Dihexa Research

Dihexa research runs from a 2013 medicinal chemistry paper to a failed human trial of its prodrug reported in 2025. Dihexa is a small molecule derived from angiotensin IV, developed for cognitive decline and reported to act through hepatocyte growth factor (HGF) and its receptor c-Met. The paper that established that mechanism was retracted in April 2025, along with two related papers from the same laboratory.

What dihexa is, and what PNB-0408 designates

Dihexa is N-hexanoyl-Tyr-Ile-(6)-aminohexanoic amide, also written Hexanoyl-Tyr-Ile-Ahx-NH2. Its CAS number, 1401708-83-5, resolves on PubChem to CID 129010512, formula C27H44N4O5, MW 504.7.

PNB-0408 is not a separate parent compound but a development code name for dihexa itself. PubChem lists PNB-0408 and ATH-1001 as synonyms of that same CID, and Wells and colleagues name the compound in print as "(PNB-0408), also known as Dihexa" ^[6]. Calling material a "PNB-0408 analog" therefore invents a lineage that does not exist. One molecule means one literature, not two.

Dihexa was built from the Nle-Tyr-Ile core of norleucine-1-angiotensin IV, with chemical caps at both ends to resist peptidase cleavage ^[1]. Calling it an angiotensin IV analog is accurate as to origin. Findings on angiotensin IV, Nle1-angiotensin IV, norleual and divalinal-angiotensin IV are not dihexa findings.

The HGF and c-Met mechanism, and the retraction of its primary evidence

The direct evidence that dihexa activates c-Met is a single retracted paper from the originating laboratory. Benoist and colleagues (2014) tied procognitive and synaptogenic effects in animal work to HGF/c-Met activation, and it was retracted in April 2025 ^[2]. Two further papers from that laboratory, both also retracted, report the opposite direction of effect for related compounds built on the same scaffold. Kawas and colleagues (2012) reported in vitro that the 6-aminohexanoic analog family binds HGF directly, blocks functional dimer formation and attenuates HGF's capacity to activate Met ^[3]. The same paper reports that a cysteine-substituted member of that family, not dihexa, suppressed lung colonization by melanoma cells in mice. An earlier paper reported that peptides mimicking the HGF dimerization domain bind HGF and block Met activation in vitro ^[4]. The indexed retraction notices state no grounds.

The discovery paper that first characterized dihexa, McCoy and colleagues (2013), has not been retracted ^[1]. It carries an Expression of Concern issued in September 2021, which remains unresolved.

The strongest independent positive animal study routes dihexa's effect through PI3K/AKT signaling downstream of angiotensin IV, without invoking HGF or c-Met ^[5]. The problem with the advertised mechanism is therefore stronger than non-replication. Its primary evidence has been withdrawn from the scientific record.

The literature is not uniformly against the HGF route. Uribe and colleagues (2015) reported that dihexa's protection of zebrafish hair cells was attenuated by the HGF antagonist 6-AH, and also by Akt, TOR and MEK inhibitors ^[7]. That paper has not been retracted. Two of its four authors held company affiliations.

Preclinical research on dihexa

The discovery paper reported that dihexa reversed scopolamine-induced and age-related maze deficits in rats ^[1]. It is also the sole source of the widely repeated claim that dihexa is orally active and crosses the blood-brain barrier. That claim sits inside a paper under an unresolved Expression of Concern.

The most substantial independent positive result is an animal study in APP/PS1 mice ^[5]. Dihexa restored spatial learning in the Morris water maze, raised neuronal and synaptophysin counts, and reduced glial activation. The PI3K inhibitor wortmannin reversed the anti-inflammatory and anti-apoptotic effects. The authors declare no conflict of interest, making it the one clean third-party positive study.

In an in vivo zebrafish study, dihexa protected lateral line hair cells from aminoglycoside toxicity ^[7]. In a rat sciatic nerve model, motor function improved in the arm combining dihexa with mesenchymal stem cells ^[8]. The arm combining mesenchymal stem cells with G-CSF, which contained no dihexa, improved to the same significance threshold. The originating investigator co-authored that paper, so it is not an independent test.

Wells and colleagues randomized 40 male Wistar rats in a 3-nitropropionic acid model of Huntington's disease ^[6]. Dihexa, named there as PNB-0408, did not protect against the weight, learning, memory or motor deficits. The authors concluded it may not be an efficacious strategy.

A systematic review of animal studies found cognitive benefit in eight of nine studies in models of deficit ^[9]. It pools dihexa with Nle1-angiotensin IV and LVV-hemorphin-7 without separating them, so it is not evidence for dihexa specifically. It also notes those peptides worked best when injected directly into the brain.

Human research on dihexa proceeds through a prodrug

A ClinicalTrials.gov registry sweep returns zero studies for dihexa, so no registered clinical trial has administered dihexa itself to people.

