ATX-304 activates AMPK directly, and is the only compound in this catalogue tested in patients

What AMPK is, and why activating it is a serious idea

Every cell in your body runs on a molecule called ATP. Using ATP spends it and turns it into ADP and then AMP, so the ratio of AMP to ATP inside a cell is a direct read-out of how depleted that cell is. AMP-activated protein kinase — AMPK — is the protein that reads that ratio. When AMP rises, AMPK switches on, and when AMPK switches on it does two things at once: it turns off the processes that consume energy to build things, and it turns on the processes that release energy by breaking things down.

In practice that means AMPK activation pulls glucose out of the bloodstream into muscle, tells cells to burn fat rather than store it, stops the liver making new fat and new cholesterol, and increases the number and quality of mitochondria. The paper that introduced O304 describes AMPK as "a master regulator of energy homeostasis" that "is activated in response to an energy shortage imposed by physical activity and caloric restriction."

Read that last clause carefully, because it is the whole appeal. Exercise activates AMPK. So does eating less. So, indirectly, does metformin, the most prescribed drug in type 2 diabetes and the one with the best long-term safety record in the entire field. AMPK is not a speculative target — it is the target that the two most reliable metabolic interventions in existence both converge on.

Which is exactly why it has been so hard to drug directly. Many compounds activate AMPK indirectly by making cells energy-stressed, which works but comes with the consequences of making cells energy-stressed. Activating the kinase directly, without poisoning the cell into doing it, is the harder problem. The 2025 liver paper states the position plainly: "Despite efforts to target AMPK, no direct activators have yet been approved for treatment of this disease."

The human trial, and exactly what it showed

The 2018 paper in JCI Insight reports both animal and human work. In diet-induced obese mice, O304 "increased glucose uptake in skeletal muscle, reduced β cell stress, and promoted β cell rest." Beta cells are the cells in your pancreas that make insulin; in type 2 diabetes they are working overtime against tissues that have stopped listening, and eventually they fail. Reducing their workload is a different and more interesting goal than pushing them harder, which is what several older diabetes drugs do.

Then the human part, in the paper's own words:

"Accordingly, O304 reduced fasting plasma glucose levels and homeostasis model assessment of insulin resistance (HOMA-IR) in a proof-of-concept phase IIa clinical trial in type 2 diabetes (T2D) patients on Metformin."

Three things in that sentence deserve attention. It was a phase IIa proof-of-concept trial, which is a small early study designed to show a signal exists, not to establish that a drug works. The patients were already on metformin, so the effect measured was on top of an AMPK-adjacent drug they were already taking. And the endpoints were fasting glucose and HOMA-IR — a calculated index of insulin resistance — rather than HbA1c, the standard measure of long-term glucose control that regulators actually require.

The vascular finding is the one that stands out most, because it appeared in both species: O304 "improved peripheral microvascular perfusion and reduced blood pressure both in animals and T2D patients." Microvascular perfusion is blood flow through the smallest vessels, and its failure is what produces the complications that actually harm people with diabetes — the eye damage, the kidney damage, the nerve damage, the foot ulcers. A metabolic drug that also improves small-vessel blood flow is addressing the thing that does the damage rather than only the number on the meter.

The heart finding, and the detail that makes it credible

The same paper reports that O304, "like exercise, activated AMPK in the heart, increased cardiac glucose uptake, reduced cardiac glycogen levels, and improved left ventricular stroke volume in mice." Stroke volume is the amount of blood the heart moves per beat, and improving it is what endurance training does to a heart.

Then one clause that matters more than it looks: "but it did not increase heart weight in mice or rats."

That is not a throwaway. A drug that makes the heart work better by making it bigger is a drug that may be producing pathological hypertrophy — the heart muscle thickening in a way that eventually stiffens it and causes failure. It is the standard concern with anything that improves cardiac performance, and it is why the authors measured heart weight in two species and reported the negative. A compound that raises stroke volume without raising heart mass is doing something structurally different from the thing you would worry about. Reporting that specifically is what a careful paper looks like.

The liver work

The 2025 study in JCI Insight tested ATX-304 in a mouse model of progressive fatty liver disease — MASLD, metabolic dysfunction-associated steatotic liver disease, which the paper notes is "the most common chronic liver disease worldwide for which there is only one approved treatment."

