Dynamic alterations of dural and bone marrow B cells in an animal model of progressive multiple sclerosis

The meninges, an anatomical buffer that has evolved to protect the central nervous system (CNS) from pathogens, are comprised of three distinct layers: the dura mater found immediately underneath the bony skull, the arachnoid mater which forms the space through which cerebrospinal fluid flows, and the pia mater which overlays the brain parenchyma. The arachnoid mater and pia mater are collectively termed the leptomeninges (LM). Whereas the LM are lymphocyte poor, the dura mater contains a variety of immune cells, including T cells, neutrophils, mature B cells, and developing B cells (Rustenhoven et al., 2021; Brioschi et al., 2021; Schafflick et al., 2021; Wang et al., 2021b; Rua and McGavern, 2018). The involvement of the dura as a source of immune cells in CNS autoimmune inflammation, such as that seen in multiple sclerosis (MS), remains unclear.

MS is an autoimmune disease of the brain and spinal cord in which immune cell subsets collaborate to cause damage to the CNS, including demyelination and axonal loss. Clinical presentation of the disease changes with age, with individuals exhibiting increasing disability without recovery (Absinta et al., 2020). This phase of the disease has been called progressive MS (PMS). The age-dependent mechanisms driving MS progression are not well understood. Pathologically, focal demyelinating lesions in the white matter (WM) are the hallmark of MS. While new WM lesions are observed in relapsing MS, they are less common in PMS. Instead, areas of grey matter (GM) demyelination directly under the pia (subpial lesions) have been associated with cortical atrophy and disease progression, particularly in aged patients (Calabrese et al., 2013). While the etiology of these GM lesions is debated, one leading hypothesis is that compartmentalized inflammation in the adjacent LM drives GM damage through soluble factors produced by LM-resident immune cells that cross through the glia limitans superficialis into the cortex, where they can have injurious properties (Magliozzi et al., 2007).

The LM are an important site of pathogenic T cell entry during the onset of experimental autoimmune encephalomyelitis (EAE) (Lodygin et al., 2019; Lodygin et al., 2013; Schlager et al., 2016). We previously showed that the LM of SJL/J mice undergo remarkable stromal cell remodeling in response to adoptive transfer of Th17 cells, enabling the formation of a niche that supports further immune cell influx and Th17 cell maintenance and polarization (Pikor et al., 2015). These LM aggregates are adjacent to regions of cortical pathology, including regions of GM demyelination, microglial activation, and glia limitans disruption, as observed in MS (Pikor et al., 2015; Ward et al., 2020). In aged mice, this form of passive EAE results in chronic disease accompanied by persistent LM immune aggregates and GM injury (Zuo et al., 2022).

The role of the dura in MS and EAE has only been recently studied. Reports using the C57BL/6 active EAE model have described an increase in dural-resident immune cells, including myeloid cells, IgG⁺ and IgM⁺ plasma cells (PCs), Th17 and regulatory T cells, while IgA⁺ PCs were decreased compared with naïve mice (Louveau et al., 2018; Rustenhoven et al., 2021; Schafflick et al., 2021; Cugurra et al., 2021). Ablation of dural lymphatics prior to EAE induction delayed disease onset and decreased disease severity, possibly due to reduced antigen and immune cell drainage to sites of activation, such as areas adjacent to dural sinuses or the cervical LNs (Louveau et al., 2018). Interestingly, neutrophils, Gr-1⁺ myeloid cells, and T cells have been found to move from the dura to the LM during EAE through arachnoid cuff exit points, which connect the dura and LM, suggesting that the dura may be a source of potentially injurious immune cells (Smyth et al., 2024). Taken together, data from C57BL/6 EAE support a role for the dura as a potential source of immune cells that invade the LM and parenchyma and/or as a conduit for antigen and immune cells out of the LM and parenchyma. However, other studies have found no difference in EAE course or immune cell infiltration into the CNS, LM, or dura following dural lymphatic ablation (Merlini et al., 2022; Li et al., 2023).

C57BL/6 EAE mice exhibit limited immune cell aggregates in the brain LM and little evidence of subpial cortical GM pathology, limiting our understanding of the consequences of immune cell movement between the brain dura and the inflamed LM during EAE. In a rat model of GM damage, the relative accumulation of T cells and myeloid cells in the dura was significantly lower than that seen in the LM or brain of the same rats (Merlini et al., 2022). To better characterize the temporal accumulation of different types of immune cells in distinct layers of the brain meninges in the context of subpial cortical pathology, we applied our model of EAE induced by adoptive transfer of encephalitogenic proteolipid protein (PLP)-primed Th17 cells into both young and aged SJL/J mice. We probed the role of age on the accumulation of immune cells in the dura, LM, brain, skull bone marrow (BM), and femur BM at multiple time points of the EAE disease trajectory. Independent of age, we found that, compared with the LM and the brain parenchyma, the dura is not a major site for accumulation of encephalitogenic Th17 cells. However, we noted that developing/immature B220^low B cells disappeared from the dura and BM during EAE and that different subsets of B cells, some of which resist anti-CD20 therapy, exhibited an age-dependent brain compartmentalization at the postacute phase.

