Put two microscope images side by side and the difference is obvious even to someone who has never looked down a lens. On the left, neuronal nuclei sit as smooth, rounded shapes. On the right, the outlines are crumpled and puckered, as though something inside had been let out. King's College London published that pair of images in June, adapted from the first figure of a Nature Communications paper, and it is as close as we get to a photograph of the question families keep asking: between the plaque on the scan and the change in the person, what actually kills the cell?
The paper, by Rebecca Casterton, Manolis Fanto, Sarah Mizielinska and colleagues at King's and the UK Dementia Research Institute, published 26 June 2026, gives that step a name: karyoptosis. Their account runs like this. When the autophagy-lysosome system, the cell's clearance and recycling machinery, cannot keep pace with misfolded protein, the resulting proteotoxic stress destabilises the nuclear lamina. LaminB1, a structural protein that normally lines the inside of the nuclear envelope, turns up as puncta out in the cytoplasm. The nuclear outline loses its regular circularity. The cell shrinks. DNA damage follows rather than leads, and nuclear material is expelled from the cell altogether. The authors describe it as non-apoptotic and non-necrotic: the stressed cells did not show caspase-3 activation or caspase-mediated PARP-1 cleavage, the classic signatures of apoptosis, and did not fit necroptosis, ferroptosis or parthanatos either. The sequence matters as much as the ending, because it puts the nucleus first and the broken DNA second.
Then they went looking for the switch. The paper reports that the p38 MAP kinase pathway regulates karyoptosis by directly phosphorylating LaminB1, and identifies serine 391 as the target residue. Pharmacological inhibition of p38 alpha reduced cytoplasmic LaminB1 puncta and nuclear deformation in cells whose lysosomal clearance had been blocked with bafilomycin A1, and it did so without clearing the p62 that marks the underlying protein jam. That is the most interesting sentence in the paper for a reader who has followed the autophagy story for years: the toxic stress stayed exactly where it was, and one downstream execution step was still interrupted. A non-phosphorylatable LaminB1 mutant (S391A) resisted the loss of LaminB1 seen in replicative senescence and rescued karyoptotic features. All of this was done in neuroblastoma cells, fibroblasts, rat primary cortical neurons expressing the toxic dipeptide repeat proteins of C9orf72-linked ALS and frontotemporal dementia, and adult fruit fly brains. None of it was done in a person.
The human part of the study is narrower, and the authors are careful about it. They stained post-mortem frontal cortex from people with frontotemporal lobar degeneration or Alzheimer's pathology, alongside age and sex matched controls, for LaminB1 and the neuronal marker MAP2, then measured nuclear circularity, nuclear area and cytoplasmic area cell by cell and fed those numbers to a k-means clustering algorithm that was blind to which diagnostic group each cell came from. Four phenotype clusters emerged in the deep layers of the cortex. One of them, defined by low LaminB1 nuclear circularity and a small cytoplasm, was over-represented by neurons from FTLD cases and from Alzheimer's cases at Braak stage 6. They then repeated the analysis on most of the same cohort with better autofluorescence quenching, which let them add cytoplasmic LaminB1 puncta as a fourth input, and again found four clusters. King's own summary of the work puts the scale at roughly 3,000 cells from 28 people, with about 35 per cent of frontal cortex cells from Alzheimer's cases showing signs of karyoptosis against about 15 per cent in aged controls.
That is a photograph, not a film. Post-mortem tissue is the end of a decades-long process, and a shrivelled nucleus in fixed tissue tells you a cell looked like that when the person died, not that karyoptosis killed it, nor when, nor how fast. First author Rebecca Casterton framed the ambition modestly in the King's release, saying the study "uncovers a new series of chemical events which can coordinate cell death in brain cells". Manolis Fanto put the therapeutic hope in equally conditional terms, suggesting that targeting the p38 and LaminB1 interaction might "slow down the process of cell death, buying time for more pinpointed therapies". Nothing here is a drug, a dose or a test. There is no p38 treatment for dementia to ask about, and the paper does not claim one.
What it does offer is a named, testable step, which is worth more to families than it sounds. Much of the last decade of thinking about proteotoxic stress has stopped at the clearance machinery, on the assumption that a cell drowning in protein simply gives up. Karyoptosis proposes something more specific and more interruptible: an execution route with its own kinase, its own substrate and its own visible hallmarks, which can in principle be blocked even while the protein burden remains. It also has a long paper trail rather than a sudden headline. The same group described nucleophagy-based LaminB1 degradation and nuclear breakdown in a mouse model of the ataxia DRPLA back in 2017, posted this work as a preprint in 2023, and only now has it come through peer review. Other laboratories have since described karyoptosis-like death outside the brain entirely, in cancer cells, which suggests the pathway is a general one that neurons happen to be badly placed to survive.