An MRI Rim, Single-Nucleus Profiles, and Lipid-Iron Metabolism in MS

Some multiple sclerosis (MS) lesions never resolve in the months after they form. They keep demyelinating and degenerating axons at their edge, and they can be found in living patients by the paramagnetic rim they show on MRI, a signal attributed to phagocytosed iron. That rim gives a rare thing in MS research, an in vivo marker of a specific pathological state, and Absinta and colleagues used it to decide where to cut. They sampled the edge of demyelinated white matter lesions at different stages of inflammation and ran single-nucleus RNA sequencing on them, yielding 66,432 transcriptomic profiles across 29,432 detected genes. Out of that came two transcriptional phenotypes they name microglia inflamed in MS and astrocytes inflamed in MS, a complement-mediated link between them, and three separate lines of evidence for complement component 1q as the mediator: human germline variation, a microglia-specific conditional knockout in mice, and antibody blockade.

Sampling Driven by the Imaging Phenotype
Tissue came from the Netherlands Brain Bank, five patients with progressive MS and three age- and sex-matched non-affected, non-dementia controls, with post-mortem interval held under 12 hours. Each patient contributed two blocks, one chronic active lesion and one chronic inactive, which keeps the comparison within patients rather than between them. Sampling covered the lesion edge at various inflammatory stages, the demyelinated core, the periplaque white matter and control white matter. After filtering, 66,432 of 72,959 nuclei survived quality control, unsupervised clustering returned 18 initial clusters labelled on known lineage markers, and the resulting populations were cross-checked in tissue with multiplexed immunofluorescence of 26 primary antibodies on sections of the same lesions. Composition shifted sharply by site: control white matter ran 81% oligodendrocytes and periplaque 89%, while the chronic active edge dropped to 60% with immune cells at 15% and astrocytes at 19% (Fisher's exact test, P = 0.002), the chronic inactive edge sat at 77%, and the hypocellular core was almost devoid of oligodendrocytes at 1% of all nuclei.

Two Phenotypes Defined by Programme Rather Than Lineage
The microglial profile at the chronic active edge, carrying TREM2, APOE, LPL, CD68, CD9, CD74, GRN, TYROBP and the C1q genes C1QA, C1QB and C1QC, partially overlaps what has been described as disease-associated microglia in other neurodegenerative conditions, which is the basis for the authors' suggestion that primary and secondary neurodegeneration share mechanisms. In MS tissue these cells split into two subclusters with distinct effector functions, a split confirmed by reanalysing raw data from two published datasets covering 3,527 of 41,998 and 1,554 of 14,763 immune cell nuclei. One subcluster is enriched for foam-cell differentiation and lipid storage, response to lipoprotein particles, lysosome and regulation of inflammatory response, consistent with myelin phagocytosis and clearance. The other carries 63 ribosomal protein genes among its top 100 differentially expressed genes, along with MHC class II (HLA-DRA, HLA-DPA1, CD74), ferritin (FTL, FTH1), Fcγ receptors (FCGRT, FCGR2A, FCGR3A), the C1 complex, and the highest IL1B expression of any immune population (z-score 1.96), an iron-handling, antigen-presenting profile that matches the iron the MRI rim detects. The astrocyte phenotype upregulates GFAP, APOE, VIM, S100B, SOD1 and the same ferritin genes, with C3 running 1.3 to 2.6 z-scores above other astrocytes.

Where the Lymphocyte Input Comes In
At the chronic active edge, immune cells broke down into activated T cells and plasmablasts or plasma cells at 6% and 0.6% of all immune cells, the inflamed microglia at 25%, monocytes or dendritic cells at 19%, and monocyte-derived and perivascular macrophages at 8%. The lymphocytes were predominantly cytotoxic CD8 T cells marked by CD2, CD8A, CCL5 and SKAP1, with some central and effector memory CD4 T cells and plasmablasts expressing IgG1, IgG3 and IgM, and in tissue sections they sat mostly perivascular. To get from co-occurrence to directed signalling, the authors used NicheNet to prioritize ligands from the immune cells and predict microglial target genes. Those targets turned out to be regulated especially by T cells, by plasmablasts and plasma cells, and by the inflamed microglia themselves in autocrine fashion. Roughly a third of differentially expressed genes in both microglial populations, 106 of 315 in the iron subcluster and 92 of 282 in the foamy one, are regulated by interferon-γ. Interactome analysis of the first 20 significant connections (z-scores 6.2 to 10.8) put the two inflamed phenotypes at a central hub wired to every other immune and glial cluster at the chronic active edge, against markedly fewer connections at the inactive edge and core, and weighted gene co-expression network analysis confirmed the module-level interaction.

