Genetic counselors who work with GRN families have long had to explain an uncomfortable asymmetry. Two relatives can carry the identical progranulin mutation, and one develops frontotemporal dementia in their fifties while the other reaches old age untouched. Since 2010 the best statistical explanation has been a common variant in a second gene, TMEM106B, which sits on almost everyone's chromosome 7 in one of two flavors. The Stanford preprint posted to bioRxiv on March 26, 2026 by Yi Zeng, Jian Xiong and colleagues in the laboratories of Aaron Gitler and Monther Abu-Remaileh puts the point starkly in its own abstract: in GRN mutation carriers, being homozygous for the protective TMEM106B haplotype appears to confer near lifetime protection against FTD. What nobody could say was why. The two genes were known to talk to each other; the conversation itself had never been overheard.
The method the Stanford team used is a way of eavesdropping on a single organelle. They tagged the lysosome, the acidic compartment where cells digest their own worn-out parts, with a handle (TMEM192 carrying three HA tags) and pulled intact lysosomes out of cells, a technique called lysosome immunoprecipitation, or LysoIP. Then they looked inside. In human neurons derived from induced pluripotent stem cells in which GRN had been knocked out, and in isogenic controls that differed only in that gene, they found a marked increase in the TMEM106B C-terminal fragment (the CTF) sitting inside the lysosomes of the GRN-null neurons. Alzforum, reporting on the preprint on April 30, 2026, put the difference on the published blot at roughly tenfold. The same buildup showed up in Grn-knockout mice. Crucially, TMEM106B messenger RNA did not change, meaning the cells were not making more of the protein; they were failing to clear a piece of it.
That detail is what makes the fragment interesting rather than merely present. The CTF is the piece of TMEM106B that cryo-electron microscopy studies have shown assembles into amyloid fibrils in the brains of people with FTD and other neurodegenerative diseases, and to a lesser degree in seniors with no diagnosis at all. In the Stanford experiments the buildup was reversible: adding recombinant progranulin back to progranulin-deficient human neurons reduced the fragment, and did so in a dose-dependent way. Then the team tested the modifier directly, using isogenic stem cell lines that differ only at position 185 of TMEM106B, the single coding change (threonine to serine, rs3173615) that rides with the risk and protective haplotypes. Neurons carrying the risk allele accumulated more fragment in an allele-dose-dependent manner, one copy worse than none, two copies worse than one, and progranulin again reversed it. Structural and genetic work pointed to a reason: TMEM106B normally pairs up into dimers, and dimerization stabilizes the protein and limits fragment formation. Engineered protein that cannot dimerize spilled out extra CTF whether granulin was present or not, which told the authors that granulin is not the thing holding the dimer together; it is the thing clearing the fragment once the dimer comes apart.
Two days later, on March 28, 2026, a second and much larger consortium led by John Replogle, Jordan Marks and Martin Fernandez, with Michael Ward, Leonard Petrucelli and others, posted a preprint reaching the same junction from a different direction. Their abstract states that TMEM106B and GRN variants converge on a single molecular intermediate, the cleaved intra-lysosomal fibril core, and that the p.T185 risk variant drives its accumulation by impairing degradation. Two independent groups, different cell models, same lesion, arriving in the same week is the kind of convergence that usually survives peer review, though neither paper has been through it yet.
For a family living with FTD-GRN, the practical value here is not a drug. It is a target you can count. Progranulin-raising therapy has already been tested at scale and come up short on the measure that matters: the ClinicalTrials.gov record for the latozinemab continuation study NCT06111014 states that it was terminated because the Phase 3 parent trial failed to meet the clinical co-primary endpoint of slowing FTD-GRN progression, a result Alector announced on October 21, 2025. Raising progranulin in the blood is easy to demonstrate and, on its own, was not enough. A lysosomal fragment that rises when granulin is lost, rises further with each copy of the risk allele, and falls when progranulin is restored is a different sort of yardstick: something inside the cell that a candidate therapy either moves or does not. The Stanford authors go further and name dimer stabilization, holding TMEM106B in its paired form, as a strategy worth pursuing. That is a hypothesis at the bench, nothing more.