Picture a misfolded tau fibril arriving at a neuron. It is swallowed the way cells swallow most things, engulfed into a bubble of membrane that matures into an endosome and then a lysosome, the acidic recycling chamber where proteins are meant to be broken down and forgotten. For the fibril, that chamber is a holding cell. If it stays inside, the story ends. If it gets out into the cytosol, where the cell keeps its own soluble tau, it can act as a template, converting healthy protein into the same misfolded shape, over and over. That single step, getting out, is the hinge on which the prion-like spread of tau through the brain is thought to turn, and it has never been watched properly.
On 10 June 2026, Eric Herrmann, Shuixia Tan, Kevin M. Rose, Richard M. Hooy and James H. Hurley, working at the University of California, Berkeley and within the Aligning Science Across Parkinson's collaborative network, posted a preprint titled "Structural Mechanism and Cellular Restriction of Tau Seeding from Endolysosomes". A preprint means it has not yet been through peer review, and that qualifier belongs in this sentence, not in a footnote. The work was done in primary neurons and astrocytes in culture, treated with pre-formed tau fibrils, and the centrepiece is cryo-electron tomography, which flash-freezes cells and reconstructs their interiors in three dimensions at near-molecular resolution. It is a way of looking at the crime scene rather than dusting it for fingerprints.
What they saw was mostly what they did not see. Unlike LLOMe, a chemical routinely used in the laboratory to damage lysosomes, the tau fibrils were not observed touching the lysosomal membrane or bending its curvature. The lysosomes in fibril-treated cells were structurally intact. And yet extensive protein aggregates of similar character sat on both sides of that intact membrane: inside the lysosomal lumen, and in the cytosol immediately outside it. The authors' reading is that tau fibrils drive co-aggregation with other cellular material inside the lysosome, and that the leak to the cytosol happens through reversible holes in the membrane rather than a rupture. Escape without a break-in, in other words.
The second half of the paper is about who is trying to stop it. Cells run a whole damage-sensing and repair service on their lysosomes, and the team tested several arms of it: the calcium sensor ALG-2, which recruits the ESCRT membrane-sealing machinery; CASM, the conjugation of ATG8 proteins to single membranes; the PITT pathway, which tethers the endoplasmic reticulum to a wounded lysosome and pumps lipids across to patch it; and VPS13C, a lipid transporter already known to genetics as a Parkinson's disease gene. The PITT pathway and VPS13C were strongly implicated in tau seeding in both neurons and astrocytes. CASM mattered in astrocytes but not neurons, and ALG-2 had a lesser role in both. The paper's own title frames these as a restriction system, a brake on escape, which fits the established picture in which ESCRT-mediated repair counteracts tau getting out.
This is a live argument rather than a settled one, and it is worth a family knowing that. In 2024, in the peer-reviewed Proceedings of the National Academy of Sciences, some of the same Berkeley researchers reported that endocytosed tau caused lysosomes to swell, deacidify and recruit ESCRT proteins, with new cytosolic tau aggregates forming on the surface of tau-burdened lysosomes, consistent with nanoscale damage. In November 2025, a separate group posted a preprint arguing something different again: in human iPSC-derived neurons, seeded tau growth stayed inside acidic, membrane-intact endolysosomes, with transient openings letting cytosolic monomers in rather than letting fibrils out. Three studies, one compartment, and a real disagreement about which side of the membrane the damage is done on.
Why any of this matters outside a microscope room: anti-tau drugs have had a mixed record, and one plausible reason is that the field has been aiming at the wrong moment in the journey. An antibody that mops up tau in the space between cells does nothing about a fibril already sitting in a lysosome. If escape is a lipid and membrane problem rather than a brute-force rupture, the targets look different, and they look like the cell's own repair machinery. None of that is a therapy today. Nobody has given a person anything based on this work, and a cultured astrocyte is not a temporal lobe. What the paper offers is a sharper map of the step everyone agrees is decisive, drawn from actual pictures of the cell.