Mural cells interact with macrophages in the dura mater to regulate CNS immune surveillance

The central nervous system (CNS) and the peripheral immune system have a delicate relationship. The CNS relies on support from peripheral immune cells for proper functioning, yet it must maintain immune tolerance to prevent provoking a destructive autoreactive immune response (Prinz and Priller, 2017; Filiano et al., 2015, 2017). The meninges are complex, multimembrane structures that have been implicated in CNS immunity. The involvement of each meningeal layer in regulating a CNS-associated immune response likely depends on the context and the extent of the immune response (Angelini et al., 2023; Eme-Scolan et al., 2023; Mapunda et al., 2022). Under homeostatic conditions, peripheral immune support is thought to be mediated via the dura, the outermost layer of the meninges, which envelopes the CNS parenchyma and houses relevant lymphatic structures (Alves de Lima et al., 2020; Antila et al., 2017; Louveau et al., 2015). Unlike blood endothelial cells in the leptomeninges (which includes the inner arachnoid and pia layers), blood endothelial cells in the dura are fenestrated, allowing for adequate immune surveillance (Mastorakos and McGavern, 2019). Cells in the dura can access cerebrospinal fluid, where CNS-derived antigens can further drain through lymphatic vessels or be presented locally by antigen-presenting cells (APCs), such as macrophages and dendritic cells, to lymphocytes along the dural sinus (Hsu et al., 2022; Pulous et al., 2022; Wang et al., 2021, 2022; Aspelund et al., 2015; Brioschi et al., 2021; Louveau et al., 2015; Li et al., 2022; Da Mesquita et al., 2018; Rustenhoven et al., 2021). This positions the dura at the interface between the peripheral immune system and the CNS at steady state. Under inflammatory conditions, such as experimental autoimmune encephalomyelitis (EAE), a mouse model for multiple sclerosis (MS), primed T cells accumulate in the leptomeninges and perivascular spaces around penetrating blood vessels prior to infiltrating the CNS parenchyma (Bartholomäus et al., 2009; Schläger et al., 2016; Merlini et al., 2022). The initial events regulating a shift from CNS immune surveillance, especially in the dura, to pathology with overt inflammation within the CNS is unclear and remains a critical barrier for understanding early events in CNS autoimmune disease.

T cells survey the healthy CNS via the dura, which contains bona fide lymphatic vessels and APCs that sample CNS-derived antigens (Rustenhoven et al., 2021; Louveau et al., 2018b). The dura is highly vascularized, but unlike the leptomeninges, blood endothelial cells in the dura lack tight junctions (Mastorakos and McGavern, 2019); thus the dura represents an anatomical site where APCs have access to CNS antigens and can present these antigens to T cells under homeostatic conditions. Once T cells migrate from circulation into tissues, they first encounter perivascular cells on the abluminal surface. Here, mural cells (which include pericytes and vascular smooth muscle cells) and macrophages play a dynamic role in dictating local immunity. To investigate if mural cells are involved in the initial steps of CNS autoimmunity, we first quantified mural cell coverage on vessels in the dura at different stages of EAE. In mice immunized for active EAE, CD13⁺ and NG2⁺ mural cells had less coverage around blood vessels in the dura at presymptomatic EAE and further declined with disease onset (Fig. 1, A and B). Although mural cell coverage decreased in presymptomatic EAE, total mural cell numbers remained stable until decreasing at the onset of symptoms (Fig. S1 A). Unlike with EAE, we failed to detect a decrease in either mural cell coverage or infiltrating myelin-specific T cells (2D2) in the dura of aged mice (95 wk old) or a mouse model of amyotrophic lateral sclerosis (SOD1-G93A transgenic), suggesting these processes may be specific to conditions with obvious CNS-associated autoimmunity in the meninges (Fig. S1, B and C). To gain a better understanding of how mural cells are associated with a CNS response during different stages of EAE, we sequenced the transcriptome of cells in the dura, leptomeninges, and brain (cortex) of naïve mice and mice immunized for EAE in the presymptomatic phase or after disease onset (Fig. 1, C and D). Mural cells were altered in each compartment yet remained transcriptionally distinct throughout EAE (Fig. 1 E). In presymptomatic EAE, mural cells in the leptomeninges had more differentially expressed genes than mural cells from the dura or the brain (Fig. 1 F), but when we compared genes differentially expressed in both the dura and leptomeninges, we found these genes changed in similar direction and magnitude and were enriched for pathways associated with leukocyte migration, cell projection organization, and protein refolding (Fig. 1, F and G). No such correlation was found for genes expressed by brain mural cells when compared with corresponding genes expressed by mural cells in either layer of the meninges (Fig. 1 G). How mural cells, especially in the dura, impact the initial stages of CNS autoimmunity is unknown.