Complement Closes the Loop Between Them
The expression pattern of the complement genes maps the microglia-to-astrocyte link in both directions. C1QA, C1QB, C1QC and CFD, which encodes a C3 activator, are expressed mainly by the inflamed microglia. The inflamed astrocytes show around 20-fold upregulation, relative to non-reactive astrocytes, of genes encoding C1q-complex activators C1S and C1R and the C1q receptors CALR and C1QBP, along with upregulation of C3. The microglia in turn express the C3 receptor C3AR. In tissue, measuring the C1q and C3d staining area across sites found both components highest at the chronic active edge (ANOVA, P = 0.0003 and P = 0.0009).

Germline Variation and a Conditional Knockout
Two genetic tests follow. On the human side, starting from previous MS genome-wide association results, the authors genotyped the cumulative presence of complement risk variants, one in C1QA, one in CR1 and three in C3, across 1,272 patients with MS (932 women, median 3 of 10 risk variant alleles), 105 of whom were classified by 3 T MRI for the in vivo presence of chronic active lesions. Cases carrying more than four paramagnetic rim lesions, an already identified clinically relevant threshold, had more complement-associated risk variants than those with fewer or none (ANOVA, P = 0.04; Tukey's multiple comparison test, P = 0.035). On the mouse side, C1q was conditionally ablated in microglia using a TMEM119-Cre driver crossed to a floxed C1qa allele, with littermate controls injected with oil or tamoxifen-treated floxed animals showing no difference between treatments. Ablation removed the vast majority of C1q immunoreactivity and mRNA from the brain, consistent with microglia being the primary CNS source, while C1q remained in peripheral infiltrating cells and liver mRNA was unchanged. In experimental autoimmune encephalomyelitis, the raised density of IBA1-positive cells, the hypertrophic and amoeboid morphologies of reactive gliosis, and expression of the disease-associated microglia marker CLEC7A were all largely attenuated in the knockout, with no change in clinical onset or score, which the authors attribute to spinal cord pathology their brain-focused analysis did not cover.

Antibody Blockade, a Trial Readout, and the Limits
The pharmacological test used a C1q-blocking antibody on reactive gliosis in neighbouring hippocampal white matter at a later disease timepoint. It reduced C1q staining intensity in the brain, and at day 42 after induction both the number of IBA1-positive cells and the proportion co-expressing FTL were significantly reduced (P < 0.05). The authors are careful that this mechanism differs from the one described for grey matter damage, which primarily involves C3 in synaptic targeting, selective pruning and circuit remodelling. For trial design they turned back to the imaging marker: among 10 patients followed yearly by 7 tesla MRI over a mean of 6.1 years, of 70 rim lesions evaluated longitudinally 27 (39%) stayed stable, 33 (47%) faded and 10 (14%) disappeared while the lesions themselves remained clearly visible on T1 and T2, giving an estimated median rim survival of roughly 7 years. A power analysis put a proof-of-concept trial at 16 patients per arm, 112 rim lesions in total, to detect a 10% difference in the proportion of fading or disappeared rims over one year at 80% power. Their limitations are stated plainly. Larger cohorts are needed to substantiate the findings on rim resolution and on the relevance of complement-gene variants. Tissue selection prioritized site and pathological stage and restricted the age range to limit the confounding effects of aging on glial subpopulations, which held down the number of autopsy cases, though reanalysis of previously published datasets supports the generalizability of the main results. Single-nucleus sequencing evaluates only pre-mRNA nuclear transcripts. And lymphocyte clonality analysis was not possible given how few of those cells were recovered.

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:
Absinta, M., Maric, D., Gharagozloo, M., Garton, T., Smith, M. D., Jin, J., Fitzgerald, K. C., Song, A., Liu, P., Lin, J.-P., Wu, T., Johnson, K. R., McGavern, D. B., Schafer, D. P., Calabresi, P. A., & Reich, D. S. (2021). A lymphocyte–microglia–astrocyte axis in chronic active multiple sclerosis. Nature, 597, 709–714. https://doi.org/10.1038/s41586-021-03892-7