Combining CSF and Serum Biomarkers to Differentiate Mechanisms of Disability Worsening in Multiple Sclerosis

Abstract

The combined use of serum and CSF biomarkers for prognostic stratification in multiple sclerosis (MS) remains underexplored. This multicenter observational study investigated associations between serum neurofilament light chain (sNfL), glial fibrillary acidic protein (sGFAP), and CSF lipid-specific IgM oligoclonal bands (LS-OCMB) with different forms of disability worsening, such as relapse-associated worsening (RAW), active progression independent of relapse activity (aPIRA), and non-active PIRA (naPIRA). A total of 535 patients with MS were included, all sampled within one year of disease onset. Biomarkers were quantified using single-molecule array and immunoblotting techniques, and CSF cell subsets were analyzed by flow cytometry. High sNfL z-scores and LS-OCMB positivity were independently associated with increased risk of RAW and aPIRA, collectively termed inflammatory-associated worsening (IAW), while elevated sGFAP levels predicted naPIRA. Patients with both high sNfL and LS-OCMB positivity had the highest risk of IAW. Among LS-OCMB-positive patients, higher regulatory T cell percentages were associated with lower sNfL levels, suggesting a protective role. Conversely, in LS-OCMB-negative patients, sNfL levels correlated with CSF C3 concentrations. These findings support the complementary role of sNfL, sGFAP, and LS-OCMB in identifying distinct mechanisms of disease worsening and may inform early personalized management strategies in MS.

Keywords: glial fibrillary acidic protein; intrathecal IgM synthesis; multiple sclerosis; neurofilament light chain; progression.

Figure A1

Gating strategy with representative images for flow cytometry analysis. The gating strategy for identifying cerebrospinal fluid (CSF) cell subsets is illustrated. Total events were initially gated to exclude debris and apoptotic cells (A, Gate P1), followed by the exclusion of doublets (B, Gate P2). Mononuclear cells were then isolated based on CD45 expression (C, Gate Mononuclear Cells). T lymphocytes were identified by CD3 expression (D, Gate T Cells), and their subdivision into CD4+ and CD8+ T cells was achieved through CD8 co-expression (G, Gates CD4+ T and CD8+ T, respectively). Further differentiation of CD4+ and CD8+ T cell subsets into naïve, central memory (CM), effector memory (EM), and terminally differentiated (TD) populations was based on CCR7 and CD45RO expression (H,J). CD4+ regulatory T cells (Tregs) were identified using CD25 and CD127 expression (K, Gate CD4+ Treg). Monocytes were gated by their CD14 expression (E, Gate Monocytes), while B cells were identified by CD19 expression (F, Gate B Cells). B cell subsets were further characterized using CD27 and CD38 expression to distinguish naïve B cells (I, Gate Naïve B), memory B cells (I, Gate Mem B), and plasmablasts (I, Gate Plasmab). Finally, CD5+ B cells were identified based on the co-expression of CD19 and CD5 (L, Gate CD19+ CD5+).

Figure 1

Multivariable Cox regressions of the risk of IAW and non-active PIRA. Estimation of the risk of IAW (A) and non-active PIRA (B) in patients categorized by CSF and serum biomarkers status. Results are presented as adjusted hazard ratios (HRs) with 95% confidence intervals (CIs). ^a Injectable/oral DMTs: glatiramer acetate, all interferon-β formulations, fumarates, teriflunomide, sphingosine-1-phosphate receptor modulators, cladribine, or azathioprine. ^b Monoclonal antibodies: natalizumab, alemtuzumab, ocrelizumab, rituximab, ofatumumab. * p < 0.05; ** p < 0.001; *** p < 0.001.

Figure 2

Differences in CSF regulatory T cells based on LS-OCMB and sNfL status. Box plots illustrate differences in regulatory T cell (T-reg) percentages across four groups categorized by the presence or absence of LS-OCMB and high or low sNfL concentrations (A). The ROC curve shows that T-reg cells effectively distinguished between high and low sNfL levels in patients with positive LS-OCMB (B). Predictive margins derived from multivariable linear regressions (adjusted by age, sex, and time from symptom onset to sampling) demonstrate that higher T-reg percentages were associated with decreasing sNfL z-score values in patients with positive LS-OCMB (C). * p < 0.05; *** p < 0.001.

Figure 3

Differences in CSF complement C3 levels based on LS-OCMB and sNfL status. Box plots show levels in complement C3 across four groups categorized by the presence or absence of LS-OCMB and high or low sNfL levels (A). C3 levels show potential to discriminate patients with high sNfL z-score values if LS-OCMB is negative (B). Predictive margins for C3 levels were estimated from multivariable linear regressions and indicate a correlation with sNfL z-scores in patients with negative LS-OCMB (C). ** p < 0.01; **** p < 0.0001.