The Cerebrospinal Fluid Proteome of Alzheimer’s Disease Patients: Differentiation and Prediction of Responders and Non-Responders to Lymphatic-Venous Anastomosis

Herein, we describe the identification of protein biomarkers in the cerebrospinal fluid (CSF) of Alzheimer's disease (AD) patients that enable the differentiation of Responders from Non-Responders to lymphatic-venous anastomosis (LVA). Discovery of these potential biomarkers was achieved by deep proteomic analysis on the pre-LVA cerebrospinal fluid in a cohort of 90 patients. A combination of two CSF processing modalities and two computational platforms allowed us to secure the largest CSF proteome ever reported (6711 proteins identified with 4506 proteins quantified at a false-discovery rate of ≤1%). Importantly, we discovered 16 CSF proteins that are expressed differentially between patients who showed documented improvements after LVA (the Responders) and those who showed no improvements (the Non-Responders). Pairing of two of the best-performing potential biomarkers─NPTX2 and IGSF10─gave an accuracy of 0.790 (AUC = 0.854) in correctly identifying Responders and Non-Responders, which improved to 0.827 (AUC = 0.880) after we added a second pair of classical AD biomarkers─p-tau181 and Aβ42─to the test. We found that the classical ELISA biomarker ratio p-tau181/Aβ42 has, by itself, a marginally acceptable accuracy of 0.644 (AUC = 0.662). Our reported observation (in 2025) of a correlation between elevated p-tau181/Aβ42 and responding to LVA is in accordance with the interpretation that Responders have a higher abundance of NPTX2, which indicates better synaptic health. Apparently, the Responders are more resilient, despite an indication of more advanced pathology. In a simulation of the performance in future testing, we performed a 5-fold cross-validation that gave an AUC of 0.859 ± 0.074 and an accuracy of 0.740 ± 0.048 for the small testing group.