To evaluate the relative accumulation of encephalitogenic Th17 cells in a setting with subpial cortical GM injury, we tracked adoptively transferred CFSE-labeled PLP-primed Th17 cells in recipient SJL/J mice. Single-cell suspensions of the dura, LM, brain, blood, and spleen were analyzed by flow cytometry at several time points following adoptive transfer (Fig. 1 A and Fig. S1 A). Comparing different levels of CFSE (see gating strategy, Fig. S1, A and B), we observed that during the pre-onset phase (days 1–5 following adoptive transfer), CFSE^high T cells were detected in the blood and spleen as early as day 1 but not in either the dura or LM (Fig. 1 B and Fig. S1 C). At disease onset (day 4 onward), CFSE^high T cells were observed in the spleen from day 4–7 but by day 11 were no longer detectable (Fig. 1 C and Fig. S1 C). As early as day 5, CFSE⁺ T cells were observed in the LM and brain of mice and continued to accumulate up until the acute phase of disease at day 11 (Fig. 1 C). Unlike the spleen, these were CFSE^low (but exhibit a CFSE signal that is above what is observed for negative controls—Fig. S1 B and C). In contrast, very few CFSE⁺ T cells were observed in the dura at any of these time points (Fig. 1 C and Fig. S1 C). The relative number of CFSE^high versus CFSE^low T cells in the spleen, dura, LM, and brain parenchyma are depicted in Fig. S1 C.

In addition, we tested whether encephalitogenic T cells infiltrate the proximal skull and/or distal femur/tibia BM over the course of EAE. While a few mice had some CFSE⁺ T cells in the femur/tibia BM, these tissues remained largely devoid of CFSE⁺ T cells at all time points (Fig. 1 C). Lastly, to determine if brain/LM-infiltrating CFSE⁺ T cells expressed the phenotype of naïve (CD44^lowCD62L⁺) or effector memory (CD44^highCD62L⁻) T cells (Sallusto et al., 1999; Aruffo et al., 1990; Bradley et al., 1992), we analyzed the expression of the activation marker CD44 and the LN entry receptor CD62L, comparing the blood, LM, and brain at days 7 and 11 following adoptive transfer. While we observed both CD44^lowCD62L⁺ and CD44^highCD62L⁻ CFSE⁺ T cells in the blood, almost all CFSE⁺ T cells in the LM and brain were CD44^highCD62L⁻ (Fig. 1 D).

Collectively, these data support previous findings indicating the LM is the primary initial site of encephalitogenic T cell accumulation during EAE (Merlini et al., 2022), and because our model invokes LM immune cell aggregates adjacent to GM lesions, these data dissociate the dura as a contributing source of encephalitogenic T cells to subpial cortical pathology.

Since both adoptively transferred and endogenous T cells contribute to LM inflammation in SJL/J EAE (Pikor et al., 2015), we next used flow cytometry to evaluate endogenous T cells at early onset (day 7), late onset (day 9), acute phase (day 11), and postacute phase (day 25) in all brain and meningeal compartments of both young and aged mice (Fig. 2 A). There was no difference in the number of CD3⁺ T cells in the dura, LM, and brain of aged compared with young mice at any time point, with the exception of an increase in CD3⁺ T cells in the brain of young mice at the acute phase of disease (Fig. 2 B). In both young and aged mice, CD3⁺ T cells decreased at the postacute phase despite their highly disparate clinical scores (Fig. 2 B). Given that absolute cell numbers in the LM and brain far surpassed those in the dura during disease, we expressed these numbers as a fold change of CD3⁺ T cells from naïve numbers, comparing the dura with the LM and brain at each EAE time point. We observed that CD3⁺ T cells at the acute and postacute phases were significantly increased in the LM and brain of young and aged mice compared with the dura (Fig. 2 C). Likewise, CD3⁺ T cells primarily accumulate in the brain and LM rather than the dura during active SJL/J EAE induced by an emulsion of CFA and PLP_139-151 (Fig. S2, A and B). Taken together, these data reveal preferential accumulation of CD3⁺ T cells in the LM and brain rather than the dura during passive and active EAE, independent of age.

