Highlights

• • GLP-1R agonists promote weight loss through their effect on feeding behavior

• • An overall modulatory effect of these agents on reward behavior has been proposed

• • However, the evidence supporting this hypothesis in humans is very limited

• • Metabolic signals, such as GLP-1, are known to modulate the reward neural circuits

• • Better understanding of the brain–body connection can offer novel insights into behavior

GLP-1 receptor agonists and behavior

The most robustly documented effect of incretin-based treatments is a sustained reduction in body weight. Weight loss is achieved when energy expenditure exceeds energy intake, that is a state of negative energy balance.Reference Hill, Wyatt and Peters¹¹ Any biological, behavioral, or environmental factor that affects body weight necessarily acts through one or more components of energy balance.Reference Hill, Wyatt and Peters¹¹ In humans, GLP-1 receptor agonists have been associated with neutral to negative effects on energy expenditure, including in volitional physical activity, thus it is likely that a decrease in caloric intake is the main driver of the negative energy balance.Reference Shoemaker, Silver and Buchowski^8, Reference Maciel, Beserra and Oliveira¹² Studies using ad libitum (ieunrestricted) food intake assessments have documented a reduction of 30–40% in energy intake, compared to a placebo, in participants using an incretin-based treatment.Reference Friedrichsen, Breitschaft, Tadayon, Wizert and Skovgaard^5–Reference Shoemaker, Silver and Buchowski⁸ This reduction in food intake has been similarly reported in healthy individuals.Reference Jalleh, Pham and Marathe^13, Reference Pinelli, Jantz and Smith¹⁴ The distribution of nutrients (% energy consumed) was not significantly different, with reductions reported for all macronutrients (ie fat, carbohydrates, and protein),Reference Shoemaker, Silver and Buchowski⁸ although separate studies have reported a decreased subjective preference for high-fat foods.Reference Blundell, Finlayson and Axelsen¹⁵

Feeding behavior is complex and multidimensional. It has been parsed in multiple constituent elements, starting with hunger and the incentive salience of food (ieanticipation and valuation), to the consummatory, including the hedonic response, and termination (ie satiation) phases, all modulated by cognitive (eg learning and memory), and decision-making processes. The effects of incretin-based treatments have been primarily assessed by exploring the subjective impression of participants on their eating behavior, using standardized questionnaires and visual analogue scales. Multiple studies have reported a reduction in feelings of hunger, and increased fullness and satiety. These effects are independent of nausea, a common side effect of GLP-1R agonists.Reference Friedrichsen, Breitschaft, Tadayon, Wizert and Skovgaard^5–Reference Heise, DeVries and Urva^7, Reference Wharton, Batterham and Bhatta¹⁶ Additionally, decreased prospective food consumption (ie how much people think they could eat), and better control of eating, with fewer and less strong food cravings, and less difficulty resisting food, have been consistently documented.Reference Friedrichsen, Breitschaft, Tadayon, Wizert and Skovgaard^5–Reference Heise, DeVries and Urva^7, Reference Wharton, Batterham and Bhatta¹⁶ Furthermore, neuroimaging studies have reported decreased anticipatory responses, to food cues or consumption of highly palatable foods, in various brain regions (eg parietal and orbitofrontal cortex, insula, putamen, and amygdala).Reference Farr, Sofopoulos and Tsoukas^17, Reference van Bloemendaal, Veltman and Ten Kulve¹⁸ In contrast, the hedonic response to food seems to be preserved, with reports of no change in palatability or general food aversion.Reference Gibbons, Blundell, Tetens Hoff, Dahl, Bauer and Baekdal⁶ A recent neuroimaging study reported that treatment with liraglutide did not result in an altered hedonic experience or neural response while consuming a high-calorie food.Reference Coppin, Munoz Tord and Pool¹⁹

