More Enzyme, Less Branching, and More MS Risk
Before genome-wide association studies arrived, candidate gene studies in multiple sclerosis (MS) had a poor record, and Mkhikian and colleagues name the reasons: very small sample sizes, no confirmation cohorts, candidate genes chosen without evidence that they regulate spontaneous disease in animals, and association testing done before any functional work. Their response was to invert the order. They ignored association testing at the outset, ran a detailed biological screen for functional variants that produce disease phenotypes in mice, selected only the single variant with the greatest biological effect, and tested association in large cohorts last. What they were testing is whether four separate MS risk factors, two cytokine receptor variants, a Golgi enzyme haplotype and a vitamin, all act on the same biochemical step. This post follows the work in the order they ran it.
The Pathway and the Measurement
N-glycan branching is carried out sequentially by the Golgi enzymes Mgat1, Mgat2, Mgat4 and Mgat5, all drawing on the same substrate, UDP-GlcNAc, supplied by the hexosamine pathway. The number of branches per protein controls binding to galectins, which form a molecular lattice at the cell membrane that governs how long glycoproteins stay there before being taken back inside. The measurement throughout is L-PHA, a plant lectin that binds specifically to the β1,6GlcNAc-branched structures Mgat5 produces, read by flow cytometry. One number sets the scale for everything after it: a change of 15 to 20% in L-PHA binding is enough to alter T cell function, growth and autoimmunity, so shifts that look modest are not.
The Cytokine Side, and a Direction That Looks Backwards
Adding interleukin-2 or interleukin-7 to resting human T cells reduced branching by 20 to 30%, while mouse T cells deficient in IL-2 or IL-7RA showed roughly 50% and more than two-fold increases, confirming that autocrine signalling regulates branching in vivo. The transcriptional explanation initially reads the wrong way round: both cytokines induce about a two-fold rise in MGAT1 and 1.5 to 3-fold decreases in MGAT5, MGAT4B, the mannosidase II and IIx genes and the mannosidase Ia and b genes. More of the first branching enzyme, yet less branching. The kinetics resolve it, because Mgat1 has roughly 250-fold higher affinity for UDP-GlcNAc than Mgat5, so raising Mgat1 under limited substrate starves the enzymes downstream of it. The IL2RA*T and IL7RA*C risk alleles, carried by about 75% of European-ancestry populations, raise secretion of soluble receptors that antagonize this signalling, and both soluble receptors reduced MGAT1 messenger RNA and CTLA-4 retention at the cell membrane.
Finding the Enzyme Variant by Screening Rather Than Scanning
Sequencing the MGAT1 coding region in 42 MS patients turned up a high frequency of variants, which were cloned and tested by transient transfection into Mgat1-deficient Chinese hamster ovary cells. The haplotype that emerged carries three polymorphisms in exon 2: one non-synonymous change encoding Arg223Gln and two synonymous changes ten nucleotides apart. Arg223Gln sits on the outside of the enzyme, far from the active site, and so is unlikely to alter enzyme activity, which pushes the explanation toward expression. That is what the data show. The haplotypes raised Mgat1 activity about two-, two- and three-fold, and in peripheral blood cells from heterozygous donors the haplotype's messenger RNA ran roughly two-fold above the common one while the genomic DNA stayed balanced, so the difference is allelic expression rather than copy number. Mass spectrometry confirmed the Mgat1 product raised relative to its substrate in cells carrying the haplotype.
Why the Same Variant Can Go Either Way
The enzyme kinetics make the direction of effect conditional rather than fixed. The Michaelis constant of Mgat5 for UDP-GlcNAc is much higher than that of Mgat1, around 11 mM against 0.04 mM, while for their respective N-glycan substrates the relationship reverses, around 0.08 mM against 2 mM. Under basal UDP-GlcNAc, raising Mgat1 lowers branching by outcompeting Mgat5 for the shared substrate. Raise UDP-GlcNAc through GlcNAc supplementation, and the extra product that the more active Mgat1 generates lets Mgat5 first draw level and then overshoot the common haplotype. Ordinary differential equation modelling predicted this pattern, and GlcNAc titration in transfected cells and in CD4+ T cells confirmed it, with the size of the rescue proportional to baseline enzyme activity.
