Identification of brain-enriched proteins in CSF as biomarkers of relapsing remitting multiple sclerosis

Abstract

Background: Multiple sclerosis (MS) is a clinically and biologically heterogenous disease with currently unpredictable progression and relapse. After the development and success of neurofilament as a cerebrospinal fluid (CSF) biomarker, there is reinvigorated interest in identifying other markers of or contributors to disease. The objective of this study is to probe the predictive potential of a panel of brain-enriched proteins on MS disease progression and subtype.

Methods: This study includes 40 individuals with MS and 14 headache controls. The MS cohort consists of 20 relapsing remitting (RR) and 20 primary progressive (PP) patients. The CSF of all individuals was analyzed for 63 brain enriched proteins using a method of liquid-chromatography tandem mass spectrometry. Wilcoxon rank sum test, Kruskal-Wallis one-way ANOVA, logistic regression, and Pearson correlation were used to refine the list of candidates by comparing relative protein concentrations as well as relation to known imaging and molecular biomarkers.

Results: We report 30 proteins with some relevance to disease, clinical subtype, or severity. Strikingly, we observed widespread protein depletion in the disease CSF as compared to control. We identified numerous markers of relapsing disease, including KLK6 (kallikrein 6, OR = 0.367, p < 0.05), which may be driven by active disease as defined by MRI enhancing lesions. Other oligodendrocyte-enriched proteins also appeared at reduced levels in relapsing disease, namely CNDP1 (carnosine dipeptidase 1), LINGO1 (leucine rich repeat and Immunoglobin-like domain-containing protein 1), MAG (myelin associated glycoprotein), and MOG (myelin oligodendrocyte glycoprotein). Finally, we identified three proteins-CNDP1, APLP1 (amyloid beta precursor like protein 1), and OLFM1 (olfactomedin 1)-that were statistically different in relapsing vs. progressive disease raising the potential for use as an early biomarker to discriminate clinical subtype.

Conclusions: We illustrate the utility of targeted mass spectrometry in generating potential targets for future biomarker studies and highlight reductions in brain-enriched proteins as markers of the relapsing remitting disease stage.

Keywords: Biomarkers; CSF; Multiple sclerosis; Proteomics.

Fig. 1

Study workflow, protein candidate identification, and subcellular localization. (A) Patient samples were collected via lumbar puncture, frozen, and sent to Mount Sinai Hospital for protein purification and analysis on LC-MS/MS. (B) Of the 63 target proteins, 54 were consistently detected by the LC-MS/MS and appropriate for analysis. Statistical testing including Wilcoxon p, Kruskal-Wallis, linear regression, logistic regression, or Mann-Whitney test refined the 54 target protein panel to a subset of 30 protein candidates. (C) Subcellular localization of the 30 protein candidates using COMPARTMENTS database indicates a predominance of membrane associated proteins.

Fig. 2

All significantly detected proteins have decreased expression in MS. (A) All 54 detected peptides by LC-MS/MS are displayed relative to control. Focused subset of proteins that are significant in either all MS patients combined vs. control (B) or relapsing remitting group vs. control (C). (D) Venn-Diagram illustrating the statistically significant proteins for the following comparisons: MS vs. Ctrl, RR vs. Ctrl, and PP vs. Ctrl. TMEM132A was the only significant protein across all three analyses. CNDP1 was the only significantly different protein in RR vs. PP analysis. Box and whisker plots of the relative protein quantification of TMEM132A (E) or CNDP1 (F) in control, MS, RR, or PP groups. Box represents median, quartile 1, and quartile 3 with whiskers to range. Bar graphs represent mean with error bars +/- SEM. *p < 0.05 by Wilcoxon for all graphs in figure.

Fig. 3

Odds ratio for each listed comparison on scaled, age-adjusted data. Odds ratio displayed for select proteins under each comparison, either MS vs. Ctrl (A), RR vs. Ctrl (B), PP vs. Ctrl (C) or RR vs. PP (D). Gray color indicates non-significance. Displayed is the OR with 95% confidence range. Reference line of OR = 1 also displayed.

Fig. 4

Relation of PRM protein panel to clinically useful biomarkers. Relation of select proteins to clinically useful biomarkers, including NF-L (A), enhancing lesions on MRI (B-E), or EDSS (F-H). (Ai) Quantification of NF-L across control, MS, or RR and PP. (Aii) Odds ratio for NF-L predicting each comparison via logistic regression. Gray bars indicate non-significant difference. Error bars to 95% CI. (Aiii) NF-L quantification between Control-MRI negative, MS-MRI negative, and MS-MRI positive enhancing lesions at time of draw. Other proteins that are differentially expressed are displayed including (B) KLK6, (C) MAG, (D) MOG, and (E) SERPINI1. R² (percent of variability explained by the model) calculated via linear regression of (F) CNDP1, (G) CAMK2A, or (H) GFAP are displayed with 95% confidence intervals for either males (blue circles) or females (red triangles). *p < 0.05, **p < 0.01, ***p < 0.001 via Kruskal-Wallis with Dunn multiple comparison corrections.

Fig. 5

Candidate proteins are highly interconnected and enriched in pathways relevant to MS disease. Analysis of the 28 candidate proteins in the String database. (A) Visual mapping of the protein interconnectedness. Color represents cluster by k-means clustering. Confidence of association between proteins is indicated by line thickness. Enrichment plots of (B) Biological Processes, (C) Cellular Component, (D) Molecular Function, (E) Associated Disease, and (F) KEGG Pathways. Bar indicates strength of association; FDR given by color; fraction indicates observed gene count / background gene count for that pathway.