P21 (P021)

Chemical identity, and a weight that matches nothing

Nothing about P021's identity is contested. A single PubChem compound record carries the name: CID 56599151, CAS 1246751-68-7, formula C27H42N6O8, average molecular weight 578.7, InChIKey LUJZBZPLIRWWGJ-ZKHAZHQOSA-N, UNII VV8CZC8PAS. The FDA Global Substance Registration System files the same UNII under the name P-21, classified as a chemical substance, with the same formula and a computed weight of 578.66. That one CID also gathers the synonyms P021, P-21, the string P-21 nootropic peptide, and the catalogue form Ac-dggl-adamatanylglycine-NH2. The laboratory designation and the catalogue synonym denote one molecule.

Structurally it is an acetylated Asp-Gly-Gly-Leu chain whose C-terminal amide bond is made to 3-amino-adamantane-1-carboxamide rather than to a further amino acid. The shorthand in the source papers is Ac-DGGLAG-NH2, with Kazim 2014 writing it as Ac-DGGL(A)G-NH2 to mark that adamantane unit as an adamantylated glycine, which is why the compound is called a tetrapeptide in some papers and a pentapeptide in others. Baazaoui and colleagues 2017 state that the cage was placed at the C-terminus to increase blood-brain barrier permeability and decrease degradation by exopeptidases, the methods naming carboxypeptidases; the lipophilicity rationale belongs instead to Liu 2019, Wei 2020, Wei 2021 and the 2022 review, which also repeat the C-terminal placement. The deposited structure agrees with all of them.

Two weights circulate for this molecule and both are defensible. The source papers give 578.3, which is the monoisotopic mass; PubChem's 578.7 is the average mass. A third figure near 470 daltons appears on aggregator pages and corresponds to neither value on the PubChem record, nor to the tetrapeptide amide stripped of its adamantane unit. Its origin could not be established.

Peptide 6 is a different molecule, and much of the record is its

Two compounds came out of the same programme. Peptide 6 is an 11-mer, Ac-VGDGGLFEKKL-NH2, corresponding to residues 146 to 156 of human ciliary neurotrophic factor and identified by epitope mapping with neutralising antibodies. Chohan and colleagues 2011 reported a plasma half-life over six hours for it, against roughly three minutes for recombinant CNTF. P021 is the truncated four-residue core of that sequence carrying the adamantane, first described by Li and colleagues in 2010.

Studies routinely cited as P21 evidence administered Peptide 6 instead. Blanchard 2010 in 3xTg-AD mice (PMID 20697724), Blanchard 2011 in the Ts65Dn trisomy model (PMID 22082658), Bolognin 2012 in an I2NTF-CTF rat model of sporadic Alzheimer's disease (PMID 22083255), Kazim 2015 in rat pups injected with sera from children with autism (PMID 25769033), and Chohan 2015 in a controlled cortical impact model (PMID 25255260) all used the 11-mer. The widely quoted figure of an 80 percent rise in newborn dentate gyrus neurons after brain injury belongs to Peptide 6.

The two have not been compared head to head. No study located this session ran both molecules in the same animals against the same endpoint, so the relationship between them is a design rationale rather than a measured one. Residue numbering also disagrees within the originating laboratory's own output: the 2017 paper places the P021 core at CNTF residues 148 to 151, the 2022 review at 147 to 150. Both point at the same DGGL segment, and the offset is unexplained.

The Cerebrolysin attribution rests on a retracted paper

Vendor and aggregator pages commonly introduce P21 as a peptide derived from Cerebrolysin, sometimes describing it as the product of reverse-engineering that preparation. Methods sections in the primary literature say something else. Every origin account in the source papers runs through ciliary neurotrophic factor: epitope mapping with neutralising antibodies raised against human CNTF, identification of the active 11-mer, truncation to the four-residue core, then adamantylation. Cerebrolysin is not mentioned in any of them.

One published experiment did put the two side by side. Rockenstein and colleagues, in the Journal of Alzheimer's Disease in 2011, compared Cerebrolysin with Peptide 6 and Peptide 6A in amyloid precursor protein transgenic mice and reported that all three raised doublecortin-positive and BrdU-positive cell counts in a region-specific pattern, with Cerebrolysin associated with reduced TUNEL staining and the peptides with raised PCNA. PubMed types that paper as a Retracted Publication. The retraction notice was printed in June 2026 at 111(4):1893, PMID 42015030.

