Abstract
OBJECTIVE: Alzheimer's disease (AD) is strongly associated with aging, yet the interactions remain unclear. This study modeled replicative senescence in patient-derived fibroblasts to compare gene expression between AD dementia and controls across senescence stages and to evaluate whether stage-specific alterations reflect disease characteristics with diagnostic implications. METHODS: Dermal fibroblasts from 13 AD dementia patients and 13 healthy controls were repeatedly passaged to induce replicative senescence and classified into young (passage 7), mid-old (passage 18), and old stages (passage 25-28). Transcriptomic profiling was performed by RNA sequencing, followed by stepwise gene extraction, machine learning-based classification, and correlation analyses with AD biomarkers. RESULTS: Fibroblasts were successfully driven into replicative senescence, validated by SA-β-gal staining, increased expression of CDKN1A and CDKN2A, and transcriptomic age acceleration. From transcriptome data, 605 senescence-associated genes were identified, enriched in extracellular matrix remodeling, chromatin organization, and immune-related pathways. Machine learning classifiers trained on these genes achieved the highest accuracy at the mid-old stage above 0.9, markedly outperforming the young and old stages. In addition, among the most consistently selected mid-old genes, H2AC18, H1-2, and LTBP1 showed significant correlations with cortical amyloid burden and plasma pTau217, linking cellular transcriptomic changes to established AD biomarkers. CONCLUSION: In summary, replicative senescence models of patient-derived fibroblasts revealed that transcriptomic differences between AD dementia and controls peak at the mid-old stage. This transitional window represents the most informative point for capturing disease-related alterations with strong biomarker relevance.