In MS brains, B cells are concentrated in the perivascular space of one or a few larger veins within the plaque centers and are rarely found around small vessels or diffusely in the lesion parenchyma. The number of B cells in lesions varies greatly, ranging from none in some progressive MS patients to four times the number of T cells in others (23). As mentioned before, proliferating B cells are found in intrameningeal follicles with a predominance of germinal-center-like B cell clusters (24), while nonproliferating plasma cells (CD138+Ki67−) are present in brain tissue (25). Related B cell clones are found in both the meninges and parenchyma of patients with MS (26).

Lymphocytic perivascular cuffs are most prominent at the periphery of active plaques and are occasionally observed in regions without evidence of demyelination or macrophage infiltration. These cuffs predominantly comprise T cells and MHC-II+ cells, with a high presence of CD8+ cells and variable numbers of CD4+ cells. The CD8/CD4 ratios within the cuffs range from 1:1 to 50:1. In normal-appearing white matter, cuffs are sparse and mainly consisted of CD8+ T cells. The distribution of T cells in the parenchyma mirrors that in the perivascular cuffs, being predominantly CD8+ cells with variable numbers of CD4+ cells (27). Despite CD4+ T cells generally being outnumbered by CD8+ T cells in brain lesions (23, 27–30), their role in triggering local pathology in MS cannot be ruled out. A significant presence of CD4+ T cells in pre-active lesion sites suggests their involvement in the early stages of lesion formation (28). However, MS donors with a low CD8/CD4 ratio in different brain lesions exhibit a higher proportion of inactive remyelinated areas compared to those with a high CD8/CD4 ratio (29).

Parenchymal infiltration is observed in active and mixed active/inactive lesions, but not in inactive lesions (29). Further characterization revealed that these cells are enriched for CD20^dim tissue-resident memory (TRM) CD8 T cells, which are often CD103-negative (30). These TRM CD8 T cells are reactivated in acute, relapsing, and progressive MS lesions (23). It has been shown that TRM CD8 T cells expressing granzyme B, but lacking perforin, can still kill neurons in MS (23, 31).

Recent epidemiological studies have provided strong evidence that MS is an uncommon complication of infection with the Epstein-Barr virus (EBV), a herpesvirus that infects more than 90% of the population (32). Other studies have demonstrated a strong association between EBV infection and MS, with CD8+ T cells (33, 34), particularly CD8 TRM (35), playing a significant role in this process. Analysis of postmortem MS brain samples revealed the expression of the EBV lytic protein BZLF-1 and interactions between cytotoxic CD8+ T cells and EBV-infected plasma cells in inflammatory lesions, suggesting that failure to control EBV infection could lead to intracerebral viral reactivation and disease relapse. These studies indicate that persistent EBV infection in the CNS stimulates a CD8+ T-cell response aimed at clearing the virus, but this response inadvertently causes CNS injury. In MS brain tissue, CD8+ T cells recognizing EBV proteins were found in white matter lesions and meninges (36, 37). Interestingly, clonally expanded CD8+ T cells reactive to autologous EBV-infected B cells are found forming immune synapses in the brain parenchyma of MS donors (34). The presence of these expanded CD8+ T cells can be detected in the CSF even at the earliest stages of MS (38, 39), suggesting that they likely play an important role in pathogenesis. The expression of immune checkpoint molecules like PD-L1 on EBV-infected B cells and PD-1 on CD8+ T cells may facilitate local immune evasion, leading to EBV persistence and ongoing inflammation in the MS brain (40).

scRNA-seq data suggest that brain macrophages exhibit an early stress response in normal-appearing white matter, indicating they might be among the first cells involved in the onset of demyelination (41). However, the exact role of this stress response is not well understood. It could also be a protective mechanism by these cells, aiming to prevent lesion formation in the affected tissue area. Similarly, microglia in normal-appearing white matter exhibit an alert state while displaying features of immunosuppression, suggesting they are responding to ongoing neuroinflammation (42). Therefore, it is not clear whether macrophages/microglia have a deleterious role, as previously suggested. In a mouse model, it has been shown that both microglia and monocyte-derived macrophages engulf equal amounts of myelin debris following demyelination. Furthermore, these macrophages can compensate for microglia depletion by enhancing their phagocytic activity. Importantly, microglia have been demonstrated to promote the recruitment of oligodendrocyte progenitor cells, facilitating their differentiation and subsequent remyelination (43).

