NAD, Fumarate, and CNS Resilience in MS

Asking why an axon or an oligodendrocyte survives in multiple sclerosis (MS) turns fairly quickly into a question about two small metabolites. One is NAD, whose level inside an axon decides whether a conserved self-destruction programme runs. The other is fumarate, a tricarboxylic acid cycle intermediate whose ester is already a licensed MS therapy and which works by switching on a transcriptional stress programme. Four papers from 2025 and 2026 approach CNS resilience from these two directions, and most of the evidence sits outside the animal model: human iPSC-derived neurons and oligodendrocytes carrying two different metabolic insults, blood cells from relapsing-remitting MS patients before and after a year on treatment, organotypic brain slices, and one dose study of a SARM1 inhibitor in experimental autoimmune encephalomyelitis.

The NAD Sensor That Decides Whether an Axon Dies
Axonal degeneration is one of the pathological features underlying neurodegeneration in MS, and SARM1 is the enzyme that executes it. SARM1 is an NAD hydrolase, and its activation drains NAD while generating adenosine diphosphate ribose, which disturbs calcium levels inside the axon and produces the damage. The switch is built out of the metabolite itself. SARM1 sits as a homomultimeric octamer whose autoinhibitory ARM domains hold the enzyme domains in an inactive shape, and both NAD and a second metabolite, nicotinamide mononucleotide, compete for an allosteric pocket in that ARM domain. During axonal injury the NMN to NAD ratio rises, NMN has the higher affinity, it pushes NAD out of the pocket, the ARM domains release, and the enzyme starts cutting. Resilience here is a ratio between two nucleotides rather than the concentration of either one.

A Dose Curve That Runs Backwards in EAE
Mani and colleagues tested what happens when you block that enzyme with a specific class of compound. Base-exchange inhibitors work as pro-drugs: SARM1 itself converts them, bonding the compound to adenosine diphosphate ribose to make the active inhibitor. So the inhibitor is made by the enzyme it is meant to stop, using NAD. Dosing mice orally twice daily from the day of myelin oligodendrocyte glycoprotein immunization, they ran RO-7529 at 2, 10 and 50 mg/kg to span subpharmacological to near-complete target coverage. At 50 mg/kg EAE symptoms fell. At 2 mg/kg disease started earlier and was worse through the acute inflammatory phase than in vehicle controls, with significantly raised serum neurofilament light chain, more inflammation and a trend toward more demyelination in the lumbar spinal cord on blinded scoring at day 28. Two mechanistically different inhibitors, the allosteric covalent EV-99 and the NAMPT inhibitor FK-866, showed no low-dose toxicity under the same conditions.

Human Neurons, Two Metabolic Insults, and a Stressor That Nothing Rescued
González and colleagues built the same question into human cells, with no immunization anywhere in the design. They differentiated iPSCs into neurons and oligodendrocytes, then picked stressors to match MS mechanisms rather than a model: rotenone, which inhibits mitochondrial complex I, and 1 mM glutamate, which produces the axonal beading that precedes transection. Rotenone cut survival to 73.65% and pretreatment with IL-4, minocycline or pioglitazone gave a mild but significant rescue. The control gives it edges: taxol, a microtubule stabilizer that kills without touching the respiratory chain, cut viability to a similar 83.61%, and nothing rescued it. Pioglitazone is a PPAR-gamma agonist and its protection was abolished by the antagonists GW9662 and T0070907, while three GSK3-beta inhibitors reproduced its effect against glutamate and did nothing against rotenone, splitting the two insults onto separate arms. Proteomics matched that split: pioglitazone downregulated 235 proteins including GSK3-beta, while minocycline moved only 21, one of them LAP3, an enzyme in glutathione biosynthesis and glutamate breakdown, and minocycline was the only compound that lowered the mitochondrial superoxide signal. In acute cuprizone mice both protected neurons and axons yet reduced neither demyelination nor oligodendroglial loss, which the authors state plainly.

