The history of GLP-1 research — from Habener and Drucker to Ozempic

What made GLP-1 interesting was its behavior. When it was cleaved from proglucagon in the gut's L-cells — as opposed to the pancreas, which cleaved the same precursor into glucagon — it turned out to stimulate insulin secretion in a manner that depended on blood glucose being elevated. This glucose-dependent insulinotropism was a pharmacologically important property: a molecule that boosted insulin when sugar was high, but not when it was normal, would theoretically be less likely to cause hypoglycemia than anything previously available. The incretin system — the idea that gut hormones coordinate insulin secretion in response to food — had been hypothesized for decades, but here was a concrete molecular candidate.

Daniel Drucker, working at the University of Toronto, became the person who most thoroughly characterized what GLP-1 actually did.

Drucker's laboratory spent the late 1980s and 1990s methodically working through GLP-1's actions: on the pancreatic beta cell, where it stimulated insulin synthesis and secretion; on the pancreatic alpha cell, where it suppressed glucagon; on gastric emptying, which it slowed; and, critically, on beta cell survival, where it appeared to have trophic effects — the possibility that it didn't just stimulate existing beta cells but might preserve or expand the beta cell mass that type 2 diabetes progressively destroys. Drucker's group also characterized the GLP-1 receptor with the precision necessary for anyone to eventually design drugs that would activate it. This was painstaking, unfashionable science that didn't have an obvious commercial endpoint. It was the kind of work that makes drug development possible twenty years later without getting any of the credit.

There was also a problem that Drucker's work made unmistakably clear: GLP-1 was useless as a drug in its native form.

The enzyme DPP-4 — dipeptidyl peptidase-4, abundant in the blood and on the surface of endothelial cells throughout the vasculature — degrades GLP-1 by cleaving two amino acids from its N-terminus within minutes of the molecule entering circulation. By the time endogenous GLP-1 reaches the pancreas after a meal, a significant fraction has already been inactivated. Continuous intravenous infusion of GLP-1 could achieve meaningful effects in clinical studies, but that wasn't a therapy. The half-life problem was real and blocking development.

Two different strategies emerged from this obstacle.

The first was to leave GLP-1 alone and instead inhibit DPP-4. If you blocked the enzyme that destroyed GLP-1, you could extend the life of whatever endogenous GLP-1 the body was already producing after meals. This became the DPP-4 inhibitor class — sitagliptin, approved as Januvia in 2006, followed by saxagliptin, alogliptin, linagliptin. DPP-4 inhibitors work modestly: they roughly double post-meal GLP-1 levels, produce about 0.5-0.8% reductions in HbA1c in clinical trials, and have clean safety profiles. They don't produce meaningful weight loss because the GLP-1 levels they achieve are still within physiological range — they're not overwhelming the system, they're just letting it run a little longer. DPP-4 inhibitors had a solid commercial run and remain in use, but they were ultimately the conservative solution to the half-life problem.

The second strategy was to build GLP-1 analogs that DPP-4 couldn't touch.

This is where the gila monster comes back into the story. John Eng's discovery of exendin-4 — a GLP-1-like peptide from gila monster saliva that was structurally resistant to DPP-4 — provided the first proof that DPP-4-resistant GLP-1 receptor agonism was achievable. Exenatide, the synthetic version of exendin-4, became the first GLP-1 agonist approved by the FDA in 2005. Liraglutide followed in 2010, built on the human GLP-1 sequence but with a fatty acid modification that slowed degradation by binding to albumin. Each generation reduced injection frequency and improved tolerability — from twice daily to once daily to once weekly — as the chemistry of half-life extension got more sophisticated.

Through this entire period — roughly 2000 to 2015 — the primary framing for GLP-1 drugs was diabetes management. Weight loss was noted. It was real. It was consistent across the trials. But it was positioned as a benefit, a bonus, not the central story. The field was organized around glycemic control. The dominant outcome measures were HbA1c and fasting glucose. Nobody was running Phase III obesity trials.

What changed was scale.

