ATX-304

Pan-AMPK activator

ATX-304 (Pan-AMPK activator) is a metabolic & fat loss research compound. Direct pan-AMPK activator — flips the cellular energy switch toward fat oxidation and glucose uptake, an exercise-mimetic small molecule.

How ATX-304 works

Direct pan-AMPK activator — flips the cellular energy switch toward fat oxidation and glucose uptake, an exercise-mimetic small molecule.

Proposed benefits

Researched for fat oxidation, appetite and energy regulation, insulin sensitivity and endurance capacity.

Where to get ATX-304

The evidence for ATX-304

Graded by what exists behind each claim.

✅ Clinically validated

• Reached phase 1 in humans — an oral direct AMPK activator, with early safety and pharmacodynamic work completed and metabolic indications targeted.
• Phase 1 tells you about tolerability at tested doses, not efficacy. No controlled efficacy result in any indication has been published.

📊 Correlative data

• Almost no research-market presence and no practical record to report. That absence is worth stating plainly rather than filling with extrapolation from other AMPK activators.

🧪 Theoretical / extrapolated

• A direct pan-AMPK activator — it binds AMPK itself rather than working indirectly. Metformin activates AMPK by inhibiting complex I and raising the AMP:ATP ratio; AICAR mimics AMP. This binds the allosteric drug and metabolite site directly.
• Direct activation is cleaner and blunter at once: no dependence on mitochondrial inhibition, but also no dependence on the cell's actual energy state, so the signal fires whether or not the cell needs it.
• Same mTOR tension as AICAR — sustained AMPK activation opposes protein synthesis, which is the predicted cost for anyone using it alongside training.

How to read these tiers: they say how much human evidence exists, not how well something works — and ✗ flags harm, never a disappointing trial. How the evidence tiers work →

What ATX-304 actually does

AMPK is not one enzyme. It is a heterotrimer — a catalytic α subunit, a scaffolding β subunit and a nucleotide-sensing γ subunit — and with two α, two β and three γ genes there are 12 mammalian complexes, expressed in different proportions in different tissues Myers 2017. "Pan" in the phrase pan-AMPK activator means the molecule turns on all of them rather than a subset. That single word is where both the promise and the problem live.

Physiologically, AMPK reads the adenylate charge: AMP and ADP bind the γ subunit's CBS domains, and phosphorylation of Thr172 on the α subunit by LKB1 or CaMKK2 switches the kinase on. A direct activator skips that. It reports an energy deficit that has not happened.

What ATX-304 is measured doing. Both published papers use the same pharmacodynamic readout: phosphorylation of acetyl-CoA carboxylase, the canonical AMPK substrate Katerelos 2024. That matters because the chemistry downstream is deterministic — phosphorylated ACC stops making malonyl-CoA, malonyl-CoA is the endogenous inhibitor of carnitine palmitoyltransferase 1, so removing it opens the gate that lets long-chain fatty acids into the mitochondrion. Fat oxidation up, lipogenesis down. Holm 2025 reports exactly that shift in mouse liver: increased fatty acid oxidation, reduced lipid synthesis, and remodeled cholesterol and lipid transport.

The target-validation experiment, which is the strongest single result on this page. In primary renal tubular epithelial cells, ATX-304 at 20 µM for 4 hours altered 66 of 126 measured metabolites and raised the basal oxygen consumption rate by 38% with maximal respiration unchanged. Its protective effect against cisplatin was completely lost in AMPK-null mouse embryonic fibroblasts Katerelos 2024. Losing the effect when you delete the target is what target engagement looks like when it is done properly.

What is not published, and should be. Neither paper reports a binding site, a dissociation constant, or isoform-selectivity ratios for ATX-304 itself. The class's best-characterized member, MK-8722, is a defined allosteric activator of all 12 complexes Myers 2017; O304 is a different chemotype with a different reported profile Steneberg 2018. Where ATX-304 sits between them is, in the public record, an assertion in a title.

Cell, rodent, human — and where it stops

Step one, in cells. Primary mouse tubular epithelial cells, ATX-304 20 µM for 4 hours before cisplatin 20 µM for 23 hours. Protection measured three independent ways — lactate dehydrogenase release, MTS viability, cleaved caspase 3 — and abolished in AMPK-null cells Katerelos 2024.

Step two, mice, acute kidney injury. C57Bl/6 mice given ATX-304 in chow at 1 mg/g for 7 days before cisplatin. Serum creatinine 0.05 ± 0.03 mM on control chow against 0.02 ± 0.01 mM on ATX-304 (P = 0.03); NGAL by western blot 3.3 ± 1.8-fold against 1.2 ± 0.55-fold (P = 0.002); histological injury score 3.5 ± 0.59 against 2.7 ± 0.74 (P = 0.03) Katerelos 2024.

Step three, mice, fatty liver disease. In a progressive steatotic-liver model, ATX-304 reduced body fat mass, lowered blood cholesterol, and mitigated steatosis and fibrosis — but the authors are explicit that the benefit came with pronounced local heterogeneity, varying between liver lobes and shifting the zonal distribution of lipid droplets Holm 2025. A drug whose effect differs by lobe is a drug whose biopsy result depends on where the needle went.

