How an Alzheimer's tau filament makes healthy human brain cells copy its shape

In Short. Tau filaments from two people who died with Alzheimer's made healthy human brain cells, grown in a dish, build filaments with the same atomic shape, and cells with only half their usual tau built none.

A healthy nerve cell keeps its long internal supports in order with the help of a protein called tau, which binds to those supports and steadies them. In Alzheimer's disease some tau changes shape, and the altered copies stack into long threads. Each thread is a filament, and the clumps of filaments that build up inside dying neurons are what doctors call tangles.

For years the shape of those filaments could be read only from the brains of people who had already died. A pair of papers posted on October 5 report something closer to the living process. A team in Cambridge took tau filaments from the brains of two people who had died with Alzheimer's and added them to human brain tissue grown in a dish. The tissue then made its own healthy tau fold into filaments with the same shape as the patients'.

The dish tissue is an organoid, a small ball of brain cells grown from human stem cells. It is far from a brain. These organoids held neurons, astrocytes and some support cells, but no microglia, the brain's resident immune cells, and no blood supply. The stem cells came from three healthy donors. The organoids were grown for two to three months and given a single dose of patient tau, usually 500 nanograms. Faint abnormal tau appeared by day 28 after dosing and grew stronger out to day 129. Raising the dose added little once it passed 500 to 1,000 nanograms.

To check the shape of what formed, the team used cryo-electron microscopy, which freezes a sample and images it with a beam of electrons to work out how a molecule is folded. By this method, called cryo-EM, they read the organoid filaments at a resolution of 4.5 angstroms. Of the filaments imaged, more than 70 percent had the paired-helical shape seen in Alzheimer's disease, and 24 percent were single strands with the same fold.

What the organoids did is called seeding, or templated copying: an added filament makes a cell's own tau take on the filament's shape, so the abnormal form copies itself from existing healthy protein. The strongest evidence that the new tau was a copy, not leftover from the dose, came from changing how much tau the cells could make. Organoids with the usual two copies of the tau gene seeded readily. Those engineered to carry one copy, with about half the normal tau, made no detectable seeded filaments, and those with no copies made none, across 7 to 12 organoids per time point. Less starting tau meant no copying.

The companion paper, from the MRC Laboratory of Molecular Biology, made the same point in mouse neurons given a human tau that carries a known alteration at position 337. When seeded with patient tau, 82 percent of the filaments the neurons built carried that marker, proving they were assembled from the cells' own tau rather than the added material. The authors note a trap: when they added far too much seed, the filaments they recovered were the original patient material, not copies, so the dose has to be kept low to tell the two apart.

That the organoid filaments match the real disease rests on independent work. In 2017 a group led by Anthony Fitzpatrick used cryo-EM on tau from one Alzheimer's brain to show that the filaments are built from two identical strands over residues 306 to 378, the paired-helical fold the new organoids reproduced.

Both new papers are preprints that have not been through peer review, and they share three senior authors, so they support each other rather than confirm each other independently.

The half-tau result points at a kind of drug already in testing. Biogen's diranersen, once called MAPTRx, lowers tau production, and in a 46-person phase 1b trial it cut tau in spinal fluid by more than 50 percent at its higher doses. It is now in a phase 2 study called CELIA. The organoid finding is a laboratory hint that lowering tau could block this copying. It is not evidence that the drug helps patients.

Nor is any of this a living brain. The organoids lack immune cells and blood vessels, and their tau is the fetal three-repeat form rather than the adult mix. What the work shows is narrower and still useful: a patient's filament can impose its shape on fresh human tau, and starving the process of tau stops it.