We usually approach drug mechanisms by asking what a molecule does — its receptor targets, its downstream pathways, its tissue distribution. But sometimes the most revealing question is simpler: does it matter how the drug gets into the body?

For semaglutide, Overgaard and colleagues (Drucker lab, Cell Reports Medicine, 2021) answered that question rigorously by comparing oral versus subcutaneous administration across the PIONEER and SUSTAIN clinical trial programs — thousands of people, propensity-matched datasets, population pharmacokinetic modeling. The question was not merely clinical. It was mechanistic: if the route of administration doesn't matter and only circulating drug levels do, that tells us something definitive about where semaglutide is working.

The result was unambiguous. At identical plasma concentrations, oral and subcutaneous semaglutide produced statistically equivalent reductions in HbA1c, equivalent weight loss, equivalent improvements in cardiovascular biomarkers (CRP, triglycerides, systolic blood pressure), and equivalent rates of nausea and vomiting. Route of administration was irrelevant. What mattered was the circulating drug level — and nothing else.

This single finding closes several open mechanistic questions at once. Oral semaglutide is absorbed through the stomach and enters the portal circulation before reaching the systemic bloodstream — it has early, high-concentration access to portal vein GLP-1 receptors and hepatic GLP-1R populations that subcutaneous semaglutide does not. If those portally activated receptors contributed meaningfully to the drug's therapeutic effect, oral and injectable semaglutide would diverge in their exposure-response curves. They don't.

Similarly, interindividual differences in GLP-1 receptor expression, receptor binding affinity, and downstream signaling activity — often invoked to explain why some patients respond more robustly than others — are not substantial determinants of outcomes once plasma level is accounted for. The exposure-response relationships hold across race, ethnicity, BMI, gender, and background medication. What shifts the curve is baseline HbA1c (for glycemic response) and gender (modest effect on weight) — neither of which implicates differential GLP-1R biology.

There is one clinically important exception: body weight. Heavier patients achieve lower average plasma semaglutide concentrations at a given dose, likely due to greater volume of distribution. This doesn't mean GLP-1R sensitivity is lower in people with higher BMI — it means they may need higher doses or injectable formulations to achieve therapeutic exposure. The exposure-response relationship itself remains intact; it's the PK that differs.

If route of administration doesn't matter and circulating drug levels drive everything, semaglutide must be acting at a target reached equally by drug delivered through the stomach or the upper arm — which means a target accessed via the systemic circulation. That points unambiguously to the brain.

A second piece of evidence nails this. When the glucagon gene is selectively deleted from intestinal cells — eliminating circulating gut-derived GLP-1 — the mice do not gain weight, do not eat more, and respond normally to a high-fat diet. The GLP-1 your L cells release after a meal is not the signal responsible for appetite control or body weight regulation. Gut GLP-1 circulates at vanishingly low concentrations systemically and appears primarily relevant for glucose homeostasis through local portal mechanisms. The weight biology operates through a completely different system: the brain GLP-1 receptor network, accessed pharmacologically through circulating semaglutide (Drucker, Mol Metab, 2022).

Peripherally administered fluorescent semaglutide reaches GLP-1R+ neurons across a distributed network: circumventricular organs, the arcuate and paraventricular hypothalamic nuclei, the NTS, the dorsomedial hypothalamus, the lateral septal nucleus, and the dorsal motor nucleus of the vagus. Beyond these canonical sites, GLP-1R signaling in the nucleus accumbens, lateral parabrachial nucleus, hippocampus, bed nucleus of the stria terminalis, and paraventricular thalamus have all been shown in gain-and-loss experiments to contribute to food intake suppression. The redundancy is not accidental — it is precisely why targeting a single population won't replicate the drug's full efficacy, and why semaglutide's anorectic effect is so robust across individuals.

For decades, hypothalamic arcuate POMC neurons were the canonical story for GLP-1's anorectic action — and they are a real part of it. Arcuate POMC/CART neurons are overwhelmingly GLP-1R-positive, and pharmacological GLP-1R injection into the ARC or PVN robustly suppresses food intake. The story seemed complete.

Two papers disrupted it. Burmeister et al. (2017) generated mice with selective Cre-lox deletion of the GLP-1R in hypothalamic neurons — including ARC POMC neurons and PVN neurons. Despite losing GLP-1R signaling in the very regions where direct agonist injection suppresses food intake, these animals showed a completely normal anorectic and weight-loss response to peripheral exenatide and liraglutide. The drugs worked as if the hypothalamic receptors had never been removed. The hypothalamus is pharmacologically sufficient — but it is not physiologically necessary for the drug's effect.

Where, then, are the obligate receptors? Huang et al. (Nature, 2024) answered this with in-vivo two-photon calcium imaging of GLP-1R+ neurons in the dorsal vagal complex (DVC) of awake mice. Selective ablation of DVC GLP-1R neurons completely abolished semaglutide-induced food intake suppression and weight loss. Ablation of hypothalamic ARC neurons did not. The DVC is both sufficient and necessary. The hypothalamus is neither — at least not for the pharmacological response to peripherally administered GLP-1 agonists.

