Humanin (MT-RNR2): An Evidence Ledger

Verified against PubChem PUG-REST property and synonym records for CID 16131438 (formula, molecular weight, CAS 330936-69-1, UNII H975EUX36G, ChEMBL4447799, DTXSID70186749, sequence designation), retrieved 17 August 2026. Fields marked “not verified” are ones we could not confirm against a primary chemical database — we leave them blank rather than guess.

What the molecule is, and how it was found

Humanin is a 24-amino-acid peptide whose open reading frame sits inside the mitochondrial 16S ribosomal RNA gene, MT-RNR2. The verified sequence is MAPRGFSCLLLLTSEIDLPVKRRA, with a molecular formula of C119H204N34O32S2 and a molecular weight near 2687.2 g/mol (PubChem CID 16131438, CAS 330936-69-1). The nuclear genome carries thirteen humanin-like paralogs, MTRNR2L1 through MTRNR2L13, of 24 to 28 residues, which is a recurring source of assay ambiguity in the literature.

Its discovery was a screen, not a hypothesis. Hashimoto et al. (2001, PNAS) used functional expression screening and recovered a cDNA encoding a short polypeptide that abolished death of neuronal cells caused by multiple familial Alzheimer's disease genes and by amyloid-beta, with no effect on death caused by polyglutamine Q79 or superoxide dismutase-1 mutants. The transfected cDNA was transcribed and the polypeptide secreted into the culture medium. The authors noted that the rescue action depended on the primary structure of the peptide.

That specificity — active against one class of insult and inert against another — is worth holding onto. It is the strongest single argument that the observed protection was not a generic cell-culture artefact, and it is also the boundary of what the founding experiment established.

The two arms the reviews describe

Review-level descriptions of humanin's mechanism consistently split into an intracellular and a secreted arm. Zhu et al. (2022) summarize the intracellular arm as binding to the pro-apoptotic Bcl-2-family proteins BAX, Bim and tBid, and to IGFBP3, associations linked in cell models to inhibition of caspase activity and mitochondrial-pathway apoptosis.

The secreted arm is described as signaling through the G-protein-coupled formyl peptide receptor-like 1 and 2 (FPRL1/FPR2) to modulate apoptosis signal-regulating kinase and c-Jun N-terminal kinase pathways, and through a trimeric CNTFR-alpha / gp130 / WSX-1 receptor complex associated with JAK2/STAT3 activation. Binding to the soluble partners VSTM2L and IGFBP3 is also reported. Hazafa et al. (2020) describe the same architecture across leukocytes, germ cells and neurons.

Two caveats belong with this. First, these are review summaries of a distributed body of primary work, not single-experiment findings — the receptor assignments were established across different laboratories in different cell systems, and this page does not treat a review statement as equivalent to a measurement. Second, and more consequentially, a large share of the underlying functional experiments used analogs rather than the wild-type 24-mer.

Wild-type versus HNG: the substitution that changed the literature

S14G-humanin, usually written HNG, replaces serine at position 14 with glycine. It is a different molecule with a different sequence, and much of what circulates as "humanin" data was generated with it.

The most recent example is explicit. Liao et al. (2026) studied humanin and MOTS-c in atrial fibrillation, and the paper is careful about which molecule went into which experiment. Human atrial tissue and plasma were assayed for endogenous humanin. But the mouse intervention — angiotensin II-induced atrial fibrillation in 36 male C57BL/6J mice — administered HNG or MOTS-c, not wild-type humanin. Secondary summaries routinely collapse that distinction and report the mouse findings as humanin findings.

Structure-activity work explains why the substitution exists at all. Hashimoto et al. (2001, J Neurosci) mapped the rescue action to the domain from Pro3 to Pro19, found seven residues essential, and identified Cys8 and Ser14 as commonly required across Alzheimer's-relevant insults. In that same study only HNG, among HNG, activity-dependent neurotrophic factor, IGF-I and basic FGF, abolished cytotoxicity from all five tested insults. Whether the wild-type peptide would have done the same at higher concentration was not the question the experiment asked.

Structure, and the cysteine problem

Humanin is largely disordered in aqueous buffer. Hirano et al. (2011) examined the wild-type 24-mer alongside the S7A and S14G analogs in the presence of negatively charged DOPG liposomes and found all three adopted an indistinguishable anti-parallel beta-sheet structure, while unrelated control peptides remained disordered with and without liposome. Since S7A is described as inactive and S14G as more active, the conversion to beta-sheet on a charged membrane cannot by itself explain the activity difference. The authors instead noted that S14G was more stable in its disordered form in buffer at physiological temperature.

