What Is ATX-304? The Small-Molecule AMPK Activator With a Mitochondrial Twist

ATX-304 goes by a chemistry-lab name, but it began as O304, a compound built to activate every isoform of AMP-activated protein kinase (AMPK) at once — the enzyme that senses when a cell is short on energy and tells it to start burning fuel. It is not a peptide; it is a small molecule, taken orally, developed originally for type 2 diabetes and now studied across liver, kidney, and vascular disease models. Preclinical data are extensive. Human data are thin. Everything below describes laboratory, animal, and early clinical research, not a product for human use.

From O304 to ATX-304 — a small molecule, not a peptide

Researchers at Umeå University and the Swedish biotech Betagenon first described O304 in 2018 as a “pan-AMPK activator” — meaning it turns on all known combinations of AMPK subunits rather than just one isoform, a property most existing AMPK-directed tools lack¹. In diet-induced obese mice, the compound increased skeletal-muscle glucose uptake, improved microvascular perfusion, and reduced blood pressure; in a small proof-of-concept study in people with type 2 diabetes already on metformin, it lowered fasting glucose and insulin resistance scores¹. The molecule was later renamed ATX-304 as it moved into the hands of Amplifier Therapeutics and its parent, Cambrian Bio, for further clinical development.

Worth stating plainly, because readers arrive here from peptide research and the mix-up is common: ATX-304 is a synthetic organic small molecule, not a peptide, and it has no structural relationship to peptide compounds sold for research such as BPC-157 or the growth-hormone secretagogues. It is closer, chemically and pharmacologically, to metabolic small molecules like metformin or the mitochondrial uncouplers discussed later in this piece.

What a pan-AMPK activator is supposed to do

AMPK is often described as the cell’s fuel gauge. When the ratio of AMP to ATP rises — during fasting, calorie restriction, or muscle contraction during exercise — AMPK switches on, shifting the cell toward burning fat and glucose and away from energy-expensive building processes². Because exercise is one of the most reliable natural activators of AMPK in skeletal muscle, drug developers have spent two decades chasing molecules that flip the same switch pharmacologically, hoping to reproduce some of exercise’s metabolic benefits without the exercise. Reviews of the field are candid that this has been harder than it sounds: most direct AMPK activators either hit only one subunit combination, do not reach the right tissues, or fail to reproduce the coordinated, whole-body response that real muscle contraction generates³. ATX-304’s selling point is that it was designed to sidestep the first problem by activating the full AMPK family rather than a single isoform. Our companion piece on SLU-PP-332 covers a different attempt at the same goal, using an entirely different receptor system, which is a useful comparison for how crowded — and how early-stage — this “exercise mimetic” category still is.

The preclinical case: diabetes, fatty liver, kidney, and blood vessels

The published animal literature on O304/ATX-304 is unusually broad for a compound this early in development. In mice with diet-induced or genetic hyperglycemia, O304 improved glucose tolerance through two separate routes at once — pushing muscle to take up glucose independent of insulin, while also reducing stress on insulin-producing beta cells — an effect the authors describe as dual and largely non-redundant with existing diabetes drugs⁴. In aged mice, the same compound reversed some age-related insulin resistance, improved cardiac function, and increased exercise capacity, which is the closest preclinical data come to supporting the “exercise mimetic” framing directly⁵.

More recent work has moved beyond metabolic disease into organ injury. A 2025 study using ATX-304 in mice with metabolic dysfunction-associated steatotic liver disease (MASLD) reported reduced oxidative stress, less liver fat accumulation, and attenuated fibrosis progression, attributing the effect to a shift in how liver cells partition fuel⁶. A separate 2024 study found that ATX-304 protected mouse kidneys against cisplatin-induced acute injury, again invoking AMPK-driven metabolic reprogramming as the mechanism, this time in kidney tubule cells under chemotherapy-induced stress⁷. And a 2024 paper linked O304 to reduced abdominal aortic aneurysm formation in a mouse model, through AMPK’s downstream effects on mTOR signaling and matrix metalloproteinase activity in vascular smooth muscle⁸. Taken together, this is a genuinely wide preclinical footprint — diabetes, fatty liver, acute kidney injury, and vascular disease — all attributed to the same proposed mechanism.

The complication: mitochondrial uncoupler or AMPK activator?