Within the DVC, Huang et al. used calcium imaging to monitor AP and NTS GLP-1R neurons simultaneously while exposing awake mice to semaglutide, nutritive stimuli, and nauseogenic stimuli. The finding was structurally important: only 6% of responding DVC neurons were activated by both nutritive and aversive inputs. The vast majority were tuned to one or the other.

NTS GLP-1R neurons → PVH (MC4R): Activated by nutritive signals. Chemogenetic activation extends inter-meal intervals and reduces body weight over time. No conditioned flavor aversion, no place avoidance, no aversive taste reactivity. This is satiety — the circuit that makes you feel done eating.

AP GLP-1R neurons → lPBN (CGRP): Activated by aversive stimuli. Chemogenetic activation suppresses food intake — but also generates robust conditioned taste aversion, place avoidance, and aversive licking responses. This is the nausea circuit.

Critically, when the AP aversion pathway was blocked — either chemogenetically or through selective receptor deletion — GLP-1 medicines retained full anorectic efficacy. The satiety circuit alone is sufficient. Nausea and weight loss do not need to travel together. This is why the prevailing assumption that "some GI side effects are inherent to the mechanism" appears to be wrong — and it suggests that drugs or formulations capable of preferentially engaging NTS satiety neurons over AP aversion neurons could deliver the therapeutic benefit without the tolerability cost.

One clinical observation fits this perfectly. GLP-1-related nausea typically peaks during dose titration and attenuates over weeks to months — even as weight loss continues or deepens. If the AP aversion circuit undergoes tachyphylaxis or desensitization while the NTS satiety circuit remains engaged, that's precisely what you'd expect. The biology may already be making this dissociation in patients who persist through titration.

The patient description "food noise went quiet" is not metaphor. In human fMRI studies, GLP-1 infusion reduces blood-oxygen-level-dependent (BOLD) activation in the amygdala, caudate, insula, nucleus accumbens, orbitofrontal cortex, and putamen — the exact regions that process food reward, emotional eating, and cue-triggered desire. Exenatide infusion in people with obesity decreased brain activation to pictures of food in the insula, amygdala, putamen, and orbitofrontal cortex; these effects were blocked by co-administration of the GLP-1R antagonist exendin(9-39), confirming receptor specificity rather than a non-specific sedative effect.

Critically, these effects were preserved in people with obesity but absent in normal-weight individuals. People with obesity show heightened food-cue-driven reward activation in these circuits — and GLP-1 pharmacotherapy specifically targets that dysregulation. This has a clinical corollary: the STEP 4 eating behavior sub-study found a ~35% reduction in ad libitum energy intake with semaglutide 2.4mg, with minimal effects on gastric emptying. The food noise is quieted by appetite biology, not delayed digestion.

When circulating semaglutide levels rise — through any route — they reach the CNS via the systemic circulation, engage GLP-1R+ neurons in the DVC and distributed forebrain reward circuits, and dampen the neurohormonal signals that drive excessive food intake in people with obesity. The gut is not the target. The brain is.

The STEP 4 randomized withdrawal trial was not designed to probe mechanism — but its design inadvertently creates one of the cleanest mechanistic windows in obesity pharmacology. For the first 20 weeks, every participant received semaglutide 2.4mg weekly plus structured lifestyle intervention (calorie-reduced diet, 150 minutes of activity weekly, behavioral tracking). By week 20, all participants had lost 10.6% of baseline body weight. Then they were randomized — double-blind — to continue semaglutide or switch to placebo. Both arms continued the identical lifestyle program through week 68.

The divergence was immediate and complete. Continued semaglutide → further weight loss, reaching approximately 18% total reduction. Placebo → rapid regain, converging toward the 3–5% net loss typical of long-term lifestyle interventions alone. The lifestyle intervention did not fail in the placebo group. Their biology reasserted. What the drug was doing — suppressing appetite, quieting food-cue reactivity, counteracting the neurohormonal drive to regain — stopped when the circulating drug level fell. The Overgaard finding and the STEP 4 withdrawal finding are two sides of the same coin: it is the plasma level, operating through the CNS, that produces and sustains the benefit.

To understand why weight returns when semaglutide stops, you need to understand what it was doing while it worked. Weight loss — through any mechanism — triggers a coordinated neurohormonal defense system evolved to protect against starvation. Sumithran et al. (NEJM, 2011) characterized this in 50 participants followed through a very-low-energy diet and one year of follow-up: ghrelin rose, PYY fell, CCK fell, amylin fell. At 52 weeks — even after much of the weight had returned — these unfavorable changes persisted. The set point had shifted. The signals hadn't normalized.

Beyond hormones: resting metabolic rate falls approximately 300–400 kcal/day below what body composition changes alone would predict when weight is maintained 10% below baseline — a finding first rigorously quantified by Leibel et al. (1995) in per-kg-fat-free-mass terms, and subsequently confirmed and extended by Rosenbaum and colleagues to show the adaptation persists for more than a year after weight loss. Fothergill et al. found a similar magnitude of adaptation six years after The Biggest Loser competition — even in participants who had fully regained. Brain reward circuitry shifts: weight-regainers show sustained food-cue reactivity even when satiated, a neurobiological vulnerability that weight-maintainers don't exhibit (Neseliler et al., 2019).