B cells are important contributors to both MS pathology and regulation of neuroinflammation (Wang et al., 2021a). We previously revealed that elimination of LM-resident B cells with anti-CD20 treatment spares the subpial GM from demyelination, synapse loss, and oxidative stress (Wang et al., 2024). To assess the dynamics of mature B cell accumulation in the brain and meninges, we quantified IgM⁺IgD⁺ versus class-switched B220^high B cells which were identified as CD19⁺B220^highCD24^lowIgM⁺IgD⁺ and CD19⁺B220^highCD24^lowIgM⁻IgD⁻, respectively (see Fig. S2 C for gating). In young mice, absolute numbers of B220^high B cells increased at the onset and during the acute phase of EAE, then decreased by the postacute phase across the brain, LM, and dura. Of note, the LM and brain of aged mice harbored significantly fewer IgM⁺IgD⁺B220^high B cells at the acute phase of disease compared with young mice at the same time point (Fig. 3 A). When expressed as a fold change compared with baseline, B220^high B cells accumulate in the young and aged LM to a significantly greater extent than they do in the dura (Fig. 3 B). In young mice, the brain also showed a significant fold change increase in IgM⁺IgD⁺B220^high B cells compared with the dura that was not observed in aged mice (Fig. 3 B).

We additionally observed that compared with young EAE mice, class-switched B220^high B cells were significantly increased in the dura, LM, and brain of aged EAE mice (Fig. 3 A). Most class-switched B cells were IgG⁺ (Fig. S2, D and E). However, like IgM⁺IgD⁺B220^high B cells, the EAE-induced fold change in class-switched B220^high B cells was very modest in the dura compared with the LM and brain (Fig. 3 B). Collectively, these data show that IgM⁺IgD⁺ and class-switched B220^high B cells preferentially increase in the LM and brain during the establishment of EAE in young mice, with a preferential accumulation of class-switched rather than IgM⁺IgD⁺B220^high B cells in aged mice.

We next hypothesized that preferential accumulation of B220^high B cells in the LM but not the dura during EAE may be in part explained by differential expression of chemotactic and survival factors. To test this, we measured concentrations of the B cell survival factor B cell activating factor (BAFF) and the B cell attracting chemokine CXCL13 in supernatants from whole tissue extracts of the dura and LM from young and aged mice throughout EAE. Compared with naïve mice, BAFF levels increased in both the LM and dura compartments throughout EAE in young mice (Fig. 3 C). A similar pattern was seen in the LM of aged mice; however, unlike young mice, BAFF levels were not elevated in the dura in response to EAE (Fig. 3 C). Like BAFF, CXCL13 also increased in the LM in response to EAE; however, there was minimal induction of CXCL13 in the dura at any time point in young or aged mice (Fig. 3 C). This suggests that an increase in CXCL13 production in the LM may explain the preferential increase in IgM⁺IgD⁺ and class-switched B220^high B cells in the LM compared with the dura during EAE. The production of CXCL13 within each compartment may be driven by the differential infiltration of Th17 T cells into these tissues. Indeed, our lab previously showed that interactions between lymphotoxin (LT) αβ on Th17 cells and LTβR on stromal cells drove CXCL13 production in the LM during passive SJL/J EAE (Pikor et al., 2015). It is therefore possible that greater Th17 cell infiltration into the LM over the dura during EAE drives increased CXCL13 expression and preferential recruitment of B220^high B cells to this compartment.

We previously conducted single-cell RNA sequencing of the LM of naïve and EAE SJL/J mice and saw many age-related differences in expression of B cell–associated genes. Of note, compared with aged mice, LM cells derived from young EAE mice showed evidence of upregulation of transcripts associated with developing/immature B cells, including the immunoglobulin superfamily gene expressed in pro- and pre-B cells Vpreb3 (Zuo et al., 2022). These findings, as well as the recent identification of developing B cells in the dura of naïve mice and nonhuman primates (Brioschi et al., 2021; Schafflick et al., 2021; Wang et al., 2021b), prompted us to examine developing/immature B cells in the dura during EAE. As such, we performed a phenotypic analysis of developing/immature B cells in various tissue compartments during EAE in young and aged mice. Using the BM as a comparator, given it is the site of B cell development, we observed two populations of CD45⁺CD11b⁻CD3⁻CD19⁺B220⁺ cells in the dura of young naïve SJL/J mice based on B220 and CD24 expression: B220^highCD24^low and B220^lowCD24^high B cells. We observed that CD19⁺B220^highCD24^low B cells were negative for IL7R but expressed MHCII and both IgM and IgD, whereas CD19⁺B220^lowCD24^high B cells were variable for IL7R and negative for MHCII and IgD (Fig. S2 C). Therefore, in addition to mature B220^highCD24^lowIgM⁺IgD⁺ B cells, we detected immature/developing B220^lowCD24^highIgD⁻ B cells in the naïve dura, which for simplicity we term B220^low B cells.

