4D marmoset brain map reveals MRI and molecular signatures for onset of multiple sclerosis–like lesions [spatial transcriptomes]

Abstract: The cellular and molecular dynamics of multiple sclerosis (MS) lesion development and repair cannot be inferred through study of postmortem tissue harvested decades after lesion onset. Employing MRI-informed spatiotemporal RNA profiling and experimental autoimmune encephalitis (EAE) in the common marmoset — a faithful model for these processes — we identified five microenvironmental groups pertinent to neural function, immune and glial responses, tissue destruction and repair, and regulatory networks at brain borders during lesion evolution. Before visible demyelination, astrocytic and ependymal secretory signals delineated perivascular and periventricular regions, later becoming demyelination hotspots where MS-like lesions frequently grow. An MRI biomarker, the ratio of proton density-weighted signal to T1 relaxation time, captured the hypercellularity phase before apparent myelin destruction. At lesion onset, there was a global shift in cellular connectivity, especially in extracellular matrix-mediated signaling. Early responses included microglia and oligodendrocyte precursor cell proliferation and diversification, replaced by monocyte derivatives and lingering lymphocytes as lesions aged. Concurrent with demyelination, reparative signaling modules were found at the lesion edge as early as 10 days post-lesion establishment. Over-representation of senescence-associated secretory phenotype at the brain borders and the formation of concentric glial barriers at the lesion edge prompted model-based identification of potential treatments to reverse EAE-associated molecular pathology. Our study highlights the marmoset platform's potential for assessing MS treatment outcomes. From 11 marmosets, we derived a transcriptomic map with spatial and single-nucleus resolution and studied the heterogeneity of tissue damage at an individual lesion level. We developed a 10x Visium-compatible staining protocol to quantitatively integrate 16 regions of interest (ROI) and spatiotemporally dissect the significance of focal and global molecular changes as lesions evolve. Mapping the temporal sequence of cellular and molecular changes during lesion evolution, we retrospectively estimated lesion age through serial MRI (Methods). To pinpoint the cellular source of spatially resolved signals, we integrated single-nucleus RNA-seq (snRNA-seq) libraries (n = 43) from 8 animals with spatial transcriptomes (ST) (n = 16 ROI) from 6 animals. WM from healthy control (n = 13) and WM with T2-hyperintense demyelinated MRI features from EAE animals (n = 14) were the most extensively sampled groups. We used additional categories, including normal-appearing WM from EAE animals (n = 2), Gd-enhancing demyelinating WM lesions (n = 2), and resolved WM lesions that were no longer T2-hyperintense on the terminal MRI (n = 3), to group the remaining WM samples. In parallel, we included leukocortical T2-hyperintense lesions (n = 2) along with matching healthy (n = 2) and normal-appearing (n = 1) control tissue, and abnormal-appearing lateral geniculate nucleus tissue on MRI (n = 2) along with matching healthy control tissue (n = 2), to explore tissue-specific and shared responses. We implemented a hierarchical workflow comparing across cell classes (Level 1 analysis, L1) and cell clusters within a class (Level 2 analysis, L2) to better realize the source of each profile change.