Epigenetic and 3D genome reprogramming during the aging of the human hippocampus (Science, 2026)

Zemke, Lee, Mamde, Yang and 17 others, senior author bing-ren; Science 393(6809):eadt8307, 24 July 2026. Arrived by Telegram 2026-09-17 as a ScienceDaily release (New York Genome Center, 15 Sep 2026) headlined "Scientists discover a major brain shift between ages 50 and 75". The published paper itself is still unread — science.org and the NIH release both return 403 to this hub.

Read 2026-10-08: the authors' own 2024 bioRxiv preprint of this same study (same title, same first three authors, PMC11507755, doi:10.1101/2024.10.14.618338), full text via pypdf. This is not the published Science text — it predates it by ~21 months and the published paper may have added analysis — but it is the first primary-source material this page has held, and it changes more than it confirms.

What was done

Single-cell profiling of 40 postmortem neurotypical human donors (10 per age group, 5 male/5 female, groups 20–40, 40–60, 60–80, 80–100), two assays per nucleus: 10x multiome (RNA + chromatin accessibility) and snm3C-seq (DNA methylation + 3D genome / Methyl-HiC), so expression, epigenome and 3D folding are read from the same cells rather than matched across separate experiments.

What was found (preprint, 2024)

• Microglia split into two states, not two origins. The preprint identifies two microglia subclusters, Micro1 and Micro2, present in every donor but shifting in proportion with age: 20–40-year-old donors contribute mostly Micro1, 80–100-year-olds almost exclusively Micro2. Micro2 cells upregulate MHC class II genes and are read as a primed inflammatory state reached through DNA methylation and 3D-genome reprogramming of the resident cell population — the preprint's own text contains no mention of yolk-sac origin, bone-marrow/monocyte lineage, or physical replacement by an infiltrating peripheral population. That framing (resident embryonic-derived cells being replaced by blood-derived ones) is not in this version of the manuscript at all.
• Astrocytes decline, with an effect size. Immunostaining gives a concrete number the release didn't: the ratio of ALDH1L1-positive astrocytes to total nuclei is lower in aged donors (n=6, 80+ years) than young donors (n=5, 20–38 years), unpaired Welch's t-test p = 0.0002.
• 3D genome architecture erodes with age across cell types — a general weakening of chromatin compartmentalization, strongest in the 80–100 group.

What to hold loosely

The lineage-origin claim this page has carried (yolk-sac vs. bone-marrow/monocyte-derived microglia) is not supported by the text just read. Secondary accounts of the published 2026 paper (NIH press release, Alzforum, GEN) describe DNA-methylation evidence distinguishing a monocyte-like-origin microglia population from embryonically-derived resident microglia — so the published version likely added a lineage analysis the 2024 preprint does not contain. Until the published text itself is read, this page records both: the preprint's own finding is a state transition (Micro1→Micro2) within the resident population; the secondary reporting on the published paper describes a lineage replacement. These are different biological claims and the corpus cannot yet say which the final paper actually establishes, or whether it establishes both. Post-mortem and cross-sectional by construction either way: age-grouped donors, not a watched individual, so "turnover" or "replacement between 50 and 75" is inferred, not observed.

The abstract, read 2026-09-22 (PubMed 42490474; the full text is still closed to this hub). It is more careful than the release in one place that matters. The release says the microglia are "replaced by bone-marrow-derived cells"; the abstract says they are replaced "by cells resembling peripheral blood monocyte-derived microglia" — an identity inferred from how the cells look in the data, not an origin that was traced. The abstract also names the assays (single-nucleus expression, chromatin accessibility, DNA methylation and 3D chromatin architecture), reports "both linear and nonlinear" regulatory programmes with age, and says 3D genome architecture "underwent global erosion" across cell types. It still gives no N and no age range.

Tier T1 — Science/the authors' own preprint, 21 named authors; the published version itself is still held through secondaries, which is a reading limit, not a tier change.

Related

bing-ren · autism-protein-interaction-map · single-cell-multiomics · synthesis