We next set out to characterize changes in B220^low B cells during EAE. Compared to naïve mice, B220^low B cells were dramatically decreased in the dura at onset and at the acute phase of EAE in both young and aged mice while increasing in the LM of young but not aged mice at disease onset (Fig. 3 D). In addition, B220^low B cells appeared in the brain parenchyma of young but not aged EAE mice at the postacute phase (Fig. 3 D). Moreover, an increased ratio of class-switched B220^high to B220^low B cells was observed in the LM and brain of aged mice (Fig. 3 F), and this was also reflected in the spleen (Fig. S2 F). Taken together, these data show that B220^low B cells virtually disappear from the dura at the same time as their accumulation in the LM and brain during EAE in young but not aged mice.

Previous studies have suggested that developing B cell populations seed the dura from the overlying skull rather than distal tibial BM (Brioschi et al., 2021). We therefore determined if B220^low B cells were altered in these two BM sites in young versus aged mice during EAE. Like the dura, compared to naïve mice, B220^low B cells decreased in the skull BM at the acute phase of EAE in both young and aged mice (Fig. 3 E). Despite its distance from the CNS, we saw the same decrease in B220^low B cells in the femur/tibia BM at the acute phase of EAE, independent of age. By the postacute phase, B220^low B cells had rebounded in the femur and skull BM of young mice to levels greater than those seen in naïve mice, but this rebound was not observed in aged mice, who at this time point still have chronic EAE (Fig. 3 E). It is conceivable that CNS inflammation may trigger a “brain-body” circuit that communicates with the peripheral BM to cause the release of B220^low B cells (Jin et al., 2024). Future experiments are required to determine the source of B220^low B cells that reach the brain LM and parenchyma (dura versus skull BM versus distal BM).

To gain further insights into B220^low B cells, we performed single-cell RNA sequencing on whole dural dissections (GEO accession no. GSE299404). After quality control and integration using Harmony, we were able to retrieve 87,715 total cells across dural isolations collected at homeostasis, acute, and postacute phase EAE from young and aged mice. Of these, we identified 3,025 cells as B cells or PCs based on expression of canonical genes (Cd19, Ms4a1, Ighm, Ighd, Prdm1, Igha, and Ighg2a). We noted a cluster of B220^low (Ptprc-low) cells based on gene expression of Cd19, Cd24a, and Ighm and low expression of Ms4a1 and Ighd (Fig. 4 A), which correlate with surface expression of their protein counterparts by flow cytometry (Fig. S2 C and Fig. S3 A). Hartlehnert et al. previously showed that Ncl, Eif2f2, and Eif4a1 (so-called “Bc-1” genes) and Hspa1a, Hspa1b, and Cd69 (so-called “FOBc” genes) are expressed in spinal cord meningeal B cells in B cell–dependent EAE but are downregulated in the absence of Bcl6 expression in Th17 cells (Hartlehnert et al., 2021). We therefore compared expression of these genes with our dural B cell data. We found that, in the dura of SJL/J mice, B220^low cells primarily express genes associated with the Bc-1 subset, as well as Cd69 (Fig. 4 B). Differential gene expression analysis comparing B220^low B cells (Cd19, Ighm^high, Ms4a1^low, and Cd24a^high) with B220^high B cells (Cd19, Ighd, Ms4a1^high, and Cd24a^low) revealed an upregulation of genes associated with B cell development in B220^low B cells, including Rag1, Rag2, Ebf1, Vpreb1, Vpreb2, Vpreb3, Pax5, Blnk, and Dntt (Fig. 4 C). Pathway analysis of genes upregulated in B220^low B cells revealed enrichment of genes associated with development and activation, such as V(D)J recombination, somatic rearrangement, and B cell activation and differentiation (Fig. 4 D).

We next performed CellChat communication inference analysis (Jin et al., 2021) at various time points of EAE to understand which cell types B220^low B cells and B220^high B cells may interact with and whether these interactions change as EAE progresses. Examining B cells as “senders”, we noted that most interactions between B cells and other dural-resident cells emanated from B220^low rather than B220^high B cells (Fig. 4 E), many of which were with fibroblast clusters. After counting the total number of inferred interactions between B cells and other dural cell clusters (Fig. 4 F, left) and interaction strengths using the CellChat probability algorithm (Fig. 4 F, right), we found that B220^low B cells were predicted to have more and stronger interactions than B220^high B cells at the postacute time point. Some unique interactions between B220^low B cells and fibroblasts were also observed: for example, B220^low cells predominantly send signals to fibroblasts via the GALECTIN pathway (Lgals9-P4hb) (Fig. 4 G). Taken together, these data suggest that the dura is a reservoir for developing B cells, that these B220^low B cells are the dominant communicators with fibroblasts at the postacute phase of disease, and that B220^high versus B220^low B cells exhibit unique mechanisms for communicating with dural fibroblasts.

