For nearly two decades, TDP-43 has been the protein at the centre of the frontotemporal dementia and ALS story: normally resident in the nucleus of a cell, where it helps decide which pieces of RNA get spliced into a working message, and in disease found clumped in the cytoplasm instead. Families hear the name early. What nobody could tell them was why the clumps show up in some neurons and not the one sitting beside it. That is selective vulnerability, the oldest unanswered question in this field, and it has mostly been described rather than explained.

A study published in Nature Communications on 9 March 2026 by Wolfgang P. Ruf, Julia K. Kuehlwein, Laura Meier and colleagues, with Karin M. Danzer as senior author at the DZNE Ulm site and Ulm University Hospital, tries to answer it by counting. The team worked with post-mortem primary motor cortex from 30 people with ALS, 20 with ALS-FTD (including sporadic cases and familial cases carrying the C9orf72 repeat expansion) and 32 people without neurological disease, with phospho-TDP-43 neuronal inclusions confirmed in every disease case. They used flow cytometry to sort neuronal nuclei according to whether TDP-43 had been depleted from the nucleus, then read chromatin accessibility and gene expression from the same individual nucleus, and added spatial transcriptomics to keep track of where in the cortical layers each cell sat. Combining the two molecular readouts resolved 75 cell clusters.

The answer that came back was not the one folklore predicts. The pathology sits overwhelmingly in excitatory neurons, the cells that pass signals on and amplify them, and within that group it concentrates in five subtypes: four intratelencephalic populations, defined by markers as L2-L3 LINC00507 FREM3, L3-L5 RORB LNX2, L3-L5 RORB ADGRL4 and L6 THEMIS LINC00343, plus one extratelencephalic layer 5 population, L5 FEZF2 NTNG1. Danzer, in her institute's own announcement, put it plainly: "The protein aggregates occur predominantly within excitatory cells." The inhibitory interneurons, which the same dataset shows carry higher levels of TARDBP messenger RNA than excitatory neurons (a moderate effect, eta-squared 0.23), are not where the aggregates gather. That is the beginning of a cell-intrinsic explanation rather than a restatement of the puzzle.

The second half of the finding matters more for anyone thinking about treatment. When TDP-43 leaves the nucleus, cryptic exons, stretches of intron that should never appear in a finished message, get spliced in. This study reports that the cryptic exons are not a uniform failure across the cortex: STMN2 cryptic exon counts were highest in layer 5 and 6 near-projecting, extratelencephalic and corticothalamic neurons, while KALRN cryptic exons clustered in layer 6 intratelencephalic neurons. In other words, the same loss of one protein produces different broken messages depending on which neuron it happens in. The disease signature was detectable across all major cell types but enriched in excitatory neurons, and was more pronounced in the ALS-FTD cases than in ALS alone.

For a family living with an FTD or ALS-FTD diagnosis, none of this changes anything about today. It changes the shape of the target. Much of the therapeutic effort against TDP-43, including antisense approaches aimed at cryptic exons, assumes a common lesion that a single agent can correct. This dataset argues that the lesion has a postcode: the cells that break, and the genes that break inside them, differ, and a therapy that rescues one population may leave another untouched. Danzer's own reading is that future treatments will need to be tailored to specific cell types. That is a harder brief than a single drug, and an honest one.