The finding: "ATX-304 diminishes body fat mass, lowers blood cholesterol levels, and mitigates general liver steatosis and the development of liver fibrosis, but with pronounced local heterogeneities."

That final phrase is the honest part, and it is the part a summary would drop. Steatosis is fat accumulation in the liver; fibrosis is scarring, and fibrosis is what turns a fatty liver into a failing one. Reducing both is the goal of the entire MASLD field. But "pronounced local heterogeneities" means the improvement was not uniform across the organ — the paper reports "variations in lipid distribution among liver lobes in response to ATX-304, and a shift in the zonal distribution of lipid droplets upon treatment." Different lobes responded differently, and fat moved between zones within the tissue. A biopsy is one needle into one lobe, so a treatment whose effect varies by lobe is a treatment whose effect a biopsy may over- or under-report. The authors flagged it themselves and still concluded the compound "holds promise as a potential treatment for MASLD."

The mechanism behind the liver result is the same metabolic switch: "increased fatty acid oxidation, reduced lipid synthesis, as well as remodeling of cholesterol and lipid transport." Burn more, build less, move it differently.

The kidney work, and what it says about breadth

A 2024 paper in Biomedicine & Pharmacotherapy tested ATX-304 against cisplatin-induced acute kidney injury. Cisplatin is a chemotherapy drug that damages kidneys as a side effect; the model is a well-defined, deliberately induced injury rather than a slow disease.

The design was specific: "Mice received ATX-304 (1 mg/g) or control chow for 7 days before cisplatin-induced AKI." Pre-treatment, not rescue. And the readouts were hard rather than inferred: "It protected against CI-AKI measured by serum creatinine (control 0.05 + 0.03 mM vs ATX-304 0.02 + 0.01 mM, P = 0.03), western blot for neutrophil gelatinase-associated lipocalin (NGAL) (control 3.3 + 1.8-fold vs ATX-304 1.2 + 0.55-fold, P = 0.002), and histological injury." Creatinine is the standard clinical measure of kidney function, NGAL is an early injury marker, and histology is looking at the tissue directly. Three independent measures agreeing is a stronger result than any one of them.

The paper also confirmed the mechanism was actually engaged: "ATX-304 increased acetyl-CoA carboxylase phosphorylation, indicating AMPK activation." Acetyl-CoA carboxylase is one of AMPK's direct targets, and its phosphorylation state is the standard way to prove the kinase is switched on rather than assumed to be.

Muscle, pancreas, blood vessels, heart, liver, kidney. Six tissues, four independent research groups across Sweden, the Netherlands and Australia, consistent direction of effect. That is a broader and more replicated evidence base than most compounds in this catalogue have, and the breadth is a real point in its favour.

How it compares to the drug it is standing next to

Metformin is the reference point, and the comparison is more useful than any of the individual findings.

Metformin has been prescribed since the 1950s, is taken by hundreds of millions of people, costs almost nothing, and has the best long-term safety record of any drug in metabolic medicine. It works substantially through AMPK — but indirectly. It inhibits part of the mitochondrial machinery, which raises the AMP-to-ATP ratio inside the cell, which activates AMPK because the cell is genuinely short of energy. That is why it works and also why it causes the gastrointestinal side effects it does, and why it carries a rare but real risk of lactic acidosis: the mechanism is mild, controlled energy stress.

A direct activator does not need the cell to be depleted. It binds the kinase and switches it on. In principle that decouples the benefit from the stress, which is the entire argument for developing one and the reason the field has spent two decades trying.

The trial design reflects this precisely: the patients were on metformin, and O304 was added. So the question asked was not "is this better than metformin" but "does this add something on top of it." That is the right question for an adjunct and the wrong question for a replacement, and the answer — a reduction in fasting glucose and HOMA-IR — is a signal that it adds something, in a small study, over a short period.

The vascular result is where it could differentiate. Metformin does not meaningfully improve microvascular perfusion. If a direct AMPK activator reliably does, it would be addressing the mechanism behind the complications that actually disable people with diabetes, and doing it alongside the drug they are already taking rather than instead of it. That is a genuinely valuable thing to be, and it is also a claim resting on one small company-run trial.