We recently showed that depletion of B cells with anti-CD20, a therapy that has been largely successful in the treatment of relapse remitting MS (Hauser et al., 2017; Hauser et al., 2008; Bar-Or et al., 2008), reduces the size of Th17 cell–induced LM immune aggregates and prevents GM demyelination and cortical injury at the acute phase of disease (Wang et al., 2024). We noted that B220^low B cells express significantly less CD20 than B220^high B cells (Fig. S3 A) and thus hypothesized that these may be resistant to anti-CD20 therapy. Using flow cytometry, we first confirmed that administration of anti-CD20, but not a matched isotype control, significantly decreased the number of IgM⁺IgD⁺ and class-switched B220^high B cells in all tissues tested in young adoptive transfer EAE mice at the postacute phase (Fig. 5, A and B), while sparing CD3⁺ T cells (Fig. S3 B). In contrast, the number of B220^low B cells in the femur/tibia BM, skull BM, and dura were unchanged between isotype and anti-CD20–treated mice during the postacute phase. Anti-CD20 treatment resulted in a modest reduction of B220^low B cells in the LM and brain, and these were not completely depleted (Fig. 5 B). The relative persistence of B220^low B cells compared with IgM⁺IgD⁺ and class-switched B220^high B cells was highly significant in the dura, LM, and brain (Fig. 5 C).

Since cortical GM pathology resolves in young mice (Zuo et al., 2022), and B220^low B cells persist in the brain at the postacute time point only in young mice, it is tempting to speculate that B220^low B cells contribute toward the prevention of the disease chronicity that we ordinarily observe in aged mice. Moreover, the conspicuous accumulation of class-switched IgG⁺ B220^high B cells at the expense of B220^low B cells in the brain and LM of aged mice, which have severe and persistent cortical GM pathology (Zuo et al., 2022), further supports this hypothesis. Indeed, memory B cells have been implicated in MS pathogenesis (Jelcic et al., 2018; Li et al., 2015). However, it is important to note that age-dependent disease chronicity in this model is likely linked to multiple facets of an aged hematopoietic system and aged glial cells. Indeed, Segal and colleagues have identified changes in microglia that correlate with exacerbated EAE (Atkinson et al., 2022), and we have shown that aged SJL/J passive EAE mice have an abundance in neutrophils in the LM (Zuo et al., 2022).

There are some limitations to our study. Although we have made progress in annotating the transcriptomes of B220^high versus B220^low B cells at multiple time points during EAE in young versus aged mice (Fig. 4), we do not yet have an identifying marker for B220^low B cells that can be used for immunofluorescence. This limits our ability to pinpoint the location of B220^low B cells in the brain and LM of postacute young EAE mice and to evaluate their functional relevance. Moreover, we did not detect any enriched pathways that met that P value cutoff when we analyzed differentially expressed genes between B220^low cells at naïve vs postacute timepoints in the dura. Due to this, we could not conclude whether postacute B220^low cells are regenerative or protective, at least in the dura. In future, it would be useful to examine cortical B220^low B cells at the postacute phase of EAE in young mice to determine whether in this location B220^low B cells have a direct or indirect role in protecting against GM injury—this will be challenging due to the relative scarcity of this population. In addition, although class-switched B220^high B cells accumulate in aged mice, we do not know what their role is in chronic age-dependent disease. Lastly, while spinal cord WM pathology in passive C57BL/6 and SJL/J EAE has a clear clinical correlate (paralysis), it is difficult to implement a correlate of GM pathology in SJL/J passive EAE mice (for example, impaired cognition) due to the paralytic state of the animals.

In summary, we found that T cells and B220^high B cells accumulate in the LM and brain during EAE to a much greater extent than the dura, potentially through the local induction of CXCL13 within the LM. Moreover, we identified a novel subset of partially anti-CD20–resistant B220^low B cells that disappear from the dura and BM yet accumulate in the brain during EAE at a time point when young but not aged mice remit from disease, suggesting a potential neuroprotective role for these cells that remains to be investigated.

Female 3- to 4-wk-old SJL/J CD45.1⁺ mice were purchased from Envigo (mouse code: 052). Mice were housed in the Division of Comparative Medicine, University of Toronto animal facility under specific pathogen-free conditions, in a closed caging system with a 12-h light/dark cycle, and had free access to a standard irradiated chow diet (2918; Teklad, Envigo) and acidified water (reverse osmosis and ultraviolet sterilized). All experiments were conducted with ethical approval and in compliance with standards for animal care as set out by the Local Animal Care Committee, University of Toronto. Of note, we opted to only use female mice because in the past our laboratory has observed no sex differences in EAE between male versus female mice (Zuo et al., 2022), and male mice present with multiple husbandry issues, including penile prolapse, resulting in an increased need for euthanasia.

