Multiple sclerosis (MS) is a chronic disease of the central nervous system. The occurrence of MS is a phased process while its cause is still unclear. Here, by combining white matter single-nucleus transcriptomic datasets from MS and control samples, we found molecular crosstalk between oligodendrocytes (OLs) and immune cells involved in MS pathology. Using a machine learning approach, we identified oligodendrocyte precursor cells (OPCs) and OL subtypes at various developmental stages. We highlighted their unique molecular characteristics and analyzed their distribution throughout development, adulthood, and in different regions impacted by MS. We also found an increased number of Pre-OPCs and newly formed oligodendrocytes (NFOLs) in normal appearing white matter (NAWM), which were scarcely detected in MS lesions. By cell communication analysis and in vitro coculture, we found the interaction between SIRPA on microglia and CD47 on stressed oligodendrocytes was significantly reduced in MS lesions compared with NAWM, potentially preventing microglial phagocytosis of OLs. In contrast, CD74-MIF signaling between microglia and OLs was increased in lesions, which may lead to their retention around OLs.