Decoding Multiple Sclerosis Progression Through Blood Metabolomics and Transcriptomics

Multiple sclerosis (MS) is a chronic autoimmune and neurodegenerative disorder in which inflammatory injury, demyelination, and neuro-axonal damage contribute to progressive neurological disability. The study by Oppong and colleagues, entitled “Blood metabolomic and transcriptomic signatures stratify patient subgroups in multiple sclerosis according to disease severity,” focuses specifically on distinguishing relapsing-remitting multiple sclerosis (RRMS) from secondary progressive multiple sclerosis (SPMS). This distinction is clinically important because progression from RRMS to SPMS is frequently recognized only retrospectively, after irreversible disability has accumulated. At present, no validated blood-based biomarker reliably separates these disease phases, despite their differing pathological characteristics and therapeutic requirements. The investigators therefore examined whether molecular changes detectable in peripheral blood could provide a more objective representation of disease severity. Their central approach combined serum metabolomics, whole-blood transcriptomics, machine-learning classification, and molecular-network analysis. This multi-omic strategy was designed not merely to classify patients but also to identify biological pathways potentially responsible for disease progression. The broader objective was to establish a molecular framework that could eventually support earlier recognition of progressive disease, more precise patient stratification, and improved selection of therapeutic strategies.

Integrating Metabolomics, Transcriptomics, and Machine Learning
The investigators analyzed serum obtained from 52 patients with RRMS, 29 patients with SPMS, 80 healthy controls, and 30 individuals with neuromyelitis optica who served as disease controls. Serum metabolomics was performed using a standardized nuclear magnetic resonance spectroscopy platform capable of quantifying approximately 250 biomarkers, including amino acids, fatty acids, apolipoproteins, and multiple classes and sizes of HDL, LDL, IDL, and VLDL lipoprotein particles. The metabolomic dataset was subsequently examined using several supervised machine-learning approaches, including logistic regression, bagged and boosted logistic regression, random forest, support vector machines, neural networks, and sparse partial least-squares discriminant analysis. Ten-fold cross-validation was incorporated into model evaluation to reduce overfitting. Importantly, the researchers also examined models that excluded age and Expanded Disability Status Scale (EDSS) scores, allowing them to determine whether the molecular data contained discriminatory information independent of conventional indicators of disease severity. Whole-blood RNA sequencing was additionally performed in a smaller subgroup of five RRMS and eight SPMS patients, enabling the investigators to compare circulating metabolic changes with alterations in gene expression.

A Metabolic Signature Capable of Distinguishing RRMS from SPMS
The machine-learning analyses demonstrated that circulating metabolites contained substantial information about MS disease phenotype. Among the models comparing RRMS and SPMS, boosted logistic regression achieved an accuracy of 92.6%, sensitivity of 89.7%, specificity of 94.2%, and an AUC-ROC of 0.965. Even when age and EDSS were excluded, metabolites alone retained meaningful discriminatory performance, supporting the proposition that the observed molecular signature was not simply a reflection of older age or greater disability in patients with SPMS. The prominent discriminatory features included glutamine, linoleic acid, acetoacetate, saturated fatty acids, omega-6 fatty acids, several cholesterol-containing lipoprotein fractions, lactate, pyruvate, citrate, sphingomyelins, alanine, tyrosine, and other amino acids. The authors then developed a simplified scoring system based on particularly informative metabolites. Five features—cholines, glutamine, saturated fatty acids, acetoacetate, and sphingomyelins—were incorporated into the final AutoScore model, which achieved an AUC-ROC of approximately 0.946 in the test dataset. Figure 1 on page 5 visually reinforces this finding by showing separation between RRMS and SPMS and by comparing the discriminatory performance of the molecular signature with individual clinical and metabolic variables.

Evidence for Altered Cellular Energy Metabolism in Progressive Disease
Beyond patient classification, the metabolomic findings suggested a fundamental alteration in energy metabolism associated with SPMS. Pathway-enrichment analysis identified significant changes involving ketone-body metabolism, glycerolipid metabolism, gluconeogenesis, glycolysis, pyruvate metabolism, the tricarboxylic acid cycle, and several amino-acid pathways. More specifically, SPMS was characterized by increased serum concentrations of lactate and glutamine, which were associated with gluconeogenic metabolism, together with elevated acetoacetate and β-hydroxybutyrate, two principal ketone bodies. Citrate was also elevated, whereas glycolysis-related metabolites such as alanine and pyruvate were reduced. In parallel, triglycerides, cholines, fatty acids, and several lipoprotein-associated lipid measures were dysregulated. Figure 2 on page 7 illustrates these differences and places them within interconnected metabolic pathways. Collectively, the results are consistent with a proposed shift away from conventional glycolytic metabolism toward greater reliance on gluconeogenesis, ketogenesis, fatty-acid utilization, and altered mitochondrial energy processing. The authors interpret this pattern as evidence of systemic metabolic stress associated with progressive disease. Importantly, the study establishes an association rather than demonstrating that these metabolic alterations directly cause neurodegeneration; nevertheless, the coordinated changes suggest that cellular energy dysfunction may be biologically relevant to the transition toward a progressive MS phenotype.

