GLP-1 and the Neurobiology of Eating Control: Recent Advances

Abstract

Obesity is now considered a chronic relapsing progressive disease, associated with increased all-cause mortality that scales with body weight, affecting more than 1 billion people worldwide. Excess body fat is strongly associated with excess energy intake, and most successful anti-obesity medications (AOMs) counter this positive energy balance through the suppression of eating to drive weight loss. Historically, AOMs have been characterized by modest weight loss and side effects which are compliance-limiting, and in some cases life-threatening. However, the field of obesity pharmacotherapy has now entered a new era of AOMs based on analogues of the gut hormone and neuropeptide glucagon-like peptide-1 (GLP-1). The latest versions of these drugs elicit unprecedented levels of weight loss in clinical trials, which are now starting to be substantiated in real-world usage. Notably, these drugs reduce weight primarily by reducing energy intake, via activation of the GLP-1 receptor on multiple sites of action primarily in the central nervous system, although the most relevant sites of action, and the neural circuits recruited remain contentious. Here we provide a targeted synthesis of recent developments in the field of GLP-1 neurobiology, highlighting studies which have advanced our understanding of how GLP-1 signaling modulates eating, and identify open questions and future challenges we believe still need to be addressed to aid the prevention and/or treatment of obesity.

Obesity, characterized clinically by a body mass index (BMI) of > 30 kg/m², is widely considered to be a chronic relapsing progressive disease (1, 2), associated with increased all-cause mortality which scales dramatically with increased body weight (3). More than two-thirds of adults in the United Kingdom and the United States are overweight (BMI > 25 kg/m²), with 27% and 42% of the UK and US populations now living with clinically defined obesity (4, 5). Increasing “Westernization” of diets and lifestyles across the globe has been accompanied by a transition from underweight to obesity as the most prevalent form of malnutrition in almost all countries, with more than 1 billion people now living with obesity due to the worldwide prevalence doubling in men, tripling in women, and quadrupling in children over the last 30 years (6, 7).

The cause of obesity is the subject of intense debate, with numerous theories and models proposed (8), although the diversity of genetic risk factors and individual responses to the obesogenic environment suggests that obesity is a heterogenous disease resulting from the interaction of multiple intrinsic and extrinsic influences (9‐11). There is reasonable consensus that excess body fat is primarily associated with excess energy intake, rather than insufficient energy expenditure. However, even at this most basic level of explanation, proponents of the 2 most prevalent hypotheses—the energy balance model and carbohydrate-insulin model—dispute the direction of causality between food overconsumption and weight gain (12).

Anti-obesity medications (AOMs) attempt to counter this positive energy imbalance via various pharmacological strategies to achieve reductions in energy intake—by decreasing food consumption or nutrient absorption, and/or by increased energy expenditure. While many compounds have been utilized for this indication, historically these have been characterized by modest weight loss (typically ≤ 5%) and side effects which are compliance-limiting, and in some cases life-threatening (13).

However, after decades of modest results, the field has now entered a new era of AOMs based on analogues of glucagon-like peptide 1 (GLP-1), that elicit unprecedented levels of weight loss in clinical trials, which are now starting to be substantiated in real-world usage (14). Notably, this drug class reduces weight primarily by reducing energy intake, via activation of the cognate GLP-1 receptor (GLP-1R), mainly on sites of action in the central nervous system (CNS) (15, 16). The basic biology of endogenous GLP-1, and the history and current pipeline of exogenous GLP-1-based drugs, have been the subject of several comprehensive recent reviews (14, 15, 17, 18). Here we provide a targeted synthesis of what we consider some of the most important recent developments in the fast-moving field of GLP-1 neurobiology, specifically relevant to obesity. In particular, we highlight studies which have advanced our understanding of how GLP-1 signaling modulates eating, and we identify what we consider important open questions and future challenges to be addressed surrounding GLP-1-based AOMs to aid the prevention and/or treatment of obesity.

GLP-1 Systems and the Neurobiology of Obesity

Endogenous GLP-1 in the Physiological Control of Eating

Endogenous GLP-1 can be classed as a gut hormone and/or neuropeptide, based on its 2 principal sources—enteroendocrine “L-cells” in the intestine, and preproglucagon (PPG) neurons in the caudal brainstem (18‐21). These 2 sources, and the populations of GLP-1R-expressing cells which are accessible to them, are considered by us and others to respectively define the peripheral and central GLP-1 systems (16, 22, 23), and recent insights into their respective roles in eating are discussed below.