Studies that depleted mural cells have demonstrated the role of these cells in vascular maintenance; however, the mechanism and their role in CNS autoimmune disease remain inconclusive. Acute experimental loss of substantial levels of brain mural cells, including pericytes and vascular smooth muscle cells, can induce a breakdown in the blood–brain barrier (BBB), neurovascular uncoupling, permeability of blood-derived molecules, and a loss of neurons (Nikolakopoulou et al., 2019; Armulik et al., 2010; Kisler et al., 2020). Surprisingly, acute mural cell depletion in the retina did not affect the blood–retina barrier (Park et al., 2017), suggesting mural cells are not solely responsible for regulating vascular permeability; yet, this function likely depends on tissue-specific, multicellular communication. Chronic mural cell depletion (90% depletion) using the genetic Pdgfb^ret/ret model caused a transcriptional response in brain endothelial cells, an upregulation of leukocyte adhesion molecules in the CNS, and increased immune cell infiltration around blood vessels in the brain (Török et al., 2021; Mäe et al., 2021). These mice were more susceptible to EAE and developed spontaneous disease when crossed with 2D2 transgenic mice that have myelin-specific T cells (i.e., they express T cell receptors restricted to the myelin oligodendrocyte glycoprotein [MOG] antigen; Török et al., 2021). To test if reducing mural cell coverage around blood vessels was sufficient to alter T cell dynamics in the dura, we generated Pdgfr-β-CreER^T2::ROSA-iDTR mice where diphtheria toxin receptor (DTR) expression was induced by tamoxifen on mural cells, which were subsequently ablated using diphtheria toxin (DTx; Fig. 2 A). After tamoxifen administration, mural cells represented over 80% of DTR⁺ cells in the dura at the mRNA level (47.3% pericytes and 33.3% smooth muscle cells; Fig. 2 B), and ∼75% of cells expressing DTR at the protein level were NG2⁺ pericytes (Fig. S2 A), with ∼20% α smooth muscle actin (αSMA)⁺ vascular smooth muscle cells (Fig. S2 A). DTR was expressed by similar cell populations in the leptomeninges and brain, with slightly more fibroblasts expressing DTR in the leptomeninges (Fig. S2 B). Notably, 11.1% of DTR⁺ cells in the dura were identified as fibroblasts; however, the expression level of DTR mRNA was lower in fibroblasts compared with mural cells (Fig. 2 B). Using a mild dose of DTx, we partially depleted mural cells in the dura (Fig. 2, C and D) without decreasing other cell populations in the dura, leptomeninges, or brain (Fig. S2 C), including fibroblasts, which is important given their critical role in CNS biology (Dorrier et al., 2022). Additionally, using a mild dose of DTx did not cause overt disruption of the BBB or blood vessel integrity (Fig. 2 E and Fig. S2, D–F). Moreover, unlike with chronic depletion of mural cells, including pericytes (Török et al., 2021), our mild depletion strategy did not induce massive peripheral immune cell infiltration into the meninges (Fig. S2 G). To monitor homeostatic patrolling of T cells, we injected unstimulated T cells intravenously (IV) 2 d after DTx. We injected 2D2 T cells to test if there was a CNS antigen-specific T cell response or OT-II T cells (in which TCRs are restricted on chicken ovalbumin [OVA]) to control for an antigen-independent T cell response. 3 d later, we detected 2D2 T cells, but not OT-II or wild-type T cells, in the dura of mural cell–depleted mice (Fig. 2, F and G; and Fig. S2 H). CD4⁺ 2D2 T cells that were retained in the dura were skewed to a T helper 17 (T_H17) subset, suggesting they may represent a population relevant for CNS pathology (Fig. 2 H and Fig. S2 J). Since the dura and leptomeninges differ in barrier properties and play different roles under homeostatic and inflammatory conditions, we analyzed each tissue separately. While transferred 2D2 T cells were retained in the dura, trafficking was selective as we did not detect 2D2 T cells in the leptomeninges or the brain parenchyma (Fig. 2 I and Fig. S2 K; and Fig. S3). This could be due to the expression of tight junctions or a lack of inflammation-induced adhesion molecules expressed within the parenchyma and leptomeninges (Ransohoff et al., 2003; Piccio et al., 2002). We did detect 2D2 T cells in the choroid plexus; however, this trafficking was independent of mural cell depletion as their presence was similarly detected in non-transgenic mice treated with DTx (Fig. S3). To test if these processes could impact CNS autoimmunity, we partially depleted mural cells and then adoptively transferred activated T_H17-skewed 2D2 T cells. Mice with partial mural cell depletion were more susceptible to EAE (Fig. 2 J).