Single-cell mass cytometry profiling revealed a periventricular accumulation of both CD8+ and CD4+ T cells as well as CD56^bright natural killer (NK) cells in multiple sclerosis. CD56^bright NK cells expressed higher levels of proteins involved in NK cell activation and migration suggesting a role in MS (44).

The origin of immune cells present in the CSF can be either peripheral or derived from nervous tissue. A study investigated whether human CSF contains TRM cells and CNS-resident myeloid cells from the parenchyma or border tissues (45). By comparing CSF immune cells from multiple sclerosis relapse with those collected during therapeutic very late antigen-4 blockade, researchers identified immune subsets from various lineages, including CNS border-associated macrophages, CD8 and CD4 TRM cells, and tissue-resident NK cells. All lymphocytic CNS-resident cells expressed CXCR6 but differed in ITGAE expression (encoding CD103). CD4 and CD8 TRM cells shared a common signature with ZFP36L2, DUSP1, and ID2 expression. The researchers also developed a user-friendly application to facilitate the transfer of cell identities from a reference atlas to new CSF scRNA-seq data (45).

In an animal model of MS, T cells from different peripheral sites exhibit distinct signatures based on their origin. Site-specific labeling revealed unique characteristics of inguinal and mesenteric T cells in the inflamed CNS, though these signatures were not clearly extrapolated to CD4+ subsets in MS CSF cells (46).

The most consistent findings regarding cell type composition among various scRNA-seq studies on the CSF of MS patients is the increase in antibody-secreting cells (ASC) (plasmablasts and plasma cells) and memory B cells (22, 45, 47–52), the expansion of B-cell-helping T follicular helper (Tfh) cells (22, 45, 48, 49, 52, 53), increased proportions of regulatory T cells (Treg) (45, 48, 49), and higher numbers of NK cells (22, 45, 47, 49).

B and plasma cells are found to be enriched in MS patients who are oligoclonal band-positive and those with active disease (51). A study integrating multiple scRNA-seq datasets of CSF cells from patients with early RRMS demonstrates that the expanded CSF B lineage cells resembled class-switched ASC (47). Interestingly, a study that performed single-cell full-length RNA-seq and BCR reconstruction in the CSF of MS patients demonstrated an association between IgG constant region polymorphisms and stereotyped B-cell responses in MS. This suggests that the intrathecal B-cell response in these patients may target structurally similar epitopes (54). Regarding NK, an increase in absolute NK cell counts has been reported in flow cytometry settings and is associated with MS activity. MS patients showed a significant increase in the regulatory/effector (CD56^bright/CD56^dim) NK cell ratio compared to other inflammatory neurological diseases (OIND) and non-inflammatory neurological diseases (NIND) groups, suggesting that regulatory NK cells are attempting to counteract adaptive immune activation (55). An increase in both CD56^bright and CD56^dim NK cells was observed in MS after scRNAseq analysis (49), but the exact role of the different NK cell subsets still needs to be clarified.

A study including two patients with RRMS and one with anti-MOG disorder, identified cells exhibiting a transcriptomic signature akin to microglia. These microglia-like cells were distinguishable from other myeloid cell populations through flow cytometry. The presence of microglia within the human CSF, detectable by surface protein expression, underscores their potential role in neuroinflammation (56). However, these microglia-like cells are also found in non-inflammatory conditions (22).

Cantoni and collaborators study included subjects with RRMS and clinically isolated syndrome, highlighting MS as a highly compartmentalized immune disorder characterized by intrathecal B cell expansion, altered T cell dynamics, and innate immune involvement ( Table 6 ) (22). MS patients exhibited a significant increase in Tfh cells in the CSF, supporting B cell activation and antibody production, along with a decrease in Th1 cells in the CSF and a marked increase in Th17 cells in the blood. Additionally, there was a higher frequency of naive CD8+ T cells in the CSF, suggesting recruitment from the periphery, while effector memory CD8+ GZMB+ T cells were absent, indicating a different mechanism of immune activation compared to neurodegenerative diseases. MS patients also showed a reduction in CD16+ monocytes (non-classical) and microglia-like cells, distinguishing MS from other inflammatory conditions, but had significantly increased AREG+ cDC2 DC in the CSF, supporting their role in disease progression. NK cell composition was altered, with a decrease in cytotoxic NK cells (CD56d^imLAG3–) and an increase in tissue-resident NK cells (CXCR6+ ITGAE+), suggesting selective retention within the CNS. Furthermore, MS was marked by a dramatic expansion of atypical memory B cells and plasmablasts in the CSF, with CD11c+ age-associated B cells indicating a chronic inflammatory state, reinforcing the dominant role of B cells in MS pathogenesis.