Endogenous peripheral GLP-1

Gut-derived (peripheral) GLP-1 is best known as an incretin hormone, released postprandially from L-cells and acting on pancreatic β-cell GLP-1R to increase glucose-dependent insulin secretion—the action underlying the therapeutic value of potentiating GLP-1 signaling for the treatment of type 2 diabetes (24, 25). Conversely, the role of gut-derived GLP-1 in the physiological regulation of eating, its site(s) of action, and the neural circuitry recruited, have been more challenging to clarify (23). Early human studies, using intravenously infused physiological concentrations of the active form of the GLP-1 peptide (7-36 amide; see (14) for details of post-translational processing of glucagon gene products), reported reduced appetite and food intake (26, 27). However, there is conflicting evidence from mechanistic studies manipulating endogenous peripheral GLP-1 signaling in rodents, with reports of impaired glucose handling, but no effect on cumulative food intake or body weight, following deletion of the glucagon gene from L-cells (28, 29). Similarly, global GLP-1R knockout mice did not show altered food intake or body weight (30, 31), consistent with results from mice with GLP-1R knockout targeted to discrete domains of peripheral, central and enteric nervous systems (32). Conversely, chemogenetic activation of colonic L-cells (33), or secretagogue-induced increases of endogenous GLP-1 levels detectable in the hepatic portal vein (HPV) (34), both improved glucose handling and decreased short-term food intake. This latter finding is consistent with reports that intra-HPV infusion of a supraphysiological dose of GLP-1 decreased meal size, and that this satiating effect was blocked by antagonism of GLP-1R in the brainstem area postrema (AP) (35). The same study reported that AP GLP-1R antagonism alone did not affect meal size, and that endogenous postprandial GLP-1 increases were detectable in HPV but not systemic circulation, consistent with observations that the vast majority of active GLP-1 is degraded by the enzyme dipeptidyl peptidase (DPP)-IV by the time it exits the HPV (14). Studies combining anatomical tracing with single cell transcriptomics suggest that a substantial proportion of GLP-1R-expressing vagal afferent neurons (VANs) innervate the HPV, but relatively few innervate intestinal villi (36‐38). Importantly, knockdown of GLP-1R in VANs increased meal size (39), and abolished the satiating effects of secretagogue-induced GLP-1 release (34). These studies, and others using less specific knockdown of peripheral nervous system GLP-1R (32, 40), suggest that under physiological conditions, endogenous gut-derived GLP-1 plays a role in satiation in rodents and humans, but compensatory mechanisms and/or existence of parallel satiation circuits prevent expression of overt body weight phenotypes following loss-of-function manipulations. This satiating action of gut-derived GLP-1 seemingly occurs primarily at peripheral sites on VAN terminals in the HPV and/or intestinal mucosa, rather than directly on GLP-1R expressing CNS neurons, at least under “normal” physiological conditions. Thus, gut-derived GLP-1 most likely contributes to the process of satiation by acting peripherally as an interoceptive input to central eating control networks (41), and/or by more directly slowing gastric emptying via vago-vagal circuitry (42).

The density of L-cells increases toward the distal end of the gastrointestinal tract, peaking in the ileum and colon—a paradoxical observation given that nutrient-sensing cell density would be expected to be highest in the proximal intestine where nutrient absorption primarily takes place (25, 33). However, while the proximal L-cell population likely mediates meal-induced GLP-1 release via direct nutrient sensing and/or under cephalic phase neuronal control (23), the distal population may primarily mediate the “ileal brake.” This visceral alarm reflex potently inhibits eating and gastrointestinal motility upon detection of malabsorbed nutrients or bile salts in the distal intestine, which can occur after rapidly consumed large meals or conditions of potentiated gastric emptying (14, 43). A recent study by Zhang et al (44) elegantly delineated the neural circuit basis for this ileal brake, mapping a GLP-1-dependent inter-organ circuit from ileal L-cells to the brain. This intestinofugal pathway connects GLP-1R-expressing enteric neurons to stomach-innervating sympathetic neurons which inhibit gastric emptying and eating by a potent nitric oxide neuron-mediated increase in gastric distension. Interestingly, rather than acting via the canonical mechanosensory VAN pathway (23), ileal GLP-1-evoked gastric distension suppressed eating via a spinal afferent pathway to the parasubthalamic nucleus and a brainstem premotor circuit controlling orofacial food rejection behavior. This study, along with pioneering applications of systems neuroscience approaches in mice to investigate vagal gut-brain GLP-1R circuitry (36, 38), demonstrates that endogenous GLP-1 released from L-cells can suppress eating via discrete vagal and spinal gut-brain pathways and CNS circuits, depending on the size and rate of the meal consumed. The extent to which these distinct physiological pathways are involved in pathophysiological conditions involving dysregulated eating or gut motility, and their relevance as targets for current or future AOMs, are translationally important open questions.

While the GLP-1R populations on which gut-derived GLP-1 act to suppress eating appear to be predominantly located in the periphery, GLP-1R-expressing neurons are highly abundant in the AP and hypothalamic median eminence, along with other circumventricular organs, which are all theoretically accessible to endogenous GLP-1 in systemic circulation (16, 45). This abundance suggests a physiological role for these GLP-1Rs, whereby under circumstances similar to those which may trigger the ileal brake, peripheral GLP-1 reaches high enough systemic concentrations to have a “supraphysiological” anorectic action via these CNS targets which lie outside of the blood-brain barrier, most plausibly the AP and/or median eminence (14, 46) (Fig. 1A). Whether this direct CNS action of peripheral endogenous GLP-1 suppresses eating due to nonaversive satiation and/or nausea, if such effects can be functionally dissociated at the level of GLP-1R populations in discrete circumventricular organs (or subpopulations within them), and to what extent exogenous GLP-1-based AOMs act via such (sub)populations, are crucial open questions which are explored further in the section “Are the Side Effects of GLP-1-Based AOMs an Obligatory Feature or a Dissociable Bug?”

