What Is FGF21? The Growth Factor That Became a Hormone

Peptide & protein science

FGF21 is the growth factor that stopped behaving like one. It belongs to the fibroblast growth factor family, whose usual job is to signal to the cell next door. But FGF21 lacks the molecular grip that keeps its relatives anchored beside the tissue that made them, so it drifts away from the liver and travels in the blood instead. That single structural defect is what turned a growth factor into a hormone, and it is the reason FGF21 analogs are now in late-stage trials for fatty liver disease while the natural molecule itself was never usable as a drug.

What FGF21 is, and why it is the odd one out

FGF21 is a small secreted protein made mainly in the liver. Its gene sits on chromosome 19 at 19q13.33, and the protein is produced as a 209-amino-acid precursor whose first 28 residues are a signal peptide that gets clipped off during secretion, leaving a mature chain of 181 amino acids weighing roughly 19.5 kDa.

Most fibroblast growth factors bind heparan sulfate, a sugar polymer studded through the matrix between cells, with high affinity. That binding is not incidental: it is what pins them in place. A classical FGF is released, immediately caught by the surrounding matrix, and acts on its immediate neighbours. It is a local instruction, not a broadcast.

FGF21 and its two closest relatives, FGF19 and FGF23, bind heparan sulfate very weakly. Nothing holds them. They leave the tissue of origin, enter the circulation, and act at a distance, which is the working definition of a hormone. What reads as a binding defect in a growth factor is precisely the property that makes an endocrine signal possible.

The catch is that heparan sulfate was not only an anchor; it also helped classical FGFs grip their receptors. Having given that up, FGF21 needs something else to make a productive receptor complex, and that something is a co-receptor called beta-Klotho.

Beta-Klotho is the address on the envelope

FGF21 signals through three receptors — FGFR1c, FGFR2c and FGFR3c — but only when those receptors are partnered with beta-Klotho, a single-pass membrane protein encoded by the KLB gene. Without beta-Klotho present, FGF21 binds too weakly to do anything useful.

This is the elegant part. FGFR1c is expressed almost everywhere in the body. Beta-Klotho is not: it is restricted largely to fat tissue, the liver, the pancreas and a small number of brain regions, including the suprachiasmatic nucleus of the hypothalamus and the dorsal vagal complex of the hindbrain. So the tissue selectivity of FGF21 is not written into the hormone and is not written into the receptor. It is written into the distribution of the co-receptor. Change where beta-Klotho is expressed and you change who can hear the message.

What happens in mice, and what does not happen in people

Almost everything exciting ever written about FGF21 began in rodents, and the human data have been consistently more modest. Keeping the two apart is the only honest way to read this literature.

In mice. FGF21 is induced by fasting through the transcription factor PPAR-alpha, which is why it is often called a starvation hormone. Signalling in fat tissue is required for its acute insulin-sensitising effect, while the increase in energy expenditure is fat-independent and appears to run through the brain, driving sympathetic activity to brown fat and browning of white fat. Mice engineered to overexpress FGF21 grow slowly, run low glucose and triglycerides, and resist age-associated obesity and diabetes. FGF21 also reduces preference for sweet tastes and for alcohol in mice and in monkeys, an effect that requires beta-Klotho in the brain.

In people. The picture narrows considerably:

• Fasting barely moves it. A two-day fast or a ketogenic diet does not reliably raise human FGF21. It takes roughly seven days of starvation to produce about a 74% rise. The starvation-hormone label is a mouse label.
• Glucose lowering has never convinced. The first engineered analog, LY2405319, improved lipids in people with type 2 diabetes but produced only a trend toward lower glucose. A second, PF-05231023, showed no significant effect on glycaemic control at all.
• Weight loss is weak. In the pegozafermin phase 2b liver trial there was no apparent effect on body weight. In the efruxifermin phase 2b trial the 50 mg group lost 2.9 kg from baseline, but only 2.3 kg more than placebo, which did not reach statistical significance. In monkeys, the weight loss seen with one analog was attributable entirely to eating less, with no measurable change in energy expenditure.
• Lipids and liver fat, on the other hand, move hard. Triglycerides fall, HDL rises, and liver fat measured by MRI drops by half or more. This is where the drug programmes ended up.

The sweet-tooth story deserves a similar caution. Human genetic studies have linked variants at the FGF21 locus to relatively higher carbohydrate and alcohol intake, which looks like beautiful confirmation of the mouse work. But a 2024 analysis found that the variant associated with macronutrient preference and the variant that actually governs circulating FGF21 levels are distinct genetic signals, which is harder to reconcile with a simple causal chain.

Why the natural hormone had to be redesigned

Native FGF21 is a poor drug. It is small enough to be filtered out by the kidney, and it is chewed up by enzymes in the blood. The most important of those enzymes is fibroblast activation protein, or FAP, which cuts the mature chain after proline 171 and releases the last ten amino acids. That short tail is what grips beta-Klotho, so losing it is catastrophic: the truncated protein has been reported to lose roughly 400-fold of its ability to activate signalling in fat cells. A second enzyme, DPP-4, trims the other end.

The specificity of the FAP cut is a nice piece of molecular detail. FAP needs a glycine and then a proline at the two positions immediately before the scissile bond. Human FGF21 has exactly that. Mouse FGF21 has a glutamate where the glycine should be, and is not cleaved by FAP at all — one more reason mouse results have translated badly.