An increase of DC has also been reported by other single-cell studies (49). One of the studies revealed that AXL+SIGLEC6+ DC (ASDCs) are particularly overrepresented in the CSF of patients with inflammatory demyelinating disease, including antibody-mediated disease and MS, compared to controls (53). ASDCs were also found in brain tissues during acute attacks, often in close contact with T cells, suggesting a role in CNS autoimmunity. However, given that the controls consist of samples from only one peripheral neuropathy and one degenerative neurological disease, it is advisable to confirm the results with more appropriate and comprehensive controls to ensure their validity. Another study reported enrichment with a rare subtype of CD8+ T cells in MS compared to other pathologies (51). This subtype is characterized by the expression of cytotoxic markers (GZMA, GZMK), upregulation of the co-stimulatory marker CD27, upregulation of inhibitory receptors (HAVCR2, TIGIT), and downregulation of chemokines (CCL4, CCL5). Regarding monocytes, a recent mass cytometry and scRNA-seq combined analysis demonstrated an enrichment of a classical monocyte population in 22% of MS patients at the time of diagnosis, that predicted a more aggressive progression of the disease (57). This monocyte population is found more frequently in MS patients carrying the HLA-DRB1*15:01 allele both in blood and CSF. It is characterized by CD206, CD209, CCR5 and CCR2 expression and exhibited antigen-presenting abilities along with a pro-inflammatory profile. In addition, another study involving blood and CSF scRNAseq analyses with sorted cells suggests that IL-11 induces NLRP3 inflammasome activation in monocytes and promotes the migration of inflammatory cells to the central nervous system (58).

Cantoni and collaborators identified several similarities between MS and other neuroinflammatory diseases, such as a strong increase in Tfh cells, B cells, and plasmablasts ( Table 6 ). However, the other neuroinflammatory diseases included in the study did not exhibit an expansion of CD8+ T cells, AREG+ cDC2 DC, or a reduction in Th1 cells, CD16+ monocytes, and microglia-like cells. Nonetheless, an overall increase in DC was observed (22). On the other hand, they did not find an increase in Tfh cells in infectious diseases or other alterations observed in MS patients, except for the expansion of plasmablasts. However, they did identify certain exclusive changes in the group of infectious diseases. These results contrast with those reported in another recently published study, although the diseases included in neuroinflammatory, and infectious diseases groups were not the same. This study that integrates single-cell gene expression with lymphocyte receptor sequencing concludes that most of the changes observed in MS are common to other neuroinflammatory diseases (52). Jacobs and collaborators included in the study the largest number of CSF MS samples to date, with 123 cases of untreated MS, 19 patients with OIND, 23 patients with infectious neurological diseases (IDs), and 36 patients with NIND. However, the study did not include a healthy control group and instead used NIND as the reference group. The study shows that clonal expansion of CSF B and T cells is observed across different neurological diseases but is most prominent in inflammatory diseases such as MS. It is important to note that no MS-specific cell subsets were identified in this study. They found that neuroinflammatory diseases, including MS, exhibit an enrichment of most of the cell subsets reported by other studies (52). For example, the rise in Tfh cells was noted across all three inflammatory groups (MS, OIND, and ID), but not in the non-inflammatory controls. They suggest that the presence of these cells in CSF might be characteristic of neuroinflammatory CNS disorders. While MS CSF cellular composition was like OIND CSF, it contained lower relative proportions of NK and Tregs compared with ID CSF. SUB1, which is involved in B cell differentiation into ASC, is the most highly expressed gene in MS and ID. The fact that, for example, the increase in ASC, B cells, and T regulatory cells is common to other neuroinflammatory diseases (51) and inflammatory demyelinating diseases (53) was reported before by other single-cell transcriptomic studies.

