Two Approaches to Iron in MS
Iron sits awkwardly in multiple sclerosis (MS) research. Imaging studies show excessive iron deposited in the grey matter and basal ganglia of patients, which points one way; studies of serum iron in patients find levels similar to or lower than in healthy people, which points the other. Tang and colleagues approached that tension from two directions at once. First they used expression data to find iron metabolism genes altered in MS, drawing on a dataset of resting and activated CD4⁺ T cells from 28 patients and 28 healthy individuals, with two further datasets held back for validation, and a reference set of roughly 520 iron metabolism genes. Then they ran Mendelian randomization on four serum iron markers, instrumented from a study of 23,986 individuals across 11 cohorts, to ask whether any of them causally affects MS risk. Reading the paper well means understanding what each half establishes on its own.
The Gene Side, and a Model That Travels
Taking the intersection of differentially expressed genes, genes from weighted correlation network analysis, and the iron metabolism reference set produced eight hub genes: IREB2, LAMP2, ISCU, CDK5RAP1, ATP6V1G1, DCUN1D1, ATP13A2 and SKP1. Gene ontology analysis put six of the eight in processes directly concerning iron: cellular iron ion homeostasis, iron ion homeostasis and iron ion transport. Most were expressed more highly in the MS group, with ATP13A2 the exception, higher in controls. A logistic regression model built from these genes reached an area under the curve of 0.83 in the discovery dataset, and then 0.98 and 0.99 in two independent blood datasets. That external performance is the most convincing part of this half, because a model that holds up in data it was not built on is doing more than fitting noise.
The Causal Side: Two Signals and Two Nulls
The Mendelian randomization tested four exposures. Transferrin came out associated with higher MS risk, at an odds ratio of 1.22 (95% CI 1.10–1.36, P = 2.18×10⁻⁴). Transferrin saturation went the other way, at 0.86 (0.75–0.98, P = 2.22×10⁻²). Serum iron at 0.96 (0.79–1.17) and ferritin at 1.04 (0.70–1.55) showed nothing. The two positive results agree with each other once the biology is unpacked: transferrin is the transport protein and rises when iron is scarce, while transferrin saturation is the fraction of it carrying iron and falls in the same situation. Both therefore say that markers of low iron availability track with higher MS risk. Reverse Mendelian randomization, running MS as the exposure against each iron marker, was null throughout, which supports reading the direction as the authors do.
Why the Two Null Results Shape the Interpretation
Ferritin is the standard clinical measure of body iron stores and serum iron the most direct measure of circulating iron, and neither shows an effect. The two markers that do are the transport protein and a ratio derived from it. That pattern is compatible with a real effect on how iron is handled and moved rather than on how much of it there is, which is a more specific claim than iron deficiency in general, and a more interesting one given the role the authors set out for transferrin: it carries iron across the blood-brain barrier, and helps oligodendrocytes, the main iron-containing cells in the brain, acquire iron from their surroundings and regulate how it is used and stored. The authors also deserve credit for not simplifying the wider picture. Their introduction reviews the evidence for iron excess and deposition in MS brains at length, and the discussion works through both excess and deficiency rather than keeping only the half that fits their result.
The Sensitivity Testing Is the Full Battery
Five estimation methods were used, with inverse-variance weighting as the primary one alongside the ratio test, weighted generalized linear regression, weighted median and MR-Egger. The diagnostics cover everything a reader would want to see. Cochran's Q tested heterogeneity, returning P values from 0.24 to 0.76. MR-Egger intercepts tested directional pleiotropy, coming in between 0.00 and 0.04 with P values from 0.20 to 0.93. MR-PRESSO tested horizontal pleiotropy, from 0.07 to 0.80. Leave-one-out analysis confirmed no single variant drove any estimate, and Radial MR found no outliers for transferrin saturation or transferrin. Every one of these passed, and reporting the full table rather than a summary sentence is what lets a reader confirm it.
How Iron Could Be Acting
The discussion sets out several routes, which matters because a causal estimate without a mechanism is hard to use. Iron promotes the differentiation and function of pathogenic T lymphocytes, with iron-dependent production of granulocyte-macrophage colony-stimulating factor linked to the binding of Fe²⁺ and the stabilization of Poly(rC) Binding Protein 1, pushing T cells toward a more inflammatory phenotype. Iron also drives glycolysis and oxidative phosphorylation in T cells, giving them more pathogenic metabolic characteristics. In the other direction, excess iron can trigger ferroptosis, a form of cell death distinct from apoptosis that begins with lipid peroxidation, and oligodendrocytes are especially vulnerable to it because they carry high iron alongside low levels of the antioxidant enzyme glutathione. The gene analysis points the same way: enrichment analysis linked the hub genes to other neurodegenerative diseases including Huntington's and amyotrophic lateral sclerosis, and to MS-related pathways including glycosphingolipid biosynthesis, B cell receptor signalling and Toll-like receptor signalling.
The Gap the Authors Name, and How to Close It
An important consideration for interpreting the findings is noted in the paper’s limitations section. Because genome-wide data were not available for the iron-related hub genes, the authors could not use Mendelian randomization to test whether those genes relate causally to MS. That is worth sitting with, because it defines what the study is. The two halves both concern iron, but they do not validate each other: the gene analysis finds expression differences in CD4⁺ T cells, the causal analysis tests serum protein markers, and neither speaks directly to the other. Read as one integrated argument the paper promises more than it delivers; read as two parallel lines of evidence converging on iron, it is a useful contribution with a clearly marked next step. Their second stated limitation points the same way, that the data come from European population databases and may not generalize. The next step is specific. Expression quantitative trait data for IREB2, ISCU and the other hub genes would let the same framework test the genes themselves, and cell-type-specific data would show whether the T cell signal and the serum signal are describing one process or two.
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:
Tang, C., Yang, J., Zhu, C., Ding, Y., Yang, S., Xu, B., & He, D. (2024). Iron metabolism disorder and multiple sclerosis: a comprehensive analysis. Frontiers in Immunology, 15, 1376838. https://doi.org/10.3389/fimmu.2024.1376838
