNTS) portion, and intestinal and post-absorptive signals enter via the caudal (cNTS) portion. The NTS can then project to dopaminergic circuits directly or via polysynaptic circuits including the parabrachial nucleus (PBN).

To compare the functions of the mesolimbic and nigrostriatal pathways in intestinal sensing, Tellez et al. measured dopamine release in both the ventral and dorsal striatum while mice licked sucralose and received intragastric infusions of either sucralose or glucose¹⁴¹. While both solutions induced dopamine release in the ventral striatum, only intragastric glucose induced dopamine release in the dorsal striatum. Suppressing taste-mediated signalling by mixing in a bitter tastant reduced ventral striatal dopamine release. And suppressing gut-mediated signalling using a sugar analogue that cannot be absorbed reduced dorsal striatal release. In addition, using both cell ablation and optogenetic activation of neurons expressing dopamine D1 receptors in the dorsal and ventral striatum, it was determined that the nigrostriatal pathway is both necessary and sufficient for the post-ingestive reinforcing effect of glucose¹⁴¹. This study demonstrates independent roles for the mesolimbic and nigrostriatal dopamine pathways in learning the value of consuming glucose. Their role in the immediate acceptance of glucose is less clear, and the relative contributions of intestinal and post-absorptive signalling on nigrostriatal dopamine release remains to be established.

Activation of mesolimbic dopamine neurons using optogenetics is sufficient to drive voluntary ingestion of water¹⁴². Moreover, in two-bottle preference tests designed to test the ‘reward value’ of dopaminergic signalling, mesolimbic activation was more rewarding than non-caloric sweetener sucralose but less rewarding than sucrose¹⁴². In this regard, mice can reverse their natural preference for sucrose over sucralose if sucralose is paired with mesolimbic activation¹⁴². Thus, the mesolimbic system assigns more value to solutions than solely taste receptor-mediated mechanisms but less value than SGLT1 and taste receptor mechanisms together.

Recently, miniscopes implanted in the brain were used to directly image individual VTA neuron activity in freely behaving mice. Intra-gastric infusion of sucrose positively modulated a greater proportion of dopaminergic neurons in the VTA than did intra-gastric infusion of sucralose¹⁴³. It was also shown that burst firing of VTA dopaminergic neurons is required for the reinforcing effects of intra-gastric sucrose in a food-seeking task. Transection of the hepatic vagus nerve blunted VTA responses, suggesting that the mesolimbic activation depends on post-absorptive signalling¹⁴³. These results suggest that midbrain neurons, specifically those mesolimbic dopaminergic neurons, assign the reward value of sugars independently of their sweet taste.

Midbrain dopaminergic centres receive inputs from NTS neurons that project directly or via the lateral parabrachial nucleus¹²⁹. In tracing experiments, rNTS neurons polysynaptically target the VTA, while cNTS neurons project monosynaptically or polysynaptically to the VTA^109,144 and polysynaptically to the SNPc¹⁰⁹. Viral tracing studies also confirmed that the nigrostriatal pathway receives polysynaptic input from the reward neurons of the right vagus¹⁰⁹. Activation of right vagal reward neurons increased dopamine release in the dorsal striatum but not the ventral striatum¹⁰⁹. Parallel studies showed that optogenetic activation of the left vagus increased activity of VTA dopaminergic neurons¹⁴³. These results imply an asymmetry within the vagus, with left vagal stimulation increasing mesolimbic activation and right vagal stimulation increasing nigrostriatal activation.

Agouti-related peptide feeding circuits

Hypothalamic neurons have long been known to be activated by intra-intestinal infusions of nutrients, including glucose¹⁴⁵. Subsequent studies identified specific neuronal populations involved in controlling food intake, including the agouti-related peptide (AgRP) neurons of the arcuate nucleus^146,147. Activity of AgRP neurons is strongly correlated with the energy state of the animal¹⁴⁸, and stimulation of these neurons with optogenetics or chemogenetics leads to appetitive behaviours, including food consumption and seeking^149,150. This consumption is dependent on the nutrient composition of the food consumed, as AgRP neuron activation leads to rapid consumption of sucrose but not saccharin pellets¹⁵¹. By contrast, ablation of these neurons leads to a loss of desire to consume food and failure to thrive¹⁴⁷. AgRP neurons project downstream to several brain areas involved in food consumption, such as the lateral hypothalamus, paraventricular hypothalamus and the bed nucleus of the stria terminalis¹⁵².

Although AgRP neurons receive synaptic inputs from diverse brain areas including the cNTS, they are also adjacent to the median eminence and thus were traditionally thought to respond to circulating humoral factors¹⁵³. Recent direct imaging of AgRP calcium activity in awake mice has shown that AgRP neurons are rapidly and strongly suppressed by nutrient consumption^154–156. The timing of the response was more rapid than possible by humoral mechanisms, suggesting that AgRP neurons may receive information via polysynaptic pathways from the viscera. The rapid suppression of AgRP neuron activity occurs in response to intragastric infusion of glucose and not sucralose, mirroring the behavioural studies, and further confirming the importance of nutrient identity¹⁵⁶. AgRP suppression by intraduodenal infusion of glucose depends on SGLT1¹²². Moreover, the responses are proportional to the caloric load of glucose and are thought to represent predicted nutrient consumption¹⁵⁵.

