Messages travel along nerve fibers in the brain faster because of myelin, a fatty coating wrapped around each fiber much like the plastic insulation on an electrical wire. The cells that build and maintain this coating are called oligodendrocytes. A study published this month in Cell Stem Cell by Dr. Leyla Akay, Dr. Li-Huei Tsai, and colleagues at MIT found that the APOE4 gene variant turns down a signaling pathway called Wnt inside these cells. With Wnt signaling turned down, oligodendrocytes store fat in small droplets and build less myelin. When the team restored the pathway with a chemical compound, myelin production recovered in human cell cultures and in mice.

APOE4 is a common gene variant linked to late-onset Alzheimer’s disease. In an earlier article I described how carrying one copy is associated with roughly three times the lifetime risk of Alzheimer’s and two copies with roughly twelve times the risk. About a quarter of people of European ancestry carry at least one copy. I carry one copy and have a family history of Alzheimer’s on both sides of my family, which is why I founded NeuroAge. The new study points to a specific molecular target in APOE4 brain cells that several existing drugs act on, providing new insights and potential drug targets for Alzheimer’s disease.
APOE4 was first tied to myelin loss through cholesterol in 2022
The cholesterol finding. In 2022 the same MIT group, with Dr. Joel Blanchard as lead author, reported in Nature that APOE4 caused cholesterol to collect in the wrong places inside oligodendrocytes. They saw this in postmortem human brain tissue, in human cells grown from stem cells, and in mice. Brains from APOE4 carriers had less myelin. In APOE4 mice, a compound called cyclodextrin that helps cells move cholesterol increased myelin and improved performance on learning and memory tests.
The open question. Fat droplets can build up inside a cell for many reasons. The 2022 work showed that APOE4 disrupted fat handling in oligodendrocytes without identifying which signal inside the cell was responsible. The new study set out to find that signal.
The new study switched off genes one at a time to find what controls the fat buildup
Turning off genes one at a time. The team grew oligodendrocytes carrying APOE4 from human stem cells, which are cells that can be directed to become any cell type in the body. They used CRISPR, a gene-editing tool, to switch off each gene that carries instructions for making a protein, one gene at a time and then measured which changes made fat droplets build up or clear. Switching off genes that promote Wnt signaling increased the droplets while switching off genes that restrain Wnt reduced them.
What Wnt signaling does. Wnt is a communication system that cells use to control growth and development. Inside the cell a protein called beta-catenin carries the Wnt signal forward. An enzyme called GSK3b shuts the signal down by marking beta-catenin for destruction.
A second method pointed to the same pathway. The researchers also measured which genes were physically open and available for use in the DNA of APOE4 and APOE3 cells, in lab cultures and in postmortem human brain tissue. Genes that promote Wnt signaling were less accessible in APOE4 samples. The gene-by-gene screen and the measurements of open DNA both pointed to Wnt.
The imbalance inside APOE4 cells. Compared with APOE3 oligodendrocytes, APOE4 oligodendrocytes had less beta-catenin and more GSK3b. That combination keeps the Wnt signal low.
Developmental genes turning up in brain aging is a long-running pattern
Wnt is a developmental pathway. Wnt signaling builds the nervous system before birth, directing which cells become which type and where they go. Genes of this kind, used to construct the brain early in life, are overrepresented among the genes whose activity changes again in the aging brain.
Gene activity measured across the lifespan. Dr. Carlo Colantuoni, Dr. Daniel Weinberger, Dr. Joel Kleinman, and colleagues measured gene activity in the prefrontal cortex, the region at the front of the brain used for thinking and planning, in 269 postmortem brains spanning fetal life through old age (Nature, 2011). A wave of expression changes during fetal development reversed after birth and appeared again during aging. The rate of change was fastest before birth, slowed through childhood and adulthood, and increased again after age 50.
Wnt in the aging brain. Dr. Tao Lu, Dr. Liviu Aron, Dr. Bruce Yankner, and colleagues reported that the developmental regulator REST, which switches off neuronal genes during development, is switched back on in aging human neurons and is lost in mild cognitive impairment, the stage of memory loss that often comes before Alzheimer’s dementia, and in Alzheimer’s itself (Nature, 2014). Wnt signaling was one of the inputs that induced REST, and beta-catenin in the nucleus increased in normally aging prefrontal cortex while decreasing in Alzheimer’s brains. Lithium and CHIR99021, the same GSK3 blockers used in the new MIT study, raised REST levels in that work.
What the new study adds. Wnt activity shifting in the aging and Alzheimer’s brain has been described for over a decade. The MIT study locates a Wnt change in one cell type, the oligodendrocyte, ties it to a specific APOE4 effect on fat handling and myelin, and shows that reversing it restores myelin in those cells.

