For decades, Alzheimer’s research has centered on neurons, the brain cells that die as the disease progresses, and on the sticky amyloid-beta plaques that accumulate between them. A study published in Nature Neuroscience in May 2026 shifts the spotlight to a different cell entirely: the astrocyte, a star-shaped support cell that outnumbers neurons in many brain regions and, it turns out, can be switched on to devour the very plaques that define the disease.
The research team found that a single protein, the transcription factor Sox9, acts as that switch. When Sox9 was artificially boosted inside astrocytes of mice engineered to develop Alzheimer’s-like pathology, the cells ramped up production of a surface receptor called MEGF10 and began engulfing amyloid plaques that had already formed. The mice not only had fewer plaques afterward; they also performed normally on memory and learning tests, a combination that many experimental Alzheimer’s treatments have struggled to achieve in animal models.
What the experiments actually showed
The researchers used App knock-in mouse models, a newer generation of Alzheimer’s mice that carry disease-linked mutations in their own amyloid precursor protein gene rather than flooding the brain with an artificial surplus of the protein. This design avoids some of the artifacts that plagued earlier mouse studies and produces amyloid buildup that more closely mirrors what happens in human brains.
Crucially, the team waited until plaques had already accumulated before switching on Sox9 in astrocytes. That timing matters. Preventing plaques from forming in the first place is a different challenge from clearing ones that are already there, and many past interventions that looked promising at the prevention stage failed when tested against established pathology. Here, plaque burden dropped and cognitive performance held steady, suggesting the astrocytes were actively dismantling existing deposits rather than simply slowing new ones.
The molecular trail led directly to MEGF10. Sox9 is a transcription factor, meaning it sits on DNA and turns genes on or off. Two publicly deposited genomic datasets back up the connection. An RNA-sequencing dataset confirmed that Sox9 manipulation changed the activity of phagocytosis-related genes in astrocytes, with MEGF10 among the most prominent. A companion ChIP-seq dataset showed Sox9 physically binding near MEGF10’s regulatory regions in hippocampal tissue from healthy, Alzheimer’s-model, and aged mice. Together, the data establish a direct line: Sox9 lands on the genome near MEGF10, turns it up, and the astrocyte gains the machinery to eat plaques.
When the researchers boosted MEGF10 alone, without raising Sox9, the mice still showed preserved cognition. That result suggests MEGF10 is not just one cog in a large Sox9-driven program but the critical downstream player responsible for the cognitive benefit.
MEGF10 was already on the radar
MEGF10 did not come out of nowhere. Earlier peer-reviewed cell-culture studies had shown that overexpressing the receptor increased uptake of amyloid-beta 42, the peptide fragment most strongly linked to Alzheimer’s toxicity, while knocking it down reduced uptake. Mutational experiments in those prior studies pinpointed specific motifs in MEGF10’s cytoplasmic tail that are required for engulfment, indicating the receptor actively drives the process rather than passively sticking to debris. (These findings come from previously published work cited by the Nature Neuroscience paper, not from the new study itself.)
Separate foundational work established MEGF10 as the receptor astrocytes use to prune unnecessary synapses during normal brain development and maintenance. Loss of MEGF10 left mice with excess synapses and disrupted circuit refinement; boosting it enhanced synaptic removal. The new study extends that known biology from synapse cleanup to amyloid cleanup, connecting two previously parallel research threads into a single pathway.
Why this has not reached patients yet
Every result described above comes from mice. No human clinical data, no postmortem brain tissue analysis, and no patient-derived cell experiments have yet confirmed that the Sox9-to-MEGF10 pathway operates the same way in people with Alzheimer’s disease. Even well-designed knock-in mouse models do not replicate the decades-long progression, the tangled tau protein pathology, the vascular damage, or the complex neuroinflammation seen in human patients.
There are also open questions about specificity. The RNA-seq data show that Sox9 does not just flip on MEGF10; it shifts the activity of a broad set of genes related to phagocytosis and metabolism. Which of those changes are helpful and which might cause harm is not yet clear. If Sox9 activates a wide pro-phagocytic program, there is a real risk that astrocytes could start stripping away healthy synapses alongside amyloid deposits, potentially destabilizing neural circuits.
Sox9 itself carries additional safety concerns. It is a powerful developmental transcription factor involved in glial cell specification and extracellular matrix regulation. Sustained or high-level activation in adult astrocytes could affect scar formation after injury, blood-brain barrier integrity, or inflammatory responses in ways that short-term mouse experiments would not reveal. Longer studies in aged animals and in models with additional health complications will be needed before anyone can assess the true risk profile.
No statements from the research team about therapeutic timelines, drug delivery strategies, or safety testing have appeared in the published sources. The deposited datasets allow other laboratories to independently verify the molecular claims, but they say nothing about how a treatment based on this pathway might actually reach human astrocytes, whether through gene therapy, small molecules that mimic Sox9’s effect on MEGF10, or some other approach.
Where this fits in the larger Alzheimer’s landscape
The FDA-approved anti-amyloid antibodies lecanemab and donanemab attack plaques from the outside, recruiting the immune system’s microglia and peripheral antibodies to break down amyloid deposits. The Sox9-MEGF10 pathway represents a fundamentally different strategy: turning on the brain’s own astrocytes to consume plaques from within. Whether the two approaches could work together, or whether they might interfere with each other by competing for the same amyloid targets or altering inflammatory signaling, is an open question the current data do not address.
The history of Alzheimer’s research urges caution. Dozens of mechanisms that looked compelling in rodent models have failed to translate into meaningful human benefit once tested in clinical trials spanning years rather than months. What makes this study noteworthy is not a promise of a cure but the quality of the mechanistic evidence: multiple independent data types (in vivo plaque clearance, behavioral testing, transcriptomics, chromatin binding) converging on a single, testable pathway.
What the Sox9-MEGF10 axis still needs to prove in human-derived systems
For now, the Sox9-MEGF10 axis is best understood as a well-supported hypothesis about how to harness astrocytes for therapeutic benefit. It needs validation in human-derived systems, such as patient-derived brain organoids or induced pluripotent stem cell astrocyte cultures, and eventually in carefully designed clinical studies. The finding does not prove that this approach will halt or reverse Alzheimer’s in people. What it does is identify a specific molecular lever in a cell type that has been largely overlooked, and it provides the raw data for other research groups to pull on it.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