Step four, in humans: zero, for this molecule. The class comparator that did reach people is O304, a different compound: a proof-of-concept phase IIa in type 2 diabetes patients already on metformin reduced fasting plasma glucose and HOMA-IR, improved peripheral microvascular perfusion and reduced blood pressure Steneberg 2018. That is genuinely encouraging for the idea of activating AMPK in a person. It is not evidence about ATX-304.

The obstacle, and it is a dose obstacle. Both ATX-304 papers dose by concentration in chow or in medium, never by plasma level. Working the chow figure through: a 30 g mouse eating about 3.5 g/day of 1 mg/g chow takes roughly 3.5 mg/day, about 115 mg/kg/day. Scaled by body surface area (mouse 3, human 37) that is around 9 mg/kg, or 650 mg/day for a 70 kg adult — six to seven times the ~100 mg this site records. Nobody has measured whether the human curve sits where that arithmetic puts it, because nobody has measured a human curve at all.

ATX-304 pharmacokinetics — how much of it actually gets in

Route: oral, by the only route ever published for it. In mice it is delivered in the chow at 1 mg/g Katerelos 2024, and in cells at 20 µM. There is no injectable formulation, no subcutaneous or intraperitoneal injection study, and no published comparison of an injected against an oral dose for this molecule.

What degrades it: unpublished, in every species. No CYP assignment, no measurement of hepatic clearance or renal clearance, no metabolite identification, no protein binding. That is a real limitation rather than a formality — a small molecule with no known clearance route has no predictable interaction profile with metformin, a statin, or anything else somebody taking it for metabolic reasons is likely to already be on.

The oral barrier, reasoned rather than measured. Continuous dietary dosing over 7 days is the delivery method both papers chose, which is what investigators do when they want steady exposure and do not have — or do not report — an oral bioavailability figure. No first-pass extraction fraction has been published, so the fraction of a swallowed dose that reaches systemic circulation is unknown. What can be said is that a compound only ever given in food for a week has never been shown to work as a single daily capsule, which is how it is actually taken.

The one exposure–response anchor that exists. 20 µM for 4 hours in a renal tubular cell produced a 38% rise in basal oxygen consumption Katerelos 2024. Any argument that a ~100 mg oral human dose does something has to end with a plasma concentration near that figure, and no such number has ever been reported.

What would have to be true, and how you would know it was not

Four predictions. The fourth is the one that argues against the compound, and it names the test nobody is running.

1. Hepatic lipogenesis should fall, so the lipid panel should move first. Inhibiting acetyl-CoA carboxylase drops malonyl-CoA and reduces de novo lipogenesis; Holm 2025 measured lower blood cholesterol in mice. Draw a lipid panel and ApoB at baseline and at 12 weeks. Triglycerides falling with ApoB flat would fit the mechanism; both flat at 12 weeks, in someone whose diet has not changed, is the exposure argument from the section above showing up as a null result.

2. Glycemia should improve before body weight does. The class result in humans — with a different molecule — was reduced fasting plasma glucose and HOMA-IR Steneberg 2018. Fasting insulin and HbA1c at baseline and 12 weeks are the read-out, and the specific claim is that insulin sensitivity, not weight, is what should shift first.

3. Liver enzymes should be stable or improve, not rise. A CMP and GGT at baseline and 12 weeks. In the mouse model the liver got better, so a transaminase rise on this compound is not an expected on-target effect and should be treated as a reason to stop and look for another cause.

4. The prediction that cuts against it, and it is about the heart. MK-8722, the best-characterized pan-AMPK activator, produced cardiac hypertrophy and increased cardiac glycogen in rodents and rhesus monkeys — without apparent functional consequences, but it produced it Myers 2017. O304, a different chemotype, specifically did not increase heart weight in mice or rats Steneberg 2018, so the effect is not inevitable across the class. ATX-304's heart has never been examined in any published experiment. The falsifiable version: an echocardiogram with left ventricular mass index before starting and at 6 months, plus high-sensitivity troponin T in the same draws. A rising LV mass index is the class effect arriving; a flat one is the first real evidence this molecule does not share it.

What nobody has tested yet

Nobody has imaged a heart on ATX-304. Given what MK-8722 did in two species Myers 2017 and what O304 did not do in two others Steneberg 2018, cardiac mass on ATX-304 is the obvious experiment, it is cheap in a mouse, and the two published papers — one on kidney, one on liver — did not report it.

Nobody has published a plasma level in any species. Chow concentration is not exposure. Until a pharmacokinetic study exists, the 1 mg/g figure cannot be converted into a human dose by anything better than the body-surface-area arithmetic in the section above, which assumes identical absorption and clearance between species and is wrong more often than it is right.