Donor 6- to 10-wk-old female SJL/J were immunized with 100 µg of PLP_139-151 (HSLGKWLGHPDKF; Canpeptide) in an emulsion of incomplete Freund’s adjuvant (BD Difco) and 200 µg of Mycobacterium tuberculosis H37 Ra (231141; BD Difco) via three 100 μl subcutaneous injections on either flank and on the back, for a total injection volume of 300 μl. Nine days following immunization, donors were humanely euthanized by CO₂ asphyxiation, and the spleens and LNs (inguinal, brachial, axillary, and cervical) were collected and processed into single-cell suspension by mashing through a 70-µm filter under sterile conditions. Cells were then resuspended in complete culture media: RPMI 1640 (Sigma-Aldrich) supplemented with 10% fetal bovine serum (Gibco), 1× penicillin-streptomycin (Sigma-Aldrich), 1× GlutaMAX (Gibco), 10 mM HEPES (Corning), 1 mM sodium pyruvate (Gibco), 1× non-essential amino acids (Gibco), and 1× β-mercaptoethanol (Gibco). Cells were restimulated ex vivo with PLP_139–151 (10 μg/ml), recombinant mouse IL-23 (R&D Systems; 10 ng/ml), anti–IFN-γ (Bioceros; 20 μg/ml), and anti–IL-4 (Bioceros; 20 μg/ml) for 72 h at 37°C. 10 million cells from culture were then transferred into young (6- to 10-wk-old) or aged (8- to 10-mo-old) female SJL/J recipient mice via i.p. injection on day 0 of EAE. For CFSE tracking experiments, culture products were first stained at 50 million cells/ml with CFSE (C34554; Invitrogen) in PBS at room temperature for 10 min with gentle mixing, then washed with PBS and transferred into recipient mice via adoptive transfer as described. Active EAE was induced in 8-wk-old SJL/J mice using the same protocol as donor immunization, and immunized mice were followed for progression of disease.

For clinical evaluation of disease progression, mice were weighed each day and given a composite clinical score out of a 16-point scoring system that reflects impairments in each limb and tail, as previously described (Galicia et al., 2018; Zuo et al., 2022). Briefly, scores are assigned in 0.5 increments, with each limb given a score out of 3, the degree of tail paralysis given a score out of 2, and the righting reflex given a score out of 2. Upon the onset of clinical disease, mice were transferred to a heat pad, provided with wet mash and Napa Nectar (SE Lab Group), and additional hydration was supplied through subcutaneous injections of Lactated Ringer’s Saline. All experiments were pre-approved by the University of Toronto Temerty Faculty of Medicine animal ethics board (protocol #20011363).

Young (6- to 10-wk-old) SJL/J mice were randomized into either treatment or control groups, where the treatment group was administered purified anti-CD20 monoclonal antibody (18B12; mouse IgG2a) generously provided by Roche, while the control group was administered isotype control mouse IgG2a antibodies (BE0085; Bio X Cell). Both treatment and control groups received 10 mg/kg of antibody dissolved in PBS via five i.v. injections on days −1, 4, 9, 14, and 19 of adoptive transfer EAE. Mice were then humanely euthanized, and tissues were collected for analysis on day 24 of EAE.

For CFSE tracking experiments, mice were i.v. injected with 3 µg of anti–CD45.1-PE antibody (12-0453-82, clone: A20; Thermo Fisher Scientific) in 200 μl of PBS three min prior to euthanization to preferentially label blood-derived lymphocytes (Ruscher and Hogquist, 2018; Zuo et al., 2022). For all other experiments, mice were immediately humanely euthanized by CO₂ asphyxiation. Blood was collected into heparin-coated microvette capillary tubes (Sarstedt) via intracardiac bleed, and mice were transcardially perfused with 60 ml of ice-cold PBS prior to collection of brain, LM, dura, or BM.

100 μl of blood from heparin-coated microvette capillary tubes was then resuspended in 2 ml of red cell lysis buffer (155 mM NH₄Cl, 12 mM NaHCO₃, and 0.1 mM EDTA) for 4 min on ice. RBC lysis was stopped with the addition of 1 ml of PBS, and samples were centrifuged and resuspended in PBS prior to downstream applications.

The skin overlying the skull was removed, and mice were decapitated to separate the skull from the spinal column. The skullcap was then carefully separated from the brain by making incisions from the occipital to the frontal bone using fine surgical scissors. The skullcap and brain were each placed into separate petri dishes with 1 ml of PBS. The dura mater was carefully scored from the skullcap, and the LM were carefully peeled from the brain under a dissection microscope. Briefly, the dura mater was scored off the skullcap using forceps to release attachment points between the dura and the skullcap, whereas the LM were removed from the brainstem, cerebellum, ventricles, hypothalamus, olfactory bulbs, and cortex. The dura and LM were then placed into 200 and 400 μl of digestion buffer, respectively (RPMI 1640 [Sigma-Aldrich] supplemented with 2% fetal bovine serum [Gibco], 1 mM Hepes [Corning], 1× penicillin/streptomycin [Gibco], and 1× GlutaMAX [Gibco]). After 30 min to an hour of sitting in digestion buffer, 200 μl of supernatant was removed from the digestion buffer of the dura or the LM and stored at −20°C to −80°C for downstream protein analysis. Sufficient digestion buffer was then added to top the dura up to 1 ml and the LM up to 500 μl, then digestion enzymes DNase I (dura: 0.5 mg/ml, LM: 60 μg/ml; Sigma-Aldrich) and collagenase P (1 mg/ml, Roche) were added. Tubes were mixed by gentle vortex, and digestion was allowed to proceed for 15 min at 37°C with gentle agitation. The digestion was stopped with 1 ml of PBS, and samples were filtered through 35-μm mesh strainer caps into 5-ml FACS tubes prior to downstream applications.

