Dihexa: what the research actually shows about the angiotensin IV analog

Dihexa, full name N-hexanoic-Tyr-Ile-(6)-aminohexanoic amide, is a synthetic peptidomimetic ^[1]. It was designed at Washington State University by Joseph Harding and John Wright in the early 2010s ^[2]. The starting point was angiotensin IV, a hexapeptide fragment of angiotensin II that has procognitive properties in animal models but can't cross the blood-brain barrier on its own. Native Ang IV also gets chewed up almost instantly by plasma peptidases.

So the WSU group did what medicinal chemists do. They stabilized the molecule, made it small and hydrophobic enough to slip into the brain, and stripped it down to two amino acid-like residues bracketed by fatty-acid caps. The result, in rats, was orally bioavailable and BBB-permeable. Molecular weight came in around 504 Da ^[3].

The compound never moved into human trials under the dihexa name. A successor molecule, NDX-1017, later renamed ATH-1017, was advanced by a WSU spin-out, M3 Biotechnology, which became Athira Pharma ^[4]. That's the molecule that reached the clinic. Dihexa itself didn't.

Most people associate angiotensin with blood pressure. The renin-angiotensin system is best known for angiotensin II and the AT1 receptor, the target of every ARB in the cardiology formulary. Angiotensin IV is a downstream fragment that does something different. It binds the AT4 receptor, which turns out to be IRAP (insulin-regulated aminopeptidase), and it appears to influence hippocampal memory consolidation in rodent models.

The catch with native Ang IV is delivery. It doesn't cross the BBB at meaningful concentrations, and its half-life in blood is measured in seconds.

Dihexa was built to solve both problems and, the WSU group claimed, to do something more interesting than agonizing the AT4 receptor directly: act on a different pathway entirely.

Most peptides marketed for cognition fall into a few buckets. Racetam-adjacent cholinergic modulators. GHRH analogs that work via downstream IGF-1. The Russian-origin nootropic peptides like Semax and Selank, which appear to nudge BDNF and various neurotransmitter systems. Cerebrolysin, a porcine brain hydrolysate with a heterogeneous mechanism nobody fully maps.

Dihexa is in a different category. The proposed mechanism is direct synaptogenesis, the formation of new functional synaptic connections, via potentiation of hepatocyte growth factor signaling at the c-Met receptor. Not modulation of an existing neurotransmitter system. Not trophic support for existing circuits. New connections.

That's the claim. Whether the claim survives contact with the retraction record is a separate question.

Here's the proposed mechanism, as laid out in the original WSU papers. Dihexa binds HGF, hepatocyte growth factor, with high affinity. It then dimerizes with endogenous HGF, and that complex potentiates activation of the c-Met receptor tyrosine kinase at picomolar to subnanomolar concentrations. C-Met activation drives downstream PI3K/AKT, MAPK/ERK, and STAT3 signaling. In neurons, that cascade is linked to dendritic spine formation and the assembly of new excitatory synapses, particularly in the hippocampus.

In cultured hippocampal neurons, dihexa markedly increased the frequency of miniature excitatory postsynaptic currents (mEPSCs). That's an electrophysiological readout for functional synapse number. A 2-fold increase is large.

There's a reason this mechanism is interesting beyond the magnitude. HGF/c-Met signaling appears to be upregulated under conditions of injury and neurodegeneration, which would mean the pathway gets engaged preferentially when something is wrong rather than during baseline cognition. The animal data line up with this: dihexa improved performance in cognitively impaired rats but did not improve cognition in rats with normal baseline function.

The retraction problem. The two foundational papers establishing the HGF/c-Met mechanism for dihexa specifically, Kawas et al. 2012 and Benoist et al. 2014 (PMID 25187433), were both formally retracted in 2025, following Notices of Concern issued in 2021. The independent neuroscience behind c-Met's role at excitatory synapses (Tyndall and Walikonis, 2006-2007) is intact ^[5]. The biochemistry tying dihexa to that pathway is not.

