What MS Genetics Finds When It Looks at Metabolism

Metabolism has become a standing theme in multiple sclerosis (MS) research, but the work supporting it arrives at very different levels of resolution. Some studies ask whether a gene's expression causes disease risk, using inherited variation as the instrument. Others work inside a single cell type and ask what one metabolic enzyme does to inflammation or to neuronal survival. Three recent papers sit at those different levels, and reading them together shows both how much metabolic involvement in MS has been demonstrated and how little of it currently joins up. Each identifies a real mechanism. None of them identifies the same one.

A Gene-Level Search Found a Mitochondrial Hit
Ye and colleagues asked which metabolically relevant genes carry genetic evidence of causing neurodegenerative disease, rather than merely accompanying it. They applied summary data-based Mendelian randomization to 936 oxidative stress-related genes, integrating genome-wide association data with expression quantitative trait loci in a multi-cohort design that used blood eQTLs for discovery and for a second round, then brain eQTLs for a further check. Colocalization verified the inferences, and DNA methylation and pathway analyses followed. Fifteen genes came out associated with neurodegenerative disease risk, and for MS the finding was TSFM, where higher genetically predicted expression tracked with increased risk. That association survived both the confirmation cohort and the brain eQTL data, which is the part that matters, since many such results do not hold up when the tissue changes. TSFM encodes a mitochondrial translation elongation factor, and pathway enrichment across the fifteen genes pointed specifically to autophagy-apoptosis and mitochondrial pathways.

Inflammation Reaching Neuronal Energy Metabolism Through the Proteasome
Woo and colleagues followed a chain from inflammation to neuronal death that runs entirely through metabolism, combining transcriptional, proteomic and functional analyses of proteasomal subunits in inflamed neurons. Interferon-γ induces the immunoproteasome subunit PSMB8, which upsets the proteasomal balance and lowers proteasome activity. Because the proteasome is what clears it, PFKFB3 then accumulates. PFKFB3 regulates how a cell routes glucose, and its build-up pushes neurons into higher glycolysis while lowering pentose phosphate pathway activity, which costs the cell its main route to antioxidant capacity. Oxidative injury and ferroptosis follow. The causal test came from two directions at once: neuron-specific genetic targeting and systemic pharmacological targeting of either PSMB8 or PFKFB3 protected neurons in culture and in a mouse model of MS.

A Metabolic Enzyme Writing Inflammatory Memory Into Chromatin
Lee and colleagues worked on a different cell type and found metabolism operating through a different route again. They identified an astrocyte subset that remembers past stimulation, mounting amplified pro-inflammatory responses when challenged a second time, using single-cell RNA sequencing, assays for chromatin accessibility, chromatin immunoprecipitation, nucleic acid-based cell sorting and cell-specific in vivo CRISPR-Cas9 perturbation. The memory turns out to be controlled by a metabolic enzyme. ATP-citrate lyase supplies the metabolite that the histone-modifying enzyme p300 requires to open chromatin, so what a cell can transcribe on second exposure depends on what its metabolism produced. Astrocytes positive for both ACLY and p300 increased in number in acute and chronic experimental autoimmune encephalomyelitis, and inactivating them genetically improved the disease. The same memory phenotype appeared in human astrocytes in culture, and both sequencing and immunohistochemistry found increased numbers of these astrocytes in chronic MS lesions.

Three Different Kinds of Metabolic Involvement
Nothing in these three results describes one pathway, and the differences are more than anatomical. A mitochondrial translation elongation factor raising susceptibility is a statement about how much protein a mitochondrion can build. A proteasome subunit diverting glucose handling is a statement about protein turnover controlling flux through two competing branches of glucose metabolism. A citrate-cleaving enzyme setting chromatin accessibility is a statement about metabolism determining which genes a cell can reach later. These are three distinct relationships between metabolism and disease, operating on different timescales and in different cell types, and only the last of them makes the metabolic enzyme a regulator of gene expression rather than a consumer or producer of energy.

Where Genetics Stops and Mechanism Begins
Read against each other, these three papers have a gap in the middle. The genetic study gives a gene with human inherited evidence behind it and no mechanism attached: TSFM arrives as a putative causal association with an expression direction and a tissue, and nothing in the paper says what altered mitochondrial translation does to an oligodendrocyte or a T cell. The two mechanism papers give the opposite. PSMB8, PFKFB3 and ACLY come with detailed pathways, genetic perturbation in vivo and human tissue confirmation, and none of the three is reported here as carrying a human genetic association with MS risk. So the gene with population-level evidence has no mechanism, and the mechanisms with experimental evidence have no reported population-level support. That is not a flaw in any of the three studies. It describes where the field currently is.

What Each Design Can and Cannot Settle
The division of labour is also a division of limits. Mendelian randomization delivers direction of effect and resistance to reverse causation, which the mechanism studies cannot offer, but it assumes linear effects, it depends on instrument strength, and it measures lifelong average exposure rather than exposure at any particular moment. An eQTL-based estimate is also a statement about expression rather than about function, which is why the authors present TSFM as putative rather than settled. The two mechanism papers carry the opposite profile. Both ran genetic perturbation in vivo and both closed the loop with human tissue, which is as close to causal demonstration as this field reaches, but a mouse model and a post-mortem lesion say nothing about how often a mechanism operates in patients, in whom, or at what point in the disease. The useful reading is that these designs answer different questions, and that connecting them is the work that has not been done yet.

Disclaimer: This blog post is based on the cited studies 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:
Ye, Z., Xu, J., Zhang, T., Zhu, Y., Lu, Y., Chen, X., Yang, Z., Yuan, Z., Huang, S., Weng, Y., & Yang, D. (2026). Multi-omics insights into oxidative stress-related genes and neurodegenerative diseases: from epigenetic regulation to pathogenic pathways. Journal of Molecular Neuroscience, 76(4). https://doi.org/10.1007/s12031-026-02601-1
Woo, M. S., Brand, J., Bal, L. C., et al., & Friese, M. A. (2025). The immunoproteasome disturbs neuronal metabolism and drives neurodegeneration in multiple sclerosis. Cell, 188(17), 4567–4585.e32. https://doi.org/10.1016/j.cell.2025.05.029
Lee, H.-G., Rone, J. M., Li, Z., Akl, C. F., et al., & Quintana, F. J. (2024). Disease-associated astrocyte epigenetic memory promotes CNS pathology. Nature, 627(8005), 865–872. https://doi.org/10.1038/s41586-024-07187-5