Reliable numbers for how long native FGF21 survives in human blood are surprisingly hard to source. What is documented is that in mice, human FGF21 disappeared with a half-life of about 48 minutes, extended to 79 minutes when a FAP inhibitor was given, and that intact FGF21 incubated with recombinant FAP was gone with a half-life near 38 minutes. Every clinical candidate has been engineered to survive far longer than that.

What the human trials actually show

Two programmes have carried FGF21 into phase 3, and both did it on liver disease rather than diabetes.

Pegozafermin is a glycoPEGylated FGF21 analog. In ENLIVEN, a 24-week phase 2b trial in metabolic dysfunction-associated steatohepatitis published in the New England Journal of Medicine in 2023, 26% of patients on 30 mg weekly improved fibrosis by at least one stage without worsening of steatohepatitis, against 7% on placebo, a 19-point difference with a p value of 0.009. Liver fat on MRI fell by about 48%. Body weight did not meaningfully change. A separate phase 2 trial in severe hypertriglyceridaemia cut median triglycerides by 57.3% against 11.9% on placebo.

Efruxifermin is a bivalent antibody-Fc fusion carrying two FGF21 chains. In HARMONY, also phase 2b in steatohepatitis, 39% of the 28 mg group and 41% of the 50 mg group improved fibrosis by a stage without worsening disease, against 20% on placebo. A detail usually lost in the coverage: in the pre-specified full analysis set rather than the biopsy analysis set, the 50 mg result did not reach significance (p=0.123).

And then the corrective. SYMMETRY tested efruxifermin in people whose disease had already progressed to compensated cirrhosis, and missed its primary endpoint: at week 36, fibrosis improvement was 19% on 50 mg versus 13% on placebo, p=0.52. Week 96 results looked better, but the investigators state plainly that no formal hypothesis testing of secondary outcomes was done, so no statistical conclusions can be drawn from them.

Safety signals worth knowing about

• Gastrointestinal effects dominate. Nausea and diarrhoea are the most common adverse events across every trial, and they scale with dose. In SYMMETRY, diarrhoea affected 54% of the 50 mg group versus 30% on placebo, and nausea 46% versus 30%.
• Increased appetite is an unusual entry on the list. It was reported in 40% of the 50 mg SYMMETRY group versus 7% on placebo. That is the opposite direction from the incretin drugs most readers will be comparing against, and it is part of why FGF21 analogs are not weight-loss drugs.
• Injection-site reactions — erythema, bruising, rash — are common, though placebo rates in these trials were also high.
• Anti-drug antibodies were detected in 65 of 78 patients in HARMONY, at low titre.
• Bone is the genuine open question. Preclinical work has raised concerns about FGF21 and bone, though those findings are confounded by growth suppression in young animals. A one-year study in skeletally mature monkeys with pegbelfermin was clean, and 24-week human trials showed no adverse DXA changes. But the longest human trial to date, SYMMETRY at 96 weeks, reported small reductions in lumbar spine and femoral neck bone mineral density on efruxifermin compared with placebo, with no difference in fractures — four in each group. This is a question phase 3 is designed to answer, not a settled finding in either direction.

Is there an approved FGF21 medicine?

No. As of August 2026, no FGF21 analog appears in Drugs@FDA or DailyMed as an approved product, and none holds an EMA marketing authorisation. Every molecule described here is investigational. The one drug approved specifically for steatohepatitis, resmetirom, received FDA accelerated approval in March 2024 and works through a completely different mechanism — it is a thyroid hormone receptor-beta agonist, not an FGF21 analog. Pegozafermin holds EMA PRIME designation, which is development support and explicitly not an approval.

Frequently asked questions

Why did a hormone that fixes diabetes in mice fail to lower blood sugar in people?

Nobody has a complete answer, but the pattern is consistent. Both early clinical analogs improved lipids and left glucose largely untouched. One contributing factor may be that the mouse effects lean heavily on increased energy expenditure through brown fat, and humans have far less brown fat than a laboratory mouse. Whatever the cause, both surviving programmes redirected from glycaemia to liver fibrosis and triglycerides, and that is where the positive trials came from.

What is beta-Klotho, and why does FGF21 need it?

Beta-Klotho is a membrane protein that partners the FGF receptor and makes FGF21 binding productive. Because the receptor itself is nearly ubiquitous but beta-Klotho is restricted to fat, liver, pancreas and a few brain regions, the co-receptor functions as the address on the envelope: it determines which tissues can respond at all.

Is FGF21 the same thing as a peptide?

Not quite. At 181 amino acids in its mature form, FGF21 sits well above the size range usually called a peptide and is more accurately described as a small protein hormone. The distinction matters practically: proteins of this size are harder to synthesise chemically, are usually produced in cell culture, and are more likely to provoke anti-drug antibodies.

Does FGF21 reduce cravings for sugar and alcohol?

In mice and monkeys, administered FGF21 does reduce preference for sweet tastes and for alcohol, and the effect requires beta-Klotho in the brain. Human genetic associations at the FGF21 locus point the same way. But the 2024 finding that the preference-associated variant and the variant controlling circulating FGF21 are separate signals means the human version of this story is not settled.

References

Informational only — not medical advice. Nothing here describes a treatment plan or a dose for any person. FGF21 analogs are investigational and are not approved for any indication. Consult a qualified healthcare professional about your own care. Intended for readers 21 and over.