Background

Immune cells play an important role in the pathogenesis of MS. However, our knowledge of the diversity of immune cell types and states in health and disease has been limited by the restrictions of traditional oligo-marker immunological approaches. Single-cell RNA sequencing (scRNA-seq) provides new means for discovery and characterization of immune cells by measuring expression levels of thousands of genes simultaneously at the single cell level.

Objectives

Here, we investigated the immune cell composition of cerebrospinal fluid (CSF), as the most accessible immune compartment in contact with the central nervous system, using scRNA-seq in MS and non-MS neuroinflammatory conditions (ONID).

Methods

Fresh CSF samples were collected at the time of diagnostic lumbar puncture from 8 untreated RRMS and 7 ONID patients. In addition, to examine the context of the reconstituted CSF after ocrelizumab treatment, we sampled 6 progressive MS patients on ocrelizumab for >1 year, 5 months after their last dose. A replication set of 3 untreated RRMS and 3 ONID are also collected.

Results

After removing doublets, RBCs, and low-quality cells, 59,288 cells were remained for analysis. Data were integrated using canonical correlation analysis. Over-clustering using ‘Seurat’ and iterative merging of clusters with more similar gene expression using ‘SCCAF’ resulted in 13 major clusters: 1 αβ-T cell cluster (consisting of 24 subclusters), 4 other T (NK, NKT, γδ-T), 5 myeloid (microglia-like monocytic, myeloid DC, granulocytes), B cells, plasmablasts and plasmacytoid DC.

Consistent with earlier smaller studies, we observed subsets of myeloid cells with microglia-like features (3% frequency). Using our reference brain-derived microglia scRNA-seq dataset (221,126 cells), we have characterized the microglial subtypes to which the CSF cells most resemble. Interestingly, these are part of the 4 myeloid subtypes which percentage were reduced in newly diagnosed RRMS relative to ONID. In addition, 3 CD4+ and 1 CD8+ memory T cell subsets were increased in RRMS. If replicated, these intriguing differences could guide our understanding of earliest stages of MS vs. other elements of the differential.

B cell component was largely reconstituted in the ocrelizumab group, and the patients had a CD4+ and CD8+ T cell pattern largely similar to ONID (vs RRMS). However, myeloid cell percentages were increased even more in this group. Further studies are needed to resolve whether these changes are due to progression vs. age vs. ocrelizumab.

Conclusions

Our data offer a new level of resolution in cell population shifts and suggest that measuring certain combinations of subsets could be useful clinically. In addition to these results, we will present on gene-level expression differences and a first draft of a CSF molecular network map across the 3 conditions.