Dihexa: The HGF/c-Met Claim and the Distance Between a Rodent Model and a Person

Dihexa is claimed to work by binding hepatocyte growth factor (HGF) and potentiating HGF's activation of its receptor, the c-Met tyrosine kinase, in the brain. The evidence for that claim is in rats and cell culture, from one laboratory. The paper that made the HGF/c-Met dependence explicit was retracted in April 2025 after a university investigation found falsified or fabricated data 34. There are no published human data on Dihexa itself: no pharmacokinetic study, no safety study, no trial.

That leaves a large gap between the rodent model and a person, and this article is about that gap. It sets out what Dihexa is, what the surviving rodent data show, what was withdrawn, what the related clinical programme found, and why the pathway it targets demands more caution than most.

From angiotensin IV to Dihexa

Angiotensin IV is the hexapeptide Val-Tyr-Ile-His-Pro-Phe, a fragment of angiotensin II produced by aminopeptidase cleavage. From the 1990s a group at Washington State University led by Joseph Harding reported that angiotensin IV and its analogues improved learning and memory in rats. The practical problem was the usual one for short peptides: fast degradation, and no passage across the gut or the blood-brain barrier.

The 2013 paper that introduced Dihexa set out to solve this 1. Earlier work had located the procognitive activity of a norleucine-substituted angiotensin IV analogue in its first three residues. The group capped that tripeptide at both ends, adding an N-terminal hexanoic acid and a C-terminal aminohexanoic amide. The result is N-hexanoic-Tyr-Ile-(6)-aminohexanoic amide, a hydrophobic molecule the authors reported to be orally active and brain-penetrant. In Sprague-Dawley rats, oral Dihexa reversed scopolamine-induced deficits in the Morris water maze and improved performance in aged rats. The paper also reported marked synaptogenic activity in cultured neurones 1.

In 2021 the journal attached a notice of concern to that paper 2. A notice of concern is not a retraction. It signals that the editors have questions about the reliability of some of the data, and readers should treat the paper's findings with that caution.

The HGF/c-Met claim and its retraction

The mechanistic step came in 2014 3. The paper claimed that Dihexa bound HGF with high affinity, and that Dihexa and its parent analogue induced c-Met phosphorylation when HGF was present at concentrations too low to act alone. It also claimed that an HGF antagonist infused into the brain ventricles blocked the maze effect of oral Dihexa. If true, that last experiment would have been the decisive one. It is the kind of blockade test that turns a correlation into a mechanism.

On 29 April 2025 the journal retracted the paper at the editor's request. The retraction notice states that an investigation by Washington State University found that Figures 1B, 2A and 2C, and data in a later erratum submission, contained falsified and/or fabricated data. It names Leen Kawas and Joseph Harding as solely responsible 4. A 2012 paper from the same group, which had first described angiotensin IV analogues as modifiers of HGF/Met, was retracted at the same time 5.

What remains of the evidence

The table shows the problem plainly. Every surviving claim comes from rodents or cultured cells, from one laboratory, in a paper the journal has flagged. The mechanism claims are withdrawn. The human column is empty. The evidence has not been independently replicated in any published form that makes up for this.

What the rodent models can and cannot show

Even taken at face value, the surviving behavioural data answer a narrower question than they are usually said to. The scopolamine model makes rats forgetful by blocking muscarinic acetylcholine receptors, and a compound that reverses the deficit has counteracted an acute, drug-induced cholinergic lesion 1. That is a legitimate screening model. It is also one with a long record of producing positive results for compounds that later failed in Alzheimer's disease, because the human disease involves progressive neurodegeneration, not a reversible receptor blockade.

The Morris water maze measures how quickly a rat learns and remembers the location of a hidden platform in a pool. It is sensitive to hippocampal function and to many other things, including motivation, swimming ability, stress and visual acuity. A change in escape latency is informative when these confounds are controlled and assessment is blinded. The published abstracts do not allow a reader to judge how well that was done here, and the notice of concern means the underlying data now need more scrutiny than a typical paper, not less 2.