Figure 1.

Gut-brain pathways and sites of action of endogenous and exogenous GLP-1 signaling. A, Schematic overview of the neuronal and hormonal/pharmacological pathways by which endogenous GLP-1 and GLP-1-based AOMs convey anorectic signals to the brain. Under physiological conditions, proximal L-cells in the upper small intestine release GLP-1 in response to ingested nutrients, signaling via paracrine and endocrine routes (blue) to act on GLP-1 receptors on chemosensory vagal afferent in intestinal villi and the hepatic portal vein (HPV), and on gut-innervating mechanosensory vagal afferents. Under supraphysiological conditions, for example, after exceptionally large and/or nutrient-dense meals, GLP-1 from L-cells may also reach sufficiently high systemic concentrations to act directly on the brain, via neuronal GLP-1 receptors in circumventricular organs such as the area postrema. Undigested nutrients reaching the ileum can trigger GLP-1 release from distal L-cells to trigger the multisynaptic gut-brain circuit (orange) comprising the ileal brake. Peripherally administered GLP-1-based AOMs reach steady-state systemic concentrations orders of magnitude higher than endogenous GLP-1, reaching and acting primarily on circumventricular organs in the CNS (magenta). AOMs also presumably bind all peripheral sites of action of endogenous GLP-1, but these pathways may be dispensable for AOM effects on eating and body weight. B, Recent mechanistic studies, primarily conducted in rodents, have identified a number of sites of direct action putatively mediating the effects of GLP-1-based AOMs on eating and body weight, including the area postrema and nucleus tractus solitarius (47‐49), locus coeruleus (50), hypothalamic arcuate (51, 52) and dorsomedial nuclei (53), and the lateral septal nucleus (54). (Created in BioRender. Brierley, D. (2024) BioRender.com/d09h242).

Endogenous central GLP-1

In addition to the peripheral source of GLP-1, this peptide is also produced and released from a small population of PPG neurons in the brainstem nucleus tractus solitarius (NTS) and adjacent intermediate reticular nucleus (19‐21), along with a small microcircuit population in the olfactory bulb (16, 55, 56). An extensive body of evidence from pharmacological studies using direct CNS injections of GLP-1R ligands supports roles for central GLP-1 signaling in eating behaviors and energy balance (reviewed in detail in (15, 16)). Recent studies using transgenic mice demonstrated that NTS PPG neurons are the primary source of endogenous central GLP-1, they project to almost all brains regions where GLP-1R are expressed, and they have roles in physiological satiation and stress-induced hypophagia (22, 45, 57‐60). NTS PPG neurons receive synaptic input from VANs and AP neurons; however, the majority of these input neurons do not express GLP-1R, and PPG NTS neurons are not activated by or required for the anorectic effects of liraglutide or semaglutide, supporting an anatomical and functional separation between peripheral and central GLP-1 systems (16, 22, 23). The principal vagal input to PPGs to drive satiation signaling likely comes from oxytocin receptor-expressing VANs, a mechanosensory vagal population largely distinct from GLP-1R-expressing VANs, which preferentially innervate the duodenum rather than stomach, consistent with the role of PPGs in mediating large meal-induced satiation (22, 23, 36, 57). This role for NTS PPG neurons in physiological satiation was recently confirmed in an elegant study by Ly et al (61), which overcame the technical challenge of performing calcium imaging of neuronal dynamics in the caudal brainstem of awake-behaving mice. This study convincingly demonstrated that distinct NTS populations of prolactin-releasing hormone (PRLH) and PPG neurons are sequentially activated during discrete phases of meal ingestion, to respectively regulate the pace of ingestion, and terminate meals upon receipt of mechanosensory feedback from the gut.

The widespread projection pattern of PPGs is largely mirrored by the distribution of GLP-1R-expressing neurons, found in virtually all CNS regions implicated in the control of eating and energy balance (20, 45). There is thus some question over whether discrete subpopulations of PPGs project to distinct GLP-1R-expressing populations to mediate different aspects of metabolic function, or if most/all PPGs project to all regions to provide coordinated modulation of multiple circuits (16, 62).