Dihexa has nevertheless reached humans as a prodrug. Fosgonimeton (ATH-1017, earlier NDX-1017) is a different molecule, the phosphate ester of dihexa's tyrosine phenol, and the FDA substance registry records dihexa as its active metabolite ^[14]. Six registered human trials administered fosgonimeton, roughly 1,180 enrollments in total ^[13]. The open-label extension enrolled participants rolling over from two blinded parent trials, so distinct individuals number closer to 750. Several hundred of those received active drug. These are prodrug trials rather than dihexa trials, but the active moiety is dihexa.

The Phase 1 program found fosgonimeton safe and well tolerated in 88 human subjects ^[10]. A frequently quoted significant ERP P300 latency change (p=0.027) compared seven active subjects against four on placebo. That sample is far too small to carry weight.

LIFT-AD was a randomized, double-blind, placebo-controlled Phase 2/3 human trial in mild-to-moderate Alzheimer's disease, with 549 people dosed and 287 in the primary analysis ^[11]. It missed its primary endpoint and every secondary endpoint at week 26. The Global Statistical Test difference was -0.08 (SE 0.10), p=0.70; ADAS-Cog11 was -0.70 (0.77), p=0.35; ADCS-ADL23 was +0.67 (0.92), p=0.61. Discontinuation for adverse events ran three times higher on drug, 14.2% against 4.6%, mostly injection-site reactions. The one nominally significant change in plasma biomarkers was in p-tau217, an exploratory rather than a secondary measure ^[11].

Two earlier human trials also failed their registered primary endpoints. ACT-AD (n=77) did not reach significance on ERP P300 latency in the full population, and only a post hoc subgroup was positive ^[12]. SHAPE, in Parkinson's disease dementia and dementia with Lewy bodies, was terminated at 28 participants against a planned enrollment of roughly 75, and it missed its primary endpoint ^[13]. The sponsor announced a nominally significant secondary cognitive result there, which LIFT-AD did not replicate. The open-label extension was terminated after LIFT-AD failed ^[13].

The remaining three trials carried no registered efficacy endpoint. The Phase 1 program and the open-label extension were registered on adverse-event incidence, and the sixth study on mass balance and pharmacokinetics ^[13]. Three of the six trials therefore had a registered efficacy endpoint to miss, and all three missed it.

All six human trials delivered the compound as a phosphate prodrug by subcutaneous injection. That choice sits uneasily with the claim that the parent molecule is orally active and brain-penetrant.

The sponsor stopped pursuing Alzheimer's disease after LIFT-AD. It announced in September 2024 that it would focus on a different oral HGF/MET modulator in ALS rather than run further Alzheimer's studies. That announcement comes from a secondary source rather than a company filing.

The c-Met safety question

MET is a well-characterized proto-oncogene and an established oncogenic driver when amplified. Sustained agonism there is a reasonable thing to ask about.

No carcinogenicity study and no long-term toxicology study of dihexa appears in the published literature. A PubMed search for dihexa and cancer returns no records at all.

The only human data bearing on that question come from LIFT-AD ^[11]. Across 26 weeks in the dosed safety population, the single event in the neoplasms system organ class was one benign lipoma, in the placebo arm. Serious adverse events were balanced or slightly favored drug.

That is a no-signal finding from six months of controlled human exposure. The trial was not designed or powered to detect carcinogenicity, and no long-term human data exist.

Regulatory status and conflicts of interest in the literature

Dihexa is not approved by any regulator, anywhere, in any indication, and neither is fosgonimeton. Field-qualified searches of the FDA label and approval datasets return nothing for either name, as do the current EMA medicines and orphan designation datasets. No marketing application was ever submitted to either agency, so the compound was never rejected, refused or withdrawn. Dihexa does hold an FDA UNII, 9WYX65A5C2, which is a substance identifier rather than regulatory recognition.

Nearly the whole positive evidence base was produced by parties with a financial stake. The originating investigators held affiliations with the company that later sponsored the trials. Every clinical paper on the prodrug is company-authored, with employment, equity and stock-option declarations. The rat nerve study declares no conflicts while listing the originating investigator as a co-author ^[8]. The independent mouse study ^[5] and the independent rat study ^[6] are the exceptions, and one of those was negative.

Where the evidence stands

This is not a compound at an early stage of investigation. A development program carried the molecule into humans as a prodrug, it did not work, and the sponsor abandoned the Alzheimer's program. Its advertised mechanism has had the primary evidence retracted. One independent animal study reported a positive result, by a different pathway than the advertised one. The one independent test in a fresh disease model was negative. Dihexa remains a research compound with no human-use authorization anywhere.

Frequently Asked Questions

Is PNB-0408 a different compound from dihexa? No, PNB-0408 is a development code name for dihexa itself, listed as a synonym of PubChem CID 129010512 and stated in print by Wells and colleagues ^[6]. ATH-1001 is a further code name for the same molecule, so material described as a "PNB-0408 analog" is misdescribed.

