Sunifiram (DM-235)

Identity, and the ring that is not there

Nothing about sunifiram's identity is in dispute. One name, one structure, and three registries that agree on it. PubChem files the molecule as CID 4223812, CAS 314728-85-3, formula C14H18N2O2, average molecular weight 246.30, monoisotopic mass 246.136827821, InChIKey DGOWDUFJCINDGI-UHFFFAOYSA-N, IUPAC name 1-(4-benzoylpiperazin-1-yl)propan-1-one. The FDA Global Substance Registration System holds the same molecule under UNII 66924E735K, with a computed weight of 246.3055. ChEMBL carries it as CHEMBL309176, same InChIKey, weight 246.31, and no maximum clinical phase recorded.

The molecule is a piperazine acylated on both nitrogens, benzoyl on one and propanoyl on the other. It contains no five-membered lactam, which is to say no 2-pyrrolidinone ring, and it is therefore not a racetam in the structural sense that word carries. The Hippocampus paper from the Sendai laboratory opens by calling it a pyrrolidone nootropic drug structurally related to piracetam; its companion in Behavioural Brain Research calls it a pyrrolidone nootropic drug without the piracetam comparison. The Florence group that made it said something narrower in 2000: that an N-acylpiperazine group can mimic the 2-pyrrolidinone ring. Mimicry of a ring in a passive avoidance assay is not possession of that ring. Manetti 2000 reported the mimicry; later descriptions restate it as structural relatedness to piracetam. One aggregator page calls it a piperazine alkaloid; it is fully synthetic and no natural source has been described.

Regulatory status is frequently misread off the registry entries. The GSRS record is marked approved, which records that the substance definition passed curation, not that any medicine was authorised. A search of the openFDA drug label and Drugs@FDA endpoints for sunifiram returned no match on 18 August 2026. The compound is not approved for human use anywhere. A text search of the 138-page list of controlled substances published by the United States Drug Enforcement Administration, retrieved the same day, returned no entry for sunifiram or for unifiram. Absence from the federal controlled-substances schedules is a statement about that list only; it does not establish that a substance may lawfully be sold or supplied for human consumption, which is governed separately.

Where the potency figure comes from

Manetti and colleagues described the molecule in the Journal of Medicinal Chemistry in November 2000, as compound 13 in a series of 4-substituted 1-acylpiperazines obtained by simplifying a bicyclic lactam scaffold. Screening was the mouse passive avoidance test. That paper reports the compound as active at 0.001 milligrams per kilogram subcutaneously, and describes it, together with the bicyclic parent DM232, as raising acetylcholine release in rat brain. The same group's May 2000 paper had reported DM232 as active at the same dose by the same route, so the headline figure was not new to the piperazine series.

Ghelardini and colleagues published the characterisation two years later in Naunyn-Schmiedeberg's Archives of Pharmacology. Amnesia induced in mice by scopolamine at 1.5 milligrams per kilogram intraperitoneally was prevented by the compound at 0.001 to 0.1 milligrams per kilogram intraperitoneally, and at 0.01 to 0.1 orally, in the passive avoidance test; the comparators in the same report were piracetam at 30 to 100 intraperitoneally, aniracetam at 100 orally and rolipram at 30 orally. Amnesia induced by mecamylamine, baclofen and clonidine was also prevented. In rats, 0.1 intraperitoneally before each daily acquisition training prevented the memory impairment produced by scopolamine at 0.8 in a Morris water maze. Rota-rod, Animex and hole board measures were unchanged at the highest effective doses.

Two multipliers circulate and both come from the originating group. The 2002 paper states that the potency is about 1,000 times higher than that of the most active piracetam-like compounds. The 2006 review in CNS Drug Reviews, carrying four of the same authors, states that the compounds are four orders of magnitude more potent than piracetam. Nothing in the retrieved text reconciles them, and the ratio depends on which pair of doses is compared, since the 2002 report gives a range for each drug rather than a single figure. Route is inconsistent too: subcutaneous in the founding paper, intraperitoneal and oral in the characterisation. Sample sizes are not stated in any retrieved abstract, and the full texts are paywalled; PubMed Central holds the 2006 review as a scanned deposit flagged not open access.