Vitamin D, and the Experiment That Makes It Mechanistic
The active metabolite 1,25-dihydroxyvitamin D₃ upregulated MGAT1 messenger RNA in human CD4+CD25+ T cell blasts, matching the direction of the MGAT1 haplotype and opposing that of the IL2RA*T and IL7RA*C risk alleles. In mice, lowering dietary vitamin D₃ reduced branching in a dose- and time-dependent way, while intraperitoneal injection raised it in CD4+, CD8+ and B220+ cells and had little effect on macrophages, dendritic cells or natural killer cells. The experiment that turns this from correlation into mechanism is a blocking one. Vitamin D₃ inhibited myelin basic protein-induced experimental autoimmune encephalomyelitis in female PL/J mice in the absence of swainsonine in the drinking water, and failed to do so in its presence, where swainsonine blocks branching downstream of Mgat1 and cut CD4+ T cell branching by about half. Swainsonine likewise reversed the vitamin D-induced hypoproliferation of T cells from immunized mice. The protection therefore runs through branching rather than alongside it.
The Association Testing, Run Last
Only then did the authors go to cohorts: two European-ancestry MS case-control sets of 2,705 and 9,132 individuals, combining to 3,280 cases and 8,557 controls, plus an African-American cohort of 758 cases and 480 controls. On point association, IL2RA rs2104286 reproduced at P = 2.44×10⁻⁶ (OR 1.18) and IL7RA rs6897932 at P = 1.05×10⁻³ (OR 1.12), and the MGAT1 haplotype, carried by only about 7.2% of people, associated at P = 1.51×10⁻⁵ (OR 1.36). CTLA-4 Thr17Ala showed no point association in either cohort. The interactions are where the design earns its keep. The MGAT1 haplotype raised MS risk when fewer than four IL2RA*T and IL7RA*C risk alleles were present (P = 3.60×10⁻⁵, OR 1.45) and did nothing at four copies (P = 0.195), giving an interaction P of 6.75×10⁻². Stratified on CTLA-4, the haplotype associated in Ala17 carriers (P = 4.95×10⁻⁶, OR 1.53) but not in Thr/Thr homozygotes (P = 0.329), with an interaction P of 1.20×10⁻², and the African-American cohort reproduced the same pattern at an interaction P of 2.29×10⁻³, the two combining to 3.16×10⁻⁴. Stratifying across all six alleles gave the sharpest contrast, with the haplotype promoting MS below six copies (P = 1.62×10⁻⁶, OR 1.46) and appearing protective at six (OR 0.64). The authors note that this interaction holds despite CTLA-4 Thr17Ala having no point association and only marginal effects on its own, which is what makes it epistatic rather than additive. They close the methodological argument with a number: the P values they report, in the range of 10⁻⁵ to 10⁻⁷, are orders of magnitude better than a genome-wide significance of 5×10⁻⁸, which corrects to roughly 0.02 after adjustment for multiple testing in a typical GWAS of 550,000 markers.
Disclaimer: This blog post is based on the cited study 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:
Mkhikian, H., Grigorian, A., Li, C. F., Chen, H.-L., Newton, B., Zhou, R. W., Beeton, C., Torossian, S., Tatarian, G. G., Lee, S.-U., Lau, K., Walker, E., Siminovitch, K. A., Chandy, K. G., Yu, Z., Dennis, J. W., & Demetriou, M. (2011). Genetics and the environment converge to dysregulate N-glycosylation in multiple sclerosis. Nature Communications, 2, 334. https://doi.org/10.1038/ncomms1333
