Humanin: The Lifespan Paper Ran 100 Mice and Found No Lifespan Effect

01What it is

Humanin is a chain of 24 amino acids, and its instructions sit inside MT-RNR2, a mitochondrial gene otherwise understood to build part of the machinery a mitochondrion uses to make proteins. (PMID 32271093) Hashimoto and colleagues at Keio University found it in 2001, screening for anything that kept neurons alive when familial Alzheimer's genes were switched on. (PMID 11371646)

Where MOTS-c has no published receptor I could find, humanin has a worked out mechanism. It binds a receptor assembly built around gp130, the shared signalling stalk of the interleukin-6 family of immune messengers (PMID 19386761), and a second receptor has had its structure solved with a formylated humanin in it (PMID 35365641). It also binds IGFBP-3, the carrier protein for insulin like growth factor 1, the growth signal the liver puts out, and lowers circulating levels of it (PMID 27082450). None of that has been shown in a person given the peptide.

02What the trials found

Yen and colleagues published "The mitochondrial derived peptide humanin is a regulator of lifespan and healthspan" in Aging in 2020. Its mouse arm took 100 female mice from a United States ageing colony at 18 months old and injected them twice a week for 14 months, into the abdominal cavity, the space the organs sit in, with the analogue S14G at 4 mg per kilogram. No significant difference in lifespan between the groups. (PMID 32575074) One hundred mice, fourteen months, nothing. The authors put that down to a low dose. It is still a null.

The lifespan in the title belongs to nematode worms, engineered to overexpress humanin, which lived an average of 19.0 days against 17.7 for wild type, roughly seven percent. Nothing was injected into a worm.

The same group's cognition paper used the same colony, analogue and dose. At 24 months the treated mice held their balance longer on a rotating rod, 24 controls against 23 treated, and four months later the trend was smaller and no longer significant. The maze results everybody quotes come from subsets of 7 to 10 mice a group. (PMID 30242290)

The closest anything has come to a person is a pig. Sharp and colleagues blocked a heart artery in Yucatan minipigs for 60 minutes, gave seven of them S14G at 2 mg per kilogram into a vein before blood flow was restored against seven on saline, and the dead heart muscle came out 41 percent smaller as a share of the tissue at risk. Lengthen the blockage to 75 minutes, closer to what a patient arrives with, and across 19 animals the benefit was gone. (PMID 32760857)

Now the searches. PubMed carries 558 records for humanin, and filtered to the clinical trial publication types it returns nine. Eight measure humanin in people doing something else, exercise, hyperbaric oxygen, extra oxygen before bypass surgery, a DNA methylation study, and the ninth is a cell culture paper carrying the tag in error. None of the nine gave humanin to anybody. ClinicalTrials.gov returns seven, all measuring it, and the European register matched nothing. Go and run those two yourself.

A claim like this usually fails at a company that never turns up in the obvious search, and there is one here. CohBar was set up to commercialise these peptides, and Pinchas Cohen is declared its consultant and stockholder on the competing interests statements of the lifespan paper, the cognition paper and the IGFBP-3 review, Nir Barzilai likewise on the lifespan paper. They disclose it themselves, so it is not an accusation, but the flagship results and the company selling the idea came from one group. CohBar's own human trial, 88 people, tested a MOTS-c analogue.

The human literature measures what people already make. Humanin rose in muscle and blood in ten untrained men after ten 60 second cycle bouts (PMID 32271093), and at rest it went up with age in 30 people (PMID 34351816) and down with age in the foundational paper (PMID 19623253). No peptide was given in any of them.

03Which humanin the evidence belongs to

A vial labelled humanin holds the 24-mer. Most of the papers do not.