Two mechanisms, and a screen that tested neither

Florence proposed AMPA. Galeotti and colleagues reported in 2003 that the compound at 0.1 milligrams per kilogram intraperitoneally reversed the amnesia produced by the AMPA receptor antagonist NBQX at 30 milligrams per kilogram intraperitoneally in the mouse passive avoidance test, and that in rat hippocampal slices it reversed the antagonism by kynurenic acid of NMDA-mediated tritiated noradrenaline release, an effect abolished by NBQX. Their conclusion was that the compounds act through activation of AMPA-mediated neurotransmission. The direct electrophysiological demonstration in that paper, an increase in excitatory synaptic transmission in rat hippocampus, was made with DM232 rather than with sunifiram.

Sendai proposed the glycine site of the NMDA receptor. Moriguchi and colleagues, in Hippocampus in 2013, reported that 10 to 100 nanomolar in mouse hippocampal slices enhanced long-term potentiation on a bell-shaped concentration-response curve peaking at 10 nanomolar, that the enhancement was blocked by 7-chlorokynurenic acid at the glycine site but not by ifenprodil at the polyamine site, and that pretreatment with 300 micromolar glycine abolished it. Their companion paper in Behavioural Brain Research used olfactory bulbectomised mice dosed orally at 0.01 to 1.0 milligrams per kilogram once daily for seven to twelve days, reporting improvement on Y-maze and novel object recognition, restoration of hippocampal CA1 long-term potentiation, and no change on tail suspension.

Binding data exists for neither account. The supplementary file to Gualtieri's 2016 review carries the screening results obtained through the National Institute of Mental Health Psychoactive Drug Screening Program: roughly forty receptors, channels and transporters, reported as percentage inhibition at 10 micromolar. Every value for sunifiram in that table is below 52 percent, and the only dissociation constants determined were at alpha-2 adrenoceptors, at 24.48 micromolar for alpha-2A, 18.65 for alpha-2B and 22 for alpha-2C. The rat NMDA site returned 1 percent inhibition at 10 micromolar. No AMPA or kainate binding site appears in the panel at all, so the AMPA hypothesis was neither confirmed nor excluded by it. The two mechanistic accounts have never been tested against each other in one experiment.

The chemist who made it wrote the warning

Gualtieri published a review in the Journal of Enzyme Inhibition and Medicinal Chemistry in 2016, first posted online in April 2015, under the title Unifi nootropics from the lab to the web: a story of academic (and industrial) shortcomings. Its abstract states that the two molecules were not protected by a patent, that some twelve years after their disclosure he found dozens of sites offering unifiram and sunifiram as drugs that improve cognition in healthy individuals, that only a few preclinical studies had been done, and that their long-term toxicity was unknown. The publisher returned HTTP 403 to a full-text request this session. The supplementary file was retrievable: the publisher deposits it under a Creative Commons Attribution 4.0 licence at figshare, DOI 10.6084/m9.figshare.2065971, as the file ienz_a_1021252_sm0214.pdf, which was downloaded and read directly.

Gualtieri's account rests on the absence of a patent, which is stated in the review's own abstract. No development programme is recorded anywhere in the retrieved sources, ChEMBL assigns no clinical phase, and the literature after 2000 is almost entirely synthetic chemistry rather than pharmacology. Whether the two are causally connected is the review author's argument rather than a measured finding, and this record does not extend it.

What followed was analogue chemistry rather than pharmacology. PubMed returns sixteen records for the term sunifiram across all fields. Seven of the sixteen are analogue-chemistry papers from the same Florence department, published between 2004 and 2017: ring expansions, ring contractions, phenyl substitutions and hydrophilic decorations, all screened in the same test. Two of those seven state that no analogue reached the potency of the parent, Scapecchi 2004 and Guandalini 2015; Guandalini 2012 and Martino 2017 make no such statement. Scapecchi's 2004 structure-activity study reported two analogues (13 and 14) with amnesic rather than antiamnesic activity, nearly equipotent to scopolamine. The most recent primary paper that has not been withdrawn is Martino's, from 2017.