So the attribution has nothing experimental behind it. The only study linking this chemical series to Cerebrolysin has been withdrawn, and it tested the 11-mer rather than P021. Retrieval of the notice text failed, the publisher returning HTTP 403, so this record states the fact of the retraction without asserting a cause. For scale: PubMed types 27 papers carrying Rockenstein as an author and 29 carrying Masliah as an author as Retracted Publication. It is the only retracted paper in this field bearing Khalid Iqbal's name.

What the rodent studies measured

The bulk of the record is a series of oral-diet studies in the 3xTg-AD mouse from one laboratory. Baazaoui and colleagues 2017 fed female mice a diet containing 60 nmol per gram from three months to 21 months of age, with 41 wild-type, 35 transgenic on vehicle and 32 transgenic on the compound, and reported that a doublecortin-positive cell deficit was reversed in treated animals. Changes in synapsin 1, synaptophysin and GluR1 were region-dependent rather than uniform, and several markers moved only as trends.

Dietary concentration varied by study rather than by dose-finding. The 3xTg-AD feeding studies located used 60 nmol per gram of feed, which those papers convert to roughly 162 nmol per animal per day at an average intake of 2.7 grams, while the Ts65Dn dam-feeding study used 200 nmol per gram. Rat work used gavage instead: Liu and colleagues 2019 gave 500 nmol per kilogram daily for 88 days to female Fisher rats aged 19 to 21 months, seven per group, and reported fewer retinal deposits and reduced Iba-1 and GFAP staining area against age-matched vehicle. No dose-ranging experiment was located in any species.

Developmental studies moved exposure earlier. Kazim and colleagues 2017 treated Ts65Dn dams from embryonic day 8 to postnatal day 21 at 200 nmol per gram of diet, across six groups of 9 to 13 mice, and reported that delays in surface righting, cliff aversion and ear twitch were shortened and hippocampal BDNF expression raised in three-week-old offspring, while PSD95 expression did not change significantly. Wei and colleagues 2020 used the same prenatal-to-weaning window in 3xTg-AD mice and reported reduced thioflavin-S plaque burden in the subiculum at 22 months.

Imaging entered the record most recently. Falangola and colleagues 2026 fed female 3xTg-AD mice the same 60 nmol per gram diet from 2.5 to 8 months and imaged at 8 months, with 29 treated, 26 vehicle transgenic and 23 normal controls after one outlier exclusion, reporting higher fractional anisotropy and axial diffusivity in the corpus callosum in treated animals than in vehicle transgenics. That paper carries a corrigendum printed the following month against page 9. Its text could not be retrieved, so what it changed is not established here.

A different model, a null result

Almost every positive finding above comes from the originating laboratory or its direct collaborators. The one full test run outside it is only partly outside. Mottolese and colleagues 2024, working on CDKL5 deficiency disorder in a laboratory in Bologna, ran the compound in both cell culture and mice. The head of the originating laboratory is a co-author on that paper, affiliated there to both his institute and the company developing the compound, and the competing-interests declaration records that he is that company's cofounder and chief scientific officer, holds several patents on the composition and use of P021, and is developing it as a therapeutic.

In SH-SY5Y neuroblastoma cells carrying a CDKL5 knockout, treatment restored proliferation, survival and maturation measures and altered GSK3beta phosphorylation, across three to five replicates per condition. The in vivo work used 39 male animals across both arms, 22 Cdkl5 knockouts and 17 wild-type littermate controls. Oral treatment at 60 nmol per gram of diet for 70 days from weaning produced no significant increase in hippocampal BDNF, no correction of the neuroanatomical deficits, and no significant memory improvement on passive avoidance. A second arm at 750 nmol per mouse by daily intraperitoneal injection for 30 days in four-month-old animals also left BDNF unchanged.

Those authors framed the outcome as a question of timing and duration rather than of mechanism, suggesting prenatal initiation or longer treatment might be required. Either reading leaves the same position on the record: this is the clearest in vivo test published by a group outside the originating laboratory, run in a model that laboratory had not worked in, and it did not reproduce the BDNF effect that the proposed mechanism runs through. The declared financial interest of its co-author cuts against reading it as an arm's-length replication attempt, not in favour.

Pharmacokinetics, and where the three-hour figure lives

A plasma figure of greater than three hours circulates widely, and the papers carrying it describe it incompatibly. The 2016 review from the originating laboratory calls it a plasma half-life of over three hours. The 2022 review from the same laboratory repeats the same wording. Baazaoui 2017, citing the same underlying reference, calls it a plasma stability of greater than three hours. A half-life and a stability window are not the same measurement, and nothing in the retrieved text reconciles them.