S14G-humanin, usually written HNG, is that chain with the serine at position 14 swapped for a glycine, reported as roughly a thousand times more potent. Follow that figure and it thins out. The 2006 structure paper asserts it without measuring it (PMID 16945331), and the review most often cited for it hangs that sentence on entry 23 of its own bibliography, which is Alonso and colleagues in PNAS 2001 on tau, the protein that clumps into the tangles seen in Alzheimer's brains. No humanin in it anywhere. (PMID 25538685) The laboratory that discovered the peptide put the gap at two to three orders of magnitude, 100 to 1000 fold. (PMID 12787071) The number everybody repeats is the top of a range, on a receipt for a different purchase.

I could not open either 2001 paper. The publisher refuses the full text and PubMed Central answers that downloading is not allowed, so anyone who goes to check the original number will hit the same locked door.

Where the swap matters most is the metabolic work. Muzumdar and colleagues infused native humanin into the brain fluid spaces of three month old male Sprague-Dawley rats, 20 micrograms in total, and insulin action improved. Into a vein, at 0.375 mg per kilogram per hour, the same native peptide did not significantly alter insulin action. Only then did they engineer HNGF6A, which did reproduce it into a vein. (PMID 19623253) HNG carried the liver fat work in mice too. (PMID 26058861) The one clear exception is a native humanin study that normalised glucose tolerance in non obese diabetic mice, route and dose not given. (PMID 19800083)

04How it compares

MOTS-c is the sibling, encoded in the same genome and sold on the same promise, and one column below needs explaining. Eight fabricated trial records sit on ClinicalTrials.gov under a sponsor called Hudson Biotech, documented last month by Morgan McSweeney, and MOTS-c is one of the eight compounds they name. Humanin is not among them, which is less a compliment than an accident: there is no fake humanin trial because I could not find a humanin trial at all.

Everywhere except evidence in people, where both are empty, they separate, mostly in humanin's favour. A 2023 analysis asked which of the eight mitochondrial peptides look like sequences natural selection has preserved. Humanin's carries the marks of a gene evolution has kept intact, a lopsided preference for particular spellings of each amino acid and a chain that has barely changed across species. MOTS-c's carries neither, though those authors add that it has a highly conserved front end and cannot be ruled out. (PMID 37644144) That July 2026 committee recommendation was not binding, and still needs formal rulemaking.

05Where the dose came from

Nowhere, and two of the three sites quoting numbers say so themselves.

One gives HNG at 100 to 500 micrograms per injection two to three times a week. A second gives 0.5 to 2 mg a day under the skin. A third describes 2.5 to 10 mg per dose, daily or every other day. A hundredfold spread for the same two molecules, a bread recipe where one cook says a teaspoon and the next says a cup. Convention, not a finding.

Two of them say outright where the numbers came from. One writes that no human clinical trials exist for exogenous humanin or HNG and that its dosing is entirely extrapolated. The other calls its protocols extrapolation from rodent studies without confirmed human safety data or pharmacokinetics, the measurements of what an injected dose does over time. Those are their words, and my own searching came back the same way. The third site claims instead that human safety and IGF-1 lowering have been demonstrated in human pharmacological studies, which is a study I could not find. No pharmacokinetic study turned up in any species either, so no measured half life stands behind those schedules.

06What could go wrong

I could find no published human safety record for administered humanin or any analogue, and no approval anywhere I looked. Empty is not the same as clean.

The cancer question was raised a year after the discovery and never closed. Maximov and colleagues called humanin a potential oncopeptide and cautioned that treating Alzheimer's with it might raise cancer risk, in a hypothesis paper rather than an experiment. (PMID 12445508) Twenty four years on, serum humanin came back higher in 75 Egyptian women with breast cancer than in 70 matched controls, none of them given anything. (PMID 41518474) It separates the groups poorly, though: at the authors' threshold it caught 78.7 percent of the cancers but cleared only 62.9 percent of the controls, not far off a coin toss on the negative side.

The analogue suppresses platelets, the cells that start a clot, though in the mice tested bleeding time was not prolonged. (PMID 32174022) What it would do alongside a blood thinner is an open question.

