Here's a question that sounds simple but has been genuinely hard to answer: when microglia in an old brain look inflamed and dysfunctional, is that because the cells themselves are worn out, or because the neighborhood they live in has gone bad?
Microglia are the brain's resident immune cells. In aged brains, they crank up interferon signaling, start expressing disease-associated genes, and generally look like they've had a rough few decades. The assumption has been that this is a cell-intrinsic program, something baked into the cells over years of accumulated damage.
This paper runs the experiment that tests that assumption directly. Young cells placed into old brains aged rapidly. Old cells placed into young brains got younger. The environment is running the show.
To pull this off, the team needed a way to swap out brain immune cells with precise genetic control, at scale, in both young and aged animals. That's not a trivial ask. Viral transduction of microglia is notoriously inefficient and triggers its own interferon response, which would contaminate exactly the signal you're trying to measure. Breeding transgenic lines takes years.
The solution: ex vivo expanded hematopoietic stem cells (eHSCs) from Rosa26-Cas9-EGFP mice, grown up to 100 million cells in culture, CRISPR-editable before transplant. Recipients were conditioned with busulfan chemotherapy, then treated with the CSF1R inhibitor PLX5662 to clear out resident microglia. Donor cells moved in and took over.
Peripheral myeloid reconstitution hit 75-100% chimerism in both young (3 mo) and aged (18 mo) recipients. The orthogonal fresh bone marrow model pushed brain chimerism above 90% in all animals. The EGFP tag let them sort donor cells cleanly from any residual host cells for sequencing. From a flow perspective, that's a clean separation โ the kind of chimerism you'd want before trusting any downstream transcriptomics.
The resulting cells, called Reconstituted Cells (ReCs), were profiled by scRNA-seq, CITE-seq (105 surface proteins), spatial transcriptomics (CosMx), bulk RNA-seq, and confocal morphology. This is a genuinely comprehensive readout stack.
method
Before testing aging, the team asked a more basic question: can peripherally-derived cells even become microglia-like? The answer is yes, with caveats. ReCs don't share the developmental origin of true microglia โ they lack the transcription factors Sall1 and Sall3 โ but they adopt a defined set of microglia-like molecular states, including homeostatic, interferon-response, and disease-associated clusters.
More striking: 285 region-specific genes that differ between cerebellar and cortical native microglia also differ between cerebellar and cortical ReCs. The cerebellum makes cells look cerebellar. The cortex makes them look cortical. This holds at the morphology level too โ cerebellar ReCs are less ramified, with reduced surface area and branch length, mirroring the rod-like shape of native cerebellar microglia.
Peripherally-derived myeloid cells adopt region-specific transcriptional, morphological, and spatial tiling profiles characteristic of resident microglia. The brain environment, not developmental origin, writes the identity.
result
Now for the central experiment. Young ReCs transplanted into aged brains were compared to young ReCs in young brains. In the cerebellum, 403 genes showed concordant age-related changes in both native aged microglia and young ReCs placed in old brains. These weren't random genes: the overlap was enriched for interferon signaling, ribosomal protein induction, and antigen presentation machinery โ the canonical hallmarks of microglial aging.
The effect was region-specific. The cerebellum drove stronger aging signatures than the cortex, consistent with what's known about accelerated aging in that region. This pattern was captured in a gene signature called the Cerebellar Accelerated Aging Signature (CAAS), which was more strongly induced in cerebellar ReCs in aged brains than in cortical ReCs, and replicated across independent published datasets including human microglia data.
The reciprocal experiment closed the loop. Old donor cells placed into young brains adopted youthful profiles. In total, 759 genes shifted expression in response to the recipient's age, independent of the donor's age, sex, or genotype. The environment is not just influencing these cells. It is rewriting them.
Young myeloid cells acquire aging phenotypes within weeks of entering an aged brain, particularly in the cerebellum. Old cells in young brains reverse these changes. Cell-intrinsic age is not the primary driver.
result
Knowing the environment drives aging is useful. Knowing the molecular switch is actionable. STAT1 was a candidate because it sits at the center of interferon signaling and was consistently upregulated in aged cerebellar microglia and ReCs.
The team knocked out Stat1 in eHSCs before transplantation using CRISPR, achieving a 98% KO score in blood cells post-reconstitution. When these Stat1-deficient young cells were placed into aged brains, they were strongly protected. Interferon response genes stayed low. Antigen presentation genes stayed low. DAM-like genes including Gpnmb, Spp1, and Itgax were dampened. The homeostatic marker P2ry12 stayed up. The CAAS score in aged cerebellar Stat1-/- ReCs was significantly lower than in controls.
Critically, Stat1 knockout did not prevent ReCs from reading regional identity cues at young baseline. The cells still knew they were in the cerebellum versus the cortex. STAT1 is specifically required for the aged environment response, not for general environmental sensing.
result
T cells have been the leading suspect for driving interferon signaling in aged microglia. This paper challenges that. Computational cell-cell communication modeling (CellChat) predicted that IFN-gamma signaling from NK cells, not T cells, was the dominant age-associated input to microglia. Removing NK cells from the analysis in silico eliminated the IFN-II signaling axis entirely.
The genetic experiments confirmed it. Rag2-/-yc-/- mice, which lack T, B, and NK cells, showed no age-related increase in interferon response genes in cerebellar tissue, while DAM-like genes like Itgax and Lpl still went up. Rag2-/- mice, which lack T and B cells but retain NK cells, showed a cerebellar interferon response nearly identical to wild-type aged mice. T and B cells are not the drivers here.
Antibody-mediated depletion of NK cells (anti-NK1.1) in aged wild-type mice, starting at 16 months, prevented the age-related increase in interferon signaling in cerebellar microglia by 17-18 months. DAM-like gene expression was unaffected, suggesting two parallel aging programs in the cerebellum: one driven by NK cell interferon signaling, one driven by other local factors like myelin debris.
result
result
The model that emerges is clean: NK cells in the aged brain secrete IFN-gamma, which activates STAT1 in microglia, which drives the interferon response and antigen presentation signatures of microglial aging. A separate, parallel program drives DAM-like gene expression, likely through white matter-derived signals like myelin debris converging on STAT1 through a different route.
The practical implication is pointed. JAK-STAT inhibition, already used clinically for inflammatory conditions, could potentially reset this feed-forward loop. NK cell depletion reversed the interferon signature in weeks. That's a fast effect for an aging phenotype.
A few honest caveats: the busulfan conditioning may itself contribute to elevated DAM-like gene expression in ReCs, which the authors acknowledge. The CAAS was defined conservatively, requiring concordance across two regions, so some true aging genes were excluded. The study used mixed-sex models across different experimental arms, and sex-specific effects can't be fully ruled out. NK cell numbers in the choroid plexus were too low to directly track their age-related changes.
None of that undermines the central finding. The aged brain is an aging machine, and now there's a scalable platform to figure out exactly how it works.
model