DNA methylation signature of cognitive reserve moderates CSF tau pathology in prodromal Alzheimer’s disease

Background Cognitive reserve (CR) refers to differences in the adaptability of cognitive processes that modify the impact of Alzheimer’s disease (AD) pathology on cognitive performance. Currently there are no established blood-based biomarkers of CR in prodromal AD. In this study, we operationalize CR as memory reserve, defined as moderation (attenuation) of the CSF pTau181-memory association. DNA methylation (DNAm) integrates genetic and environmental influences and may capture biological processes that mitigate the impact of AD pathology on memory. We aimed to identify blood DNAm loci that moderate the association between cerebrospinal fluid (CSF) phosphorylated tau (pTau181) and memory in mild cognitive impairment (MCI). We also sought to determine if a DNAm-based signature of memory reserve predicts future memory decline. Methods We analyzed 92 amyloid positive MCI participants from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) with blood DNAm, CSF pTau181, and memory scores (PHC_MEM) collected at the same visit. We first regressed memory scores on covariates (age, sex, number of APOE4 alleles, estimated major immune cell type proportions) and used the residuals as covariate-adjusted memory scores. At each CpG, we then fitted linear models of memory on DNAm, pTau181, and their interaction. Inflations were corrected using the bacon method. We identified differentially methylated regions (DMRs), assessed pathway enrichment, and performed integrative analyses incorporating external resources including expression quantitative trait methylation (eQTM), methylation quantitative trait loci (mQTL) databases, AD genome-wide association study summary statistics, and blood–brain DNAm correlations. A methylation score was constructed and evaluated in linear mixed-effects models of longitudinal memory in 88 participants with follow-up information. Results After removing CpGs with low variability, we identified 6 CpGs with suggestive significance for DNAm×pTau181 interaction ( P- value < 1×10 ^-5 ) and 11 DMRs that passed multiple comparisons correction. These loci mapped to genes involved in synaptic function, vascular and blood-brain barrier integrity, amyloid clearance, immune and metabolic regulation. Almost all showed no strong marginal associations with pTau181 or memory, supporting a moderating rather than mediating role. Pathway analysis revealed enrichment of adipocytokine signaling and adipose metabolic pathways, and a number of CpGs associated with mQTLs overlapped with AD genetic risk loci. A higher baseline MRS attenuated the pTau-memory association and significantly associated with slower future memory decline, independent of age, sex, education, APOE ε4, and baseline pTau181. Conclusions Blood DNAm patterns that moderate the pTau-memory relationship capture biology underlying memory reserve involving synaptic, vascular, immune, and metabolic pathways, and can be summarized into an MRS that predicts longitudinal memory trajectories in MCI. These findings support blood DNAm as a promising, non-invasive biomarker of cognitive resilience to AD pathology.