And the blood level does not run in a straight line. Bolignano and colleagues followed 94 people on dialysis and found risk raised at both ends, worst below 450.7 and above 759.5 picograms per millilitre, then reproduced the curve in 83 more. (PMID 39102184, PMID 41849628) Same group both times, so not an independent replication. More is not obviously better.

07What nobody knows

Whether native humanin does anything at an injectable dose.

Which humanin a blood test measures. Thirteen humanin like genes sit in the nuclear genome rather than the mitochondrial one, and I could find no paper showing the commercial antibody assays tell them apart. (PMID 19477263) Every human number here rests on those assays.

And humanin's status on the 2026 World Anti-Doping Agency list, which I could not open at the source.

08My take

What gets me is the gap between a title and its own methods section. A hundred mice, fourteen months of injections, a null on the outcome the title names, and I did not find it mentioned once in the secondary write ups I read. The authors put the null down to a low dose, which may be right. The experiment is still to be run.

If you have seen humanin sold, did the page say humanin or S14G, and did the studies it cited match what was in the vial?

Frequently asked

Has humanin been tested in humans?+

I could find no published study, and no registered one, in which humanin or any humanin analogue was given to a person by any route. The seven records on ClinicalTrials.gov all measure humanin in blood or tissue rather than administering it, and the European register returned nothing at all.

Did humanin extend lifespan in the study that says lifespan in its title?+

Not in its mice. That paper injected 100 female mice with the analogue S14G at 4 mg per kilogram twice weekly for 14 months, into the abdominal cavity, which is an animal dose by an animal route with no published human equivalent I could find, and reported no significant difference in lifespan between treated and untreated animals. The lifespan result belongs to nematode worms genetically engineered to make more humanin, which lived about seven percent longer, 19.0 days against 17.7.

Is S14G-humanin the same thing as humanin?+

No. S14G, also written HNG, has one amino acid of the 24 replaced, and reviews describe it as roughly a thousand times more potent, a figure I could not trace to a primary dose response experiment. The review most often cited for it attaches that number to a bibliography entry which turns out to be a paper about tau tangles in Alzheimer's brains. Most of the mouse and pig work used the analogue, while the material sold in vials is usually native humanin.

Where do the humanin doses on vendor sites come from?+

From extrapolation, which two of the three sites quoting numbers state themselves. The figures circulating as community convention rather than as findings run from 100 micrograms to 10 milligrams per dose under the skin, a hundredfold spread for the same two molecules, and I could find no human dose finding study and no pharmacokinetic study behind any of them. The third site claims human pharmacological studies I could not find.

Is humanin safe?+

I could find no published human safety record for administered humanin or any analogue, meaning no tolerability data, no dose ceiling and no adverse event count, and empty is not the same as clean. What the literature does have is worth knowing. The possibility that a peptide whose job is preventing cell death could favour tumours was raised in 2002 and never settled. The analogue suppresses platelet activation in mice, though bleeding time in those mice was not prolonged. And in dialysis patients both low and high blood levels of humanin went with raised cardiovascular risk.

How does humanin compare to MOTS-c?+

Both are encoded inside mitochondrial DNA and neither has a human administration study I could find. Humanin has the better established mechanism, including a receptor complex and a solved receptor structure, and it passes a test of whether the sequence has been conserved by natural selection that MOTS-c does not, though the authors of that analysis say MOTS-c cannot be ruled out. An FDA compounding advisory committee recommended MOTS-c in July 2026, which is a recommendation rather than a binding decision and still requires formal rulemaking before anything changes, and humanin was not on that agenda at all.

---

This content is for educational and informational purposes only and is not medical advice. Peptides discussed are research compounds and may not be approved for human use. Nothing here should be used to diagnose, treat, cure, or prevent any disease. Always consult a qualified healthcare professional before starting any peptide, supplement, or protocol. Individual responses vary. Do not self-administer compounds without proper medical supervision.