Has dihexa been tested in humans, and did it work? Not as dihexa, since a ClinicalTrials.gov sweep returns zero registered studies of dihexa itself. Six registered human trials administered fosgonimeton, a prodrug that the FDA substance registry records as releasing dihexa ^[13][14]. Three of those six carried a registered efficacy endpoint, and all three missed it: LIFT-AD missed its primary endpoint and every secondary endpoint ^[11], and ACT-AD and SHAPE missed theirs ^[12][13].

References

1. 1

   McCoy AT, Benoist CC, Wright JW, et al. Evaluation of metabolically stabilized angiotensin IV analogs as procognitive/antidementia agents. J Pharmacol Exp Ther. 2013 Jan;344(1):141-154. PMID 23055539. Expression of Concern: PMID 34551989. (2021), unresolved.
2. 2

   Benoist CC, Kawas LH, Zhu M, et al. The procognitive and synaptogenic effects of angiotensin IV-derived peptides are dependent on activation of the hepatocyte growth factor/c-met system. J Pharmacol Exp Ther. 2014 Nov;351(2):390-402. PMID 25187433. RETRACTED (notice PMID 40312093., 2025).
3. 3

   Kawas LH, McCoy AT, Yamamoto BJ, et al. Development of angiotensin IV analogs as hepatocyte growth factor/Met modifiers. J Pharmacol Exp Ther. 2012 Mar;340(3):539-548. PMID 22129598. RETRACTED (notice PMID 40312092., 2025).
4. 4

   Kawas LH, Yamamoto BJ, Wright JW, Harding JW. Mimics of the dimerization domain of hepatocyte growth factor exhibit anti-Met and anticancer activity. J Pharmacol Exp Ther. 2011 Nov;339(2):509-518. PMID 21859930. RETRACTED (notice PMID 40312094., 2025).
5. 5

   Sun X, Deng Y, Fu X, et al. AngIV-analog dihexa rescues cognitive impairment and recovers memory in the APP/PS1 mouse via the PI3K/AKT signaling pathway. Brain Sci. 2021 Nov 11;11(11):1487. PMID 34827486.
6. 6

   Wells RG, Neilson LE, McHill AW, Hiller AL. Effects of an angiotensin IV analog on 3-nitropropionic acid-induced Huntington's disease-like symptoms in rats. J Huntingtons Dis. 2024;13(1):55-66. PMID 38489193.
7. 7

   Uribe PM, Kawas LH, Harding JW, Coffin AB. Hepatocyte growth factor mimetic protects lateral line hair cells from aminoglycoside exposure. Front Cell Neurosci. 2015 Jan 28;9:3. PMID 25674052.
8. 8

   Weiss JB, Wagner AL, Do BH, et al. Stem cell, granulocyte-colony stimulating factor and/or dihexa to promote limb function recovery in a rat sciatic nerve damage-repair model. Ann Med Surg (Lond). 2021 Nov;71:102917. PMID 34703584.
9. 9

   Ho JK, Nation DA. Cognitive benefits of angiotensin IV and angiotensin-(1-7): a systematic review of experimental studies. Neurosci Biobehav Rev. 2018 Sep;92:209-225. PMID 29733881.
10. 10

    Hua X, Church K, Walker W, et al. Safety, tolerability, pharmacokinetics, and pharmacodynamics of the positive modulator of HGF/MET, fosgonimeton, in healthy volunteers and subjects with Alzheimer's disease. J Alzheimers Dis. 2022;86(3):1399-1413. PMID 35180125.
11. 11

    Porsteinsson AP, Moebius HJ, Sabbagh MN, et al. Fosgonimeton in mild-to-moderate Alzheimer's disease. J Alzheimers Dis Rep. 2025;9:25424823251405817. PMID 41393340.
12. 12

    Moebius HJ, Bernick C, Winner P, et al. ACT-AD: fosgonimeton in mild-to-moderate Alzheimer's disease, first results of a randomized, placebo-controlled, 26-week Phase 2 proof-of-concept trial. Alzheimers Dement. 2022;18(S10). DOI: 10.1002/alz.061572. Conference abstract, not PubMed-indexed.
13. 13

    ClinicalTrials.gov registry records: NCT04488419 (LIFT-AD), NCT04886063 (open-label extension, terminated), NCT04491006 (ACT-AD), NCT04831281 (SHAPE, terminated), NCT05511558, NCT03298672 (Phase 1, NDX-1017). Retrieved via ClinicalTrials.gov API v2, 2026-08-28.
14. 14

    FDA Global Substance Registration System, substance record for fosgonimeton, UNII H91OA9858J. Structured relationship "METABOLITE ACTIVE->PRODRUG" to DIHEXA, UNII 9WYX65A5C2. Retrieved 2026-08-28.