Transcriptomics Supports a Systemic Molecular Shift in SPMS
The transcriptomic component provided an independent molecular layer supporting the metabolomic observations. RNA sequencing of whole blood identified 1,052 differentially expressed genes between SPMS and RRMS, of which 948 were upregulated and 104 were downregulated in SPMS. These genes were sufficiently distinct to cluster the two patient groups separately in principal-component analysis. Pathway analysis showed enrichment of processes involving RNA metabolism, cellular responses to stress, lipid metabolism, cellular respiration, immune signaling, and regulation of immune-effector functions. The authors then integrated transcriptomic and metabolomic data into a gene–metabolite interaction network. Twenty-six upregulated and four downregulated genes interacted with eight metabolites from the SPMS-associated signature, including creatinine, citrate, pyruvate, lactate, phenylalanine, tyrosine, glycine, and linoleic acid. Genes involved in lipid handling and energy metabolism, including SCD5, PLA2G12A, CARM1, SLC25A1, and GOT2, contributed to this network. The resulting molecular architecture linked amino-acid metabolism, lipid metabolism, pyruvate metabolism, the TCA cycle, gluconeogenesis, and ketogenesis. Thus, the transcriptomic findings did not merely reproduce individual metabolite associations; they indicated coordinated regulation across different biological levels, strengthening the hypothesis that SPMS is accompanied by widespread metabolic and cellular adaptation.

Potential Implications for Biomarkers and Precision Medicine
The clinical significance of these findings lies in the possibility that metabolomic profiles could complement existing neurological assessment and imaging when identifying progressive disease. The study demonstrated that the combined metabolite signature outperformed several individual metabolites and conventional variables such as age and EDSS in discriminating RRMS from SPMS. Acetoacetate was particularly notable: the article reports an AUC-ROC of 0.87 for this metabolite alone, compared with 0.70 for EDSS and 0.83 for age, although a single metabolite would not be sufficient to establish a clinical diagnostic test. More broadly, the standardized NMR platform used in the study represents an important translational advantage because it is relatively rapid, reproducible, cost-effective, and compatible with large-scale biomarker studies. If validated prospectively, a blood-based metabolic score might eventually help identify individuals approaching a progressive disease state before substantial irreversible disability becomes evident. Such biomarkers could also improve clinical-trial stratification, define biologically distinct MS endotypes, and potentially identify metabolic pathways suitable for therapeutic investigation. The authors nevertheless present these applications as future possibilities rather than currently established clinical uses; the metabolic score remains a research tool requiring validation before it could influence diagnosis or treatment decisions.

Limitations, Future Validation, and Scientific Significance
Despite its promising findings, the study contains important limitations that constrain immediate clinical interpretation. The metabolomics groups were unequal in size, while the transcriptomic analysis involved only 13 participants, substantially limiting the statistical power and generalizability of the RNA-sequencing component. Differences in demographic characteristics between RRMS and SPMS were also present, although age and EDSS were incorporated into the machine-learning analysis and additional models were constructed without these variables. Furthermore, patients with SPMS had previously received disease-modifying therapy before progressing from RRMS, whereas the RRMS cohort was recruited before first treatment, introducing a potential treatment-related source of variation. Metabolomic measurements can additionally be influenced by factors such as age, diet, hormonal status, and analytical platform, while transcript abundance does not necessarily correspond directly to protein abundance or biological activity. Consequently, longitudinal studies following patients through the RRMS-to-SPMS transition and replication in larger independent cohorts will be essential. Nevertheless, the study provides compelling evidence that MS progression is accompanied by coordinated alterations in circulating metabolites and gene-expression pathways, particularly those governing lipid metabolism, cellular respiration, gluconeogenesis, and ketogenesis. Its principal contribution is therefore both diagnostic and mechanistic: it identifies candidate blood biomarkers while simultaneously offering a molecular model of the metabolic stress associated with progressive multiple sclerosis.

Disclaimer: This blog post is based on the provided research article and is intended for informational purposes only. It is not intended to provide medical advice. Please consult with a healthcare professional for any health concerns.

References:
Oppong, A. E., Coelewij, L., Robertson, G., Martin-Gutierrez, L., Waddington, K. E., Dönnes, P., ... & Jury, E. C. (2024). Blood metabolomic and transcriptomic signatures stratify patient subgroups in multiple sclerosis according to disease severity. IScience, 27(3).