Global Metabolomic Profiling of PSP and Healthy Controls Yields CSF and Serum Biomarkers
LeWitt PA, VandeVrede L, Boxer A, Li J, Zhang J, and Hong L. Global Metabolomic Profiling of PSP and Healthy Controls Yields CSF and Serum Biomarkers. Mov Disord 2022; 37:S15-S16.
Objective: Global metabolomic profiling of biochemicals for diagnostic differentiation of CSF and serum from 34 healthy controls (CN) and 44 patients with progressive supranuclear palsy (PSP).
Background: Though PSP lacks evidence for specific metabolic disturbances, Mori et al (PLoS ONE 2019; 14(9):e0223113) reported alteration of small-molecular weight compounds in serum specimens, yielding a diagnostic biochemical profile.
Methods: Ultrahigh-performance liquid chromatography linked to tandem-mass spectrometry measured relative concentrations of biochemicals <1.5 kDalton molecular weight. Authentic standards facilitated all chemical identifications. Analytes underwent univariate analysis with false discovery rate-adjusted p-value (≤0.05) determinations. For characterizing selected biochemical panels discriminating PSP from CN, the multivariate analysis used several biostatistical methods.
Results: Metabolomic profiling determined 441 CSF and 1,042 serum compounds. CSF univariate analysis (moderated t-test) selected 28 compounds (27 PSP-specific); serum provided 3 PSP-specific compounds. CSF multinomial regression analysis selected a 12- compound panel (diagnostic accuracy: 87.8%). Sensitivity and specificity for CN were 94.1% and 92.2%; for PSP, 90.9% and 88.9%. The 5-fold cross-validation ROC- curve-AUC for PSP versus CN was 83.6%. For serum, multinomial regression analysis selected 6 compounds with a diagnostic accuracy of 77.0%. Sensitivity and specificity for CN were 88.6% and 88.2%; for PSP, 82.5% and 78.7%. The 5-fold cross-validation ROC-AUC for PSP versus CN was 72.4%.
Conclusion: In CSF (and to a lesser extent in serum), global metabolomic profiling helps with diagnostic differentiation between PSP and CN. The compounds selected in uni and multivariate biochemical panels mostly lacked shared biochemical properties and didn’t map to canonical metabolic pathways. We couldn’t replicate the PSP-specific serum biochemical profile or any components as reported by Mori (2019). Among the 28 chosen CSF constituents, 4 with the highest LASSO regression coefficients were unrelated compounds (1-methylhistidine, creatine N6-succinyladenosine, and nicotinate ribonucleoside). Metabolomic profiling can guide searching for diagnostic and progression biomarkers of PSP and may inform the analysis of pathogenesis.
Funding: Sastry Family Foundation Endowment, Wayne State University School of Medicine.