Blocking GSK3b restored myelin in lab models
Pushing the pathway in both directions. In APOE3 oligodendrocytes, a chemical called Dif3 that increases GSK3b activity increased fat droplets. In APOE4 oligodendrocytes, a GSK3b blocker called CHIR99021 reduced them.
Brain organoids. The team also grew organoids, small three-dimensional clusters of human brain cells that reproduce some features of brain tissue, combining APOE4 oligodendrocytes with neurons. Blocking GSK3b in these organoids increased myelin and promoted the growth of nerve fibers.
Mice. The final test used mice that carry human APOE4 and are engineered to develop tau tangles, one of the protein buildups found in Alzheimer’s brains. Mice treated with CHIR99021 had fewer fat droplets and more myelin in the hippocampus, a brain region involved in forming memories, compared with untreated mice.
GSK3b also acts on tau
One enzyme linked to two features of Alzheimer’s. GSK3b is one of the main enzymes that attaches phosphate groups to tau, a change that leads tau to clump into the tangles seen in Alzheimer’s brains. Dr. Tsai and Dr. Akay have said they plan to investigate the link between GSK3b and tau in their APOE4 models. If the same enzyme contributes to both in APOE4 carriers, a drug that blocks it could affect more than one part of the disease process.
Several drugs that block GSK3 have already been given to people
CHIR99021 is a laboratory research compound that has not been tested in people. Drugs that block GSK3 and have been studied in people include lithium and tideglusib. The 2022 study points to a separate route through cholesterol transport.
Lithium blocks GSK3. Lithium has been prescribed for decades to treat bipolar disorder. One of its known actions inside cells is blocking GSK3. I covered the broader lithium evidence in Low-Dose Lithium for Brain Health?
Lithium and myelin in mice. In 2025 Dr. Bruce Yankner’s group at Harvard Medical School reported in Nature that lithium levels were lower in the brains of people with mild cognitive impairment. Lowering brain lithium by about half in mice increased the buildup of tau and of amyloid, a protein that collects into deposits called plaques in Alzheimer’s brains. The same mice lost myelin, axons, and the synapses that connect neurons. The team traced these effects at least partly to increased GSK3b activity. Lithium orotate, a form of lithium that binds less to amyloid plaques, prevented these changes in Alzheimer’s mouse models and in aging mice. Lithium carbonate, the form prescribed for bipolar disorder, had little effect at the same low doses in those mice. The Harvard and MIT studies used different approaches and both connect GSK3b activity to myelin loss.
Randomized trials of lithium in people with memory loss. Dr. Orestes Forlenza and colleagues at the University of São Paulo randomized 45 people whose mild cognitive impairment mainly affected memory to low-dose lithium or placebo for 12 months. The lithium group had lower levels of the altered form of tau in spinal fluid and better scores on the cognitive portion of a standard Alzheimer’s rating scale (2011 study). In a longer version of the trial with 61 participants treated for two years, the placebo group declined while the lithium group stayed stable (2019 study). The 2026 LATTICE trial led by Dr. Ariel Gildengers at the University of Pittsburgh randomized 80 people aged 60 and older with mild cognitive impairment to low-dose lithium carbonate or placebo for two years (JAMA Neurology). It did not meet any of its six main outcomes. Verbal memory declined at about half the rate in the lithium group, although the difference was small enough to have arisen by chance. About a quarter of LATTICE participants had evidence of amyloid buildup at the start of the trial.
Population studies of lithium. In Denmark, people whose drinking water contained more than 15 micrograms of lithium per liter had a 17% lower rate of dementia diagnosis than people whose water contained 2 to 5 micrograms per liter (JAMA Psychiatry). Dementia rates did not fall steadily as lithium concentrations increased, and the intermediate exposure group had higher rates than the lowest group. The authors noted that other differences between municipalities could explain part of the association. In a UK study of 29,618 patients aged 50 and older using mental health services, 9.7% of those prescribed lithium were diagnosed with dementia compared with 11.2% of those who were not (PLOS Medicine). Lithium use remained linked to lower dementia risk after the researchers accounted for smoking, other medications, and other illnesses. Most lithium users in that study had bipolar disorder, which itself raises dementia risk.
A lithium orotate trial is planned. Johns Hopkins University has registered a trial of lithium orotate in Alzheimer’s disease called LiO-AD (NCT07459959). The registry lists it as not yet recruiting, with an estimated start in December 2026 and completion in 2029. The human dose of lithium orotate for brain protection has not been established. At the higher doses used for bipolar disorder, lithium requires blood level monitoring because of its effects on the kidneys and thyroid.
