What the design carries
Belk and colleagues label every clone of cells by the somatic mutations it has accumulated, then ask whether cells carrying a marrow clone's label turn up in brain tissue. Applied to 20 aged individuals, the approach found evidence of an influx in every one of them, and single-cell lineage tracing on mitochondrial DNA variants showed that the infiltrating cells resemble microglia and can make up a large fraction of the microglial pool. The comparison that gives this its weight comes from mice, where microglia are seeded during embryogenesis and maintained with minimal input from adult blood production.[1]
Shared mutations are a strong marker of shared ancestry, and the origin half of the claim is the half they settle. Identity is a separate question. The paper's own wording is that the infiltrating cells are similar to microglia, and a competing reading of the same data is that a share of them are perivascular or transiently resident myeloid cells carrying the marrow label without joining the parenchymal microglial pool. Tissue taken after death shows where cells ended up, and it cannot show how long they stayed.[1]
Where the release moves past the sample
The Stanford release says immune cells from the blood begin entering the brain as early as middle age, and calls the process a distinctly human feature of aging, absent in mice and non-human primates. What the paper reports is the method applied to 20 aged individuals. A sample drawn from the old end of the age range can establish that an influx has happened by then; it fixes no start point, and the start point is the sentence most likely to be repeated.[1]
The Alzheimer's thread travels the same way. What the paper reports is a protective association between most types of clonal hematopoiesis and Alzheimer's disease in human cohort data. The release moves from that association to engineering peripheral immune cells that would clear amyloid and tau aggregates, given to people preventively. Cohort associations of this shape stay open to confounding by survival and by who reaches a diagnosis at all, and the paper keeps the word association where it belongs.[1]
What a follow-up would have to measure
The timing claim is testable with the method already in hand. If the same somatic-mutation tracing is applied to donors spanning middle age rather than only the aged, and the marrow-derived fraction of the microglial pool is reported by age band, then as early as middle age becomes a measured quantity with an error bar around it. Published fraction-by-age-band figures before the end of 2027 would settle it; without them the timing stays an extrapolation from 20 aged donors.[1]
Credit where it is due: an influx found in all 20 individuals examined, labelled twice over — by accumulated somatic mutations and independently by mitochondrial DNA variants — is hard to dismiss as an artefact, and it breaks cleanly from the mouse picture that has organised this field for years. The questions land on the two extensions instead: a start date the sample cannot date, and a protective link that stays associational in the paper's own words. Both are repairable with work well within reach of the same group.[1]