ATX-304 (O-304)

Verified against PubChem PUG REST, CID 50923806 — returned formula C16H11Cl2N3O2S, MW 380.2, IUPAC name 4-chloro-N-[2-[(4-chlorophenyl)methyl]-3-oxo-1,2,4-thiadiazol-5-yl]benzamide, InChIKey WEDWLYRQKUTOAX-UHFFFAOYSA-N, CAS 1261289-04-6, UNII SPS2TLH4CM. Fields marked “not verified” are ones we could not confirm against a primary chemical database — we leave them blank rather than guess.

Two names, one molecule

ATX-304 and O-304 designate the same compound. PubChem CID 50923806 resolves it as 4-chloro-N-[2-[(4-chlorophenyl)methyl]-3-oxo-1,2,4-thiadiazol-5-yl]benzamide, CAS 1261289-04-6, molecular formula C16H11Cl2N3O2S, molecular weight 380.2 g/mol, InChIKey WEDWLYRQKUTOAX-UHFFFAOYSA-N, UNII SPS2TLH4CM. Structurally it is a 1,2,4-thiadiazol-3(2H)-one ring carrying a benzyl substituent and a chlorobenzamide. It is a synthetic small molecule, not a peptide, and the analytical and handling conventions that apply to lyophilized peptides do not transfer to it.

The naming split matters for anyone reading the literature. The O-304 designation dominates papers published between 2018 and 2021, nearly all originating from Umeå University and the Swedish company Betagenon AB. Papers from 2024 onward use ATX-304. A literature search on either name alone misses roughly half the published work, which is one reason secondary summaries of this compound tend to be incomplete in one direction or the other.

No ChEMBL parent identifier was located for this compound during verification — PubChem lists only SCHEMBL deposition identifiers, which are automated structure deposits rather than curated bioactivity records. That absence is itself informative: it indicates the compound has not accumulated the curated cross-referenced bioactivity dataset that better-studied kinase modulators carry.

The mechanism described in the primary literature

AMPK is a heterotrimeric kinase whose activity depends on phosphorylation of a threonine residue, Thr172, on the catalytic alpha subunit. The mechanistic account given across the retrieved papers is that this compound does not phosphorylate Thr172 itself but suppresses its dephosphorylation, so the pool of kinase already in the active state persists longer. The description used is pan-activation: action across AMPK heterotrimer combinations rather than selectivity for one isoform pairing.

The standard readout in these papers is phosphorylation of acetyl-CoA carboxylase, a direct AMPK substrate. Katerelos and colleagues reported increased ACC phosphorylation in mouse tissue and cultured tubular epithelial cells as their confirmation that the kinase had been engaged (PMID 38749175).

That study also carried the most useful mechanistic control in the retrieved record. Protection against cisplatin injury in cultured cells was abolished in AMPK-null murine embryonic fibroblasts. A compound producing an effect that disappears when its putative target is deleted is a substantially stronger mechanistic claim than one supported by downstream phosphorylation alone, and this control is worth noting because most of the other papers do not include an equivalent.

A separate reported action is increased cellular respiration through mitochondrial uncoupling, measured directly as oxygen consumption rate in the same study.

The single human trial in the peer-reviewed record

One human trial has been published. Steneberg and colleagues described TELLUS, an exploratory proof-of-concept phase IIa study: randomized, parallel-group, double-blinded and placebo-controlled, running 28 days in 65 adults with type 2 diabetes who were already stable on metformin (PMID 29925691). Over that period the treated group recorded lower fasting plasma glucose and lower HOMA-IR than placebo, along with improved peripheral microvascular perfusion and reduced blood pressure.

Several features bound what this trial can support. Twenty-eight days is short for glycaemic endpoints; HbA1c reflects roughly three months of glycaemia and cannot move meaningfully in four weeks, which is why the reported endpoints are fasting glucose and a calculated insulin-resistance index rather than the endpoint diabetes drugs are ordinarily judged on. The population was already on metformin, so the finding is an add-on effect in treated patients, not a standalone one. And it is explicitly labelled exploratory and proof-of-concept by its own authors.

No registration record for this trial could be located. A search of ClinicalTrials.gov by intervention name across O304, O-304 and ATX-304 returned zero records, as did a sponsor search across Betagenon and its successor entities. This is documented further in the unsourced claims below.