The two forms of lithium have not been tested the same way. The form that worked in the Harvard mouse work was lithium orotate, and the form that failed in those same mice was lithium carbonate. Every randomized trial in people so far, including Forlenza’s trials and LATTICE, used lithium carbonate or another conventional salt. No randomized trial of lithium orotate in people has reported results. That leaves two separate gaps. The salt with mouse evidence behind it has no human trial data, and the salt with human trial data did not work in the mouse experiments that prompted the current interest. Lithium orotate is sold over the counter as a supplement, and in my opinion the evidence does not yet support taking it for brain health.
Tideglusib. Tideglusib is a drug that blocks GSK3, developed for neurological disease. In the ARGO trial, a mid-stage test of whether a drug works, 306 people with mild to moderate Alzheimer’s dementia took tideglusib or placebo for 26 weeks (Journal of Alzheimer’s Disease). The drug was acceptably safe and did not improve the main cognitive outcome or the secondary measures. Participants already had dementia when treatment began. In The Alzheimer’s Pipeline Is Finally Catching Up to the Biology I described how drugs aimed at upstream biology have tended to show their clearest signals in people at earlier stages of disease.
Cholesterol transport. In the 2022 MIT study, cyclodextrin increased myelin in APOE4 mice by helping oligodendrocytes move cholesterol.
Cyclodextrins are ring-shaped sugar molecules. The ring is water-friendly on the outside and greasy on the inside, so it can trap a fat-soluble molecule such as cholesterol in its center and carry it out of a cell. The form used in the research is 2-hydroxypropyl-beta-cyclodextrin, often shortened to HPBCD. In the 2022 MIT work it pulled cholesterol out of APOE4 oligodendrocytes and improved myelin and memory in mice.
Availability for brain use. No cyclodextrin is approved as a treatment for any neurological condition. The closest thing is Niemann-Pick disease type C, a rare inherited disorder where cholesterol gets stuck inside cells.
Rafael Holdings and Cyclo Therapeutics submitted a New Drug Application to the FDA in October 2026 for intravenous HPBCD, now named remibetadex (Trappsol Cyclo), for Niemann-Pick type C. The pivotal Phase 3 study missed its primary efficacy endpoint overall while showing a benefit in a prespecified subgroup. The FDA has not yet decided. financialcontentallsci
An earlier version, adrabetadex (VTS-270), was given into the spinal fluid. Mallinckrodt discontinued its clinical development in January 2021 after the FDA concluded the trial data were not sufficient to establish effectiveness. npukpatientworthy
Cyclo Therapeutics has also described plans to develop intravenous Trappsol Cyclo for Alzheimer’s.
What is actually purchasable. Alpha-, beta-, and gamma-cyclodextrin are sold as food ingredients and as over-the-counter fiber supplements, and HPBCD is sold as a laboratory and pharmaceutical-grade excipient, the inactive carrier used to dissolve other drugs. Those are not the same as a sterile intravenous formulation given at a therapeutic dose under medical supervision, and swallowing cyclodextrin is a different exposure from infusing it. The intravenous and intrathecal routes have caused hearing loss in NPC patients, which is part of why dosing happens in a clinical setting.
The new screen also pointed to specific steps in fat processing that change when Wnt signaling is low. Dr. Tsai has said the screen could provide new targets for drug development.
Turning up Wnt signaling throughout the body raises safety questions
Wnt and cell growth. Wnt signaling controls cell growth in many tissues. Mutations that leave the pathway permanently switched on drive most colorectal cancers. A drug that strongly activates Wnt signaling throughout the body for years would need careful study for cancer risk.
GSK3 has many jobs. GSK3 also helps regulate how cells store sugar as glycogen and takes part in several other signaling pathways. Approaches that act mainly in oligodendrocytes, or that target the fat-processing steps downstream of Wnt, could limit effects in other tissues. Lithium comes with decades of safety data in people. The safety of a newly developed GSK3 blocker taken for years has not been established.
The myelin results so far come from cells and mice
Where the evidence stands. The myelin rescue in the new study was shown in cultured human cells, brain organoids, and one mouse model. The human data come from postmortem brain tissue, which shows differences in APOE4 brains without testing whether restoring Wnt signaling changes the course of disease in a living person. Answering that question would take trials that enroll APOE4 carriers and use MRI to measure white matter, the myelin-wrapped wiring that connects brain regions, alongside memory testing.
What the study adds for APOE4 carriers. The study places Wnt signaling in oligodendrocytes alongside the cholesterol and inflammation pathways that earlier research has linked to APOE4. Lithium, a drug that acts on this pathway, is moving into new Alzheimer’s trials. The gene screen behind the MIT study may also point to further drug targets as the authors follow up on the fat-processing steps it identified.