The synaptogenesis data come from cultured neurones, where dendritic spine density is counted after exposure to the compound. Cell culture is where the retracted HGF experiments were also done 34. More spines in a dish are consistent with a pro-plasticity effect. They are not evidence that the same thing happens in a living brain, or that it would improve cognition if it did.

How the claim outlived its evidence

Descriptions of Dihexa in secondary sources still commonly state as fact that it is an exceptionally potent synapse-forming agent and that it works through HGF. Both claims trace back to the originating laboratory's own papers, including the one now retracted 3. A retraction removes a paper from the scientific record. It does not remove the sentences that were copied from it into reviews, summaries and supplier descriptions over the following decade. Readers meeting a confident mechanistic description of Dihexa should check whether it cites a paper published before April 2025, and if so which one.

The related clinical programme

The closest thing to human evidence comes from a different molecule. The Washington State University work was spun out into a company later renamed Athira Pharma, which developed fosgonimeton, a prodrug described as a positive modulator of HGF/MET signalling. Alzforum records the compound as possibly related to Dihexa, not identical to it 7. Fosgonimeton missed its primary endpoint in the ACT-AD trial, with 77 participants, reported in 2022. It then missed the primary endpoint, a combined cognition and function score at 26 weeks, and its key secondary endpoints in the phase 2/3 LIFT-AD trial in mild-to-moderate Alzheimer's disease, announced in September 2024 7.

Alzforum also records that Athira's chief executive, Leen Kawas, was placed on leave in June 2021 over research-integrity questions. An independent investigation found altered images in her dissertation and at least four publications from her graduate work, and she resigned in October 2021 7. None of this is evidence about Dihexa's pharmacology in humans. It shows the one route by which that pharmacology might have been tested clinically, and that the targeted hypothesis did not produce a positive result there.

Why the target raises a specific concern

HGF and c-Met are not a neurotrophic system that happens to be expressed in tumours. They are a core oncogenic signalling axis. c-Met drives cell proliferation, motility, invasion and survival. Aberrant activation, whether through amplification, mutation or excess HGF, is established in many human cancers and in metastatic spread 6. Several approved cancer drugs work by inhibiting it.

A compound whose proposed purpose is to amplify HGF/c-Met signalling, taken systemically and orally, therefore raises a question that has not been asked of it in any published study: what does chronic potentiation of this pathway do outside the brain? The effect would not be confined to neurones. No carcinogenicity study, long-term toxicology study or tissue-distribution study of Dihexa appears in the literature. The absence is not a finding of harm. But the mechanism itself, if real, predicts this is the risk that most needs measuring.

Regulatory position in 2026

Dihexa is not approved as a medicine by any major regulator. The FDA sorts bulk substances nominated for compounding under section 503A into categories, and Category 2 holds substances for which it has identified significant safety risks 8. On 15 April 2026 the FDA announced that twelve peptides, including Dihexa acetate, would be removed from Category 2. Dihexa was among five set aside for a later advisory committee review expected before February 2027, not the July 2026 meeting 9. The removal followed withdrawal of the nominations. It did not place Dihexa in Category 1 and does not in itself make it eligible for compounding 10.

What would be needed

1. Independent replication, by an unconnected laboratory, of the HGF-binding and c-Met potentiation claims, with raw data available.
2. Independent replication of the rodent behavioural results with blinded assessment.
3. Pharmacokinetics in animals and then humans: oral bioavailability, brain exposure, half-life and tissue distribution.
4. Long-term toxicology and carcinogenicity studies matched to the proposed mechanism.
5. Only after these, a phase 1 study in humans.

In summary, the originating group reported that Dihexa improves maze performance in rats 1. The claim that explained how it did so has been retracted 45. The related clinical compound failed 7. The pathway it is said to amplify is oncogenic 6. What has been demonstrated is a behavioural effect in rodents, in a flagged paper, and there is no human data at all.