In Parkinson's and Alzheimer's disease, some of the brain's own proteins stack up into long, extremely stable fibres. This page shows what those fibres look like, how they spread, and how we photograph single molecules to find out.
Everything in 3D here is a real structure solved by cryo-electron microscopy. Drag to rotate, pinch or scroll to zoom.
A protein is a chain of building blocks called amino acids, strung together like beads. There are 20 kinds, and each has a different side group sticking off the chain. What matters most is its character: oily ones avoid water and like to hide together, charged ones attract or repel each other, and a few special ones bend or lock the chain.
Ubiquitin's 76 letters mix oily and charged ones along the whole chain. When it folds, the oily ones tuck inside, away from water, and hold the shape together. Lysine 48 (K48) is where the cell links ubiquitins into the chains that mark a protein for disposal.
The start is rich in positive lysines (K) and grips cell membranes. The middle stretch (NAC, residues 61–95) is about half oily with almost no charge: this is the part that stacks into the fibre core. The tail carries 15 negative charges (D, E) and stays floppy.
Compare the two in 3D. Ubiquitin, a small protein found in every cell (it's the "dispose of this" tag from section 3), always folds into the same compact shape of one helix packed against a β-sheet. The α-synuclein shapes are eight snapshots from a published ensemble (Protein Ensemble Database entry PED00024, 576 structures), and no two look alike, which is exactly what "disordered" means.
A useful picture is a landscape of free energy, a measure of how stable each shape of the chain is. The chain wanders over it like a ball, drifting toward lower ground: the lower the point, the more stable that shape. Compare ubiquitin, a typical folded protein, with α-synuclein:
A sketch, not a calculation. Real landscapes have thousands of dimensions; this squeezes them onto one axis.
Most proteins fold into a compact, specific shape and do one job. α-synuclein, the protein in the model above, is different: in a healthy nerve cell it has no fixed shape at all and floats around as a flexible chain.
Occasionally, copies of it lie down flat, one exactly on top of the next, like sheets of paper in a ream. Each new layer is held to the one below by a zipper of hydrogen bonds. The result is a fibre thousands of layers long that cells find very hard to break down. The protein highlighted in orange above is the last layer at one end of the fibre, the spot where the next copy would join.
The same protein chain can flatten into several different folds. Which fold appears seems to depend on the disease. Fibres taken from patients with multiple system atrophy look different from those in Parkinson's disease, and both differ from fibres grown in a test tube.
The same pattern holds for tau, the protein behind Alzheimer's disease: Alzheimer's, Pick's disease and chronic traumatic encephalopathy each have their own tau fold. That makes fibril shape a possible fingerprint of the disease.
After a protein is made, the cell can attach small chemical groups to it: a phosphate, a sugar, a small protein called ubiquitin. These post-translational modifications (PTMs) act like sticky notes that change how the protein behaves. Fibres from patients carry a lot of them.
Only the middle section of α-synuclein is built into the fibre core. The two ends stay floppy and form a fuzzy coat around it. Whether a tag lands in the core or the coat changes what it can do.
Starting a fibre from scratch is slow and rare. Once one exists, it copies itself. Four steps, shown here as a sketch:
Rarely, a few loose copies happen to stack into a stable nucleus.
Loose protein that meets an end is pulled into the same fold.
Fibres snap. Every break makes two new growing ends.
Fragments leave the cell and seed the same fold in a neighbour.
The curve is what researchers measure in the lab with thioflavin T, a dye that glows when it binds fibres. Adding a few ready-made fibres (seeds) skips the slow nucleation step, which is why seeds are so potent. The same property is used diagnostically: seed amplification assays detect tiny amounts of α-synuclein seeds in spinal fluid by letting them template a large reaction.
Dozens of different proteins can form amyloid, and they are behind very different diseases, from dementia to heart failure to diabetes. Every structure below was solved by cryo-EM, most from tissue taken from patients. Pick a disease to see its fibre cross-section.
Fibres are far too small for a light microscope, so we freeze them in a thin film of glass-like ice and photograph them with electrons. The catch is that electrons destroy protein, so each picture uses a very low dose and comes out extremely noisy.
The trick is averaging. A fibre is the same all the way along, so we cut every picture into thousands of short segments, line them up, and average. The noise cancels out and the structure remains. Try it with the fibre from section 1.
Before any averaging, someone has to freeze the sample and photograph it. The experiment has four steps. A top-end cryo-electron microscope is about as tall as a room and sits on its own vibration-damped floor.
A tiny drop (about 3 µL) of fibres goes onto a 3 mm metal grid covered in a thin film full of holes.
Excess liquid is blotted away and the grid is plunged into liquid ethane at about −180 °C. The water freezes so fast that it turns into glass-like ice instead of crystals.
Inside the microscope, a beam of electrons accelerated to 300,000 volts passes through the ice. The column is under vacuum and kept cold, and lenses magnify the image onto a camera.
Each picture is a short movie of about 40 frames, so any drift can be corrected before the frames are added up. Thousands of movies are collected over a day or two.
From there, the computer takes over: finding the fibres in every picture, cutting them into segments and averaging, as in the demo below.
Change the viewing angle and the projection changes. The layer stripes along the fibre are 0.48 nm apart; they only come through once enough segments are averaged. A real reconstruction uses tens to hundreds of thousands of segments.
Cryo-EM gives a sharp snapshot of the fibre, but it can't say how a change to the protein would affect its stability; for that we turn to computer simulation. A molecular dynamics simulation follows every atom, updating their positions every two femtoseconds (two millionths of a billionth of a second), for millions of steps.
A special kind of simulation allows us to answer the question: would adding a phosphate to serine 87 make the fibre more or less stable? Measuring that directly would mean simulating fibres falling apart, which takes far too long. Instead we cheat: in the computer, we slowly turn the serine into a phosphoserine, once in the loose protein and once inside the fibre, and measure the work it takes each time. Atoms appearing out of nothing is impossible in reality but perfectly fine in a simulation, which is why this is called alchemy.
Each switch is a short simulation that forces the change in a fraction of a nanosecond. Because it is rushed, every run costs a slightly different amount of work. We run many switches in both directions: serine to phosphoserine (red) and back again (blue, with the sign flipped so both can be compared).
A result from statistical physics, the Crooks Fluctuation theorem, says the true free energy change ΔG sits where the two distributions cross, even though no single rushed switch gives it. With a few runs the estimate wobbles; with more it settles. Doing the same in the loose protein and subtracting gives ΔΔG. A positive value means the tag makes the fibre less stable. The numbers here are illustrative, not results from a real calculation.