What it does inside the body, practically

ATX-304 is an orally available small molecule, not a peptide, and it is described as peripherally restricted — meaning it is designed not to cross into the brain. That is a deliberate design choice and a meaningful one. AMPK in the hypothalamus regulates appetite in the opposite direction to AMPK in muscle: activating it centrally tends to increase food intake, which is precisely what you do not want in an obesity drug. Keeping the compound out of the brain avoids fighting itself.

The mechanism is a metabolic switch rather than an appetite suppressant, which distinguishes it sharply from tesofensine and from the GLP-1 drugs. It does not make you want to eat less. It changes what your tissues do with the fuel they receive: more glucose pulled into muscle, more fat burned, less fat and cholesterol synthesised in the liver. The liver paper describes the same shift in the organ's own terms — "increased fatty acid oxidation, reduced lipid synthesis, as well as remodeling of cholesterol and lipid transport."

Two consequences follow for anyone reading the research. Effects on body composition through this route are slower than appetite suppression, because they depend on a sustained change in fuel handling rather than on a calorie deficit imposed within days. And because AMPK activation increases glucose uptake into muscle, the interaction that matters most is with anything else that lowers blood glucose — insulin and the sulfonylureas above all, where additive glucose lowering is how hypoglycaemia happens. The trial deliberately used metformin, which does not cause hypoglycaemia on its own, as the background therapy.

The conflict of interest, stated because it is in the record

The 2018 human paper carries a substantial conflict of interest declaration. The senior author is "a cofounder, shareholder, and consultant of the unlisted biotech company Betagenon AB." Five further authors are "all employed by, and hold shares in" Betagenon. Two more are shareholders or consultants.

This does not make the trial wrong. Almost every early clinical trial of a novel compound is run by people with a stake in it, because nobody else has a reason to fund the work. It does mean the phase IIa result is a company-run proof-of-concept study reported by people who own part of the company, and that independent replication carries weight it would not otherwise need to carry. The 2024 kidney work and the 2025 liver work come from groups without that relationship, which is part of why they matter.

Where it stands, and what remains unknown

ATX-304 has no regulatory approval anywhere. Cambrian Bio has described it as entering phase 2 development in obesity and cardiometabolic disease and has presented phase 1b human translational data, but a completed, published, peer-reviewed phase 2 obesity trial does not exist in the literature. The published human evidence is still the 2018 phase IIa study in diabetes patients on metformin.

So the honest position is this. The target is the best-validated metabolic target there is, converged on by exercise, caloric restriction and metformin. The compound engages it directly, which nothing approved does. It has produced consistent effects across six tissues and four research groups. And it has been given to human patients with a measurable, published result — which is more than SLU-PP-332, 5-amino-1MQ or tesofensine can say in their respective ways.

What is not known is what a phase 2 will show. Fasting glucose and HOMA-IR in a small proof-of-concept study are not HbA1c in a powered trial. There is no published long-term human safety data. There is no published human dose-response. AMPK sits in every cell in the body, and a direct activator of a master metabolic regulator is a systemic intervention by definition — the mouse and rat heart-weight data is reassuring on the specific concern it addresses and says nothing about the others. And the compound is being developed by a company, on a timeline set by that company, which means the absence of a published phase 2 is information about the state of the evidence and not about the state of the science.

The two names, and why it matters more than a naming quirk

It is worth returning to this because it changes what you can find out.

The molecule was published as O304 by Betagenon and Umeå University. The 2018 human trial, the 2023 follow-up on muscle glucose effectiveness and beta-cell preservation, and the earlier mechanistic work are all under that name. ATX-304 is the name it acquired on licensing, and the 2024 kidney paper and the 2025 liver paper use it.

So a search for one name returns roughly half the literature. Worse, it returns a systematically biased half: search ATX-304 alone and you find the preclinical organ-protection work and the company's phase 2 announcements, and you miss the phase IIa human trial entirely — the single most decision-relevant study that exists on this compound. Search O304 alone and you find the human data and miss the liver and kidney work.

This is not a curiosity. It is the practical reason that pages about this compound are so inconsistent about whether human data exists, and it is why anyone reading further should search both. The 2018 paper is PMID 29925691. The 2025 liver paper is PMID 40197369. The 2024 kidney paper is PMID 38749175.

Everything above is what the published record supports. None of it is medical advice, and ATX-304 is sold for research use only, not for human consumption.