Semaglutide counteracts each of these forces directly. It suppresses ghrelin-driven hunger, enhances postprandial satiety signaling, and dampens the reward-circuit overactivation that drives food-cue hypersensitivity. When circulating levels fall, these counter-regulatory forces re-emerge unchecked — exactly as the weight trajectory data predict. Tzang et al.'s 2025 meta-analysis of 18 RCTs and 3,771 participants confirmed the magnitude: 7.31 kg rebound at more than 26 weeks post-cessation. Semaglutide rebounded more than liraglutide (8.21 vs. 4.29 kg) — the more potent the drug, the stronger the counter-regulatory response when it is removed, likely because the system had been more completely suppressed.

The pharmacokinetics, the mechanism studies, and the withdrawal trial converge on a unified picture. GLP-1 receptor agonism reduces body weight by engaging a distributed CNS network — obligately through the hindbrain DVC, redundantly through the hypothalamus and forebrain reward circuits — to suppress the neurohormonal drive to eat that is dysregulated in people with obesity. It does this through appetite suppression, not thermogenesis; through brain circuits, not gut receptors; through plasma drug levels, not route of administration.

The nausea is separable from the efficacy — anatomically, functionally, and potentially pharmacologically. The Huang et al. data suggest that a next-generation molecule capable of preferentially engaging NTS satiety circuitry over AP aversion circuitry could achieve equivalent weight loss with substantially better tolerability. That's not speculation about a distant future. It's a design target, informed by circuit-level mechanistic understanding that didn't exist a decade ago.

And the STEP 4 weight trajectory is not, at its core, a story about drug dependency. It's a story about what was already there. The counter-regulatory system — the ghrelin surge, the leptin fall, the suppressed satiety hormones, the reward-circuit hyperactivation — was present before treatment, was masked by pharmacological GLP-1R engagement, and re-emerged when that engagement ceased. The drug didn't create a problem by stopping. It revealed the problem that was there all along.

That is the clearest argument for understanding obesity as a chronic neurobiological disease. Not because the medications are addictive or impermanent. But because the biology they're treating is persistent, redundant, and — for most people — not resolved by a finite course of treatment. What semaglutide teaches us about GLP-1 is also what it teaches us about obesity: the mechanism runs deep, and so must the treatment.

REFERENCES

1. Overgaard RV, et al. Levels of circulating semaglutide determine reductions in HbA1c and body weight in people with type 2 diabetes. Cell Rep Med. 2021;2(9):100387.

2. Drucker DJ. GLP-1 physiology informs the pharmacotherapy of obesity. Mol Metab. 2022;57:101351.

3. Huang KP, et al. Dissociable hindbrain GLP1R circuits for satiety and aversion. Nature. 2024;632(8025):585–593.

4. Burmeister MA, et al. The hypothalamic GLP-1 receptor is sufficient but not necessary for regulation of energy balance. Diabetes. 2017;66(2):372–384.

5. Friedrichsen M, et al. The effect of semaglutide 2.4mg on energy intake, appetite, control of eating, and gastric emptying in adults with obesity. Diabetes Obes Metab. 2021;23(3):754–762.

6. Rubino DM, et al. Effect of continued weekly subcutaneous semaglutide vs placebo on weight loss maintenance: the STEP 4 randomized clinical trial. JAMA. 2021;325(14):1414–1425.

7. Sumithran P, et al. Long-term persistence of hormonal adaptations to weight loss. N Engl J Med. 2011;365(17):1597–1604.

8. Leibel RL, Rosenbaum M, Hirsch J. Changes in energy expenditure resulting from altered body weight. N Engl J Med. 1995;332(10):621–628.

9. Rosenbaum M, Hirsch J, Gallagher DA, Leibel RL. Long-term persistence of adaptive thermogenesis in subjects who have maintained a reduced body weight. Am J Clin Nutr. 2008;88(4):906–912.

10. Fothergill E, et al. Persistent metabolic adaptation 6 years after "The Biggest Loser" competition. Obesity (Silver Spring). 2016;24(8):1612–1619.

11. Neseliler S, et al. Neurocognitive and hormonal correlates of voluntary weight loss in humans. Cell Metabolism. 2019;29(1):39–49.

12. Tzang CC, et al. Metabolic rebound after GLP-1 receptor agonist discontinuation: a systematic review and meta-analysis. eClinicalMedicine. 2025;90:103680.

13. Wilding JPH, et al. Weight regain and cardiometabolic effects after withdrawal of semaglutide: the STEP 1 trial extension. Diabetes Obes Metab. 2022;24(8):1553–1564.

14. Drucker DJ. GLP-1-based therapies for diabetes, obesity and beyond. Nat Rev Drug Discov. 2025. doi:10.1038/s41573-025-01183-8.