The largest genome-wide association study yet conducted on dementia lesions rather than dementia diagnoses has found that APOE, the best known common risk gene in Alzheimer's disease, is associated with almost every lesion type examined at autopsy except those in the brain's blood vessels. The study, published in PLOS Genetics on 29 June 2026 by Brenna Cholerton, Thomas Montine and colleagues with the Alzheimer's Disease Genetics Consortium, pooled harmonised neuropathology and genetic data on 12,509 brain donors after quality control. The dataset was 53.5 per cent female, with a mean age at death of 81.7 years; 37 per cent carried one APOE e4 allele and 9 per cent carried two. The authors report genome-wide significant APOE associations with the hallmark Alzheimer's lesions and with other lesions of Alzheimer's disease and related dementias, but not with cerebrovascular arteriolosclerosis, cerebrovascular atherosclerosis or vascular brain injury.
Most genetic studies of dementia use the clinical label, which in later life usually reflects several coexisting brain diseases at once. This study instead used what pathologists scored under the microscope: amyloid plaques, neurofibrillary tangles, cerebral amyloid angiopathy, Lewy body disease, TDP-43 inclusions, hippocampal sclerosis, and separately the vascular findings of arteriolosclerosis, atherosclerosis, infarcts and lacunes. The abstract reports 12 genome-wide significant loci across 10 lesion phenotypes, five of them already implicated in clinical dementia genetics (variants at BIN1, PICALM/EED, TMEM106B, GRN and SNCA/SNCA-AS1) and seven not previously reported at genome-wide significance (EPHA5, PSMG1, LINC00276, VAPA, LINC00290, DOCK4 and SLAIN2/SLC10A4). Estimated SNP heritability was high for amyloid plaque severity (up to 78 per cent), intermediate for TDP-43 inclusions (41 per cent), and low for tangle severity and the cerebrovascular measures (11 to 26 per cent).
The vessel findings are not empty of genetics; they are simply not APOE. The authors report separate genome-wide significant signals for atherosclerosis on chromosome 2 near LINC00276 and near VAPA, and a signal near DOCK4 for the presence of infarcts or lacunes. Read together, the pattern is consistent with the vascular contribution to mixed dementia having its own genetic architecture rather than being a downstream consequence of amyloid, though this is an association study in donated brains and cannot establish cause or sequence.
The authors set out the limits themselves. The sample, while the largest of its kind, is small beside clinical Alzheimer's GWAS that now exceed a million people, so a missing association may reflect low statistical power rather than a real absence. Effect sizes varied across contributing sites and genotyping arrays. And the cohort lacks racial and ethnic diversity, which the authors say means variant frequencies and relevance may differ in non-European populations, limiting how far the results generalise.