In the study by Jacobs and collaborators, an increase in markers of tissue residence, antigen presentation, cytotoxicity, and proliferation were observed across cell types and cohorts (52). Gene Set Enrichment Analysis (GSEA) revealed enrichment of interferon γ (IFN-γ) signaling and complement pathways in expanded clones across various cell types. The genes defining these clones were consistent across cohorts, suggesting a general transcriptional program linked to clonal expansion rather than an MS-specific occurrence. The level of clonality in the CSF T cell compartment showed no significant difference between patients with MS and those with NIND. Interestingly, it was lower in MS patients compared to those in the ID cohort. The TCR repertoire in CSF was polarized toward memory and effector T cell subsets, with clonally expanded T cells present in both inflammatory and non-inflammatory conditions. Clonally expanded cells in the CSF were predominantly of a CD8+ tissue-resident effector memory phenotype and were enriched for EBV-specific and CMV-specific CDR3 sequences in MS cases and NINDS. The authors accurately note that the clonal expansion of T cells recognizing these viral antigens, identified through bioinformatic antigen-prediction tools, may be influenced by the biased nature of public databases where these epitopes are predominant. This bias may not precisely reflect the true proportion of antigens recognized by the clonally expanded T cells.

Jacobs and collaborators conducted cis-expression quantitative trait loci (eQTL) mapping in CSF CD4+ T cells, CD8+ T cells, and B cells focusing on genes that were upregulated in CSF and have been implicated in MS pathogenesis through GWAS (52). They found substantial evidence that the MS risk allele alters the expression of several genes (ORMDL3, ANKRD55, FCRL3, AHI1, EAF2, GDPD5, and ZC2HC1A) in the CSF. They identified CSF-eQTLs, including ETS1 in CD4+ T cells and EAF2 in B cells, suggesting mechanisms for T cell migration to the CNS and B cell proliferation regulation.

A previously published study suggests that alterations in viral control mechanisms may play a crucial role in the development of MS. Altered activity in pathways responsible for controlling inflammatory and type 1 interferon responses in both T cells and myeloid cells was found in MS patients. Additionally, a systematic search for expression eQTLs highlighted two eQTLs in CD8+ T cells, fine-mapped to MS susceptibility variants in the viral control genes ZC3HAV1 (rs10271373) and IFITM2 (rs1059091) (51).

A very promising strategy for studying the mechanisms involved in the onset of MS is to examine identical twins. Identical twins have a 25% risk of developing the disease if their sibling has already been diagnosed with MS (59). Among individuals without MS, some exhibit no signs of brain inflammation, while others have subclinical indicators of inflammation that do not meet the criteria for an MS diagnosis. Those with subclinical indicators may be in a very early, or prodromal, stage of the disease. In these subjects, clonally expanded TRM-like CD8+ cells were found at a higher degree compared to controls, and expanded plasmablasts were observed in both MS and prodromal individuals with oligoclonal bands (60). Single-cell transcriptomic analysis of peripheral CD8+ T cells unveiled disease-associated transcriptional features that persist across both prodromal and clinically manifest stages of MS (61). These findings were substantiated by brain tissue analysis, which underscored the preservation of these immunological and metabolic signatures at sites of neuroinflammation. Using scRNA-seq and scTCR-seq analysis, T cell clonotypes with high migratory potential were found in both blood and CSF. Indeed, it has been previously demonstrated that brain-infiltrating CD8+ T cells persist as clonal expansions in both the CSF and blood (15). These clonotypes were part of T cell subsets expressing gene profiles indicative of potent effector function, activation, and a high capacity for energy production (61). Altogether, this suggests that CD8+ T cells are involved in the initiation and progression of the disease.

Publicly available data not only allows for the reanalysis of scRNA-seq data from different perspectives but has also been proposed as a tool for exploring drug repositioning. To achieve this objective, a study demonstrated that AIM2 inflammasome, SMAD2/3 signaling, and complement activation pathways are activated in MS within various CSF and peripheral blood mononuclear cells (PBMC). By leveraging genes from the most activated pathways, researchers identified several promising small molecules capable of reversing the transcriptomic signatures of MS immune cells. Notably, among these molecules, they detected the MS drug Mitoxantrone, which validates the approach (62).

Anti-CD19/CD20 monoclonal antibody B cell depletion has shown therapeutic efficacy in autoantibody-associated autoimmune diseases, including NMOSD, but a subset of patients remains refractory to current therapies. Thus, chimeric antigen receptor (CAR) T cell therapy is a promising alternative. A single-cell transcriptome, TCR, and BCR data were generated from PBMCs and CSF cells from NMOSD patients treated with BCMA CAR T cells to elucidate the intricacies of this immunotherapy (18). Expanding cytotoxic-like CD8+ CAR T cell clones were pinpointed as the primary agents in autoimmunity. Anti-BCMA CAR T cells with improved chemotaxis successfully traversed the blood-CSF barrier, eradicated plasmablasts and plasma cells in the CSF, and reduced neuroinflammation. The early memory phenotype expressing CD44 in infusion products was potentially linked to CAR T cell persistence in autoimmunity. Additionally, CAR T cells from NMOSD patients exhibited unique characteristics of diminished cytotoxicity compared to those from hematological malignancies, offering mechanistic insights into CAR T cell function in neurological autoimmune diseases. In NMOSD patients, CD8+ CAR T cells exhibited an increased transition to NK-like T cells without significant up-regulation of the exhaustion phenotype. However, high expression of predysfunctional genes in CD8+ CAR T cells in late remission suggests early-stage dysfunctional cells (18).