That's a meaningful gap. It doesn't mean dihexa doesn't activate c-Met. It means the published evidence that it does has been pulled, and the field is currently relying on a single 2021 paper from an independent Chinese group (Sun et al., PMID 34827486) that confirmed PI3K/AKT involvement in an Alzheimer's mouse model ^[6]. That paper used dihexa, observed an effect, and traced it to a c-Met-adjacent pathway. It did not re-establish the binding biochemistry.

The bulk of dihexa's evidence comes from a handful of papers, most from one lab.

McCoy et al. 2013 (PMID 23055539) reported that dihexa reversed scopolamine-induced spatial memory deficits in aged Fischer 344 rats in the Morris water maze ^[7]. Picomolar systemic doses produced effects. The same paper reported increases in dendritic spine density and synaptic markers (VGLUT1, synapsin, PSD-95) in hippocampal tissue. This paper carries a Notice of Concern but has not been formally retracted.

Benoist et al. 2014 confirmed the HGF binding and c-Met activation story and showed that intracerebroventricular injection of an HGF antagonist blocked the procognitive effects of oral dihexa. This paper was retracted in April 2025.

Sun et al. 2021 is the most important independent replication. The group, working at China Pharmaceutical University and Nanjing Medical University, tested dihexa in APP/PS1 transgenic Alzheimer's mice ^[8]. They reported restored Morris water maze performance, increased neuronal density on Nissl staining, increased synaptophysin, reduced neuroinflammation (lower IL-1β and TNF-α, higher IL-10), and, critically, reversal of effects when the PI3K inhibitor wortmannin was co-administered. That last point is the cleanest mechanistic evidence in the dihexa literature post-retraction.

Two more studies are worth knowing about. Uribe et al. 2015 in zebrafish showed dihexa protected lateral line hair cells from aminoglycoside ototoxicity at 1 µM ^[9], a tissue-protection finding outside the cognitive frame. Weiss et al. 2021 from Madigan Army Medical Center tested 2-4 mg/kg dihexa in combination with mesenchymal stem cells and G-CSF in a rat sciatic nerve transection model, reporting significant sensory improvement over the course of the follow-up period.

A 2018 systematic review on Ang IV and analogs pulled together dozens of animal studies examining cognitive effects. Most cognitive-deficit studies and a majority of normal-animal studies showed benefit. Intracerebroventricular administration was the most consistent route.

That's the corpus. Roughly six experimental papers of direct relevance, one of them retracted, one carrying a Notice of Concern, the strongest independent replication coming from a single 2021 paper.

In animal models, dihexa has been tested for spatial working memory (Morris water maze), passive avoidance learning, scopolamine-induced cognitive impairment, age-related cognitive decline in aged rats, and cognitive deficits in APP/PS1 Alzheimer's mice. The pattern across studies is consistent: effects show up under conditions of impairment, not in cognitively normal animals.

What hasn't been studied in any peer-reviewed animal model: attention, executive function as measured by anything analogous to human tests, mood-related endpoints, addiction or reward circuitry, or sleep. The biohacker community's framing of dihexa as a broad cognitive enhancer maps onto a much narrower preclinical story about hippocampal-dependent memory in injury and disease models.

This is where dihexa was always pointed. The WSU group framed the molecule from the start as an Alzheimer's candidate, and the successor compound ATH-1017 carried that framing into the clinic under Athira Pharma. The pitch goes like this: in AD, synapse loss precedes and predicts cognitive decline. If you can drive new synapse formation pharmacologically, you potentially address the disease at a structural rather than purely symptomatic level.

The Sun 2021 APP/PS1 mouse paper is the most compelling preclinical AD finding. Cognitive recovery, neuronal density increases, reduced neuroinflammation, mechanism traced to PI3K/AKT. For an animal study, it's well-constructed.

It's also one study. And the molecule's commercial successor, ATH-1017, has had a difficult clinical trajectory in human Alzheimer's trials, which is worth knowing if you're trying to extrapolate from rodent data to what's likely to happen in humans.

