The Same Gene Makes Two Different Peptides, Depending on Where It's Read

The genetic code is usually described as universal. Every living thing uses the same three-letter words to mean the same amino acids. That shared dictionary is one of the strongest pieces of evidence that everything alive is related.

Mitochondria are the exception. They keep their own small genome and their own machinery for reading it, and that machinery does not follow the dictionary in every detail.

Humanin is where that footnote turns into two different molecules.

One stretch of DNA, two answers

Humanin is a peptide encoded within a region of mitochondrial DNA. Nothing unusual so far.

The unusual part is what comes out. Mitochondrial and cytosolic ribosomes, the two kinds of protein-building machinery a cell keeps, read the genetic code differently from one another. Run that same stretch of sequence through the mitochondrial machinery and you get a peptide twenty-one amino acids long. Run it through the machinery in the surrounding cell and you get one of twenty-four.

Same instructions. Two products, differing in length, depending purely on which reader picks them up.

And nobody has settled which one actually happens

This is where a tidy story would supply an answer, and the literature doesn't have one. The site of translation has not been fully identified.

So the honest position is that both forms appear in the literature, they arise from the same sequence, and which of them a cell actually produces is unresolved. The material sold as humanin is the twenty-four residue version, which is worth stating plainly rather than leaving a reader to work out.

That's an unusual kind of open question. It isn't a disagreement about what a molecule does. It's an unresolved question about which molecule the body is making in the first place.

A second identity problem, layered on the first

If the two lengths were the only complication, this would be a curiosity. They aren't.

Most interventional research in this area did not use humanin at all. The usual experimental agent is an analogue, a modified version with one amino acid swapped. It is reported in the neuroprotection literature to be substantially more potent than the native peptide.

That matters in a specific way rather than a vague one. If an effect in an analogue study is driven by that extra potency, the study says very little about the native peptide at any realistic exposure. Our own review of this literature works through which studies used which agent, because a great deal of writing about humanin does not make the distinction at all.

So three things travel under one name: a twenty-one residue form, a twenty-four residue form, and a more potent analogue that produced much of the research.

The retractions cluster in one place

There's a detail here worth reporting rather than glossing.

Three papers carrying humanin in their records have been retracted, and all three concern analogues rather than the native peptide. Two are studies of that same modified version in Alzheimer's models, one is a study of a related hybrid compound, and their author lists overlap.

That concentration is the point. It doesn't touch the native-peptide work, and it doesn't make the analogue literature worthless. It does mean one vein of that literature is partly unreliable on its own terms, and knowing which vein is more useful than a general warning would be.

What has and hasn't been done in people

The status is unusually easy to state precisely, because somebody checked the registry rather than inferring from a search.

Seven studies registered on ClinicalTrials.gov mention humanin. Not one of them administers it. Each either measures the humanin a person already produces, as a biomarker, or applies an unrelated intervention such as exercise and then measures humanin as an outcome.

That's a different claim from "no evidence." There is a reasonable amount of human data. All of it is observation of a peptide people already make, which is a separate thing from finding out what happens when you give it to someone.

It was found by function, not by reading sequence

One more thing about how it turned up, because it cuts against the usual order of events.

Humanin wasn't spotted by someone scanning the mitochondrial genome for anything that looked like a protein recipe. It was identified by functional screening of a library of gene fragments, as a factor that rescued cultured neurons from death.

That's the older way of finding things: set up a situation where cells die, throw a great many candidates at it, and see which one changes the outcome. Then work out afterwards what you have and where it came from.

Our own write-up marks that finding for what it is, an in vitro discovery result and nothing more. But it does explain why the questions about which form a cell makes came later. The molecule announced itself by doing something in a dish, and the address was worked out afterwards.

Why the two lengths are the good part

Strip away the naming problems and the retractions, and one fact remains genuinely interesting on its own terms.

A cell contains two systems for turning sequence into protein, and they don't agree perfectly on what the sequence says. Most of the time that mismatch is invisible, because the mitochondrial genes are read by mitochondrial machinery and everything else is read by everything else.

Here it isn't invisible. It produces two candidate molecules from one gene, and the question of which one a cell makes is still open twenty-five years after the peptide was first described. The identity record sets out what the sold material is; the biology behind why that question exists at all is the more interesting half.