Ultrasensitive Saliva-Based Detection of Early Alzheimer's Disease Biomarkers via Nanoparticle-Enhanced Evanescent...

Non-invasive biomarkers for early Alzheimer's disease (AD) screening remain a critical unmet need. Current cerebrospinal fluid (CSF) assays, while highly informative, are invasive and unsuitable for large-scale or repeat testing, whereas blood-based biomarkers, despite recent diagnostic advances, still face challenges related to assay standardization, analytical complexity, and sophisticated instrumentation requirements. Saliva represents an attractive alternative matrix due to its accessibility and minimal burden on patients; however, the extremely low abundance and instability of amyloid-β (Aβ) peptides have thus far limited the development of reliable salivary diagnostics. We developed and validated a nanoparticle-enhanced total internal reflection scattering (TIRS) microscopy platform for ultrasensitive, real-time quantification of salivary Aβ proteins. Metallic nanoparticles functionalized with anti-Aβ antibodies were used to amplify scattering signals and enable robust detection at sub-picogram concentrations. The assay was evaluated in two established AD mouse models, APP_sl and 5xFAD, in comparison with wild-type controls (n = 33 and n = 34, respectively). Since the validation of Aβ levels in saliva is not feasible with current state-of-the-art technologies, we validated the findings by measuring Aβ levels in the cortex and hippocampus via immunohistochemistry and ELISA. The TIRS assay demonstrated high analytical sensitivity and specificity for Aβ detection in saliva. In both APP_sl and 5xFAD models, salivary Aβ₄₂ concentrations were significantly elevated in transgenic mice and showed strong correlations with brain amyloid deposition. Logistic regression and support vector machine (SVM) classifiers were applied to quantify diagnostic performance and threshold-based discrimination based on salivary Aβ₄₂, identified as the most discriminative Aβ form in descriptive analyses. In APP_sl mice, logistic regression and SVM models achieved 92% classification accuracy with balanced sensitivity and specificity. These findings establish nanoparticle-enhanced TIRS as a rapid, accurate, and non-invasive tool for salivary Aβ quantification. By overcoming historical limitations of saliva-based biomarker detection, this technology provides a foundation for future translational development, including validation in human cohorts and optimization toward scalable and point-of-care diagnostic implementations.