The one recent paper was retracted

Agha and colleagues published a synthesis route and a set of sunifiram-carbamate and anthranilamide hybrids in the Journal of Enzyme Inhibition and Medicinal Chemistry in 2022. The reported results were in vitro: hybrid 3f inhibited acetylcholinesterase with a half-maximal inhibitory concentration of 18 plus or minus 0.2 nanomolar against 29.9 plus or minus 0.15 for donepezil, three compounds raised acetylcholine release in the A549 cell line, and blood-brain barrier penetration was a prediction from computed ADMET properties rather than a measurement. Those figures are recorded in this document only as circulating figures, in the unsourced array, for the reason set out next.

That paper was retracted on 7 August 2026. The statement of retraction records that editors and publisher identified several references containing incorrect information or unable to be verified, that the authors on being asked acknowledged fundamental errors, and that retraction followed because the validity of the references could not be established. It also records that the authors disagree with the retraction. PubMed now types the 2022 article as a Retracted Publication, and the article remains online marked as such.

Checking the author clusters rather than the single citation returns more. PubMed types three papers carrying Ibrahim TS as an author as Retracted Publication, the sunifiram hybrid paper among them, the other two in unrelated synthetic and formulation work. On the Florence side, an author search returns three retracted papers under Ghelardini C and one under Romanelli MN, covering melanoma histone deacetylase work, a murine colitis model, an oxaliplatin neuropathy study and an acetyl-L-carnitine study. None is a sunifiram paper. No paper cited in this record carries a retraction, expression of concern or erratum other than the 2022 hybrid study.

What was never measured

Pharmacokinetics do not exist for this molecule. A PubMed search combining sunifiram with pharmacokinetics returned zero records on 18 August 2026, and no absorption, distribution, elimination, plasma half-life or brain tissue concentration figure was located for it in any species, by any route, in any of the sixteen indexed papers. The bell-shaped concentration-response curve in the slice work is a bath concentration and says nothing about exposure in a living animal.

Toxicology is equally absent. A PubMed search combining DM235 with toxicity, toxicology or LD50 returned one record, the 2017 analogue-chemistry paper, which reports no toxicological assessment. No acute lethality figure, no repeat-dose study, no genotoxicity, reproductive or carcinogenicity study was located. The behavioural batteries in the 2002 and 2003 reports checked motor coordination, spontaneous motility and inspection activity at the doses tested, which is a screen for confounds in a memory assay rather than a safety study.

Human data does not exist either. ClinicalTrials.gov returned no registered study for sunifiram, for DM-235 or for unifiram when queried on 18 August 2026. No case report, poisoning report or clinical observation was located in PubMed. No analytical paper documenting detection of the compound in a marketed product was found; the published survey work on unapproved drugs in cognitive-enhancement supplements covers centrophenoxine, omberacetam, phenibut and the modafinil analogues, and sunifiram does not appear in it.

What is not known

No human data exists. ClinicalTrials.gov returned no registered study for sunifiram, DM-235 or unifiram, no case report or poisoning report was located in PubMed, and no analytical paper documenting the compound in a marketed product was found. Absorption, distribution and elimination are entirely uncharacterised: no plasma concentration curve, no half-life, no bioavailability figure and no brain tissue measurement exists in any species, so the relationship between the nanomolar concentrations that altered long-term potentiation in slices and the microgram-per-kilogram doses that altered behaviour in mice is unknown. Toxicology is absent in the same way. The lethality, repeat-dose, genotoxicity and reproductive studies that would establish a margin were not performed, and the chemist who led the originating group said so in print in 2016. Mechanism remains contested between two laboratories that never ran a comparison, and the one binding screen available covered neither of the receptors under discussion. Sample sizes were not stated in any retrieved abstract, and the paywalled full texts could not be read this session, so the effect sizes behind the behavioural findings are known only by direction and significance. Almost the entire non-retracted record originates from one department in Florence, in one strain-unspecified mouse assay, and the one recent paper from outside that group was retracted in August 2026. The compound is not approved for human use anywhere; the FDA registry entry records chemical identity, not authorisation.

Questions