Diagnostic biomarkers from proteomic characterization of cerebrospinal fluid in patients with brain malignancies

Abstract

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.

A connectivity signature for glioblastoma

“…Remarkably, we have recently identified CHI3L1 as a key cerebrospinal fluid proteomic biomarker in GB in an independent study (13) . The mRNA level and protein level of CHI3L1 in GB were highly correlated (n = 93, R = 0.85, Supplementary Figure 1A).…”

Leveraging the CSF proteome toward minimally-invasive diagnostics and biological characterization of brain malignancies

“…Next, we applied a 70% intra-group coverage threshold to obtain a dataset of 755 proteins for downstream analyses, representing an improvement in proteome coverage over the 506 proteins recovered by Schmid and colleagues in a comparable cohort using an identical coverage threshold ( Fig S2A-D ). 16 Tumor suppressor genes (e.g., PTEN, NF1, TP53) and oncogenes (e.g., EGFR, BRAF, PDGFRA) implicated in GBM, BM and CNSL were not reliably recovered by our protocol, which we attributed to our approach preferentially detecting secreted, rather than intracellular or membrane-bound proteins ( Table S1 ). 21 Importantly, UMAP representation of CSF proteomes revealed coherent separation of each diagnostic group, suggesting that CSF proteomics are suitable for diagnosing malignancy, and discriminating between different brain neoplasms ( Fig 1B ).…”

“…Functional annotation revealed that malignancy was associated with apoptotic signaling, glycolysis and heme metabolism, whereas non-malignancy was associated with elastic fibres- and extracellular matrix (ECM)-associated proteins, and neuronal and glial processes ( Fig 2C, D ; Fig S3A ). Importantly, using published proteomic CSF data from GBM, 16 BM, 16 CNSL, 16 Alzheimer’s disease (AD), 22 amyotrophic lateral sclerosis (ALS), 23 and clinically-isolated syndrome of demyelination [CIS; first attack of multiple sclerosis (MS)], 24 we demonstrated that our malignancy and non-malignancy signatures are significantly more deregulated in neoplastic disease (GBM, BM, CNSL) than non-neoplastic disease (AD, ALS, CIS/MS) ( Fig 2E, F ).…”

“…(E, F) Malignant ( E ) and non-malignant ( F ) index differences (Δ) between disease and control samples. Discovery cohorts include data from current study, and validation cohorts include previously published proteomic data from Schmid et al (GBM, BM, CNSL) 16 , Bader et al (AD) 22 , Bereman et al (ALS) 23 , and Stoop et al (CIS/MS) 24 . Difference between neoplastic (ND) and non-neoplastic disease (NND) in validation cohorts was assessed by Student’s t-test.…”

“…Difference between neoplastic (ND) and non-neoplastic disease (NND) in validation cohorts was assessed by Student’s t-test. (G) ROC curve assessing sensitivity and specificity of non-malignant and malignant indices as diagnostic biomarkers in current (discovery) cohort and Schmid (validation) cohort 16 . The indicated performance metrics correspond to Model 1 in Table 1 .…”