# Human microglial transitions at the Aβ-tau inflection point associate with divergent pathways to dementia and resilience.

- **DOI:** [10.1038/s41591-026-04393-8](https://doi.org/10.1038/s41591-026-04393-8)
- **Rating:** Limited (2.8 out of 5)
- **Verdict:** A substantive spatial atlas identifies microglial states at the amyloid–tau interface, but donor-level inference leaves its mechanistic conclusions materially uncertain.
- **Framework version:** v5.9
- **Pipeline version:** v0.4
- **Evaluation date:** 2026-10-08
- **Evaluation page:** https://www.nabu.science/paper/10.1038%2Fs41591-026-04393-8

## Key claims

1. We identified six distinct tissue domains representing a spatial pathological continuum of AD, with a key inflection point marked by a shift from Aβ-associated inflammatory changes to tau-associated cellular programs.
   - Support: Supported
   - Type: Primary finding
   - Basis: spatial clustering and pathology correlations support the pattern, but donor-nested inference and cross-sectional ordering limit certainty
   - Location: Abstract; Results, Fig. 2
2. This transition was accompanied by a change in microglial states, from early inflammatory to late antigen-presenting phenotypes, termed early and late plaque-induced gene (PIG) programs.
   - Support: Supported
   - Type: Secondary
   - Basis: Visium, snRNA-seq, Xenium and external biopsy data converge, though the programs were derived from cross-sectional associations
   - Location: Abstract; Results, Figs. 3–4
3. Resilient individuals showed distinct pathological patterns: octogenarians without dementia lacked late PIGs, whereas centenarians showed late PIG activation that was uncoupled from tau accumulation.
   - Support: Tentative
   - Type: Secondary
   - Basis: cross-cohort protocol differences, heterogeneous centenarian cognition and limited donor-level inference weaken the resilience interpretation
   - Location: Abstract; Results, Fig. 5
4. This uncoupling demonstrates that late PIG activation is not intrinsically pathogenic but that its downstream consequences depend on the surrounding cellular and pathological context.
   - Support: Tentative
   - Type: Secondary
   - Basis: the observational cross-cohort design cannot establish absence of pathogenicity or context-dependent causal effects
   - Location: Discussion, paragraph 3

## Strengths

- **Spatial pathology and transcriptomics directly linked** Aβ and pTau were quantified on the same sections used for Visium, allowing local pathology to be related to transcriptional states rather than relying only on donor-level staging.
  - Dimension: methodological rigour
  - Location: Results, Fig. 1e; Methods, Image Analysis
- **Convergent multimodal validation of PIG programs** The early and late PIG patterns were examined through snRNA-seq state mapping, an external cortical-biopsy dataset, Xenium and immunofluorescence, providing multiple forms of corroboration.
  - Dimension: contribution
  - Location: Results, Figs. 3l–n and 4
- **Rare centenarian cohort extends the evidence base** Spatial profiling of 20 centenarian brains supplies an uncommon comparison in which late-PIG expression occurs with little detected pTau, creating a focused hypothesis about microglial–tau uncoupling.
  - Dimension: contribution
  - Location: Results, Fig. 5

## Limitations

- **Donor dependence is not modeled** Core pathology–expression and tissue-domain tests use spots, cells or hexbins as analysis units, but the Methods do not describe donor-clustered, mixed-effects or donor-pseudobulk inference. Reported significance may therefore be materially overstated.
  - Dimension: methodological rigour
  - Location: Methods, Differential Gene Expression Analysis; Statistical Analysis, Enrichment Within TDs
- **Centenarian comparison carries technical confounding** OCT and CEN Visium libraries used different permeabilization times, protocol revisions, PCR annealing temperatures and AT8 procedures. These differences bear directly on the conclusion that late PIGs are induced by Aβ alone in centenarians.
  - Dimension: methodological rigour
  - Location: Methods, Immunofluorescence, Image Acquisition and Visium Spatial Gene Expression
- **Causal conclusions exceed observational design** The Discussion says late PIG activation is not intrinsically pathogenic and that persistent Aβ enables tau pathology, although the Limitations section states that causality cannot be inferred.
  - Dimension: positioning
  - Location: Discussion, paragraphs 3 and 6; Limitations

## Reliability checks

Caveats (4 of 4 checks).

- **Results check:** Noted
  - This transition was accompanied by a change in microglial states, from early inflammatory to late antigen-presenting phenotypes, termed early and late plaque-induced gene (PIG) programs: "Weighted gene co-expression network analysis (WGCNA)50 on 43,169 gray matter Visium spots identified 16 modules" (Results, Neuroinflammatory signatures defined by co-expression modules) vs "Signed co-expression networks were constructed using WGCNA (R, v.1.72.5)77 on Visium spatial transcriptomic data from gray matter regions of the OCT cohort (17,271 genes, 41,943 spots)" (Methods, Gene co-expression analysis of Visium spatial gene expression). The Results and Methods sample sizes for OCT WGCNA differ by 43,169 − 41,943 = 1,226 spots, with Fig. 3a also reporting 41,943, but the supplied text does not establish which observations actually entered the network. → Effect on result unclear
  - Minor reporting notes (1):
    - Descriptive: "The Visium transcriptomic analysis was performed on the CEN cohort and the OCT cohort with the same workflow, generating 60,129 and 61,739 annotated Visium spots, respectively." (Results, Spatial and single-nucleus transcriptomics across octogenarian and centenarian brains to dissect resilience) vs "The curated dataset comprised 60,129 spots from 24 individuals (mean 1,831 genes, IQR: 1,168–2,352). The same approach yielded 61,739 spots from 20 centenarian individuals." (Methods, Visium spatial transcriptomics processing). The cohort order in the Results sentence assigns 60,129 spots to CEN and 61,739 to OCT, whereas Methods and the cohort-specific Results sections assign those totals to OCT and CEN, respectively. → Conclusion holds
- **Reporting and transparency:** Clean. No reporting or transparency concern is established by the supplied concerns and article text.
  - Ethics approval: declared
- **Reference integrity:** Clean. 20 of 20 checkable references verified
  - 20 references in manuscript
  - 1 reference confirmed by manual review
  - References not verified automatically are resolved manually within 24 hours.

## Post-publication record

- **Status:** Clean
- **Summary:** No retraction or expression of concern found in Crossref (includes Retraction Watch).
- **Last checked:** 2026-10-17
