Integrated Genetic and Multi-Omic Analyses Reveal Oral Microbial Associations with Brain Structure and Alzheimer’s Disease
TL;DR
This cross-scale atlas prioritizes microbial and anatomical targets and provides a translational foundation for mechanistic validation and non-invasive neurodegenerative risk markers.
Abstract
Oral microbial dysbiosis is linked to systemic inflammatory, vascular, and metabolic perturbations, but species-level relationships with human brain structure remain poorly defined. Using two-sample Mendelian randomization, we mapped 438 of 439 oral species with genome-wide association study instruments to 1325 adult brain imaging-derived phenotypes. In total, 87 associations involving 51 taxa and 79 traits passed Benjamini–Hochberg correction across taxa within each imaging-derived phenotype. The childhood and metabolomic screens identified 196 nominal taxon–volume associations and 1118 candidate metabolite-linked paths, respectively, as exploratory findings. Spatial and single-nucleus transcriptomics localized candidate regional and cell-associated variation in brain-structure gene programs. In total, 35 taxa showed nominal genetic associations with Alzheimer’s disease. Integration with full-length oral 16S ribosomal ribonucleic acid gene sequencing from a 51-participant case–control cohort prioritized Selenomonas infelix, Capnocytophaga granulosa, and Dialister pneumosintes. In a secondary exploratory analysis, the three-taxon salivary model incorporating educational attainment yielded a leave-one-out cross-validated area under the receiver operating characteristic curve of 0.843. This post-selection estimate is subject to optimism, and the model remains a proof of concept requiring independent external validation. This cross-scale atlas prioritizes microbial and anatomical targets and provides a translational foundation for mechanistic validation and non-invasive neurodegenerative risk markers.