Most of what's known about dihexa pharmacokinetics comes from rodent studies, and most of those trace back to the WSU group. Radiolabeled tracer studies confirmed brain accumulation after oral and parenteral dosing in rats. The molecule's design, small, hydrophobic, peptidase-resistant, predicts BBB penetration, and the experimental data, where it exists, supports that.

The half-life numbers are striking. In vitro rat serum half-life: several hours, based on reported measurements. Pharmacokinetic modeling of IV dosing in rats yielded a half-life of 12.68 days ^[10]. Intraperitoneal dosing: several days. The likely explanation is extensive tissue distribution driven by high lipophilicity. The molecule partitions into fat and membrane compartments and releases slowly back into circulation.

A multi-day functional half-life with a c-Met-activating mechanism is exactly the combination you want to understand before chronic dosing in a species you've never tested it in. Nobody has.

There's no published human pharmacokinetic data. None.

Oral dosing works in rats. That was a core design goal, and the original papers reported procognitive effects after oral administration. Subcutaneous, intraperitoneal, intracerebroventricular, and transdermal routes have all been used in various animal protocols. ICV was the most reliably effective route across the Ho and Nation systematic review.

Biohacker communities have improvised intranasal and transdermal protocols based on the molecule's hydrophobicity and rodent oral data. None of these routes have human pharmacokinetic validation. The assumption that oral or transdermal dosing in humans produces brain exposure comparable to what's been measured in rats is exactly that, an assumption.

The honest answer here is short. There are no formal clinical adverse event data for dihexa in humans. Even if scattered signals appeared in the FDA's FAERS database, FAERS for an unapproved gray-market compound is close to uninterpretable. Reports are unverified, identification is unreliable, and use contexts are unknown.

What's missing is more telling than what's present. No published carcinogenicity studies. No genotoxicity assays. No long-term toxicology in any species. No reproductive or developmental toxicity data. No drug interaction studies. The molecule has been in research circulation for over a decade, and the standard package of safety pharmacology that any IND-stage compound would have generated by now simply doesn't exist.

This is the part of the dihexa story that should make any thoughtful reader stop. C-Met is a validated oncology target. Not "potentially involved in cancer." Validated. HGF/c-Met activation is implicated in proliferation, invasion, angiogenesis, and metastasis across a long list of malignancies. Roughly half of hepatocellular carcinomas show HGF/c-Met pathway activation ^[11]. Lung, breast, colon, gastric, and glioblastoma tumors also use the pathway. The pharmaceutical industry has spent considerable money developing c-Met inhibitors, drugs designed to block precisely the receptor dihexa is designed to activate.

A systemically bioavailable c-Met agonist with a multi-day tissue half-life has never been given to humans under controlled conditions and followed for cancer endpoints. Nobody knows what happens.

The theoretical risk profile includes promotion of occult tumors not yet clinically apparent, acceleration of precancerous lesions, and enhancement of metastatic possible in patients with prior cancer history. The mechanism is well-characterized enough that "we don't know" is not a reassuring answer. It's a flag.

The counterargument the WSU group and downstream advocates have offered is that c-Met activation is preferentially engaged in injury and disease states rather than constitutively driven. That may be true at the level of cognitive effect. It's not a satisfying answer at the level of oncogenesis, where c-Met activation in any tissue with a precancerous clone is the concerning event.

Dihexa is not FDA-approved for any indication. No IND has been filed. No NDA has been filed. The compound is not on the FDA's bulk drug substance candidate list for either 503A or 503B compounding pharmacies.

The FDA has gone further than mere non-listing. Dihexa acetate appeared on the FDA's list of substances that may present real safety risks when used in compounding ^[12], and the agency stated explicitly that it "has not identified any human exposure data" and "lacks important information regarding any safety issues" for the compound. That's unusually direct regulatory language.

That review now has a shape. Dihexa acetate came out of Category 2 in April 2026, in the action that moved twelve peptides out of the prohibited bucket — a procedural consequence of the original nominators withdrawing their submissions, not a safety finding. Seven of the twelve went before the Pharmacy Compounding Advisory Committee on July 23-24, 2026; dihexa was not among them, and its own review is scheduled to follow before February 2027, alongside GHK-Cu, melanotan II, cathelicidin (LL-37) and PEG-MGF.