Nobody has tested whether the γ2 subunit carries the cardiac effect — and this is the specific hypothesis worth stating. Mutations in the AMPK γ2 gene PRKAG2 cause a hereditary human cardiomyopathy with glycogen accumulation, and what MK-8722 produced was cardiac hypertrophy with increased cardiac glycogen Myers 2017. The resemblance is close enough to be worth a direct test: give a pan-activator to a γ2-deficient mouse and see whether the hypertrophy disappears. If it does, the class problem is an isoform problem, and a γ2-sparing activator becomes a rational design target rather than a hope. Labeled as what it is — an inference from two facts that have never been put in the same experiment.

Nobody has tested it against exercise. AMPK is the enzyme muscle contraction activates. Whether a direct activator adds anything to a person who already trains, or merely occupies a pathway that is already maximally engaged for an hour a day, is a two-arm study with a VO2 test and a muscle biopsy, and it has never been done for any compound in this class.

ATX-304 — its own safety story, not its class's

This compound's own adverse-event record is empty, and that has to be said before anything else. Two published papers, one species, both reporting protection. No repeat-dose toxicology study, no maximum tolerated dose, no human exposure, no adverse-event table anywhere Katerelos 2024 Holm 2025. Everything below this sentence is class reasoning, explicitly labeled as such, because class reasoning is all that exists.

The class's known problem is cardiac and it is specific. MK-8722 gave rodents and rhesus monkeys cardiac hypertrophy with increased cardiac glycogen Myers 2017. That is not a vague signal; it is the reason the systemic pan-activation strategy stalled despite producing robust, insulin-independent glucose uptake in skeletal muscle with no hypoglycemia. The counterweight is that O304 was reported to activate AMPK in the heart, increase cardiac glucose uptake, reduce cardiac glycogen and improve left ventricular stroke volume in mice without increasing heart weight in mice or rats Steneberg 2018. Two pan-AMPK activators, opposite cardiac results. That is precisely why ATX-304's silence on the heart is a gap and not a reassurance.

The second risk is the one its own data show. Four hours at 20 µM moved 66 of 126 measured metabolites in a renal tubular cell Katerelos 2024. A molecule that reorganizes half of a cell's measured metabolome in an afternoon is not a gentle intervention, and the fact that the reorganization was protective against a platinum chemotherapy does not tell you what it is in a healthy kidney over a year.

What is on the card, and what is behind it. This site lists roughly 100 mg orally. There is no published human dose for ATX-304, so that figure cannot be traced to a trial, a label or a pharmacokinetic study — it is a number in circulation, not a number in the literature. Anyone taking it should treat the dose itself as the untested variable.

Sources read for this page

• Katerelos M, Gleich K, Harley G, Loh K, Oakhill JS, Kemp BE, de Souza DP, Narayana VK, Coughlan MT, Laskowski A, Ling NXY, Murray-Segal L, Brink R, Lee M, Power DA, Mount PF. The AMPK activator ATX-304 alters cellular metabolism to protect against cisplatin-induced acute kidney injury. Biomed Pharmacother 2024 · PMID 38749175
• Holm E, Vermeulen I, Parween S, Lopez-Perez A, Cillero-Pastor B, Vandenbosch M, Remeseiro S, Hornblad A. AMPK activator ATX-304 reduces oxidative stress and improves MASLD via metabolic switching. JCI Insight 2025 · PMID 40197369
• Myers RW, Guan HP, Ehrhart J, Petrov A, Prahalada S, Tozzo E, Yang X, Kurtz MM, Trujillo M, Gonzalez Trotter D, Feng D, Xu S, Eiermann G, Holahan MA, Rubins D, Conarello S, Niu X, Souza SC, Miller C, Liu J, Lu K, Feng W, Li Y, Painter RE, Milligan JA, He H, Liu F, Ogawa A, Wisniewski D, Rohm RJ, Wang L, Bunzel M, Qian Y, Zhu W, Wang H, Bennet B, LaFranco Scheuch L, Fernandez GE, Li C, Klimas M, Zhou G, van Heek M, Biftu T, Weber A, Kelley DE, Thornberry N, Erion MD, Kemp DM, Sebhat IK. Systemic pan-AMPK activator MK-8722 improves glucose homeostasis but induces cardiac hypertrophy.. Science 2017 · PMID 28705990
• Steneberg P, Lindahl E, Dahl U, Lidh E, Straseviciene J, Backlund F, Kjellkvist E, Berggren E, Lundberg I, Bergqvist I, Ericsson M, Eriksson B, Linde K, Westman J, Edlund T, Edlund H. PAN-AMPK activator O304 improves glucose homeostasis and microvascular perfusion in mice and type 2 diabetes patients.. JCI Insight 2018 · PMID 29925691

ATX-304 — safety, predicted from mechanism

ATX-304 — interference & stacking

Bloodwork to run alongside ATX-304

Baseline first, then again at 8–12 weeks.

The Metabolic Health & Prediabetes panel covers these in one order — 8 markers, $81.45 with the discount applied.

Check results you already have → · All 103 markers A–Z

ATX-304 — frequently asked questions

ATX-304 inside a finished plan

One arm of 1 Protocol Blueprint, free to read in full.

What ATX-304 is used for

ATX-304 appears under 3 goals in the goal router.

Related Metabolic & Fat Loss compounds

← Explore the full Protocol Vault

↑ Back to on this page