APOE4 turns the cells that brace brain vessels into scar-formers, two studies find

More than 7 million older adults in the United States live with Alzheimer's disease, according to the Alzheimer's Association 2026 Facts and Figures report. The gene APOE4 is the single largest inherited risk factor for that disease, and for decades researchers have not known exactly how it does its damage. Two new studies from a Mount Sinai-led team, published on September 24 in Cell01069-X), point to a specific answer. APOE4 turns the cells that brace the brain's smallest blood vessels into scar-formers, thickening vessel walls, weakening the blood-brain barrier, and trapping amyloid protein around the vessels themselves. The team reports that the resulting damage can be interrupted in human tissue, though not yet in patients.

The full Cell article is paywalled; the version readable in full on PubMed Central is an earlier preprint of the same work, so the final published numbers may differ from what is summarized here.

The mechanism centers on pericytes, the cells that wrap the brain's capillaries and help keep the blood-brain barrier intact. In a cerebrovascular atlas covering roughly 64,000 nuclei from 220 individuals, half carrying APOE3/3 and half carrying APOE4, the team found that APOE4 brain tissue had markedly fewer pericytes. The missing cells were replaced by myofibroblasts, the scar-forming cells familiar from wound healing. Vessel walls thickened. Amyloid, the protein that forms Alzheimer's plaques, piled up around the outside of those vessels.

Lineage tracing pinned the swap on APOE4. When the researchers followed the cells over time, the myofibroblast-like cells had started out as pericytes. APOE4 had pushed them across a state boundary into a scar-forming identity. That finding reframes the vascular damage: it is not a downstream consequence of dying neurons, it is a parallel injury that may be driving the disease forward.

Two molecular handles held the new identity in place. The first is TGF-β, a signaling protein long associated with fibrosis in lung, liver and kidney disease. The second is fibronectin, an extracellular matrix protein that scaffolds scar tissue. When the researchers blocked either handle in APOE4 human-tissue models, myofibroblast counts fell, pericyte coverage rebounded, and the build-up of vascular amyloid dropped back to APOE3 levels. The intervention worked, in the dish.

The human-tissue experiments are the new piece. The team built what they call vascularized miBrains, a stem-cell-derived platform that grows human brain vasculature from induced pluripotent stem cells. miBrains give researchers a way to test vascular and protein-clearance interventions in human tissue, with a human genetic background, before a clinical trial. The same platform was used to run the TGF-β and fibronectin experiments, which is what allows the team to say the damage was reversed rather than merely slowed.

The caveats are real. The atlas integrates postmortem human tissue, aged humanized mice, and iPSC-derived models. No patient took a drug. No clinical efficacy, safety, dosing, delivery, or cost has been established. TGF-β inhibitors exist as approved cancer drugs, but re-targeting them to brain vasculature in elderly patients is its own risk. Fibronectin knockdown is even further from a clinical lever.

A separate companion paper, summarized in the Mount Sinai release on ScienceDaily, points at a second vulnerability in APOE4 astrocytes, the cells that maintain cholesterol balance and lysosomal function in the brain. That work implicates alpha-synuclein, a protein better known from Parkinson's disease research, and is not part of the vascular story. The two studies run in parallel: the gene injures brain tissue from more than one direction at once.

The bigger shift is methodological. For the first time, researchers can name a specific cell-state change caused by APOE4, watch it happen in human tissue, and reverse it. That is a target map, not a treatment. It is also a testbed. If a candidate drug is going to work against APOE4-driven vascular damage, miBrains are now the place to find out before a trial.

What to watch next: a fibronectin-directed or pericyte-state-restoring compound entering a translational program, an independent group replicating the TGF-β reversal in miBrains, and any move to test existing TGF-β inhibitors in vascularized human-tissue models of Alzheimer's.