Lipidomics Biomarkers: CSF Lipid Predicts MS Progression
- 5 days ago
- 5 min read
Diagnosing multiple sclerosis is the easy part. Telling a newly diagnosed 30-year-old whether they will still be walking unaided in ten years is not. A retrospective study in Molecular Neurobiology tested whether lipidomics biomarkers taken once, from cerebrospinal fluid drawn at the diagnostic lumbar puncture, carry that missing prognostic signal. The group quantified a single lipid, 18:2 cholesterol ester, in CSF from 115 treatment-naive patients and 57 matched controls, then tracked disability for a decade.
Key Takeaways
One CSF lipid measured at diagnosis, 18:2 cholesterol ester, tracked with how severe multiple sclerosis became over the next ten years.
Lipidomics biomarkers work best as prognostic layers, not standalone tests: this lipid scored an AUC of 0.69 alone and 0.92 inside a clinical model.
The marker did not separate patients from controls, making it a prognostic candidate rather than a diagnostic one.
Targeted LC-MS/MS quantified the lipid from 10 microlitres of CSF, a volume that suits biobanked cohorts.
Why Lipidomics Biomarkers Suit a Ten-Year Prognosis Question
Roughly 70% of adult brain cholesterol sits in myelin, and remyelination depends on clearing cholesterol from damaged membranes, esterifying it, and storing it in droplets until it is needed again. Stall that recycling and esterified cholesterol should pile up somewhere measurable. The authors chose targeted lipid profiling over a discovery sweep: 10 microlitres of CSF, single-phase butanol/methanol extraction with internal standards, then LC-ESI-MS/MS on a triple quadrupole running scheduled multiple reaction monitoring. Patients came from one MS unit in Lleida, Spain, were treatment-naive at sampling, and were followed with biannual visits, annual MRI, and certified EDSS scoring. The ten-year endpoint was binary: mild (EDSS 3 or below, no sustained worsening) versus severe (EDSS 6 or above, or a decade on high-efficacy therapy).
Key Findings
No diagnostic separation: CSF 18:2 cholesterol ester was indistinguishable between the 115 patients and 57 controls, and between relapsing-remitting and primary progressive subtypes.
Oligoclonal bands shifted the lipid: Band-positive patients carried higher levels (95% CI 16.48 to 21.49 nM) than band-negative patients (13.21 to 18.33 nM), tying it to intrathecal immune activity.
Baseline levels tracked the ten-year course: Severe-phenotype patients had more of the lipid at diagnosis (95% CI 16.95 to 21.81 nM versus 13.5 to 18.49 nM), and baseline values correlated with EDSS a decade later.
Discrimination came from the combined model: The lipid alone gave an AUC of 0.69 for loss of NEDA-3 status; adding relapse-associated worsening, lesion location, early EDSS, and band status lifted it to 0.92.
Relapse-independent progression stayed invisible: Neither PIRA nor relapse-associated worsening associated with the lipid, so it is not reading out the smouldering component the authors hoped to capture.
Figure 1. Cerebrospinal fluid 18:2 cholesterol ester across diagnostic groups. Panel a compares controls with the full multiple sclerosis cohort and finds no separation (95% CI 13.52 to 19.91 nM versus 16.01 to 19.41 nM). Panel b compares relapsing-remitting with primary progressive disease, again without a significant gap. Panel c shows the one positive contrast: oligoclonal band-positive patients carried more of the lipid than band-negative patients (p < 0.05, unpaired t-test). Adapted from Torres et al. (2026), Molecular Neurobiology.
What the Cholesterol Ester Signal Probably Reflects
Cholesterol freed from damaged myelin is esterified by lecithin:cholesterol acyltransferase and shuttled on apoE particles, or stored as droplets inside microglia. Berghoff and colleagues (2021) showed that sterol synthesis in those phagocytes is required for repair after acute demyelination, which gives a mechanistic reason to expect esterified cholesterol to build up when repair falters. An elevated CSF level at diagnosis is compatible with limited remyelination capacity from the outset, though this study measured one species and cannot separate excess production from defective uptake.
Reading the Prognostic Numbers Carefully
An AUC of 0.92 looks strong until you notice what carries it. The lipid on its own manages 0.69, and the other five model inputs are clinical and radiological variables a neurologist already has. This is a useful confirmatory finding more than a discovery: 18:2 cholesterol ester adds an objective, operator-independent term to models otherwise built on visual MRI reads. The severe phenotype is also defined partly by having received high-efficacy therapy, which entangles the outcome with decisions clinicians made while watching these same patients. Direction of effect appears disease-specific too, since Alzheimer's work of the kind behind this blood lipid panel reports lower CSF cholesterol ester alongside excess in vulnerable tissue.
In Practice: Running a Single-Species Assay Well
Pulling a low-nanomolar cholesterol ester out of 10 microlitres of CSF is unglamorous but exacting work, and the internal-standard and batch-correction choices decide whether a 3 nM group difference survives at all. We run the same class of targeted CSF and plasma lipid panels at Dalton, usually alongside protein and metabolite readouts pulled from the same aliquot, because a cholesterol-ester number is easier to interpret next to the transport proteins that move it. Teams scoping a multi-omics biomarker discovery program on banked CSF should plan that pairing before the first injection; you can't retrofit a second omics layer onto an exhausted sample set.
Frequently Asked Questions
What are lipidomics biomarkers?
Lipidomics biomarkers are individual lipid species, or panels of them, whose concentrations in blood, CSF, or tissue track a disease state or outcome. Mass spectrometry measures them, either as an untargeted survey of hundreds of lipids or as a targeted assay quantifying a short defined list.
Can lipidomics biomarkers predict multiple sclerosis progression?
They can contribute. CSF 18:2 cholesterol ester at diagnosis was higher in patients who reached a severe phenotype ten years later, though the lipid alone gave modest discrimination (AUC 0.69). Its value came from joining existing clinical and MRI predictors, lifting the combined model to 0.92. Prospective multi-centre validation is still missing.
How much sample does a targeted lipid assay need?
Very little. These authors extracted lipids from 10 microlitres of CSF before LC-MS/MS analysis. Small volume requirements are one reason targeted lipid profiling suits biobanked cohorts, where aliquots are limited and shared across assays.
Conclusion
A cholesterol-handling readout taken at diagnosis carries real information about where a patient lands ten years on, and that much is believable on this evidence. What isn't established is whether the lipid shifts a treatment decision, or whether the effect holds outside one Spanish MS unit. For groups building prognostic panels, the practical move is to bank enough CSF at diagnosis to measure it later.
Related Reading
See how we run these analyses in one lab: Dalton's multi-omics CRO services.
Citation
Torres, P., Sánchez, A. G., Sancho-Saldaña, A., Quibus, L., San Pedro Murillo, E., Ruiz-Fernández, E., Peralta, S., Solana, M. J., Brieva, L., & González-Mingot, C. (2026). 18:2 cholesterol ester is a novel prognostic biomarker of disease progression in multiple sclerosis. Molecular Neurobiology, 63(1), 698. https://doi.org/10.1007/s12035-026-05915-8
Note
This blog post summarizes findings from the above-cited research. Figures are adapted from the original publication. For full details, please refer to the source article.
By Seungjun Yeo, CEO at Dalton Bioanalytics. Specializing in multi-omics mass spectrometry for drug discovery and biomarker research.