The SCALE trials for liraglutide at 3 mg — the dose that became Saxenda — showed meaningful but not spectacular weight loss, roughly 8% of body weight over a year. The Ozempic trials for diabetes showed something more interesting: 14-15 pound weight losses in people who were being treated for diabetes, not for obesity. When Novo Nordisk ran the STEP program — a dedicated Phase III obesity trial series for semaglutide at higher doses — the results came in at approximately 15% mean body weight reduction over 68 weeks. That was more than double what Saxenda had achieved. More than the FDA's threshold for what counts as clinically meaningful. More than anything that had been produced outside of surgical intervention.

The STEP 1 results published in the New England Journal of Medicine in early 2021, and suddenly GLP-1 biology that had been developed over thirty-five years of patient, incremental science was in the center of a conversation that had nothing to do with science.

The Ozempic moment — roughly 2022 through 2023 — is a phenomenon worth trying to understand honestly, because it illuminates something real about how medical knowledge travels. The drug had been approved for diabetes in 2017. It had been available for five years. The obesity trial data was public. What changed wasn't the molecule and wasn't the data; what changed was the social network through which it spread. Celebrity mentions, before-and-after posts, the visual evidence of dramatic weight loss in people who were recognizable — these moved information at a velocity that clinical publication never could. Within months, Ozempic became one of the most searched medical terms on the internet, and Novo Nordisk faced supply shortages that affected patients who had been using it for diabetes management for years.

This created two problems simultaneously. First, the drug most people were asking for — at the doses being used for obesity — wasn't approved for obesity, and the drug approved for obesity (Wegovy, the same molecule at a higher dose under a different brand name) had its own supply constraints. Second, the cultural conversation outpaced the scientific one in ways that produced serious misinformation: that it was safe with no side effects, that it was a shortcut with no lifestyle implications, that the weight would stay off without continued use. None of this was true. The nausea and GI side effects are real. The weight largely returns when the medication is stopped, because the underlying biology — the reduced GLP-1 tone, the adipose tissue signaling, the hypothalamic regulation — returns to its pre-treatment state. The drug works while you take it, not after.

The slower, more important story that the Ozempic moment obscured was the one that had been running since 1987: that GLP-1 is a central regulator of metabolic homeostasis, that its actions extend well beyond the pancreas into the brain, the cardiovascular system, the liver, and the kidneys, and that we are still in the early stages of understanding what happens when you provide sustained, pharmacological GLP-1 receptor activation over months and years. The cardiovascular outcome trials — LEADER for liraglutide, SUSTAIN-6 and SOUL for semaglutide — showed reductions in major cardiovascular events that weren't expected from a diabetes drug and that suggest GLP-1 receptor agonists are doing something anti-inflammatory, anti-atherogenic, or cardioprotective beyond glycemic control. The renal outcomes. The potential neurological effects. The MASH data. These findings are accumulating in a way that suggests the biology is far more interesting than the weight loss number.

Habener's 1987 PNAS paper didn't predict any of this. Drucker's painstaking receptor characterization in the 1990s was done without a specific disease model in mind, mostly because the degradation problem made therapeutic application seem remote. The researchers who identified the DPP-4 cleavage site didn't know they were mapping an obstacle that would eventually be solved by a gila monster's saliva. The scientists who spent the 2010s running obesity trials didn't know that their efficacy data would eventually drive a global drug shortage.

Translational science doesn't move on a schedule. It moves the way knowledge actually moves — slowly and then all at once, with long periods of accumulation before the moment when enough pieces are in place that the whole thing becomes visible. From 1987 to 2021 is thirty-four years. From Habener's proglucagon paper to Wegovy's approval. From a laboratory curiosity to a cultural rupture.

The thirty-four years are not wasted time. They are the actual work.

What this history suggests about the science still ahead is that we are probably somewhere in the middle of another accumulation period. The triple agonists — retatrutide, which adds glucagon receptor activation — and the amylin combinations and the long-duration small-molecule agonists are all in development now, and the full picture of what they do won't be visible from inside it. Somewhere in the current literature there are probably findings that look modest today and will look essential in fifteen years. The researcher who notices them may not know what they have yet.

That's not a failure of the scientific process. That's what the scientific process looks like from the inside.