MicroRNAs in Human Cerebrospinal Fluid as Biomarkers for Alzheimer's Disease

Background: Currently available biomarkers of Alzheimer's disease (AD) include cerebrospinal fluid (CSF) protein analysis and amyloid PET imaging, each of which has limitations. The discovery of extracellular microRNAs (miRNAs) in CSF raises the possibility that miRNA may serve as novel biomarkers of AD. Objective: Investigate miRNAs in CSF obtained from living donors as biomarkers for AD. Methods: We profiled miRNAs in CSF from 50 AD patients and 49 controls using TaqMan (R) arrays. Replicate studies performed on a subset of 32 of the original CSF samples verified 20 high confidence miRNAs. Stringent data analysis using a four-step statistical selection process including log-rank and receiver operating characteristic (ROC) tests, followed by random forest tests, identified 16 additional miRNAs that discriminate AD from controls. Multimarker modeling evaluated linear combinations of these miRNAs via best-subsets logistic regression, and computed area under the ROC (AUC) curve ascertained classification performance. The influence of ApoE genotype on miRNA biomarker performance was also evaluated. Results: We discovered 36 miRNAs that discriminate AD from control CSF. 20 of these retested in replicate studies verified differential expression between AD and controls. Stringent statistical analysis also identified these 20 miRNAs, and 16 additional miRNA candidates. Top-performing linear combinations of 3 and 4 miRNAs have AUC of 0.80-0.82. Addition of ApoE genotype to the model improved performance, i.e., AUC of 3 miRNA plus ApoE4 improves to 0.84. Conclusions: CSF miRNAs can discriminate AD from controls. Combining miRNAs improves sensitivity and specificity of biomarker performance, and adding ApoE genotype improves classification.