Central GLP-1 System Targets of Endogenous GLP-1 and GLP-1RA Anti-Obesity Medications

Local injection of GLP-1 and/or GLP-1R agonists (GLP-1RAs) elicits anorectic effects in virtually all GLP-1R-expressing CNS regions investigated (reviewed in (15, 16)). This raises a similar question of what the physiological roles of endogenous GLP-1 signaling are in these CNS regions, and how relevant and/or uniquely essential they are for the effects of GLP-1RA AOMs. The site(s) of action and neural circuits recruited by GLP-1RA AOMs for their therapeutic effects is an open and contentious question. The field has, to some extent, been muddied by a long history of using these drugs as tool compounds, which are applied (ex vivo) or injected (in vivo) directly into specific GLP-1R-expressing regions in the CNS to infer their physiological roles and/or necessity for therapeutic effects of peripherally administered AOMs. However, several recent studies have compared the CNS distribution of peripherally administered fluorescently tagged GLP-1R ligands, GLP-1RA-induced cFos expression, and/or labeled GLP-1R neurons. These studies have demonstrated that the principal sites of action accessible to these drugs are GLP-1R neurons in circumventricular organs, particularly the AP and ME, and those in immediately adjacent regions—the NTS and arcuate nucleus of the hypothalamus (16, 51, 63‐66). Pertinently, the majority of these studies used acute doses of GLP-1RAs, but after 5 days of dosing, fluorescently labeled semaglutide was additionally reported in some GLP-1R-expressing regions protected by the blood-brain barrier, including other hypothalamic nuclei and regions adjacent to the ventricles, including parts of the lateral septum (63). This apparently slower or cumulative access could reflect extravasation and diffusion from AP and/or median eminence to nearby regions, and/or facilitated transport by GLP-1R-expressing tanycytes found in these circumventricular organs (52, 63). Thus, some GLP-1R neuron populations in the CNS are seemingly only accessible to endogenous PPG-derived GLP-1 and are thus unlikely to be relevant to the mechanism of action of peripherally administered GLP-1RAs, given that these drugs do not recruit PPGs (which themselves do not express GLP-1R) and so could not be indirectly acting via this relay (22, 67). GLP-1R-expressing neuron populations in the AP, and at least some of the NTS, are clearly activated by GLP-1RAs and appear to mediate at least some of the effects of these drugs on satiation, satiety, and nausea/malaise (47‐49), likely by recruiting circuits which are endogenous targets of (supraphysiological) gut-derived GLP-1. These populations can be considered separate from the PPG-defined central GLP-1 system, demonstrated by their ability to be recruited in parallel for additive suppression of eating (22, 68). Finally, a subset of GLP-1R neuron populations behind the blood-brain barrier likely represents common targets, which are both accessible to peripherally administered GLP-1RAs (probably only upon chronic dosing) and to endogenous GLP-1 released from PPGs via synapses and/or cerebrospinal fluid–mediated volume transmission (15, 16, 69). Fully characterizing which GLP-1R regions are points of exogenous and endogenous GLP-1 signal convergence is a translationally important question, with relevance to the potential of stimulating endogenous central GLP-1 release as an adjunct to improve the efficacy of GLP-1RA AOMs, demonstrated by recent proof-of-concept studies (22, 68). While the hypothalamic arcuate nucleus is the convergence region most investigated to date (20, 51, 52), several recent studies have provided evidence for roles in physiological and GLP-1RA AOM-mediated eating control for GLP-1R populations in the locus coeruleus (50), dorsomedial hypothalamus (53), and lateral septum (54) (Fig. 1B). The individual necessity of distinct GLP-1R neuron populations for the anorectic and weight loss effects of GLP-1RAs has now been claimed by multiple studies using various loss-of-function approaches in rodents (47, 50‐54, 69). Reconciling these apparent discrepancies by identifying which populations are truly necessary for the effects of the GLP-1RAs now in clinical use as AOMs, and which populations mediate nonaversive satiation and/or satiety vs undesirable side effects such as nausea-induced anorexia, are crucial challenges for the field in the years ahead.

Open Questions and Future Challenges in the New Era of GLP-1-Based Anti-Obesity Medications

The use of preclinical systems neuroscience approaches to study GLP-1 neurobiology continues to provide invaluable insights into how this peptide modulates eating, and how GLP-1RA-based AOMs interact with an array of neural circuits to lower food intake and body weight. These studies have arguably also raised as many questions as they have answered, and in this section we explore some of those we consider most pertinent to the ongoing efforts to address the obesity epidemic.

Do GLP-1-Based AOMs Need to Be Any More Effective, and How Can This Be Achieved?

The ideal AOM should be clinically efficacious for weight loss and obesity comorbidities, broadly tolerated, and safe for chronic usage. Prior to the 2014 approval of 3-mg liraglutide for this indication (70), most AOMs barely achieved the ≥ 5% threshold for clinically meaningful weight loss (71‐73). However, the potent gastrointestinal side effects and requirement for daily injection also limited the tolerability for this regimen, such that typical weight loss efficacy was not substantially improved over existing non-GLP-1RA AOMs (13). As such, the gold standard in weight loss therapy has long been bariatric surgery, which routinely achieves 25% to 30% loss (71). While this first iteration of GLP-1RA AOMs successfully established the principle of their utility and safety (17), the potential of this drug class for obesity treatment only became evident with the development of the long-acting GLP-1RA semaglutide, licensed for obesity in 2021 (74, 75). This longer-acting GLP-1RA pharmacotherapy is one of several fatty acid acylated analogues of human GLP-1, which work primarily by lowering energy intake through increased satiation and satiety (76). Semaglutide represented a step-change in the efficacy of AOMs, achieving previously unseen AOM-mediated weight-loss effects of up to 15% (77‐79), thus approaching the gold standard of bariatric surgery. More recently, GLP-1R-based multi-agonists, such as tirzepatide (co-agonist of GLP-1R and glucose-dependent insulinotropic polypeptide receptor [GIPR]), and retatrutide (tri-agonist of GLP-1R, GIPR, and glucagon receptor), are achieving unprecedented placebo-subtracted weight loss of ≥ 20% in clinical trials (80, 81). Other combinations, such as semaglutide with the amylin analogue cagrilintide (CagriSema), are showing great promise, achieving ∼15% body weight loss in shorter-term phase 1 and 2 clinical trials (82, 83). These headline data from GLP-1-based multi-agonist/combination therapies beg the question—has the quest for bariatric surgery–level efficacy of obesity pharmacotherapy now been achieved?

