Genome‐wide mapping implicates 5‐hydroxymethylcytosines in diabetes and Alzheimer’s disease
Zhou Zhang, Alana Beadell, Ana W. Capuano, David A. A Bennett, Chuan He, Wei Zhang, Zoe Arvanitakis- Psychiatry and Mental health
- Cellular and Molecular Neuroscience
- Geriatrics and Gerontology
- Neurology (clinical)
- Developmental Neuroscience
- Health Policy
- Epidemiology
Abstract
Background
Diabetes mellitus (DM) is a recognized risk factor for dementia including Alzheimer’s disease (AD). Because DM is a potentially modifiable risk factor, a greater understanding of the mechanisms linking DM to the clinical expression of AD may provide insights into the much‐needed dementia therapeutics or intervention approaches. Epigenetic analysis offers a powerful approach, and previous studies including our team’s exploration in circulating cell‐free DNA (cfDNA) suggested that the under‐investigated 5‐hydroxymethylcytosines (5hmC) is now emerging as a promising measure to investigate in DM‐related conditions.
Method
We used samples and data from the Rush Memory and Aging Project. Using the 5hmC‐Seal technique, we performed genome‐wide profiling of 5hmC in cfDNA samples (n = 74) from antemortem blood samples and brain tissue (n = 75) genomic DNA from postmortem prefrontal cortex tissue from 80 deceased individuals across four groups: AD (neuropathologically defined), DM (clinically defined Type 2 DM), DM with AD (AD+DM), and non‐AD/non‐DM controls. Differential analysis and machine learning approaches were used to investigate 5hmC signatures associated with different conditions and summarize epigenetic scores to evaluate diagnostic potential for AD in cfDNA. Pathway enrichment analysis was performed on differential 5hmC signatures to explore biological mechanisms underlying DM‐associated AD.
Result
We identified distinct 5hmC signatures and biological pathways, including Wnt signaling and inflammatory mediator regulation of TRP channels, as well as amino sugar and nucleotide sugar metabolism, that are associated with AD or AD+DM compared to non‐AD/non‐DM controls or DM alone. We further demonstrated the potential diagnostic value of 5hmC profiling in circulating cfDNA. Specifically, a 4‐gene model showed capacity for distinguishing AD+DM from DM (AUC [area under the curve] = 82.7%; 95% CI [confidence interval], 67.6‐97.9%), while an 11‐gene model distinguished AD from non‐AD/non‐DM controls (AUC = 90.0%; 95% CI, 80.4‐99.6%).
Conclusion
Genome‐wide 5hmC profiling uncovered epigenetic features and biological pathways related to nucleotide sugar metabolism that link DM to DM‐associated AD. Our findings also demonstrated the potential of utilizing 5hmC in circulating cfDNA as diagnostic biomarkers or disease monitoring tools, with the ultimate goal of preventing or ameliorating DM‐associated AD dementia and improving clinical outcomes. This study was supported partially by the NIH (RF1AG074549, R01AG17917).