Accumulating preclinical evidence has indicated that the administration of a GLP-1 receptor agonist results in a reduction in the intake of alcohol and drugs of abuse.Reference Klausen, Thomsen, Wortwein and Fink-Jensen⁹ However, evidence from clinical studies, albeit very limited, is more mixed. A recent non-controlled study documented a lower self-reported intake of alcohol and decreased frequency of binge drinking episodes in obese individuals on semaglutide or tirzepatide.Reference Quddos, Hubshman and Tegge²⁰ Dulaglutide was reported to reduce alcohol intake in a secondary analysis of a randomized clinical trial (RCT) testing this GLP-1 receptor agonist as a therapy for smoking cessation.Reference Probst, Monnerat and Vogt²¹ The primary outcome of this RCT was negative, as dulaglutide was not more effective than a placebo at promoting abstinence from smoking.Reference Lengsfeld, Burkard and Meienberg²² In the only published RCT in patients with alcohol use disorders, exenatide was not superior to placebo at reducing the number of heavy drinking days and total alcohol intake, although an exploratory analysis suggested a significant effect of the treatment in a subgroup with comorbid obesity.Reference Klausen, Jensen and Moller²³ In patients with a cocaine use disorder, acute treatment with exenatide, when compared to placebo, did not change cocaine self-administration, self-reported euphoria, or wanting of cocaine.Reference Angarita, Matuskey and Pittman²⁴

Hitherto, 2 studies assessed the effects of liraglutide on reward behavior (ie the responses to positive stimuli) using objective tasks.Reference Hanssen, Kretschmer and Rigoux^25, Reference Hanssen, Rigoux and Kuzmanovic²⁶ Hanssen et al. (2021)Reference Hanssen, Kretschmer and Rigoux²⁵ evaluated how the drug affected the willingness to exert physical effort for food and monetary rewards, using a task whereby varying amounts of food and money could be earned by squeezing a handgrip device. The results indicated that liraglutide increased the motivation to work for both food and monetary rewards in insulin-resistant participants, restoring it to a similar level when compared to insulin-sensitive individuals. Hanssen et al. (2023)Reference Hanssen, Rigoux and Kuzmanovic²⁶ probed adaptive learning, through a paradigm that assessed the ability of participants to learn associations between auditory cues and subsequent visual outcomes, which shifted in predictability throughout the experiment. The key finding was that insulin-resistant participants exhibited a reduced adaptive learning rate (ie the extent to which participants learned from their errors), which was then normalized by a one-time liraglutide administration.

Therefore, the emerging picture is that incretin-based agents affect food intake, primarily through a modulation of the anticipation and valuation of foods, rather than a hedonic action. Evidence on consummatory behavior of non-foods is more uncertain, but with potential effects on motivation to work and learning reported, specifically in subgroups with metabolic dysfunction (ie obesity and/or insulin resistance). Whereas the connection between incretin function and food intake is more direct, if not necessarily straightforward, it is worth questioning further why and how would this system also affect the consumption of non-food rewards, which have vastly different physiological roles (or, in the case of drugs, are thought to work by “hijacking” the reward neurocircuitry, but still through targets that are not necessarily involved in the signaling of nutrient availability).

GLP-1 physiology and behavior

Incretins are multifaceted peptides. Endogenous GLP-1 acts both as a gut hormone and as a neuropeptide within the central nervous system (CNS)Reference Muller, Finan and Bloom²⁷. The central and peripheral GLP-1 systems are considered to be, at least partially, functionally separate.Reference Trapp and Brierley^28, Reference Brierley, Holt and Singh²⁹ GLP-1 originates from 2 separate locations, the enteroendocrine L cells of the intestine and the preproglucagon (PPG) neurons in the nucleus tractus solitaries (NTS).Reference Trapp and Brierley^28, Reference Williams³⁰ Gut-derived hormonal GLP-1 is synthesized and secreted after nutrient ingestion,Reference Nauck and Meier³¹ it stimulates pancreatic insulin secretion and biosynthesis in a glucose-dependent manner, in addition to having numerous regulatory effects.Reference Nauck and Meier³¹ Central GLP-1 secretion is stimulated by gastric distension, or endocrine factors, such as leptin and oxytocin.Reference Muller, Finan and Bloom²⁷ The central GLP-1 system is probably not directly activated by peripheral endogenous GLP-1, as PPG neurons do not express GLP-1Rs,Reference Brierley, Holt and Singh²⁹ and active GLP-1 in the circulation is rapidly metabolized to inactive by the enzyme dipeptidyl peptidase 4 (DPP-4); it is unlikely then that GLP-1 released by the intestine reaches the brain.Reference Holst and Deacon³² PPG neurons are the main source of brain GLP-1.Reference Holt, Richards and Cook³³ These are projecting neurons; their axons containing GLP-1 vesicles are present in varied regions of the brain. Neuronally produced GLP-1 is transported to the axon terminals and stored in synaptic vesicles until its eventual release.Reference Zheng, Stornetta, Agassandian and Rinaman³⁴ GLP-1Rs are widely expressed in the CNS and are thought to modulate, directly or indirectly, multiple regions. These include “homeostatic” feeding areas (eg brainstem–hypothalamic circuit), associative networks (eg parietal cortex) as well as regions relevant to motivation (eg mesolimbic pathway) and general cognitive function (eg hippocampus).Reference Farr, Sofopoulos and Tsoukas^17, Reference Alvarez, Martinez and Roncero^35–Reference Alhadeff, Rupprecht and Hayes³⁸

