The research team, led by investigators at the University of South Florida and the Rush Alzheimer’s Disease Center, took advantage of a uniquely rich resource: five prospective clinical-pathologic cohorts of aging that have followed thousands of older adults for decades, collecting detailed clinical data during life and brain tissue after death. From these cohorts, the team obtained 1,250 samples of dorsolateral prefrontal cortex, a brain region critically involved in the cognitive decline of Alzheimer’s disease. Rather than relying on the older HumanMethylation450 array, which covers a limited slice of the methylome and has dominated previous epigenome-wide association studies, they used the newer Infinium MethylationEPIC v2.0 platform, interrogating more than 935,000 CpG sites, the cytosine-guanine dinucleotides where methylation most commonly occurs.

What sets this study apart from its predecessors is the breadth of pathological measures it examined. Instead of focusing on a single trait such as Braak staging of neurofibrillary tangles, the researchers ran parallel analyses against three distinct outcomes: a neuropathologic diagnosis of Alzheimer’s disease under the National Institute on Aging–Alzheimer’s Association guidelines, the burden of amyloid-β plaques, and the density of paired helical filament tau tangles. Each of these measures was quantified through immunohistochemistry across eight brain regions, from the hippocampus to the calcarine cortex, providing an unusually comprehensive pathological picture for each donor. Of the 1,250 participants, 826, or 66.1 percent, met criteria for a neuropathologic diagnosis of Alzheimer’s disease, and the mean age at death was 89.5 years.

The epigenome-wide association scan, adjusted for age at death, sex, race, postmortem interval, neuronal cell proportions and technical artifacts, identified 107 differentially methylated positions associated with Alzheimer’s neuropathology at a false discovery rate below 0.05. Nine of these sites were linked to pathologic diagnosis, 65 to amyloid-β load and 46 to tau tangle density. Strikingly, twelve probes were associated with two or more pathologies, and a single probe, cg08550059, was associated with all three. This overlap hints at shared molecular machinery underlying the different lesions of Alzheimer’s disease, which are known to co-occur and likely arise from common processes such as inflammation, immune activation and cellular stress.

Replication is where many epigenome-wide studies falter, but this one held up remarkably well. In an independent set of 139 brain samples from the Netherlands Brain Bank, 38 of the 107 discovered sites remained significant after multiple testing correction, annotated to 67 unique genes, and every single one showed the same direction of effect as in the discovery cohort. The team then built a weighted methylation score from these 38 sites and tested it in two further datasets: brain tissue from the Banner Sun Health Research Institute, spanning inferior frontal and superior temporal gyri, and an independent subset of the Religious Orders Study and Memory and Aging Project, spanning prefrontal cortex and striatum. In both, the score was significantly elevated in individuals with pathologic Alzheimer’s disease compared with those without, demonstrating that the signature transcends cohort, brain region and array platform.

The most provocative result came from the mediation analysis. Among the 38 replicated sites, 23 were associated with carrying the APOE ε4 allele, and 20 of those partially mediated the effect of ε4 carrier status on Alzheimer’s diagnosis, each accounting for between 1.8 and 5.9 percent of the association. When the researchers combined the 23 APOE-linked sites into a single methylation score, that score mediated 8.6 percent of the total effect of APOE genotype on neuropathologic diagnosis, a statistically robust estimate with a confidence interval of 4.3 to 14.6 percent. Notably, none of these methylation sites sit on chromosome 19, where APOE itself resides. This suggests that APOE-driven biology, such as altered lipid metabolism, microglial activation or vascular processes, remodels methylation at distant, functionally related loci across the genome, a trans-acting epigenetic effect that could represent a previously underappreciated route from genetic risk to brain pathology.

To understand what these methylation changes actually do, the team integrated their epigenetic data with single-nucleus RNA sequencing from 437 of the same brains, allowing them to examine gene expression cell type by cell type. Of the 67 genes harboring replicated methylation sites, 39 were differentially expressed across major brain cell populations, including excitatory and inhibitory neurons, astrocytes, microglia, oligodendrocytes and their precursors. A clear pattern emerged: most of these genes were downregulated in neurons but upregulated in microglia and oligodendrocyte lineage cells in Alzheimer’s cases, mirroring the well-documented shift in the diseased brain toward glial activation and neuronal vulnerability. Thirteen genes, among them LGMN, LHPP and SPNS2, were associated with all three pathological traits at the expression level, and the gene RIN3, which overlaps a known Alzheimer’s genome-wide association locus, showed genetic coregulation of methylation and expression specifically in microglia.

Several of the implicated genes carry real biological weight. ANK1, CAMTA1, RHOB and LPP had been flagged in previous cortical methylation studies, lending independent credibility to the new findings. Four genes, RIN3, WDR81, TMEM121 and FOXF1, overlap established Alzheimer’s disease GWAS loci, and seven, including GFAP, RET and RIN3, have been nominated as potential drug targets in the Agora therapeutic database. Genes such as JADE2, which has been linked to modulation of the protective APOE ε2 allele and to microglial inflammation, and CHI3L2, an inflammatory marker related to the neurodegeneration-associated CHI3L1, reinforce the emerging picture that immune dysregulation and neuroinflammation are central, not peripheral, to Alzheimer’s pathology. Enrichment analysis further showed that the disease-associated methylation sites cluster in active regulatory elements, including accessible chromatin, transcription start sites, enhancers and activation-associated histone marks, indicating they are positioned where gene regulation actually happens.

The authors are candid about the limitations. Bulk tissue methylation can mask cell-type-specific signals, the replication sample was comparatively small, and the participants were predominantly of European ancestry and highly educated, which may limit generalizability. Most fundamentally, postmortem studies are cross-sectional, so it remains impossible to say whether the methylation differences drive the pathology or result from it. Still, the convergence of evidence, from discovery to replication to cross-region validation to single-cell transcriptomics, makes a strong case that DNA methylation is deeply entwined with Alzheimer’s disease biology. If future longitudinal and experimental work confirms a causal role, the roughly 8.6 percent of APOE risk channeled through methylation could represent a modifiable layer of genetic risk, opening the door to epigenetic biomarkers for early detection and, perhaps one day, to interventions that intercept Alzheimer’s disease before plaques and tangles take irreversible hold.

Subject of Research: Brain DNA methylation as a mediator between APOE genotype and Alzheimer's disease neuropathology

Article Title: Brain DNA methylation, APOE genotype and Alzheimer’s disease neuropathology in a large postmortem cohort

Article References: Brain DNA methylation, APOE genotype and Alzheimer’s disease neuropathology in a large postmortem cohort. (n.d.). https://doi.org/10.1038/s43587-026-01243-7

Image Credits: AI Generated

DOI: 10.1038/s43587-026-01243-7

Keywords: Alzheimer's disease, DNA methylation, epigenetics, APOE, neuropathology, postmortem brain, amyloid-beta, tau tangles, microglia, gene expression, EWAS, Nature Aging