The Metabolic Switch in Progressive MS, and How Much of It Is Tested
Most of the attention this study attracts goes to its classifier, the machine learning models that separate relapsing-remitting from secondary progressive multiple sclerosis (MS) on serum metabolites. The other half of the paper is doing something different. Oppong and colleagues also sequenced whole blood RNA from the same patients and built a network joining differentially expressed genes to the altered metabolites, and out of that they propose a mechanism: progressive disease involves a metabolic switch away from glycolysis and toward gluconeogenesis and ketogenesis, that switch imposes metabolic stress, and the stress triggers stress-response pathways and subsequent neurodegeneration. That is a chain of four claims. It is worth walking through what carries each link, because the evidence behind them is not evenly distributed and the paper draws them all with the same confidence.
Pyruvate and Ketone Bodies Carry the Strongest Enrichment
Metabolite set enrichment gives ratios as well as p-values, and the ratios are the more informative half. Pyruvate metabolism came out highest at 42.74, followed by synthesis and degradation of ketone bodies at 34.31. Glycerolipid metabolism and alanine, aspartate and glutamate metabolism both sat at 12.20, butanoate metabolism at 11.36, aminoacyl-tRNA biosynthesis at 10.68, glyoxylate and dicarboxylate metabolism at 10.67, and the citrate cycle at 10.64. The individual metabolites line up behind those pathways with their own numbers: glutamine at p = 2.0E-08, cholines at 2.8E-08, total fatty acids at 4.3E-06, citrate at 1.3E-04, pyruvate at 1.1E-03, acetoacetate at 7.7E-04, alanine at 2.1E-02, lactate at 1.5E-02, and beta-hydroxybutyrate at 3.3E-02. The pathway network the authors draw from a second database even carries a Warburg effect node, which is the right label for cells running lactate up while pyruvate falls.
The Transcriptome Agrees, and One Pathway Appears on Both Lists
Sequencing returned 1,052 differentially expressed genes, 948 up and 104 down. The upregulated set enriched for membrane trafficking, regulation of apoptotic signalling, autophagy, metabolism of lipids, regulation of cellular response to stress, TNFA signalling via NFKB, cytokine signalling in the immune system, metabolism of RNA, and neutrophil degranulation. The downregulated set enriched for metabolism of RNA, tRNA processing, glyoxylate metabolism and glycine degradation, cellular response to DNA damage, cysteine and methionine metabolism, response to ketone, and regulation of immune effector process. Metabolism of RNA sits on both lists. That happens when a broad pathway holds genes moving in opposite directions, and it is a reminder that a pathway name on its own tells you a process is involved without telling you which way it went. The 90-to-10 split between upregulated and downregulated genes is also worth keeping in view, since a strongly one-sided distribution is one of the patterns that batch effects produce.
The Network, and One Node That Did Not Come From This Data
Combining all 1,052 genes with the metabolomic signature produced a network of 31 genes around 8 metabolite nodes, in four blocks. The citrate cycle block sets OGDH, IDH2, SLC25A1, PDHX, PPIF, MRPL4, APRT, PGLS, HGS, and UNG around raised citrate. The pyruvate, gluconeogenesis and ketogenesis block sets SLC16A3, HAGH, MMP9, QDPR, RPL9, and YARS around reduced pyruvate and raised lactate. The amino acid and aminoacyl-tRNA block sets GAMT, CST3, SLC47A1, ACTN4, PPBP, ADM, and GOT2 around reduced glycine and creatinine. The lipid block sets SCD5, PLA2G12A, CARM1, SERPINE1, and KIF5B around raised linoleic acid. One caveat lives in the figure caption rather than the text: phenylalanine appears as a node although this study's own metabolomics did not identify it, and it was included because earlier reports find it lowered in MS. A node imported from other work is a different kind of evidence from one this dataset produced, and the figure draws them identically.
GOT2 Is the Hinge, and Its Position Is the Argument
The one downregulated gene the authors single out is GOT2, at a log2 fold change of −0.617. What makes it interesting is not the size of the change but where it sits. GOT2 spans amino acid metabolism, aminoacyl-tRNA biosynthesis, lipid metabolism, pyruvate metabolism, gluconeogenesis, and ketogenesis, which is close to the full list of processes this study flags, and it has been linked to MS before. In the lipid block, the upregulated SCD5 (stearoyl-CoA desaturase 5), PLA2G12A (phospholipase A2 group XIIA), CARM1 (coactivator-associated arginine methyltransferase 1) and SLC25A1 (solute carrier family 25 member 1) sit alongside raised linoleic acid, which the authors read as possible impairment of linoleic acid conversion into omega-3 fatty acids in progressive disease. Anyone trying to rebuild the gene list should note a mismatch between figures: the volcano plot caption gives a log2 fold-change cutoff of 1.5, while the validation panel uses 0.585, and a value of −0.617 clears only the second.
What Was Checked Against Someone Else's Data
The strongest support in this half of the paper is external. The authors tested their pathway list against an independent gene expression dataset covering whole blood, myeloid cells, lymphocytes, and oligodendrocyte precursor cells from patients with secondary progressive versus relapsing-remitting disease. Pathways appeared in both, including metabolism of RNA, regulation of cellular response to stress, cytokine signalling in the immune system, mitochondrion organisation, unfolded protein response, protein folding, protein processing in the endoplasmic reticulum, and positive regulation of cell death. Separately, network analysis of the cellular respiration pathway returned an association with amyotrophic lateral sclerosis, which the authors read as support for dysregulated cell metabolism contributing to neurodegeneration in progressive MS. That whole blood here overlaps with brain-derived cells there is the best available argument that this is a systemic alteration rather than something particular to circulating cells.
The Mechanism Is a Hypothesis the Design Cannot Test
Return to the four-link chain. The metabolic switch is supported by convergent metabolomic and transcriptomic data. Metabolic stress is inferred from the switch rather than measured. Stress-response pathway activation is observed. Neurodegeneration is not measured anywhere in this study. Every sample was collected at a single time point, so nothing here separates a switch that drives neuronal damage from one that follows it, and enriched stress-response pathways are equally consistent with both readings. The transcriptomics rest on 8 patients with progressive and 5 with relapsing disease, which the authors list as a limitation alongside the unbalanced metabolomics groups, and they note that RNA changes do not always translate into protein expression or activity, and that batch variation makes omics data hard to reproduce without proper controls. Their hypothesis does have a testable form, and it is specific: if the switch drives the damage, GOT2 expression and the pyruvate-to-lactate ratio should move before serum neurofilament light rises in the same patient, and afterwards if the switch is a consequence. That requires serial sampling through the transition, which is the one thing a cross-sectional design cannot deliver.
Disclaimer: This blog post is based on the cited 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.
Reference:
Oppong, A. E., Coelewij, L., Robertson, G., Martin-Gutierrez, L., Waddington, K. E., Dönnes, P., Nytrova, P., Farrell, R., Pineda-Torra, I., & Jury, E. C. (2024). Blood metabolomic and transcriptomic signatures stratify patient subgroups in multiple sclerosis according to disease severity. iScience, 27(3), 109225. https://doi.org/10.1016/j.isci.2024.109225
