MIT researchers have traced the brain's biggest genetic risk for Alzheimer's to an unglamorous place: the fatty wrapping on the wiring, and the molecular switch that tells the wrapping cells to get on with it.
The work was published on October 15 in Cell Stem Cell by Leyla Akay, Li-Huei Tsai and colleagues at the Picower Institute, and it concerns APOE4, the variant carried by a large minority of people and the strongest common genetic risk factor for late-onset Alzheimer's disease.
Start with the thing the gene breaks. Neurons send long cables, axons, across the brain, and cells called oligodendrocytes wrap those cables in a sheath of fat called myelin, the way you would lag a pipe. Insulated cable carries a signal fast and clean. Think of the half-second pause before a familiar name arrives: the brain runs on timing, and timing runs on insulation.
In 2022, the same lab reported in Nature that APOE4 oligodendrocytes mishandle cholesterol, stranding fat inside the cell in droplets instead of shipping it out to the sheath, with myelination suffering as a result. That paper named the symptom. The new one goes after the cause.
The method is brute-force and rather beautiful. The team grew human oligodendrocytes from induced pluripotent stem cells carrying APOE4, then used CRISPR to knock out every protein-coding gene in the genome, one gene per cell, and sorted the cells by how much fat they had stockpiled. Whichever genes turned up over-represented in the fattest cells, and in the leanest, were the levers. "This CRISPR screen allowed us an unbiased way to test every single possibility," Akay told MIT's news office.
The levers clustered. Canonical Wnt signaling genes, including FZD10 and WNT16, pushed lipid up when they were removed. Knock out the brakes on Wnt, DKK1 and NKD1, and the droplets fell. Chromatin sequencing in cell cultures and in postmortem human brain tissue, comparing APOE4 carriers with people carrying the risk-neutral APOE3, pointed the same way: in APOE4, the Wnt program is turned down.
Downstream of Wnt sits an enzyme with a long name and a long rap sheet, glycogen synthase kinase-3 beta, GSK3B. When Wnt signaling falls, GSK3B runs hot. The team drove GSK3B activity up in APOE3 cells with a compound called Dif3 and watched droplets appear. They inhibited it in APOE4 cells with CHIR99021 and watched droplets shrink.
In three-dimensional brain organoids combining APOE4 oligodendrocytes with neurons, inhibiting GSK3B increased myelination and encouraged axons to grow. In mice carrying human APOE4 and the PS19 tau mutation, doses of CHIR99021 reduced lipid droplets in the hippocampus, the brain's memory workshop, and improved myelination.
Here is where a reader should slow down. CHIR99021 is a laboratory tool compound, sold for bench work. And GSK-3 inhibition has already had its turn in people: the ARGO trial randomized 306 patients with mild to moderate Alzheimer's to tideglusib or placebo for 26 weeks, and reported acceptable safety with no clinical benefit. That result is eleven years old and it still sets the bar. Hitting the enzyme is easy; hitting it in the right cell, at the right time, in a person, is the hard part nobody has yet done.
What changes today is the map. For thirty years the APOE4 story has been told mostly through amyloid, with blood vessels as the supporting cast. This line of work puts white matter, the brain's insulated cabling, in the frame with a specific pathway, a specific enzyme, and two measurable readouts: fat droplets and myelin thickness. Readouts are what trials need.
For the daughter reading this between calls to a neurologist, the practical content is small and worth saying plainly. There is no treatment here. There is no test to ask for. There is a mechanism that was a black box in 2022 and has a named switch in 2026, and that is the sort of progress that shows up in clinics about a decade later, if it shows up at all.