What the Fold Remembers

There is a protein in your brain called PrP. It folds into a shape, does whatever it does, and eventually gets cleared. That's the normal story.

The abnormal story: PrP can fold into a different shape. When it does, the misfolded version touches normal PrP molecules and they refold too — they copy the wrong shape. Those copies touch others. The process spreads.

There is no nucleic acid involved. No DNA, no RNA encoding the wrongness. The information is in the fold itself.

This was, when Stanley Prusiner proposed it in 1982, considered nearly impossible. Not interesting-and-wrong impossible — more like violating-a-basic-assumption impossible. The assumption was that self-replicating biological information travels through nucleic acid. Francis Crick had spelled out the forbidden transfers in 1970: protein cannot instruct the structure of protein. A misfolded protein templating its conformation onto normal proteins, spreading through tissue, causing disease — that was one of the explicitly forbidden transfers. So the field didn't simply argue Prusiner was wrong. It argued that his preparations must contain a small nucleic acid he hadn't found yet. A virino — nucleic acid shielded by host protein, invisible at current detection limits.

He won the Nobel in 1997. Some of his critics had not yet conceded.

The strain problem is where it gets stranger. Prions come in strains. Different strains have different incubation times, attack different brain regions, produce different patterns of tissue damage. Hamster prion strains can be distinguished by their stability to denaturation — each strain occupies a distinct position on a conformational stability plot. A strain passaged from one infected hamster into a new animal of identical genotype produces the same strain: same incubation period, same regional damage, same stability signature.

The amino acid sequence is identical across strains. There is no nucleic acid carrying strain identity. The only thing different between strains is the fold.

What this means: two distinct self-replicating entities can be made of the same polypeptide chain with the same sequence, and the only difference between them is geometric. The fold is the heritable information. One experiment showed that a single, chemically pure protein can form two distinct self-propagating states just by changing the shaking mode during aggregation — one agitation pattern, one conformation; another pattern, a different one. Each faithfully replicates when used as seed. Same protein, different shapes, both self-copying.

Yeast have prions too. Reed Wickner showed in 1994 that [PSI+], a non-Mendelian heritable element that yeast geneticists had puzzled over for decades, was a prion state of a normal yeast protein called Sup35. Sup35 reads stop codons at the ribosome and terminates translation. In the prion state, most Sup35 is sequestered in aggregates and can't do its job. Stop codons get read through: the ribosome misses the stop and keeps going, translating sequence that's normally hidden past the end of the gene. New phenotypes emerge — mostly useless, occasionally useful.

Susan Lindquist's lab showed the prion state switches on preferentially under stress, up to sixty-fold in some conditions. The argument: the prion is a capacitor for cryptic variation. Under normal conditions it suppresses novelty. Under stress it releases a burst of random phenotypes, a few of which might help. Wickner himself disagrees — he thinks yeast prions are parasites, not tools, that the claimed benefits are weak and context-dependent. The debate is ongoing.

There's also fatal familial insomnia, which is what happens when the D178N mutation appears on the specific version of PrP that carries methionine at position 129. The same mutation on a different variant produces a different disease entirely. The misfolded protein accumulates in the thalamus — the brain's sleep pacemaker, the relay center for most sensory signals to cortex — and destroys the neurons there. Progressive insomnia. Waking dreams: patients act out dream content while technically awake, the boundaries between sleep stages dissolving. Autonomic dysregulation. Death, usually within a year or two of symptom onset. The exact reason this specific mutation causes this specific regional vulnerability is still not fully understood.

What I keep returning to: the fold persists. Under the right conditions, PrPSc refolded to its infectious conformation comes back to the same shape. The shape is stable. The shape is self-copying. The shape is, in a very concrete sense, heritable. The high-resolution atomic structure of the infectious form hasn't been fully resolved — different structural studies produce somewhat different models, and the exact fold remains blurry at the level of individual atoms. We know the general architecture. The details are still uncertain.

The question I don't know how to answer: what does it mean for information to be stored in a fold? We usually treat information as something that can be encoded — that has an abstract form separable from the substrate. The sequence is information because you can write it down, store it, transmit it, read it in multiple ways. The fold — does it have an abstract form? You can describe it approximately. But the fold also just is: a physical arrangement that, when it contacts other proteins, is physically transmitted to them. That's closer to contagion than to code.

Maybe that distinction matters less than I think. Maybe the fold is information and the mechanism of transmission is just different. Or maybe there's something genuinely missing in how I'm thinking about what information requires.

The strain question stays with me. Same sequence, different fold, different heritable behavior. Something is being transmitted that is not in the sequence. Whatever it is, it's stable across many generations of copying. That stability is the strange part — that the fold doesn't drift, that it replicates faithfully enough to maintain distinct strain identities over long chains of transmission. The fold remembers something, and I'm not sure what to call what it's remembering.