The effects of GLP-1 on food intake are considered to result from the activation of its receptors in the CNS.Reference Trapp and Brierley^28, Reference Williams^30, Reference Boer, Hay and Tups³⁹ Evidence indicates that the peripheral and central GLP-1 systems suppress food intake independently, through distinct behavioral mechanisms.Reference Brierley, Holt and Singh^29, Reference Williams³⁰ Postprandial release of intestinal GLP-1 is thought to convey a satiation signal, primarily mediated by vagal afferent neurons,Reference Brierley and de Lartigue^40, Reference Krieger⁴¹ thus reducing ad libitum energy intake.Reference Verdich, Flint and Gutzwiller⁴² In contrast, physiological central GLP-1 signaling does not seem to be involved in ad libitum feeding, glucose tolerance, or long-term energy balance, but is instead activated by different forms of metabolic and psychogenic stressors, including unusually large meals and prolonged fasting, as well as acute stress, modulating stress-induced hypophagia.Reference Brierley, Holt and Singh^29, Reference Holt, Richards and Cook³³ Systemically administered GLP-1R agonists are degradation-resistant and long-acting, and are thus considered to mimic the action of postprandial gut-derived GLP-1. These agents additionally access GLP-1Rs within the CNS, although the extent to which they access the areas protected by the blood–brain barrier (BBB) is still unclear.Reference Salameh, Rhea, Talbot and Banks^43–Reference Gabery, Salinas and Paulsen⁴⁵ Therefore, it is unclear if GLP-1R agonists are also correlated with central GLP-1 actions. Preclinical evidence indicates that liraglutide and semaglutide do not activate PPG neurons, nor require them for their feeding suppression effects.Reference Brierley, Holt and Singh²⁹ Interestingly, concurrent activation of PPG neurons was shown to augment semaglutide’s effects on eating behavior, indicating that brain-specific GLP-1 action might not be fully reflected by existing GLP-1R agonists, and could be additional pharmacological targets.Reference Brierley, Holt and Singh²⁹

A reasonable synthesis of the extant literature is that peripheral GLP-1 is involved in day-to-day energy balance, serving as a meal termination signal. Elevating the tone of peripheral GLP-1 signal with long-acting and degradation-resistant GLP-1R agonists would then promote feelings of satiety, facilitating the cessation of feeding (improving subjective feelings of control), and decreasing the subjective value of food (reducing the anticipation and valuation of food rewards), which are all consistent with the reported behavioral effects of incretin-based treatments. On the other hand, central GLP-1 is more context-dependent and is activated by stressful situations, whereby the termination of feeding behavior is more urgent. Based on the reported inputs to NTS PPG neurons, these can be related to somatic signals (e.g. excess gastric distention), but also to metabolic stress and perceived imminent threats.Reference Petrovich⁴⁶ Conceivably, these could also be disease states and/or intermediate phenotypes, such as the excess accumulation of adipose tissue or insulin resistance. Evidence indicates that central GLP-1 signaling can promote feelings of satiety even in conditions of negative energy balance,Reference Brierley, Holt and Singh²⁹ suggesting that it is activated in situations where responding to a threat, internal or external, takes priority over replenishing energy resources.Reference Petrovich⁴⁶ Nonetheless, despite interest in the role of the GLP-1 system in the overall response to stress,Reference Guerrero-Hreins, Goldstone, Brown and Sumithran⁴⁷ the behavioral correlates, in humans, of these actions have not been directly explored. If, and to what extent, the currently documented behavioral effects of GLP-1R agonists in humans are attributable to central versus peripheral GLP-1 activation is still to be determined.