Macrophages throughout the meninges and within the perivascular spaces of the CNS homeostatically express markers typically associated with immunosurveillance and tissue healing (Galea et al., 2005; Holder et al., 2014; Zeisel et al., 2015). In MS/EAE and other neuroinflammatory diseases of the CNS, macrophages associated with the CNS, including dural macrophages, become proinflammatory and upregulate MHC class II, co-stimulatory molecules, and chemokines (Lapenna et al., 2018; Li et al., 2023). The signals that maintain perivascular and meningeal macrophages in surveillance mode or skew them to a proinflammatory phenotype are unknown. In presymptomatic EAE, genes differentially expressed in dural macrophages were enriched for pathways associated with antigen presentation and the activation of lymphocytes (Fig. 3 A). These pathways were not enriched in macrophages from the leptomeninges or the brain until disease onset, suggesting early events in the dura may be regulating a T_H cell response.

To investigate if macrophages in the dura interact with neighboring mural cells, we generated Pdgfr-β-CreER^T2::ROSA-tdTomato::Cx3cr1^GFP mice. In the dura, this mouse line expresses GFP in CX3CR1⁺ macrophages and tdTomato expression is induced via tamoxifen in platelet-derived growth factor receptor β (PDGFRβ)⁺ mural cells (Fig. 3 B). Using two-photon microscopy to image the dura of living mice, we found macrophages in proximity to mural cells near blood vessels (Fig. 3 C). Macrophages in the dura physically contacted mural cells and, moreover, we detected mural cell–derived tdTomato signal within GFP⁺ macrophages (Fig. 3 D). Using flow cytometry, we confirmed that dural macrophages took up mural cell–derived tdTomato protein (Fig. 3 E and Fig. S4 A), and this was not due to off-target expression of tdTomato mRNA (Fig. S4 B). The greatest portion of tdTomato⁺ cells were CD206⁺ macrophages, followed by CD206⁻ macrophages, and a small number of dendritic cells (Fig. S4, A–D). Interestingly, this phenomenon was not isolated to the dura, as populations of macrophages in the leptomeninges and brain also contained tdTomato (Fig. 3 E). Once CD4⁺ T cells infiltrate into tissue, they must be reactivated by APCs prior to inducing effector function (Schetters et al., 2018). In the dura, macrophages express low levels of MHC class II and thus can capture CNS antigens and present these antigens to patrolling T cells (Kivisäkk et al., 2009; Rustenhoven et al., 2021); however, under homeostatic conditions, T cells do not undergo pathological reactivation. Under a robust inflammatory condition like EAE, deleting MHC class II on microglia, as well as macrophages in the leptomeninges (both of which arise from yolk-sac progenitors) did not affect disease (Jordão et al., 2019; Mundt et al., 2019; Wolf et al., 2018). Little is known how mural cells and macrophages (and other APCs) in the dura regulate CNS immune surveillance under homeostatic conditions and in early CNS autoimmune pathology. We found that dural macrophages and dendritic cells that engulfed mural cell–derived tdTomato expressed lower levels of MHC class II (Fig. S4, C and D). Conversely, when we partially depleted mural cells, APCs in the dura (macrophages and dendritic cells) upregulated MHC class II and costimulatory molecules OX40L and ICOSL (Fig. 3 F and Fig. S4 E).