In answering this, it is important to consider that these GLP-1RAs have only been in use for obesity treatment in real-world settings for a limited time, with most available efficacy data currently derived from clinical trials, which tend to overestimate weight loss effects (84). To counteract this, recent GLP-1-based AOM trials, such as STEP 5 and SURMOUNT 3 (77, 81), implemented intention-to-treat methods such as co-primary endpoints (≥ 5% baseline weight loss outcome, and significant placebo-subtracted weight loss) and treatment-focused estimands from all randomized patients, providing average weight loss estimates which better reflect real-world compliance (85). Such real-world studies of the most commonly prescribed AOM, semaglutide, provide conflicting evidence for effectiveness (Table 1), with studies reporting only 4% to 5% weight loss (96). Conversely, the first multicenter real-world cohort study reported an average 13.4% loss (97), comparable to the STEP 1 clinical trial data (79), implying a variable response to semaglutide in different trial and real-world settings. Furthermore, the 4-year SELECT clinical trial—the longest AOM study of semaglutide to date—reported 8.7% placebo-subtracted weight loss, markedly lower than 12.5% in STEP 1 at 68 weeks (79, 95). This substantial difference could be explained by differences in patients’ motivation to maintain diet and lifestyle changes between a weight-loss trial (STEP 1) vs cardiovascular health trial (SELECT) with weight loss as a secondary outcome (13). Taken together, these results indicate that despite these efforts to improve accuracy of clinical trial data for this indication, real-world effectiveness will likely still be lower due to lower tolerance and adherence. Key drivers identified for reduced adherence to GLP-1RAs are decreasing BMI class (95), and common adverse effects of nausea, emesis, and diarrhea, which affect ∼60% of patients (17). Looking beyond headline weight loss results, substantial response variability exists within patient populations, with the  STEP and SURMOUNT trials reporting that 10% to 15% of individuals lose <5% on semaglutide or tirzepatide, respectively, and are thus classed as “nonresponders” (77, 79, 80, 94, 100). More dramatically, 32% of SELECT participants were nonresponders (95), and in a real-world setting 19% of patients did not achieve 5% loss after 12 months on semaglutide (97). It can thus be inferred that for up to a third of people deemed overweight or obese, current GLP-1RAs may not enable clinically meaningful weight loss. A likely contributing factor for this is the heterogeneity of obesity, with previous studies demonstrating that tailoring obesity treatments to obesity phenotypes significantly improves weight-loss outcomes (101). Notably, GLP-1RA efficacy is lessened in participants with a lower starting weight/BMI or with prior exposure to GLP-1RAs (95, 102), and it is reduced by a third in type 2 diabetics and in male individuals (79, 93, 95‐97, 99). Further supporting a need for precision medicine approaches for people living with obesity, some patients may have one of the loss-of-function mutations identified in the human GLP-1R gene, which can differentially affect cell surface receptor expression or signal transduction efficacy and may thus be predictive of treatment response or the necessity of co-agonist or allosteric modulator approaches (103).

Table 1.

Summary of key body weight loss efficacy in trials vs real-world of GLP-1RAs licensed for obesity

