Abstract
BACKGROUND: Neuropsychiatric systemic lupus erythematosus (NPSLE), with various morbidities and multiple manifestations in the central nervous system, remains a limited standard for diagnosis. Our study was to discover novel biomarkers for improving the diagnostic efficiency for NPSLE. METHODS: We performed a quantitative planar protein antibody microarray to screen 1000 proteins in cerebrospinal fluid from controls, systemic lupus erythematosus (SLE, non-NPSLE) patients, and NPSLE patients. Differentially expressed proteins (DEPs) as candidate biomarkers were developed into a custom multiplexed protein antibody array for further validation in an independent larger cohort. Subsequently, we used least absolute shrinkage and selection operator regression (LASSO) analysis and multivariable logistic regression analysis for optimizing feature selection and constructing a diagnostic model. A receiver operating characteristic curve (ROC) was generated to assess the effectiveness of the models. RESULTS: The expression of 29 proteins in CSF was significantly altered in the comparison of the three groups. We selected 17 proteins as candidate biomarkers in accordance with protein interaction analysis. In the larger cohort, we identified 5 DEPs as biomarkers for NPSLE, including TCN2, CST6, KLK5, L-selectin, and Trappin-2. The diagnostic model included 3 hub proteins (CST6, TCN2, KLK5) and was best at discriminating NPSLE from SLE patients. These CSF biomarkers were also highly associated with disease activity. In addition, there were 6 molecules with remarkable changes in NPSLE CSF and hippocampus, which indicated the consistency of the environment in the brain and the promising molecular targets in the pathogenesis of NPSLE. CONCLUSIONS: The dual-chips screening strategy demonstrated KLK5, L-selectin, Trappin-2, TCN2, and CST6 as CSF biomarkers for diagnosing NPSLE.