A flow cytometry-based study revealed an increase in activated CD8+ and CD4+ T cells in the CSF of mild AD and MCI patients, with CD8+ T cell activation linked to neuropsychological deficits and parahippocampal damage (92). No changes in T cell frequencies were noted in AD patients with impaired daily activities, but a reduction in TCR diversity among CD4+ T cells indicated greater clonal expansion (93). CD4+ T cell frequencies decreased significantly in severe AD (94), consistent with observations in aMCI patients with amyloid-beta (Aβ) accumulation (95, 96). Both AD and MCI patients showed a shift from central memory to late-stage effector T cells, with an increase in naive CD8+ T cells (94). Increased CD8+ T cells in aMCI patients with higher Aβ burden suggest a pathogenic role (95). Opposing roles of CD4+ and CD8+ T cells in Aβ deposition were noted. Increased expression of CD139, TREM2, and CD33 in CSF DN T cells suggests an anti-inflammatory effect of these cells in AD (88). Conversely, NKT and NK cells in the CSF were correlated with cognitive impairment and phosphorylated tau levels (94).

Among the various neuronal synucleinopathies, flow cytometry studies have only been conducted with CSF from PD patients. These studies are scarce and have utilized very few markers. Nevertheless, activation of both CD4+ T cells and CD8+ T cells has been demonstrated, with the latter showing a higher proportion of activated cells (HLA-DR+) (107). An increase in γδT cells has also been reported, although it was not determined whether these were DN or CD8+ T cells (108). γδT cells are one of the cellular subtypes that do not manifest according to the clustering and annotation methods employed to identify different cellular subtypes. In this context, the original scRNA-seq study which included cases of LBD, did not annotate this cellular subtype (77). A CSF flow cytometry study, reported a shift from classical monocytes (CD14+/CD16−) to non-classical monocytes (CD14+/CD16+) (107). Additionally, another study found an increase in the frequencies of both populations (109). No differences were observed in the proportions or absolute numbers of granulocytes and B cells (107).

Cantoni and collaborators reanalyzed data from AD, PD, and dementia with DLB, along with MCI as a prodromal stage of AD, grouping them collectively as neurodegenerative diseases ( Table 6 ) (22). These conditions share common neuroinflammatory features but exhibit distinct immune alterations. Neurodegenerative diseases are characterized by an increase in Th1 cells in the CSF, suggesting a persistent inflammatory state. Additionally, there is an enrichment of CD8+ effector memory T cells expressing granzyme B in the CSF. These cytotoxic T cells may contribute to neuronal death as some studies with postmortem brain tissue suggest (99, 102). A notable feature of neurodegenerative diseases is the increase in FN1+ and SPP1+ microglia-like cells, which express risk genes associated with AD and PD. Unlike MS and other inflammatory diseases, CD16+ monocytes are not elevated, suggesting that inflammation in neurodegenerative diseases is primarily driven by resident immune cells. While B cells are not significantly expanded, some patients with AD and DLB exhibit a mild increase in plasmablasts, hinting at a subtle humoral immune component (22).

Single-cell transcriptomics can also be useful for studying treatment responses. Investigation of genomic and transcriptomic features associated with immunotherapy response in LM patients found that immune checkpoint inhibitors initially increased CD8+ T cell activity, but this response was transient due to immune evasion mechanisms. The study highlighted the need for combinatorial therapeutic strategies (140). The investigation of the immune microenvironment of melanoma brain metastases and LM using scRNA-seq found distinct immune landscapes, with melanoma brain metastases samples showing diverse immune infiltration with active T cells, while LM samples exhibited severe T cell dysfunction and immune suppression. An exceptional LM survivor responded well to PD-1 therapy, displaying higher T cell and DC levels (141). More recently, a study highlighted an immunosuppressive lipid-associated macrophage subtype involved in osimertinib resistance and LMs, suggesting a potential therapeutic target (139).