Anti-obesity medication	Study	Dose + admin. route	% BW loss (placebo-subtracted)	Timepoint	Additional key study information	Reference
	SCALE Maintenance	3.0 mg Daily SC injection	−6.2 (−6.1)	56 weeks	≥30 kg/m2 or ≥27 kg/m2 with comorbidities No DM + low-calorie diet ITT	Wadden et al, 2013 (86)
Liraglutide (GLP-1RA)	SCALE	3.0 mg Daily SC injection	−7.9 (−5.3)	56 weeks	BMI ≥ 27 kg/m2 + dyslipidemia or hypertension or BMI ≥ 30 kg/m2 No T2D ITT	Pi-Sunyer et al, 2015 (70)
	SCALE Insulin	3.0 mg Daily SC injection	−5.8 (−4.3)	56 weeks	BMI ≥ 27 kg/m2 T2D + basal insulin and ≤2 oral antidiabetic drugs + intensive behavioral therapy ITT	Garvey et al, 2020 (87)
	SCALE IBT	3.0 mg Daily SC injection	−7.5 (−3.4)	56 weeks	BMI ≥ 30 kg/m2 No DM ITT	Wadden et al, 2020 (88)
	Real-world	3.0 mg Daily SC injection	−7.1	26 weeks	BMI ≥ 30 kg/m2  or BMI ≥ 27 kg/m2 + ≥ one weight-related comorbidity (inc. T2D)	Wharton et al, 2019 (89)
	Real-world	3.0 mg Daily SC injection	−5.9	26 weeks	BMI ≥ 25 kg/m2 + ≥ one weight-related disease (inc. DM)	Park et al, 2021 (90)
	Real-world	3.0 mg Daily SC injection	−7.1	≥ 52 weeks	BMI ≥ 30 kg/m2  or a BMI of ≥27 mg/m2 + ≥ 1 treated weight-related comorbidity No T2D	Haase et al, 2021 (91)
	Real-world	3.0 mg Daily SC injection	−12.4	44 weeks	BMI ≥ 35 kg/m2, or BMI ≥ 28 kg/m2 + ≥ 1 metabolic comorbidity No T2D	Santini et al, 2023 (92)
Semaglutide (GLP-1RA)	STEP-1	2.4 mg Once-weekly SC injection	−14.9 (−12.5)	68 weeks	BMI ≥ 30 kg/m2 or ≥27 kg/m2 + ≥1 weight-related coexisting condition No DM ITT	Wilding et al, 2021 (79)
	STEP-2	2.4 mg Once-weekly SC injection	−9.6 (−6.2)	68 weeks	BMI ≥ 27 kg/m2 T2D ITT	Davies et al, 2021 (93)
	STEP-3	2.4 mg Once-weekly SC injection	−16.0 (−10.3)	68 weeks	BMI ≥ 27 kg/m2 + ≥ 1 comorbidity or BMI ≥ 30 kg/m2 No DM + intensive behavioral therapy + a low-calorie diet for initial 8 weeks	Wadden et al, 2021 (94)
	STEP 5	2.4 mg Once-weekly SC injection	−15.2 (−12.6)	104 weeks	BMI ≥ 30.0 kg/m2 or ≥27.0 kg/m2 + ≥ weight-related comorbidity (treated or untreated) No DM ITT	Garvey et al, 2022 (77)
	SELECT	2.4 mg Once-weekly SC injection	−10.2 (−8.7)	208 weeks	BMI of ≥27 kg/m2 and established CVD No DM ITT	Ryan et al, 2024 (95)
	Real-world	0.25-2 mg Once-weekly SC injection	−4.4	≥ 26 weeks	Inc. DM	Powell et al, 2023 (96)
	Real-world	0.25-2.4 mg Once-weekly SC injection	−13.4	52 weeks	BMI ≥ 27 kg/m2 Inc. DM	Ghusn et al, 2024 (97)
Tirzepatide (GIPR/GLP-1RA)	SURMOUNT-1	15 mg Once-weekly SC injection	−20.9 (−17.8)	72 weeks	BMI ≥ 30 kg/m2 or ≥27 kg/m2 + ≥ one weight-related complication No DM ITT	Jastreboff et al, 2022 (98)
	SURMOUNT-2	15 mg Once-weekly SC injection	−14.7 (−11.5)	72 weeks	BMI ≥ 27 kg/m2 T2D ITT	Garvey et al, 2023 (99)
	SURMOUNT-3	15 mg Once-weekly SC injection	−18.4 (−20.8)	72 weeks	BMI ≥ 30 kg/m2 or BMI ≥ 27 kg/m2 + ≥ one weight-related complication No DM + intense lifestyle intervention ITT	Wadden et al, 2023 (81)

Abbreviations: BMI, body mass index; BW, body weight; CVD, cardiovascular disease; DM, diabetes mellitus; GLP-1RA, glucagon-like peptide-1 receptor agonist; ITT, intention-to-treat; T2D, type 2 diabetes.

In summary, while the newer GLP-1RAs represents a step change in AOM efficacy, many underserved patient populations remain, and thus there is an ongoing clinical need for compounds with improved real-world effectiveness, tolerance, and adherence. This position will require reappraisal as longer trials report and more real-world data become available for the newer GLP-1RA-based combination therapies in the pipeline (13, 14). However, the current clinical options, and prevalence and heterogeneity of obesity phenotypes, provide a strong rationale for ongoing preclinical discovery research to identify novel targets or combinations for more effective AOMs with fewer side effects.

Are the Side Effects of GLP-1-Based AOMs an Obligatory Feature or a Dissociable Bug?

The most common side effects of GLP-1RAs are on the gastrointestinal system, with nausea, emesis, diarrhea, and constipation occurring in 50% to 60% of subjects (17). Gastrointestinal side effects are typically classed as mild or moderate, and reported to subside in most patients following titration up to maintenance doses (77, 79, 104), although they still lead to permanent discontinuation of treatment in 4% to 10% of participants (77, 79, 95). The loss of lean body mass is another common side effect of GLP-1RAs, with studies reporting that lean mass loss accounts for ∼45% and ∼25% of total loss for semaglutide (79, 105) and tirzepatide (98), respectively. There is currently insufficient data available to determine whether such loss of lean body mass loss is primarily adaptive, or a precursor for sarcopenia, particularly in older patients (17, 106).