Following removal of the LM, the remaining brain tissue was mashed through 70-μm mesh filters into 5 ml of digestion buffer in a 6-well plate. Digestion enzymes DNase I (60 μg/ml; Sigma-Aldrich) and collagenase P (1 mg/ml; Roche) were then added prior to digestion for 30 min at 37°C with gentle agitation. Following digestion, brain homogenates were dissociated by gentle pipetting through the same 70-μm mesh filters. Brain pellets were then resuspended in 30% Percoll (Cytiva) and centrifuged at 2,000 rpm for 20 min with the brake off to separate lymphocytes from fat and myelin debris. Single-cell suspensions of brain were then washed with PBS prior to use for downstream applications.

Following removal of the dura, the skullcap was cleaned of muscle and then cut into small pieces so that pockets of BM were exposed. Skull pieces were then placed into a punctured 0.6-ml PCR tube placed within a larger 1.5-ml microtube (Sigma-Aldrich). For isolation of the leg BM, the skin of the mouse’s leg was removed, and the leg was dislocated from the hip joint. Muscle tissue was cleaned off the femur and tibia, and the ends of each bone were cut so as to expose pockets of BM. The femur and tibia were then also placed into a punctured 0.6-ml PCR tube placed within a larger 1.5-ml microtube. Skull and leg samples were then pulse spun within a microcentrifuge such that the BM from each sample exited the bone and pelleted at the bottom of the 1.5-ml microtube. Samples were washed with 1 ml of PBS and filtered through 35-μm mesh strainer caps into 5-ml FACS tubes prior to downstream applications.

Single-cell suspensions were first stained with a 1:1,000 dilution of Live/Dead Fixable Aqua (L34965; Invitrogen) for 30 min on ice. Surface markers were then stained in 2% fetal bovine serum in PBS (FACS) in the presence of anti-mouse CD16/CD32 Fc block (BE0307; Bio X Cell) at a concentration of 1/100 for 30 min on ice (Table S1). Where a biotin-conjugated antibody was used in surface staining, samples were subsequently stained with a streptavidin-conjugated fluorophore in 2% fetal bovine serum in PBS. Samples were then fixed and permeabilized in Foxp3/Transcription Factor Staining Buffer (00-5523-00; eBioscience) fixation/permeabilization diluent for 30 min on ice before transferring to FACS buffer overnight. Samples were then mixed at known volumes of FACS with Cell Bright Plus Absolute Counting Beads (C36995; Invitrogen) to allow for back calculations of cell numbers. Samples were acquired on a BD Fortessa X-20 or BD FACSymphony A3 using the FACSDiva software, and flow cytometry data were analyzed using FlowJo v10.9.0.

Single-cell suspensions from dural dissections were generated as stated above, then cells from 4 to 6 dura were pooled for staining and sorting. After sorting for live cells using Calcein violet (65-0854-39; Thermo Fisher Scientific) staining, cells were resuspended to the appropriate concentration and submitted to Princess Margaret Genomics Centre for sequencing using 10X Genomics 3′ Next-GEM chemistry. Count and barcode matrices were generated using the output from CellRanger (v.7.2.0) Count pipeline provided by 10X Genomics. Data were loaded into Rstudio (v.2022.12.0, Build 353), filtered for ambient RNA and doublets using the SoupX (v.1.6.2) (Young and Behjati, 2020) and DoubletFinder (v2.0.3) (McGinnis et al., 2019) packages, respectively, and then downstream analysis using the Seurat package (v.5.1.0) (Hao et al., 2024). Data were normalized using the SCTransform function in Seurat and then integrated with Harmony (Korsunsky et al., 2019) using SCT normalized counts. Unsupervised clustering of all dural cells using the Seurat package identified 22 clusters (9 stromal, 10 immune, 2 neural, and 1 RBC cluster). Further reclustering of the high-level B cell revealed 1 cluster corresponding to B220^low cells and 5 clusters of B220^high cells. For DEG and CellChat analysis, all IgM⁺IgD⁺ B220^high cells were combined into one cluster. These designations were re-annotated into the global dura Seurat object for downstream analyses. Pathway analysis was performed using the clusterProfiler package (v4.10.0) (Yu et al., 2012). Cell–cell interactions were inferred using the CellChat package (v2.1.2) (Jin et al., 2021). Data are available in Gene Expression Omnibus (GEO accession number GSE299404).

