Semax (ACTH(4-7)PGP)

Chemical identity, and a naming discrepancy that persists

Semax carries no International Nonproprietary Name. The name resolves instead to a single PubChem record: CID 9811102, CAS 80714-61-0, molecular formula C37H51N9O10S, molecular weight 813.9, InChIKey AFEHBIGDWIGTEH-AQRCPPRCSA-N, UNII I5FAL2585H. The sequence is Met-Glu-His-Phe-Pro-Gly-Pro, usually written MEHFPGP. The synonyms on that record run to laboratory and catalogue designations, a WHO Drug Dictionary entry filed under the bare sequence, and the descriptive form ACTH(4-7), Pro-Gly-Pro.

The molecule is a graft. Its first four residues are taken from adrenocorticotropic hormone and the last three are the tripeptide Pro-Gly-Pro, appended to slow proteolysis. Human ACTH, as annotated in UniProt P01189, runs SYSMEHFRWGKPVGKKRRPVKVYPNGAEDESAEAFPLEF, so residues 4 to 10 are Met-Glu-His-Phe-Arg-Trp-Gly. Semax retains only 4 to 7. Arg-Trp-Gly is gone. What replaced it was chosen for metabolic stability, and that substitution is why the peptide is described across this literature as non-hormonal, a characterisation stated in reviews rather than demonstrated by any corticosteroid-release measurement retrievable for this page.

That detail matters because a large share of the literature is titled wrongly. Dolotov and colleagues 2006 call it an analog of ACTH(4-10); Eremin and colleagues 2005 do the same; so do several Russian-language papers indexed since the 1990s. Later work from the same institutes uses ACTH(4-7)PGP, which is what the structure supports. Both names denote CID 9811102. Anyone reading the older titles at face value would infer three residues of shared sequence with ACTH that are not there.

Where an intranasal dose goes, and what it becomes

Shevchenko and colleagues 2006 prepared Semax tritium-labelled at the C-terminal proline at 56 Ci/mmol and gave 50 micrograms per kilogram in 20 microlitres intranasally to outbred white rats of 200 to 250 grams. At two minutes, 0.093 per cent of the total administered radioactivity was recovered per gram of brain. Eighty per cent of that radioactivity was intact peptide; the remainder was metabolites. Across the biological samples, the tripeptide Pro-Gly-Pro predominated over Semax and its other fragments.

Degradation has been mapped from two directions. Dolotov and colleagues 2004, working with plasma membranes from rat forebrain basal nuclei, reported a half-life longer than one hour in that preparation and identified dipeptidylaminopeptidases as the principal enzymes, cleaving successively to the pentapeptide HFPGP and then to PGP. Zolotarev and colleagues 2006, using peptide labelled at every residue, found that loss of the N-terminal ME and the C-terminal GP dominated in the presence of plasma membranes and of glial and neuronal cultures, and that the product patterns differed between glial and neuronal cells.

The metabolite is not inert, which complicates attribution. Dmitrieva and colleagues 2010 treated rats after permanent middle cerebral artery occlusion with either Semax or PGP alone and found that both raised transcription of neurotrophins and their receptors in the cortex, with partially overlapping profiles. Semax acted selectively in ischaemic cortex; the influence of PGP was described as mainly unspecific, appearing in operated and unoperated animals alike. No human pharmacokinetic study of Semax was located in PubMed.

The rodent neurotrophin and transcriptome record

Dolotov and colleagues 2006 gave a single 50 microgram per kilogram application to rats and measured the hippocampus. BDNF protein rose by a maximum of 1.4-fold, tyrosine phosphorylation of trkB by 1.6-fold, exon III BDNF mRNA by 3-fold and trkB mRNA by 2-fold. The same animals showed more conditioned avoidance reactions than controls. Those four multipliers are the origin of most of what circulates about this peptide and BDNF, and they were measured in rat hippocampus, not in people.

Most of the mechanistic work since has been transcriptomic, and nearly all of it comes from one institute cluster using one lesion. Filippenkov and colleagues 2020 ran RNA-Seq on rat brain 24 hours after transient middle cerebral artery occlusion and identified 394 differentially expressed genes above a 1.5-fold threshold under Semax relative to saline, with inflammatory genes suppressed and neurotransmission genes activated. Follow-ups in 2023, 2024 and 2025 extended the same design to the penumbra-associated frontal cortex and to the striatum, and added a second peptide, ACTH(6-9)PGP, as a comparator.

Sudarkina and colleagues 2021 took the same model to the protein level, reporting upregulation of active CREB in subcortical structures, downregulation of MMP-9 and c-Fos in the adjacent frontoparietal cortex, and downregulation of active JNK in both tissues at 24 hours. The retrieved abstracts of these papers do not state group sizes. A body of work this internally consistent and this concentrated in one laboratory network has not been independently replicated at anything like the volume of publication.

Binding, ion currents and a two-order concentration gap

Dolotov and colleagues 2004 reported saturable binding to plasma membranes of rat forebrain basal nuclei: time-dependent, specific, reversible, calcium-dependent, with a dissociation constant of 2.41 plus or minus 1.02 nanomolar and 33.5 plus or minus 7.9 femtomoles of binding sites per milligram of protein. The site was characterised by its kinetics rather than identified as a named receptor, and that remains the position twenty years later.

Electrophysiology has produced sharper numbers. Sharonova and colleagues 2018, using concentration-clamp on neurons isolated from rat brain, found that 1 micromolar Semax augmented GABA-activated currents in cerebellar Purkinje cells by 147 plus or minus 13 per cent, while 0.1 and 1 micromolar reduced glycine-activated chloride currents in hippocampal pyramidal neurons to 68 and 43 per cent of control. Both effects developed slowly and reversed poorly, which the authors read as implicating second messengers rather than direct channel block.