Arakawa, Niikura and Kita (2014) tested that stability hypothesis against a second inactive analog and it did not hold. C8A-humanin, which has no neuroprotective activity, proved as structurally stable as the potent S14G analog: both remained largely disordered at 37 °C, while the wild-type peptide and S7A converted to a beta-sheet-like structure that was largely irreversible on cooling. The authors concluded that instability does not account for C8A's activity loss and that an alternative explanation is required.

The chemistry has a practical consequence. The peptide contains a single cysteine at position 8, and Cys8 is one of the two residues identified as essential. Free-thiol oxidation and disulfide-mediated aggregation are therefore plausible degradation routes for any preparation — plausible, but not characterized by any published time-course found for this search.

The human record: measured, never administered

ClinicalTrials.gov holds exactly three studies naming humanin. None of them gives humanin to anyone.

NCT03431844, sponsored by the University of Tartu with Tartu University Hospital and North Estonia Medical Centre, was a prospective observational study of humanin isoforms in right atrial appendage tissue and blood plasma around on-pump coronary artery bypass grafting, with 30-day mortality, myocardial infarction, acute kidney injury and stroke as endpoints. It enrolled 106 participants and completed in August 2019. It has posted no outcome data.

NCT06105229 (Guangdong Provincial People's Hospital) is an observational study of plasma humanin as a candidate marker in acute kidney injury, 60 participants, with a listed status of UNKNOWN and a completion date of June 2024. NCT07678073 (University of Gaziantep) is interventional, but the interventions are sevoflurane general anesthesia and combined spinal-epidural anesthesia in 68 kidney-transplant recipients; humanin and MOTS-c are measured by ELISA as outcome variables at two perioperative timepoints. It began March 2026.

The distinction is the whole point. Circulating humanin has been observed in humans. Humanin has not been given to humans in any registered study.

Where the biomarker direction comes from

The claim that humanin declines with age does have a primary source, which distinguishes it from most age-decline claims in this category. Muzumdar et al. (2009) reported that detectable humanin decreased with age in rodent hypothalamus, skeletal muscle and cortex, and that circulating humanin levels decreased with age in humans and mice. The same study used hyperinsulinemic clamp technology in rats, found that intracerebroventricular humanin infusion improved insulin sensitivity, and showed that the hepatic effect was abolished by inhibiting hypothalamic STAT-3 and absent in primary hepatocytes.

What that study does not supply is a numeric reference range. No age-stratified concentration table for circulating humanin in humans was located, and none of the three registry studies has posted outcome data that would supply one. The direction is sourced; the magnitude is not.

Liao et al. (2026) add a matched-cohort human observation: humanin and MOTS-c were downregulated in atrial tissue from patients with atrial fibrillation, their levels inversely correlated with the extent of fibrosis, and plasma MOTS-c was decreased in the 39 atrial fibrillation patients relative to 39 sinus-rhythm controls, correlating inversely with NT-proBNP. That is an association in a small matched cohort, and the authors say so, calling for validation in larger cohorts.

What is not known

Humanin has never been administered to a human being in a registered study. All three ClinicalTrials.gov records naming it measure the peptide as an outcome variable: two are observational biomarker studies and the third administers anesthetic agents, not humanin. There is consequently no human pharmacokinetic data, no dose-ranging, no exposure-response relationship, no adverse-event profile and no route characterization of any kind in humans. A second and less obvious gap runs through the preclinical file: a substantial share of the in-vivo work described under the humanin heading used S14G-humanin (HNG) or other analogs rather than the wild-type 24-mer, including the mouse arm of the most recent atrial fibrillation study, so the animal evidence is not cleanly attributable to the molecule this page identifies. Assay specificity is a further open question — thirteen nuclear humanin-like paralogs (MTRNR2L1 to MTRNR2L13) share sequence homology with the mitochondrial peptide, and ELISA-based measurements in the observational studies do not resolve which species they detect. Handling chemistry is uncharacterized: no published stability time-course, no aggregation limit, no oxidation profile for the single free cysteine at position 8. Populations across the animal work skew heavily male (the atrial fibrillation model used male C57BL/6J mice; the metabolic model found the phenotype predominantly in males), and no long-term exposure data exist in any species. The reviewed role in cancer biology is explicitly described in the literature as debated and context-dependent, with reports pointing in opposite directions.

Questions