Given the suppression of AgRP neurons in response to nutrients, it was postulated that AgRP activation could provide an instructive negative valence signal to prevent future consumption. Concordantly, continuous stimulation of AgRP neurons was sufficient to condition flavour and place avoidances¹⁵⁴. By contrast, the rapid suppression of AgRP neuron activity, as seen with physiological infusion of sugar, was shown to have a positive valence and was sufficient to condition flavour-nutrient preferences¹⁵⁷. In addition, AgRP neuron activation potentiates the dopamine release in the nucleus accumbens and VTA dopaminergic neuron activity in response to food stimuli^158,159. Notably, AgRP responses can also occur to visual, olfactory and conditioned cues that precede nutrient delivery, suggesting that one of the functions of AgRP neurons is to predict imminent nutrient availability^155,160. These responses diminish rapidly without continual intraintestinal stimulation¹⁵⁴.

The identity of the visceral sensors and inputs for glucose-mediated AgRP suppression remains unclear. Recent work suggests that AgRP suppression is unaffected by vagotomy and is reduced by transection of splanchnic nerves¹²². Moreover, several studies have implicated the role of peptides such as CCK and PYY in reducing AgRP activity^155,156. However, the signal for glucose-mediated changes is still unidentified, and the neural route from the intestine or post-absorptive sites to the arcuate nucleus is an area of active research¹⁰⁵.

Integrated view

The gut’s ability to detect, distinguish and transduce stimuli from individual nutrients like glucose is a fundamental sense. This sense is critical for animals to recognize the contents that they are actively transporting from the external environment into their internal milieu. The current evidence thus leads us to propose a unified model of detecting sugar that involves continuous sensing along the pre-absorptive epithelium (Fig. 4).

Fig. 4 |. A model of neuroepithelial circuits for sugar sensing.

Chemosensation through synaptic neural circuits occurs via rods and cones in the eye, olfactory receptor neurons in the nose, taste receptor cells in the tongue, and neuropod cells in the intestine. This forms a continuous sensory system that monitors sugar from the process of its foraging to absorption.

In this model, the sensory experience of sugar begins before it is even ingested. Visual, olfactory and even tactile cues trigger central feeding circuits that predict the value of its ingestion. Once it enters the oral cavity, it can activate taste receptor cells that signal through cranial nerves and the rNTS into the gustatory cortex for conscious perception as well as into mesolimbic dopaminergic circuits for hedonic[G] value.

As sugar is ingested and contacts the stomach and proximal duodenum, neuropod cells sense it through SGLT1, and release glutamate onto vagal afferents that project to the caudal NTS, probably onto proenkephalin-expressing neurons. These neurons send projections to the nigrostriatal dopamine system, as well as the mesolimbic dopamine system and also likely engage rapid suppression of AgRP neuron activity. Simultaneously, other enteroendocrine cells begin releasing humoral and paracrine factors like GIP, GLP1 and CCK.

Finally, enterocytes actively absorb sugar via SGLT1 and deposit it into the portal venous circulation, where hepatic sensors can become activated and signal the absorbed quantity of glucose to the CNS via circulating factors and left vagal activation. The combination of intestinal epithelial signalling and portal venous signalling converge on higher-order dopaminergic and homeostatic circuits to drive acceptance of the solution and subsequent reinforcement learning.

Recognizing the combined role of the entire sensory system will probably improve future efforts to combat dietary imbalances. Artificial sweeteners cannot improve dietary choices alone because while artificial sweeteners can trick oral taste-receptor signalling, they do not trigger the intestinal SGLT1-dependent glutamatergic signalling pathway. Gut sensations are also increasingly linked to other important nervous system functions. Arousal circuits and sleep patterns can be disrupted by visceral sensation¹⁶¹. Emotional states can be modulated by intestinal microbial and nutrient stimuli¹⁶². Even memories formed in the hippocampus can be altered by intestinal signalling¹⁶³. Each of these functions involves the gut sense. Unraveling these manifold abilities will allow us to harness the gut to enjoy pleasurable foods, foster healthy habits and promote emotional well-being. After all, the instinctive wisdom of the body is the reason why, when it relates to sugar, we go with our gut.

GLOSSARY

Artificial sweeteners

Chemicals produced to mimic sugar by binding to the sweet taste receptor while also not being capable of being absorbed

Exteroceptive

Sensitivity to stimuli originating outside the epithelial barrier of the body

Interoceptive

Sensitivity to stimuli originating inside the epithelial barrier of the body

Satiation

A process that suppresses nutrient intake and stops current consumption

Satiety

A sense of fullness that persists after eating that suppresses future consumption

Reinforcement

A procedure that results in the frequency or probability of a response being increased in such a way

Appetition

A process that promotes nutrient intake

Operant

A response that produces a consequential effect on the environment

Flavour–nutrient conditioning

A form of classical conditioning where flavours of a solution are paired to infusions of chemical stimuli into the gut

Sweet

The perception of taste associated with sugar, typically triggered by activation of sweet taste receptors in the tongue

Enteroendocrine cell

Sensory epithelial cell of the gut that releases hormones in response to various stimuli

Neuropod cells

Sensory epithelial cells of the gut that synapse with nerves

Nodose ganglion

The sensory ganglion of the vagus nerve located in the jugular foramen

Bouts

Instances of food consumption

Palatability

The degree of pleasantness or reward associated with stimulating the oral palate

Mesolimbic system

A network of dopaminergic neurons consisting of the ventral striatum that receives input from the ventral tegmental area and is related to emotion and reward

Nigrostriatal system

A network of dopaminergic neurons consisting of the dorsal striatum that receives input from the substantia nigra pars compacta and is linked to action initiation

Hedonic

A degree of pleasantness or reward induced by an interaction or thought