That tidy, single-mechanism story ran into a serious complication in 2025. A paper in Chemico-Biological Interactions examined O304 directly at the mitochondrial level and reported that it functions as a mitochondrial uncoupler — a compound that lets protons leak back across the inner mitochondrial membrane instead of flowing through ATP synthase, so the cell burns fuel and generates heat without capturing the usual amount of usable energy. In that study, O304 extended lifespan in C. elegans and produced vasorelaxation in isolated rat mesenteric arteries, effects the authors tie to uncoupling and downstream autophagy rather than to AMPK activation as the primary event⁹. Uncoupling and AMPK activation are not mutually exclusive on paper: dissipating the mitochondrial proton gradient lowers ATP output relative to demand, which raises the AMP-to-ATP ratio — the exact signal AMPK is built to detect. So a molecule could uncouple mitochondria first and activate AMPK as a downstream consequence, rather than binding AMPK directly as its primary target. Whether that is what is actually happening with ATX-304, or whether direct AMPK-subunit binding remains the dominant effect at physiological doses, is not settled in the published literature.

This is not a new tension in the field. BAM15, a differently-structured but mechanistically similar mitochondrial uncoupler, has been studied for a decade as a tool for raising energy expenditure without the toxicity of older uncouplers like 2,4-dinitrophenol¹⁰, and more recent reviews frame BAM15 explicitly as a therapeutic candidate for obesity and metabolic disease through uncoupling alone, with no AMPK-first framing at all¹¹. We cover that compound’s evidence base separately in our BAM15 article. The point of setting these two side by side is not to declare ATX-304 “really” a BAM15 clone — the compounds are chemically distinct, and no head-to-head potency comparison has been published — but to show that two legitimate, peer-reviewed mechanistic explanations currently exist for the same molecule’s effects, and a supplier who tells you it is settled would be overstating the literature.

Two peer-reviewed lines of evidence describe the same molecule in different languages — one calls it a kinase activator, the other calls it a proton leak. Both cannot be the whole story, and neither paper claims it is.

Human data: one small, unpublished trial

Almost everything above comes from cells and mice. The human evidence picture is much narrower. In June 2026, Cambrian Bio presented data at the American Diabetes Association’s 86th Scientific Sessions from a phase 1b study of ATX-304 in adults with obesity and prediabetes: a randomized, double-blind, placebo-controlled design with an optional open-label extension, enrolling approximately 23 participants over an eight-week treatment period¹².

8% increase in resting metabolic rate reported in that trial, alongside statistically significant reductions in liver fat (by MRI-PDFF), visceral adipose tissue, and circulating triglycerides, plus an increase in adiponectin — but the data come from a company press release tied to a conference presentation, not a peer-reviewed publication, and weight loss itself was minimal at the exposure level tested.

That distinction matters. A conference poster and an accompanying press release are not the same evidentiary weight as a paper that has gone through peer review, with a methods section and reviewer scrutiny available to the public. The reported numbers are specific and the direction is consistent with the proposed mechanism, which is a reasonable basis for continued development — the sponsor has stated plans for two further Phase 2 studies — but it is not yet independently verifiable data, and the sample size (roughly two dozen people, split between active drug and placebo) is too small to rule out chance findings on several of the secondary endpoints measured.

Honest read: what the evidence actually supports

Pull the threads together and three separate claims need to be evaluated on their own merits, not treated as one bundle. First, ATX-304 reliably changes metabolic parameters in mice across several disease models — that part of the record is broad and consistent across independent labs. Second, the mechanism behind those changes is genuinely contested in the peer-reviewed literature, with AMPK activation and mitochondrial uncoupling both having direct experimental support and no published work yet reconciling the two into a single account. Third, the human translation story — the part that would matter most to anyone hearing “exercise mimetic” — currently rests on one small, unpublished dataset. None of that makes ATX-304 uninteresting as a research compound; if anything, an oral molecule with this breadth of preclinical activity and an open mechanistic question is exactly the kind of thing worth continued laboratory study. It does mean any framing of ATX-304 as a settled, mechanism-understood “exercise in a pill” is running well ahead of what has actually been published.

Three compounds studied under the “exercise mimetic” label, none of them peptides. ATX-304 is the only one with any reported human dosing data, and that data has not yet cleared peer review.

All compounds named here are discussed as reference materials for laboratory research use only. They are not medicines, are not approved by the EMA, FDA or any other regulator for any indication, and are not intended for human or veterinary use. Nothing above is medical advice, a therapeutic claim or dosing guidance.

Condor Research · Scientific desk
Atrio Sciences s.r.o., IČO 57 669 651, Nitra (SK) · info@condorresearch.com