The citation chain behind the figure does not converge. Baazaoui 2017 points to Kazim and colleagues 2014 in Neurobiology of Disease; the 2022 review cites three sources at once for the same sentence, a 2016 doctoral thesis from the originating laboratory, Kazim 2014 and Wei 2021. Kazim 2014 is not open access, its indexed abstract makes no mention of plasma kinetics, and Europe PMC serves no full text for it; the thesis record was located but its file could not be downloaded. The gastrointestinal stability figures carried alongside are stated inconsistently across the laboratory's own papers: one primary paper gives above 95 percent in artificial gastric juice over 30 minutes and approximately 100 percent in intestinal fluid over 120 minutes, while four others give approximately 90 percent gastric, with intestinal variously above 95 percent or approximately 100 percent. A competing half-life of under one hour also circulates, with no citation anywhere it was found.

Barrier permeability is asserted rather than quantified, but a study is asserted to exist. Liu 2019, Wei 2020 and Wei 2021 each carry the same sentence, that barrier studies on the compound were carried out through a commercial contract laboratory and demonstrated that a sufficient amount crossed, all three citing Kazim 2014; Kazim 2017 refers to the same contract work in adult C57Bl/6 mice. No numeric result from that assay has ever been published: no brain-to-plasma ratio and no brain tissue concentration was located for this compound in any species. The strongest published evidence remains indirect, in Kazim 2017, where treating dams reduced STAT3 phosphorylation at Tyr705 in offspring hippocampus, which the authors read as showing that material crossed the placenta, entered milk and then reached brain. That is an inference from a downstream marker, not a measurement of exposure.

Registry and regulatory position

A ClinicalTrials.gov search returned no registered study of this compound. Intervention searches on P021, P-21 peptide, Ac-DGGL and CNTF peptide mimetic returned nothing, and a general term search on P021 returned five records in which the string is an internal protocol identifier belonging to unrelated studies. There is no registered first-in-human study, no registered pharmacokinetic study and no registered safety study.

There is also no approval anywhere. The FDA substance registry entry records a UNII, a structure and a formula, which is a registration of chemical identity rather than a marketing authorisation. The developer named in the literature, a small business listed in the National Institute on Aging's SBIR showcase, has described pre-IND work and a projected first clinical study. A stated development plan is not a registration, and nothing had appeared in the registry as of 18 August 2026.

Safety statements in the literature are bounded by what was measured, and the two reviews from the originating laboratory do not agree on how much exposure was covered. The 2016 review states that up to one year of administration induced no adverse effect in 3xTg-AD mice or wild-type controls; the 2022 review from the same laboratory states up to 18 months, both citing the same underlying work. Several studies report no significant change in body weight over the treatment period. Those are observations in small rodent groups under a single dosing regimen. Anorexia, muscle pain and hyperalgesia were the effects that stopped recombinant CNTF in people; whether the truncated adamantylated fragment avoids them in a human has not been tested.

What is not known

There is no human data of any kind. A ClinicalTrials.gov search across P021, P-21 peptide, Ac-DGGL and CNTF peptide mimetic returned no registered study, so no pharmacokinetic, dose-ranging, safety or efficacy work in people has been registered, let alone reported. The preclinical record is also narrower than its volume suggests: nearly every positive finding originates from one laboratory and its collaborators, every 3xTg-AD feeding study located used the same single dietary concentration, no dose-response experiment exists in any species, and the mouse work is almost entirely in female animals of one transgenic line. The one in vivo test published by a group outside that laboratory, in a CDKL5 deficiency model, did not reproduce the BDNF increase on which the proposed mechanism depends, and even that paper carries the compound's patent holder as a co-author with a declared financial interest. Absorption, distribution and elimination remain uncharacterised: the plasma figure that circulates is described as a half-life in two reviews and as a stability window in a primary paper, the gastrointestinal stability percentages are stated at two different values across the same laboratory's own papers, and the citation chain behind all of them runs to a reference that could not be retrieved and to a doctoral thesis whose file could not be downloaded. A contract-laboratory barrier study is cited repeatedly but no numeric result from it has ever been published; no brain concentration measurement, no brain-to-plasma ratio, and no immunogenicity, reproductive, carcinogenicity or non-rodent toxicology study was located. The compound is not approved for human use anywhere; the FDA registry entry records chemical identity, not authorisation.

Questions