As nausea—the most common GLP-1RA side effect—is a state almost always accompanied by suppression of eating (47), there is a question of whether GLP-1RA-induced anorexia and weight loss is partly driven by activation of nausea circuits. As discussed above, the site(s) of action and neural circuit basis for GLP-1RA effects on eating and weight loss are still unclear. It is well-established, however, that GLP-1RAs activate neurons in the AP and the NTS (22, 63) and induce differential gene expression in the AP (63, 107), findings relevant because of the well-established role of the AP in sensing nausea-inducing stimuli (48, 108). Indeed, a recent rodent study has posited the AP and NTS as indispensable targets for the anorectic and weight loss effects of semaglutide, and that GLP-1R neurons in the AP, but not NTS, drive GLP-1RA-mediated nausea responses (47). This is consistent with some prior studies demonstrating a role for NTS GLP-1R neurons in physiological satiation/satiety and for some of the anorectic effects of various GLP-1RAs, with some reporting an absence of evidence of proxies for nausea (109‐113). Several sequencing studies have shown that GLP-1R neurons in the AP are a heterogeneous population, with subsets co-expressing Gfral (48, 107), encoding the GDF15 receptor known to be involved in the neural circuitry for nausea (114), or Prlhr, encoding the receptor for prolactin-releasing peptide hormone shown to mediate satiety without aversion (115). Hence it appears that there are populations of GLP-1R neurons within the AP and NTS that facilitate nausea and satiation/satiety, likely through functionally discrete neurocircuits (47).

These provocative findings imply that if it were possible to develop a GLP-1RA-based AOM which preferentially activates NTS GLP-1R neurons over those in the AP, this should bias the effect profile toward satiety while mitigating nausea. One such mechanism for tuning circuit activation within brainstem GLP-1R neuron populations is thought to contribute to clinical observations that the GLP-1R/GIPR co-agonist tirzepatide elicits fewer discontinuation-inducing adverse gastrointestinal events than semaglutide (98, 99), likely in part due to GIPR-expressing GABAergic AP neurons inhibiting AP and/or NTS neurocircuits which mediate GLP-1RA-induced nausea (114, 116, 117). This particular tuning mechanism was identified retrospectively; however, it is possible that similar preclinical circuit mapping approaches could be used for prospective identification of novel targets for combination therapies to enhance the functional specificity of GLP-1R agonism in the brainstem.

GLP-1R neurons in both the NTS and AP exhibit substantial molecular heterogeneity, including in cell surface receptor expression (48, 107, 118, 119). The NTS population may only be partially accessible to peripherally administered GLP-1RAs (16, 64), and GLP-1RAs apparently only activate a subset of brainstem GLP-1R neurons, particularly in the NTS (65, 117). Therefore, crucial questions for rationally designed pharmacological strategies for tuning GLP-1RA action in the brainstem include: what specific GLP-1R subpopulations, and non-GLP-1R-expressing neurons within the same local circuitry, are actually recruited by these drugs; and are these apparently functionally dissociable subpopulations which mediate nausea vs satiation/satiety differentially druggable? Such promise may be realized by some of the new combination therapies currently under development (14), or by creative new approaches for refining GLP-1RA specificity using bimodal unimolecular ligands (120), but will likely require additional preclinical discovery research to identify optimal subpopulations/circuits to target.

How Serious a Problem Is Weight Regain Following Cessation of GLP-1-Based AOMs, and Is There a Solution?

In the short time since semaglutide and tirzepatide were licensed as AOMs and broke through into the public consciousness, one of the most frequently raised “problems” with this drug class is that people tend to regain much of the weight they lost following cessation of treatment. That this phenomena occurs is certainly supported by evidence from clinical trials, with reports that 50% to 65% of total weight loss is regained up to a year after semaglutide or tirzepatide cessation (100, 121, 122). What is more debatable is whether this represents a fundamental flaw in the therapeutic value of this drug class, as implicitly or explicitly stated by many commentators. This position likely reflects stigma prevalent within popular culture that obesity is simply a failure of willpower, rather than a chronic disease requiring personalized and often long-term treatment (11, 123), given that similar criticisms are not typically leveled at other lifelong treatments for chronic disease, such as insulin for type 1 diabetes.

Semaglutide is only approved and recommended for weight management treatment for a maximum of 2 years in the United Kingdom, alongside a reduced-calorie diet and increased physical activity (124). Supplementing GLP-1-based treatments with lifestyle change has been investigated as a possible long-term strategy to ameliorate weight regain, with some studies demonstrating enhanced weight loss and reduced weight regain following cessation of GLP-1RA-based therapies (100, 125). Promising preclinical evidence has just been published supporting a related “lead-in” strategy, showing that 7 or 14 days of calorie restriction in DIO mice prior to commencement of semaglutide or tirzepatude treatment lowered the weight loss plateau of these drugs (126). Such an approach to lowering the weight loss plateau could conceivably help mitigate the extent of weight regain. However, the STEP3 trial did include an 8-week low-calorie diet prior to semaglutide initiation (94), which did not result in notably greater weight loss than the STEP1 trial which did not include such a lead-in (79), hence the efficacy of such an approach remains to be demonstrated in humans.