BAFF and CXCL13 concentrations in dura and LM supernatant samples were quantified using a customized mouse cartridge for the Simple Plex Ella microfluidics platform (Protein Simple). Samples were thawed and plated neat or diluted at a ratio of 1:2 in manufacturer-supplied diluent. 50 μl of each neat/diluted sample was loaded on the cartridge, which was then placed on the Ella instrument. All samples were run in triplicate. Samples were acquired on the Ella using factory-generated standard curves corresponding to each cartridge lot. The SimplePlex Explorer (v4.1.0.22; Bio-Techne) was used to retrieve and analyze assay results. The lower limit of quantification (LOQ) and upper LOQ were 10.8 pg/ml and 16,478 pg/ml for BAFF and 2.62 pg/ml and 4,000 pg/ml for CXCL13.

Unless otherwise stated, all statistical tests were conducted using GraphPad Prism v9.0, and only P values <0.05 were considered significant. The specific statistical tests used in each experiment are indicated in the figure legend text. All quantification data first underwent a Shapiro–Wilk normality test. Flow cytometry data were analyzed using FlowJo v10.9.0. Absolute cell numbers were derived from flow cytometry by recording the number of events in the gate representing the population of interest as well as the number of events recorded in the gate representing counting beads, which were added at a known volume and concentration and allowed for back calculation of cell numbers as described in the user guide for Cell Bright Plus Absolute Counting Beads (C36995; Invitrogen). Statistical analysis for flow cytometry data was conducted using a Student’s t test or using the following tests without assuming normality: Mann–Whitney test for comparison between two unpaired groups or a Kruskal–Wallis test for comparison between more than two unpaired groups followed by Dunn’s multiple comparison test. Statistical analysis for flow cytometry data comparing two variables was done using either a two-way ANOVA or a mixed effects model (if there were missing values in paired data), followed by the following post hoc tests: Bonferroni’s multiple comparison test for dependent comparisons, Sidak’s multiple comparison test for independent comparisons, and Dunnett’s multiple comparison test for independent comparisons to a control group (in this case, the dura).

Fig. S1 shows the flow cytometry gating strategies as pertaining to CFSE⁺ T cells used in core Fig. 1 and quantifies CFSE high versus low populations throughout EAE. Fig. S2, A and B, shows the consequence of dural T cell accumulation during active EAE and relates to core Fig. 2; and Fig. S2, C and D, provides the B cell gating information that relates to core Figs. 3 and 5, and Fig. S2, E and F, shows the impact of age on B cell populations in the CNS and spleen and relates to core Fig. 3. Fig. S3 A shows the level of CD20 on B220^low versus B220^high B cells in young versus old SJL/J mice in the dura and in the periphery, and Fig. S3 B shows the impact of anti-CD20 depletion on T cells in the BM, LM, brain, and dura. Fig. S3 relates to the core Fig. 5. Table S1 lists the antibodies used in flow cytometry.

All data are available in the main text, in the supplementary materials, or upon request. Data used in single-cell transcriptomic analysis of Fig. 4 are openly available in GEO using the GEO accession number GSE299404.

We would like to acknowledge the expertise and assistance of Dr. Nathalie Simard at the University of Toronto Faculty of Medicine flow cytometry facility as well as the staff of the Division of Comparative Medicine at the University of Toronto Faculty of Medicine for their animal husbandry. We would also like to acknowledge the Princess Margaret Genomics Centre and Dr. Troy Ketela for helping design the single-cell RNA-sequencing experiments and the technicians for their help in preparing libraries, running the sequencer, and pre-processing data.

This work was supported by grants from the MS Canada (EGID 1032820) and the Canadian Institutes for Health Research (Foundation grant 15992) to Jennifer L. Gommerman and the Canadian Institutes of Health Research Vanier PhD scholarship granted to Alexandra Florescu.

Author contributions: Alexandra Florescu: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, validation, visualization, and writing—original draft, review, and editing. Michelle Zuo: conceptualization, data curation, formal analysis, investigation, methodology, project administration, resources, software, validation, visualization, and writing—original draft, review, and editing. Angela A. Wang: formal analysis, investigation, methodology, validation, and writing—review and editing. Kevin Champagne-Jorgensen: investigation and methodology. Mohammed A. Noor: investigation and writing—review and editing. Lesley A. Ward: investigation. Erwin Van Puijenbroek: resources. Christian Klein: resources and writing—review and editing. Jennifer L. Gommerman: conceptualization, funding acquisition, project administration, supervision, and writing—original draft, review, and editing.