Kolbaev and colleagues 2025 applied 1 micromolar to hippocampal and cerebellar slices and reported a higher frequency of spontaneous intracellular calcium fluctuations in CA1 pyramidal layer cells, with no significant effect on proton-induced calcium rise in cerebellar granule cells. Their conclusion was negative and useful: whatever the neuroprotective mechanism is, it does not appear to run through acid-sensing ion channels in those cells.

Concentrations across this literature do not sit comfortably together. Storozhevykh and colleagues 2007 needed 100 micromolar Semax to delay calcium dysregulation and mitochondrial depolarisation in cultured cerebellar granule cells under glutamate toxicity, improving survival by around 30 per cent on average. That is two orders of magnitude above the concentration active on GABA currents and roughly five orders above the reported binding constant.

The human record, and why it is difficult to weigh

A ClinicalTrials.gov search returned no registered study of Semax under any of the names checked, whether searched as an intervention or as a free-text term. Not one. The human literature nonetheless exists and is substantial. Of 231 PubMed records for the term, 192 also match Russia or Russian, and the clinical reports sit almost entirely in Russian-language journals indexed with English abstracts and no retrievable English full text. What follows is therefore drawn from abstracts, which is a real limit on how closely any of it can be read.

Gusev and colleagues 1997 studied 30 patients in the acute period of hemispheric ischaemic stroke against a control group of 80 patients on conventional therapy, with EEG mapping and somatosensory evoked potentials as objective measures. Allocation was not randomised and no placebo is described. Gusev and colleagues 2018 examined 110 patients after ischaemic stroke, 43 men and 67 women of mean age 58.0 plus or minus 9.7 years, split by rehabilitation timing and by whether they received the peptide, and reported that plasma BDNF rose and stayed elevated, correlating with Barthel index score. PubMed types that report as a Clinical Trial; its abstract describes no blinding.

Two placebo-controlled human studies were located, and both are imaging. Lebedeva and colleagues 2018 scanned 24 healthy volunteers, 14 given intranasal Semax and 10 given placebo, before and at 5 and 20 minutes, and found a larger rostral subcomponent of the resting-state default mode network in the treated group. Panikratova and colleagues 2020 studied 52 healthy participants across Semax, Selank and placebo and reported differences in functional connectivity between the right amygdala and the right temporal cortex. Neither measured a behavioural or clinical endpoint.

Beyond stroke the clinical record spreads thin across specialities. Ivanikov and colleagues 2002 reported healing of refractory peptic ulcers at day 14 in 89.5 per cent of the treated group against 30.8 per cent of controls, without stating group sizes. Other reports cover cerebrovascular insufficiency in 187 patients, optic-nerve disease, and motor neuron disease in 27 patients, the last typed by PubMed as a Randomized Controlled Trial while its own abstract calls it an open-label trial in sequential groups. PubMed returns no systematic review and no meta-analysis. Searches of the openFDA drug label endpoint and of DailyMed returned no matches, so no US-marketed product exists; the papers themselves describe the peptide as in clinical use in Russia.

The acetylated, amidated variant sold alongside it

A second molecule circulates under a name built from this one. PubChem holds it as CID 172638603, N-acetyl semax amidate, CAS 2920938-90-3, formula C39H54N10O10S, molecular weight 855.0. The modifications are an acetyl cap on the N-terminal methionine and an amide at the C-terminal proline, both standard moves for slowing exopeptidase attack. Its entire synonym list on that record is the name, a shortened form of the name, the CAS number and two chemical-catalogue codes. No literature-derived synonym, UNII or regulatory identifier appears on it.

PubMed returns zero results for N-acetyl semax, zero for Ac-MEHFPGP and zero for NA-Semax. The molecule has no published pharmacology, no published pharmacokinetics, no published toxicology and no published comparison against the parent peptide in any species or preparation. Every quantitative statement in circulation about it, on half-life, relative potency or barrier penetration, describes a compound on which no experiment has been published. The absence is not a gap in indexing; the compound record itself carries no citation.

One paper has tested an N-terminally acetylated Semax, and it did not test amidation. Magri and colleagues 2016, in the Journal of Inorganic Biochemistry, characterised copper and zinc coordination by Semax and by Ac-Semax. At physiological pH the acetylated peptide formed a distorted CuN3O chromophore where the parent formed CuN4, with a more positive formal redox potential. In SH-SY5Y neuroblastoma cells, acetylation did not protect against copper-induced toxicity, and the authors concluded that the free N-terminal amine was required for that protection. The single published comparison of an acetylated Semax against its parent found the modification removed an activity.

What is not known

No registered trial of Semax exists. A ClinicalTrials.gov search across intervention and free-text fields returned zero studies, so no dose-ranging, pharmacokinetic, safety or efficacy work in people has been registered, and PubMed holds neither a systematic review nor a meta-analysis of the compound. The human reports that do exist are almost entirely Russian-language, are indexed with English abstracts and no retrievable English full text, and are with two imaging exceptions unblinded and without placebo; group allocation, blinding, endpoint definitions and effect sizes cannot be checked from an abstract. Nothing in the retrieved literature establishes human pharmacokinetics: no plasma concentration-time curve, no bioavailability figure for the intranasal route, no half-life measured in a person. The binding site identified in rat forebrain membranes in 2004 has never been assigned to a named receptor, and the concentrations at which effects appear in different preparations span five orders of magnitude without a stated reconciliation. Most of the mechanistic literature since 2010 comes from one Russian institute cluster applying one lesion model, and has not been independently replicated. No published study characterises chronic administration, reproductive or developmental effects, carcinogenicity, immunogenicity, or behaviour in any non-rodent species. The acetylated and amidated derivative sold under a name derived from this one has no published literature at all.

Questions