While targeted behavioral interventions combined with AOMs show promise and warrant continued investigation, such changes alone are unlikely to mitigate the weight regain most patients will experience following voluntary or mandatory cessation of GLP-1-based AOMs, hence additional strategies are required (127). In patients for whom (in)tolerability of side effects is the major driver of AOM cessation, one strategy could be switching to a lower maintenance dose, sufficient to maintain lost weight but with reduced severity of side effects and/or frequency of administration. Maritide, the GIPR antagonist/GLP-1RA, resulted in ∼16% body weight loss that was maintained up to a 70% level (hence just 30% of lost weight regained) for 5 months after just 3 monthly doses (128), suggesting potential for maintenance therapy with infrequent dosing. Maintenance therapy using orally active formulations of GLP-1RAs such as Rybelsus (oral semaglutide) (129); or the small molecule GLP-1R partial agonist orforglipron (130) may also represent a viable strategy. While early trial data suggest these compounds have somewhat lower efficacy than injectable AOMs, this efficacy may be sufficient for weight loss maintenance, and the greater convenience and potential lower cost (of small molecule drugs) may make them more palatable for long-term use. It has also been proposed that weight loss and weight regain may be biologically dichotomous challenges, which may benefit from distinct pharmacotherapies which are largely ineffective for weight loss but specifically efficacious for prevention of weight regain, such as leptin-based treatments (131).

Is Endogenous GLP-1 Signaling Involved in the Pathogenesis of Obesity?

In contrast to the substantial body of research investigating the physiological roles of endogenous GLP-1 in energy balance, and translational and clinical studies using GLP-1RA-based drugs for pharmacotherapy, there are relatively fewer studies investigating the contribution of endogenous peripheral or central GLP-1 to the pathophysiology of obesity. There is a limited body of evidence supporting a modest reduction in gut GLP-1 release in people living with obesity, which may be partly heritable (14, 132‐134). Given the role of GLP-1R-expressing VANs in physiological satiation (23, 39), reduced gut GLP-1 secretion and/or decreased vagal sensitivity to GLP-1 could conceivably be part of a vagal contribution to the impaired satiation reported in people living with obesity (135). Prolonged consumption of obesogenic diets impairs vagal satiation signaling via multiple pathophysiological mechanisms, including withdrawal of vagal fibers from the NTS (23, 136, 137). It is thus plausible that blunted vagal input to PPG neurons contributes to obesity via impairment of central GLP-1-dependent satiation signaling. This input-driven pathophysiological mechanism could potentially be overcome by direct pharmacological activation of PPG neurons, which may be feasible given recent reports that 5-HT_2CR-expressing PPG neurons are a necessary target for the 5-HT_2CR agonist AOM lorcaserin, and that they can be activated for additive anorectic effects with GLP-1-based AOMs (22, 68). A recent functional genetics study addressed the question of whether genetic variability in GLP-1R impacts people's susceptibility to obesity and type 2 diabetes, and it identified rare loss-of-function mutations which were associated with increased adiposity in a 200 000-participant UK BioBank cohort (103). However, in vitro functional assays of the identified mutations only investigated GLP-1-dependent insulin secretion, hence their impact on GLP-1-dependent satiation/satiety signaling remains to be determined.

Conclusion

Recent years have witnessed a revolution in the clinical impact of GLP-1-based pharmacotherapies for obesity, which was driven by pioneering preclinical pharmacology studies investigating how GLP-1R ligands modulate eating and body weight. This clinical success has in turn stimulated a concerted recent effort using modern systems neuroscience tools to try and unravel the complexity of endogenous central GLP-1 signaling and the targets of GLP-1R AOMs. These studies have already yielded many valuable insights; however, further basic and translational efforts are required to address knowledge gaps in the field to overcome the challenges of weight loss effectiveness, weight regain, and tolerance-limiting side effects. The increasingly sophisticated genetic and molecular tools now available should provide the specificity required to dissociate and selectively target the neural circuits recruited by GLP-1RA AOMs for their therapeutic and adverse effects. This should enable the rational design of next-generation pharmacotherapies to meet the diverse clinical needs of people living with obesity, and to aid preventative efforts to reverse the prevalence of the global obesity epidemic.

Abbreviations

AOM

anti-obesity medication

AP

area postrema

BMI

body mass index

CNS

central nervous system

GIPR

glucose-dependent insulinotropic polypeptide receptor

GLP-1

glucagon-like peptide-1

GLP-1R

glucagon-like peptide-1 receptor

GLP-1RA

glucagon-like peptide-1 receptor agonist

HPV

hepatic portal vein

NTS

nucleus tractus solitarius

PPG

preproglucagon

VAN

vagal afferent neuron

Funding

Work funded by Wellcome Trust fellowship 223279/Z/21/Z to D.I.B.

Disclosures

The authors have nothing to disclose.

Data Availability

Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.