Astrocyte-related proteins mediate the association of YWHAG with Alzheimer's pathology and enhance its diagnostic value.

Recent advances have identified YWHAG as a promising synaptic biomarker, with evidence showing that the YWHAG:NPTX2 ratio strongly predicts cognitive decline and Alzheimer's disease (AD) progression independent of amyloid and tau pathology. However, the links between YWHAG and astrocytic processes-key regulators of amyloid clearance, tau phosphorylation, and neuroinflammation-remain poorly understood. A total of 530 participants were included. Levels of YWHAG and a panel of biologically relevant astrocyte-derived proteins were measured using a proximity extension assay and validated immunoassay platforms. Associations with AD biomarkers and cognition were examined using multivariable regression, longitudinal mixed-effects models, and mediation analyses. Path analysis was performed to explore the potential pathways from YWHAG through astrocytic proteins to AD pathology and cognition. We evaluated whether combining YWHAG with astrocyte-related proteins improves its predictive performance, by comparing the area under the curve (AUC) of the combined model with that of YWHAG alone. YWHAG was positively associated with glial fibrillary acidic protein (GFAP, β = 0.558, p < 0.001), vimentin (β = 0.329, p < 0.001), aquaporin-4 (AQP4, β = 0.097, p = 0.044), thrombospondin (THBS) -1 (β = 0.470, p < 0.001), and THBS2 (β = 0.285, p < 0.001), while showing negative associations with gap junction alpha-1 protein (GJα1, β = -0.161, p < 0.001) and serpin family A member 3 (SERPINA3, β = -0.350, p < 0.001). Mediation analysis indicated that certain astrocyte-related proteins may be involved in the association between YWHAG and AD pathology. Additionally, path analysis suggested a potential pathway involving YWHAG, GJα1, Aβ42, and cognitive function. The combination of YWHAG with SERPINA3 and THBS1 achieved an AUC of 0.981, outperforming YWHAG alone (AUC = 0.885). YWHAG is associated with astrocyte-related proteins, and combining them enhances its predictive accuracy for AD, highlighting its potential utility in early clinical screening.