A Fumarate Ester, GPX4, and Patients on Treatment
Fischer and colleagues worked the fumarate side against ferroptosis, iron-dependent cell death built on peroxidation of polyunsaturated phospholipids and the membrane rupture that follows it. GPX4 is the only mammalian enzyme that reduces lipid hydroperoxides to harmless alcohols, and it needs glutathione as cofactor. Diroximel fumarate and monomethyl fumarate rescued myelin in organotypic cerebellar slices from erastin, and raised GPX4 transcripts and protein in oligodendrocytes within three hours, back to baseline by twelve, which matches the twice-daily dosing used in patients. The human arm is the part that needs no model at all: in blood cells from twelve relapsing-remitting patients, GPX4 ran at 65% of healthy donors, and in the five sampled again after a year of dimethyl fumarate it rose to 175% while HMOX1 rose to 160%. The authors also report what failed. Neither fumarate rescued iron-induced myelin damage, where only an iron chelator worked, and neither lowered lipid peroxidation itself.

Where the Same Programme Did Not Fire
The control that gives this its edges is a second autoimmune disease. In synovial fibroblasts from rheumatoid arthritis and osteoarthritis patients, the fumarates raised HMOX1 enormously, to 572% and 1077%, so the Nrf2 arm fired harder than it ever did in oligodendrocytes. GPX4 went the other way, down to roughly 40% to 65% of untreated levels on protein. The upstream switch worked and the anti-ferroptotic output did not follow, which is why the authors call the mechanism organ- and disease-specific rather than a general antioxidant effect. De Oliveira's review of dimethyl fumarate and mitochondria describes the same shape inside the CNS: Nrf2 activation with NQO-1, HO-1 and the biogenesis markers PGC-1alpha, NRF1 and TFAM is the consistent part, but the consequence splits, improving respiration in neurons and oligodendrocytes while possibly impairing it in microglia, T cells and vascular cells under inflammatory stress.

Dose, Insult, and Cell Type
Three parameters flip a result across these four papers, and none of them is the metabolite. Dose: the same SARM1 compound in the same EAE protocol cut symptoms at 50 mg/kg and worsened onset and serum neurofilament light chain at 2 mg/kg, while two inhibitors with other mechanisms did neither, which pins the reversal on the base-exchange route rather than on SARM1 blockade in general. Insult: in human neurons the same three compounds rescued rotenone and glutamate but not taxol, and the GSK3-beta arm covered glutamate alone, so protection tracked the metabolic insult rather than cell death as such. Cell type: fumarates raised GPX4 in oligodendrocytes and in patients' blood cells while lowering it in arthritis fibroblasts whose Nrf2 response was five to ten times larger, and the same drug may run mitochondria in opposite directions in a neuron and in a T cell sitting in the same lesion.

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:
Mani, A., Mendel, M., Westwood, P., et al., Giroud, M., & Keaney, J. (2025). SARM1 base-exchange inhibitors induce SARM1 activation and neurodegeneration at low doses. npj Drug Discovery, 2(1). https://doi.org/10.1038/s44386-025-00016-3
González, R. G., Yilmaz, E. N., Albrecht, S., Fucito, M., Pieragostino, D., Schmidt, A., König, S., FitzGerald, U., & Kuhlmann, T. (2025). Development of a Human Preclinical Platform for the Identification of Neuroprotective Compounds. European Journal of Neuroscience, 62(10), e70328. https://doi.org/10.1111/ejn.70328
Fischer, K., Thewes, L., Prozorovski, T., et al., Aktas, O., & Berndt, C. (2025). Fumarate-based drugs protect against neuroinflammation via upregulation of anti-ferroptotic pathways. Journal of Neuroinflammation, 22(1), 241. https://doi.org/10.1186/s12974-025-03592-3
de Oliveira, M. R. (2026). Dimethyl fumarate and mitochondrial physiology: implications for neurological disorders. Frontiers in Pharmacology, 17, 1748360. https://doi.org/10.3389/fphar.2026.1748360