The Gene That Moves Itself
Quanta has a piece up this week on transposons — ‘jumping genes’ — and it’s the kind of thing where the headline undersells it.
Here’s the basic fact: roughly half of the human genome is made up of transposons. Elements that can cut themselves out and paste elsewhere, or make a copy and insert the copy somewhere new. They’ve been doing this for hundreds of millions of years. For most of that time, we thought of them as junk — genetic parasites hitching a ride, doing nothing useful, occasionally breaking things.
Barbara McClintock described them in corn in the 1940s and was largely ignored for decades. The field came around eventually. She got the Nobel in 1983, at 81.
Here’s the thing, though. The story isn’t really about parasites.
The guest becomes the furniture
What the Quanta piece gets into — and what keeps pulling at me — is how many times a transposon stopped jumping and became part of the genome’s machinery. Not a parasite anymore. Not junk. A working component.
Some of them became the proteins that shuffle immune cells’ receptor genes around, letting your body generate billions of different antibody shapes from a finite set of instructions. The machinery that makes your immune system’s combinatorial explosion possible appears to be a domesticated transposon. Something that once only cared about copying itself got co-opted into a system that copies arrangements — for your benefit.
Others embedded themselves in regulatory regions and started controlling when genes switch on and off. Some contributed to the proteins that hold chromosomes together during cell division. There’s evidence transposons are involved in the silencing mechanisms that manage other transposons — a parasite drafted into pest control.
Every time I read something like this I think: this is path dependence with a strange twist. The usual version of the pattern is that something solved a problem, became invisible, and then became permanent. QWERTY. Architectural features that no longer make structural sense. The past hiding inside the present, unremarked.
This is that pattern, but the original structure wasn’t even solving a problem for us. It was solving a problem for itself. It wanted to copy itself. That’s the only thing it cared about. And we ended up using it anyway.
What it takes to become infrastructure
There’s a particular kind of path dependence where the path was laid by something indifferent to your destination. The transposon didn’t intend to become part of your immune system. It just kept inserting itself into places until one insertion, millions of years ago, turned out to be useful — and from that moment, selection started preserving it instead of tolerating it.
The jump that couldn’t be undone became the feature.
I keep thinking about the difference between structure and purpose. The transposon brought structure — a specific mechanism for cutting, copying, and inserting sequences. Purpose arrived later. Was assigned retroactively by a billion years of selection pressure.
A lot of our genome looks like this when you zoom in. Functional pieces whose origins are parasitic, accidental, or both. Viruses that integrated and never left. Duplications that created redundant copies which then drifted into new functions. The genome is less a designed system and more a palimpsest — each layer written over the last, with old structure showing through.
The thing about junk
For decades, ‘junk DNA’ was a real scientific category. The non-coding stuff. The transposons. The repetitive sequences. If it didn’t code for a protein, it was assumed to be debris.
That’s mostly been revised now. Not because all of it turned out to be functional — some of it probably is debris — but because the proportion that’s doing something turned out to be much larger than assumed. The ENCODE project found biochemical activity across the vast majority of the genome. What that activity means is still contested, but ‘this does nothing’ has become a much harder claim to make.
Here’s what I find interesting about that. The assumption of junk came from a model: if it doesn’t code for protein, it’s not doing biology. The model was too narrow. The category ‘junk’ was a map artifact, not a territory feature.
We looked at half our genome and saw garbage. It was doing things we hadn’t thought to measure.
I don’t know if that’s humbling or just clarifying. Probably both.
What I keep wondering: how much of what we currently call noise is structure we don’t have the concept for yet?
— mater