Characterising the effect of therapeutic intervention to discover pathogenic immune interactions in multiple sclerosis

Thesis

Multiple sclerosis (MS) is a chronic inflammatory disease of the central nervous system (CNS), in which current therapies incompletely address the mechanisms underlying sustained disability progression. A major unresolved question is how peripheral immune dysregulation relates to compartmentalised inflammatory processes within the CNS over the disease course. In this thesis, I used high-efficacy therapies as in vivo perturbational probes to study pathogenic immune interactions in MS.

To address the challenge of resolving subtle treatment-related shifts in continuous cellular phenotypes, I developed an analytical pipeline for multimodal single-cell data integrating manifold-based perturbation analysis, latent transcriptional program inference, and explainable machine learning. This enabled robust identification of treatment-perturbed immune cell states and their quantification in orthogonal datasets spanning tissue compartments, disease stages, and therapeutic contexts.

By applying this pipeline to longitudinal peripheral blood profiling during B cell depletion with ocrelizumab, I identified broad reprogramming of the immune landscape, including secondary depletion of a distinct proinflammatory, cytotoxic, tissue-homing memory T cell phenotype (Tc1-TH). This cell state was enriched in the blood of donors with highly active relapsing MS and was not reduced in a treatment non-responder with persistent breakthrough disease.

Across independent public datasets, the same transcriptional signature was enriched in T cells from cerebrospinal fluid and at the leading edge of chronic MS brain lesions, supporting a link between peripheral immune activation and compartmentalised CNS inflammation. Complementary immune profiling during natalizumab treatment demonstrated enrichment of circulating memory B cells and Tc1-TH cells during integrin 𝛼4-mediated migration blockade, providing convergent evidence for a common pathogenic axis disrupted by two mechanistically distinct high-efficacy therapies.

Collectively, this work supports a model of a pathogenic B–T cell axis in MS across tissue compartments and disease stages, centred on interactions between memory B cells and a chronically activated, CNS-homing cytotoxic T cell state. Further work to delineate the precise molecular pathways mediating these cellular interactions within chronically inflamed tissue niches may provide a framework for the rational development of more efficacious, safer, and durable therapies for MS.

Authors: Farooq, R

• DOI: 10.5287/ora-dazxvjvpj
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: machine learning; single-cell transcriptomics; T cells; B cells; neuroimmunology
• Subjects: Immunology; Computational biology; Multiple sclerosis; Neurology