Nothing about the compound's legal position has changed in the meantime. It is neither prohibited nor permitted, committee recommendations are non-binding, and any real change requires FDA rulemaking that has not begun. Our report on the July meeting has the tallies and what they do not settle.

In practice, dihexa is sold through gray-market vendors under "research use only" or "not for human use" labeling. There's no legal prescription pathway in the United States. International regulatory treatment varies but is broadly similar, unapproved everywhere we can find documentation for.

There isn't any.

No published human trials. No registered trials on ClinicalTrials.gov under the dihexa name ^[13]. The successor compound ATH-1017, developed by Athira Pharma after the WSU spin-out, did advance into human Alzheimer's trials. Dihexa itself was bypassed in that development path, which is worth thinking about. If the original molecule were the obvious clinical candidate, it would have been the molecule advanced.

The biohacker framing of dihexa runs ahead of the science in specific, identifiable ways.

The "seven orders of magnitude more potent than BDNF" claim circulates widely. It traces to comparisons in the original WSU papers of picomolar dihexa concentrations against nanomolar BDNF concentrations in synaptogenesis assays. The comparison is real in the papers, though those papers have large credibility problems now, but the framing collapses several distinct issues. Synaptogenic potency in a hippocampal culture is not the same as cognitive potency in an intact organism. BDNF is a 27-kDa protein with poor pharmacokinetics; dihexa is a small molecule with rodent oral bioavailability ^[14]. A direct potency comparison between molecules with totally different delivery profiles is not informative about clinical effect.

The "new synapses" claim is technically what the preclinical data showed, but the framing in community discussions tends to imply broad cognitive enhancement. The animal data don't support that. Effects in normal-cognition rats were null. The pattern is injury-state recovery, not enhancement above baseline.

The Alzheimer's promise framing assumes translation from APP/PS1 mice to human AD that the field's track record doesn't support. The graveyard of failed AD drug candidates is full of compounds that worked in mouse models.

The safety framing in community discussion is the part that troubles us most. The c-Met oncology concern is not speculative. It's mechanistically grounded in decades of cancer biology, and it has not been addressed by any chronic toxicology data in dihexa specifically. The framing that "no adverse effects have been reported" treats the absence of formal safety reporting infrastructure as evidence of safety. It isn't.

What remains genuinely open in the dihexa literature, in our reading:

Whether the HGF/c-Met mechanism, as originally described, holds up under independent biochemical replication post-retraction. The Sun 2021 paper supports PI3K/AKT involvement but doesn't re-prove the binding chemistry.

Whether the rat pharmacokinetic profile, particularly the multi-day functional half-life, translates to humans, and at what doses brain exposure becomes meaningful through oral or transdermal routes.

Whether chronic c-Met activation in a systemically dosed peripheral context promotes tumor formation or progression. This is the single most important unanswered question and the one least likely to be resolved through informal use data.

Whether dihexa produces cognitive effects in humans at all, given that mouse-to-human translation of cognitive endpoints has a poor track record and the molecule has been bypassed in favor of its successor by the only company that ever seriously tried to develop it.

Whether the FDA's 2027 review changes the compounding status meaningfully, or whether the agency's existing posture, explicit identification as a substance with real safety risks, hardens.

The honest read of the dihexa file in mid-2026 is this: a molecule with interesting preclinical data, foundational mechanistic papers now retracted, one independent replication supporting a related pathway, no human evidence of any kind, an unaddressed oncology concern grounded in well-established cancer biology, and a commercial successor that displaced it in clinical development. The biohacker enthusiasm tracks the 2012-2014 papers. The retraction record, the c-Met problem, and the empty human-evidence column track the present.

That's the gap worth knowing about.

Is Dihexa safe? The c-Met concern. This is the central flag: c-Met is a validated cancer target, and a systemically bioavailable c-Met agonist with a multi-day tissue half-life has never been given to humans and followed for cancer endpoints. There's no human toxicology, carcinogenicity, or long-term safety data.