The rodent record, and where it concentrates

The animal literature is broader than the human literature and consistently points in one direction. In diet-induced obese mice, the same 2018 paper reported increased skeletal-muscle glucose uptake, reduced beta-cell stress and promotion of beta-cell rest, alongside cardiac findings: increased cardiac AMPK activation and glucose uptake, reduced cardiac glycogen and improved left-ventricular stroke volume without an increase in heart weight in either mice or rats (PMID 29925691). The heart-weight observation is a deliberate negative control, since compounds that improve cardiac output by inducing hypertrophy are a known failure mode.

Ericsson and colleagues extended the work to aged mice, reporting prevention and reversal of age-associated hyperinsulinemia and insulin resistance together with improved cardiac function and exercise capacity (PMID 34795407). Holm and colleagues studied a choline-deficient high-fat mouse model of progressive fatty liver disease and reported reduced body fat mass, lowered blood cholesterol, mitigated steatosis and fibrosis progression, and a shift in liver metabolism toward fatty-acid oxidation — with the explicit caveat, stated by the authors, of pronounced local heterogeneity between and within liver lobes (PMID 40197369).

López-Pérez and colleagues reported that the compound largely prevented high-fat-diet-associated gene expression changes in pancreatic islets of CBA mice, accompanied by remodelling of active and repressive chromatin marks (PMID 34949756).

Recent human data that has not been published

The most striking figures now circulating about this compound do not come from journals. Between 2025 and 2026, press releases and conference presentations attributed to Cambrian Bio described phase 1b human translational data in adults with obesity and prediabetes, reporting changes in liver fat, visceral adipose tissue, triglycerides, adiponectin and resting metabolic rate, and separately a weight-loss figure with a stated fat-to-lean-mass split. Two phase 2 programmes were named.

None of this has been located in a peer-reviewed journal, and no corresponding registration was found on ClinicalTrials.gov under any sponsor or intervention name searched. Conference abstracts and company announcements are legitimate scientific communication, but they are not the same evidentiary object as a published trial: they are not peer-reviewed, they do not carry a full statistical analysis plan, adverse-event tables, or dropout accounting, and their headline figures are selected by the party announcing them.

The practical consequence is that the strongest numbers now attached to this compound in general circulation are also the least verifiable ones. They are logged in the unsourced table below rather than the ledger, which is the distinction this page exists to draw. If those studies are published, the figures move up into the ledger; until then they remain announcements.

How concentrated this evidence base is

A structural feature of this record deserves attention independently of any individual finding. Four of the five papers in the ledger below share overlapping authorship and institutional origin at Umeå University, with the corresponding author Helena Edlund appearing on the 2018, 2021 metabolic-aging and 2021 islet-epigenetics papers, and Andreas Hörnblad on the islet and 2025 liver papers. Betagenon AB, the company developing the compound, contributed co-authors to the foundational 2018 study.

This is normal for a compound at this stage — a molecule is usually characterised by the group that discovered it before anyone else takes an interest — and it is not an accusation of error. But it does mean that the mouse metabolic findings have not, in the retrieved literature, been independently replicated by an unaffiliated laboratory.

The clearest exception is the 2024 kidney work by Katerelos and colleagues at Austin Health and Monash University in Australia (PMID 38749175), a fully independent group working on a different organ system and a different injury model, which reported AMPK-dependent metabolic reprogramming and included the AMPK-null control. Independent replication of the metabolic endpoints specifically — glucose uptake, insulin sensitivity, body composition — was not found.

What is not known

The human evidence consists of one exploratory 28-day trial in 65 people who were already taking metformin. Nothing is established about effects beyond four weeks, about use without background metformin, or about HbA1c, which cannot respond within a 28-day window. No published human data exists for people without type 2 diabetes, and none for the endpoints most often discussed informally in connection with this compound — body composition, endurance, or ageing — where the entire evidentiary basis is mouse work or unpublished company announcements. No published human pharmacokinetic profile, no dose-ranging study, no long-term safety dataset, and no adverse-event table were located during verification. The mouse metabolic findings originate almost entirely from one research group and its commercial partner and have not been independently replicated by an unaffiliated laboratory; the one clearly independent study addressed kidney injury, not metabolism. Reproductive, developmental, hepatic and carcinogenicity toxicology were not found in any retrieved source. The compound has no approval from any regulator in any jurisdiction, and the fact that a mechanism is well described is not evidence that engaging it in humans produces a durable benefit — AMPK activation has been pursued pharmacologically for two decades without an approved direct activator emerging.

Questions