Application of Plasma Neuron-Derived Exosome-Based Multimodal Biomarkers in the Early Diagnosis of Alzheimer's Disease

Objectives Alzheimer disease (AD) pathology begins 15–20 years before clinical symptoms, necessitating non - invasive biomarkers for early diagnosis. Neuron - derived exosomes (NDEs) carry intraneuronal cargo into peripheral blood, offering a window into early AD pathology. This study aimed to develop a multimodal biomarker model integrating NDE markers, plasma free biomarkers, and imaging indicators for preclinical AD diagnosis using retrospective data. Methods This retrospective study analyzed data from 400 participants: preclinical AD (A + T+N -, n = 120), AD dementia (A + T+N+, n = 80), MCI due to AD (MCI - AD, A + T+N+, n = 80), and cognitively normal controls (A - T - N -, n = 120). Plasma NDEs were isolated using L1CAM immunocapture. NDE levels of Aβ42, p - tau181, p - tau217, GAP43, SNAP25, miR − 132 , and miR − 212 were quantified. Plasma free p - tau181, p - tau217, Aβ42/40 ratio, and NfL were measured, along with Aβ - PET and hippocampal volumetry. LASSO regression selected key biomarkers; logistic regression and random forest constructed multimodal models. Results Preclinical AD showed significant NDE alterations: elevated Aβ42, p - tau181, p - tau217; decreased GAP43 and SNAP25; and downregulated miR − 132/212 . LASSO identified 7 key markers. The multimodal model achieved AUC (Area Under the Curve) 0.94 (95% CI 0.91–0.97) for preclinical AD detection, with 88.3% sensitivity and 86.7% specificity, outperforming single - modality models (p < 0.01). Validation yielded AUC 0.91 (0.87–0.95). NDE - p - tau217 correlated with Aβ - PET (r = 0.72) and predicted 3 - year cognitive decline (r = 0.58) in participants with available longitudinal data. Discussion Plasma NDEs capture intraneuronal AD pathology at the preclinical stage. A multimodal model integrating NDE markers, plasma free biomarkers, and imaging enables high - precision early AD diagnosis. Classification of Evidence: This study provides Class II evidence that a multimodal model incorporating NDE markers, plasma free biomarkers, and imaging accurately distinguishes preclinical AD from cognitively normal individuals.