For most of the past forty years, almost every serious question about human prions ended the same way: inject mice, then wait. Wait while a transgenic colony incubates. Wait 250 days, or 500, or longer, for the animals to declare an answer about whether a surgical instrument was truly clean, whether a drug touched the agent, whether one patient's prions behaved like another's. In 2003 Peter Klohn and Charles Weissmann showed that mouse scrapie prions could be counted in cultured neuroblastoma cells with roughly the sensitivity of a mouse bioassay, ten times faster and far cheaper. Human prions declined the invitation. Dividing cells simply diluted them away, and a goal set decades ago stayed open.
It is now, at least partly, closed. In PNAS on 30 June 2026, Akin Nihat and colleagues at the MRC Prion Unit at UCL and the UCL Institute of Prion Diseases describe EKV, a dividing cell line that propagates real sporadic Creutzfeldt-Jakob disease prions taken from patients' brains. The recipe is patient rather than glamorous. They began with CAD5, a mouse neuronal line long used for animal prions, took the KDB3 subline in which mouse prion protein is knocked down, and put in the human PRNP open reading frame carrying valine at codon 129. Then they cloned single cells, challenged them with sCJD brain homogenate, kept the survivors that lit up, and did it again. Four generations of this produced EKV, which accumulated five to fourteen times more disease-associated prion protein than the first-generation clone, 81F9, across multiple patient samples.
The hard part of any such claim is proving that what grows in plastic is still the pathogen and not a laboratory artefact. Here the authors went back to the animals one last time. Lysates from sCJD-exposed 81F9 cells, passaged six times to dilute the original inoculum, were inoculated into Tg152-congenic mice that overexpress human valine-129 prion protein on a mouse Prnp-null background. All 18 mice given cell lysate developed prion disease, with a mean incubation period of 251 days (standard deviation 6), and the paper reports disease clinically and neuropathologically indistinguishable from inoculation with sCJD brain tissue. That is the sentence the field will quote: the cells were making infectious human prions, not merely misfolded protein.
What follows is the practical payoff. The team built what they call the Human Prion Assay, reporting reliable detection across inoculum dilutions from one in a thousand to one in a hundred thousand, with a mean lower limit of detection at a 1.02 x 10-5 dilution of crude brain homogenate and a dose-response fit above R2 0.97, and relative titres differing reproducibly by up to a hundredfold between patient samples. They also made persistently infected clones, iEKV: clone 1G2 running at a mean 50 percent infected cells, clone 1E10 at 59 percent, stable over at least 15 passages spanning 40 to 55 days, freezable and revivable, producing protease-resistant prion protein with glycoform patterns comparable to the original patient inoculum. In one experiment, established infection in those cells was cleared by the anti-prion protein monoclonal antibody ICSM18, the mouse parent of PRN100, the antibody given to six patients in the first-in-human treatment programme reported in The Lancet Neurology in 2022. That history is worth holding steady: curing a flask is not curing a person, and John Collinge is a director of D-Gen Ltd, the spin-out that supplied the antibodies used here.
The boundary is written into the paper itself, and families should read it before hope gets ahead of the data. EKV carries valine at codon 129, and it propagated prions only from patients with at least one valine allele; prions from methionine homozygous patients did not take, which the authors attribute to a genotype mismatch too inefficient to outrun cell division. Methionine homozygotes are heavily over-represented among people who develop sporadic CJD, so a large share of real cases sits outside this model for now. The authors also flag that six cell passages may not clear every trace of the original inoculum, which matters for how such experiments are controlled. The published version, received 5 January 2026 and accepted 6 May 2026, is a tightened relative of the bioRxiv preprint posted on 26 November 2025: the preprint led with propagation and curing, the journal version leads with scalability and quantification, and cites a companion PNAS paper from 11 June 2026 in which a CRISPR knockout strategy produced dividing cells susceptible to variant CJD.
For a family living with a CJD diagnosis, nothing about today's care changes. What changes is the price of asking questions. Whether a decontamination protocol truly destroys human prion infectivity, how much infectivity sits in tissues outside the brain, whether a candidate drug touches the human agent rather than a mouse stand-in: each of those has been rationed for decades by the cost and duration of animal bioassay. If EKV holds up in other laboratories, those experiments become weeks of bench work rather than years of waiting.