How could then the GLP-1 system affect the intake of non-foods, as well as overall reward behavior? One of the underlying concepts is that responses to food and non-food rewards converge on a common neural network, the mesolimbic pathway.Reference Volkow, Wang, Tomasi and Baler⁴⁸ Mesolimbic regions, such as the ventral tegmental area and the nucleus accumbens, express GLP-1R and receive projections from NTS PPG neurons.Reference Muller, Finan and Bloom²⁷ Activation of GLP-1R in the mesolimbic pathways has been shown to modulate dopaminergic neurotransmission, one of the key molecular mediators of reward response, in preclinical models.Reference Klausen, Thomsen, Wortwein and Fink-Jensen⁹ Conversely, studies have reported that the effects of GLP-1 on alcohol and drug intake in animals are mediated by central rather than peripheral mechanisms.Reference Klausen, Thomsen, Wortwein and Fink-Jensen⁹ Notably, the effects of GLP-1 on alcohol intake seem to be related to modulation of its rewarding/reinforcing properties, rather than a byproduct of GLP-1 overall influence on nutrient and fluid intake, as a study showed that a GLP-1R agonist was able to attenuate intravenous ethanol self-administration in mice.Reference Sorensen, Caine and Thomsen⁴⁹

Nonetheless, findings from human studies have been mixed. Two studies failed to document an effect of GLP-1R agonists on an indicator of dopamine function in vivo, the availability of striatal dopamine transporter (DAT), measured using molecular neuroimaging (e.g. positron emission tomography); Jensen et al. (2020)Reference Jensen, Galli and Thomsen⁵⁰ following the acute administration of exenatide in healthy volunteers and Athauda et al. (2017)Reference Athauda, Maclagan and Skene⁵¹ following 48 weeks of exenatide treatment in patients with Parkinson’s disease, despite improvement in motor symptoms. In contrast, Klausen et al. (2022)Reference Klausen, Jensen and Moller²³ reported a decrease in striatal DAT availability after 26 weeks of treatment with exenatide in patients with alcohol use disorder. These discrepant findings might be a result of methodological limitations from the aforementioned studies, including relatively small sample sizes. It is also worth considering the possibility that these might reflect the distinct roles of the central and peripheral GLP-1 system. The more context-dependent nature of central GLP-1 signaling indicates that its activation, which is likely necessary for the potential broader effects on reward behavior, might only be meaningful in specific situations, determined by certain physiological and/or environmental conditions. Indeed, reinforcing this hypothesis, the effects of GLP-1R agonists on non-food reward in humans, including alcohol intake and response to monetary rewards, have hitherto only been shown in individuals with obesity and/or insulin resistance.Reference Klausen, Jensen and Moller^23, Reference Hanssen, Kretschmer and Rigoux^25, Reference Hanssen, Rigoux and Kuzmanovic²⁶

An additional explanation for the potential role of the GLP-1 system on non-food reward might be in the non-specific properties of GLP-1 signaling. Activation of GLP-1 receptors has been consistently associated with neuroprotective effects in preclinical models, including preventing or reversing the effects of a range of toxic conditions on neuronal survival, by decreasing apoptosis and increasing neurogenesis, angiogenesis, and cerebral blood flow.Reference Erbil, Eren, Demirel, Kucuker, Solaroglu and Eser^52–Reference Kong, Wu and Dai⁵⁴ Some of these actions seem to be mediated by the modulation of oxidative stress and inflammatory processes; recent work documented that central GLP-1 signaling was required for the anti-inflammatory effects of GLP-1R agonists.Reference Wong, McLean and Baggio⁵⁵ These actions have substantiated the hypothesis that GLP-1R agonists may also have pro-cognitive effects, which has been tested in diverse clinical populations.Reference Chai, Liu, Yu, Yang and Sun^56–Reference Mansur, Ahmed and Cha⁵⁹ Conversely, the GLP-1 system has been implicated in the regulation of cerebral glucose metabolism. The human brain is highly dependent on glucose as its primary substrate and is considered to be particularly vulnerable to fluctuations in glucose supply and consumption. Evidence from human studies indicates that GLP-1 modulates the transport of glucose across the blood–brain barrier and the cerebral metabolic rate of glucose, in a glucose-dependent manner.Reference Nilsson, Gjedde, Brock, Gejl and Rungby⁶⁰ In a 6-months RCT testing the use of liraglutide as a treatment for Alzheimer’s disease (AD), the GLP-1R agonist was shown to prevent the decline of cerebral glucose metabolism, a pathological feature of AD’s progression, although the trial failed to document treatment effects on its primary outcome, cognitive function.Reference Gejl, Gjedde and Egefjord⁶¹

In an attempt to reconcile these broad and relatively disparate functions attributed to the GLP-1 system, it should be considered that GLP-1, and the concerted effects of physiological or pharmacological GLP-1 activation, can also function as a signal of energy availability. Extensive literature indicates that energy substrates (ie glucose) and its accompanying network of regulatory peptides, including, but not limited to, insulin, leptin, and ghrelin, affect human behavior and decision-making, beyond food-related rewards.Reference Orquin and Kurzban^62–Reference Schlogl, Janssen and Fasshauer⁶⁸ Importantly, in manipulation studies, the effects of glucose on motivated behavior seem to be a direct result of glucose ingestion, with the consequential elevation in blood glucose and counterregulatory hormonal response, rather than a hedonic or perceptual process, as artificial sweeteners or glucose mouth rinses, had either none or opposite effects on decision making.Reference Wang, Reed, Baugh and Fercho^69, Reference Wang and Dvorak⁷⁰ Indeed, the mesolimbic pathway is known to be modulated by a variety of homeostatic signals.Reference Hsu, McCutcheon and Roitman⁷¹ Replicated evidence indicates that oral glucose administration stimulates dopaminergic activity in the striatum.Reference Blum, Thanos and Gold⁷² Furthermore, mesolimbic dopamine neurons also express receptors for insulin,Reference Liu and Borgland⁷³ leptin,Reference Fernandes, Sharma, Hryhorczuk, Auguste and Fulton⁷⁴ and ghrelin.Reference Perello and Dickson⁷⁵

The precise mechanisms whereby energy availability and its signals affect reward response remain unclear, but it has been suggested that more demanding behaviors and actions are thought to require greater energetic resources, and thus higher glucose availability and utilization.Reference Kennedy and Scholey^76, Reference Scholey, Harper and Kennedy⁷⁷ Circulating glucose and regulatory hormones also have a signaling function, as indicators of the body’s “energy budget.” Furthermore, recent work has highlighted the role of dopaminergic neurotransmission in the mesolimbic pathway in cost–benefit decision-making. Every action has an energetic cost, actual and/or opportunity. Dopamine signaling has traditionally been implicated in arousal, motivation, and psychomotor activation (that is incentive salience models), and reward learning/reinforcement (that is prediction error models).Reference Beierholm, Guitart-Masip and Economides^78–Reference Hamid, Pettibone and Mabrouk⁸⁰ Recent work has suggested that dopamine can potentially signal relative value (that is the available reward for a certain cost of effort), thus integrating cost and benefit factors, which can then be employed for both learning and motivational functions.Reference Hamid, Pettibone and Mabrouk^80–Reference Soutschek, Jetter and Tobler⁸² Conceptually, homeostatic input to striatal dopamine neurons can then serve as signals of the current and anticipated energy resources. Therefore, even complex behaviors can potentially be affected by each individual’s internal energy milieu, and their perception of their energy environment.Reference Wang^83, Reference Beeler and Mourra⁸⁴