We hypothesized resident macrophages are critical to retain CNS antigen-specific T cells that infiltrated into the dura after partial depletion of mural cells. To test this, we depleted resident macrophages with PLX3397, a colony-stimulating growth factor 1 receptor antagonist (Fig. 3 G). As expected, PLX3397 depleted CD206⁺ macrophages in the dura while Ly6C^hi macrophages were not affected (Fig. S4 F). Likewise, PLX3397 depleted microglia in the brain, as well as resident macrophages and dendritic cell populations in the dura, deep cervical lymph nodes, and the spleen (Fig. S4 F). After depletion, we found reduced numbers of 2D2 T cells infiltrating into the dura (Fig. 3, H and I). These data demonstrate that resident macrophages, and potentially other APCs, in the dura contribute to retaining CNS antigen-specific T_H cells after partial mural cell depletion.

To further study the cellular interactions between mural cells and macrophages and the implications they have for regulating meningeal T cells, we utilized an in vitro assay with cultured CNS-associated mural cells (i.e., mural cells from the brain or the dura; Fig. S5, A and B). Mural cells cultured from either brain or dura were positive for canonical pericyte markers NG2, CD13, Nestin, and PDGFRβ (Birbrair et al., 2013; Kunz et al., 1994; Ieronimakis et al., 2013; Winkler et al., 2010) and for pericyte and vascular smooth muscle cell marker αSMA (Kumar et al., 2017), but did not express the endothelial marker CD31, mesenchymal stromal cell (MSC) marker collagen I, or the fibroblast marker PDGFRα (Fig. S5, A and B). Cultured brain mural cells suppressed the proliferation of CD4⁺ T cells within a bulk population of splenocytes; however, they failed to suppress the proliferation of isolated T cells, suggesting that mural cells could not directly suppress a T_H cell response alone, but rather they depended on other cell types for this function (Fig. 4 A). Much like how cytoplasmic components of mural cells transferred to macrophages in the dura (Fig. 3, C–E), cytoplasmic components from live cocultured brain mural cells were detected within splenic macrophages (Fig. 4 B). Macrophages that took up cytoplasmic components of brain mural cells downregulated genes important for antigen presentation and the costimulation of T cells (Fig. 4 C and Fig. S5 D). Similarly to our in vivo data in the dura and in vitro data using splenocytes, we found cytoplasmic components of cultured brain mural cells in CD11b⁺/CD206⁺ macrophages and, to a lesser extent, CD206⁻ macrophages and dendritic cells cocultured with mural cells after isolation from the dura (Fig. S5 E). Once CD11b⁺ dural cells interacted with mural cells, they retained the ability to suppress T_H cells, even after mural cells were removed (Fig. 4, D and E). Like mural cells cultured from the brain, mural cells cultured from the dura preconditioned CD11b⁺ dural cells to suppress the activation of T_H cells (Fig. S5 C). Unlike mural cells, cultured brain endothelial cells lacked the ability to affect a T_H cell response (Fig. S5 F). Although the majority of cells in the dura that engulfed cytoplasmic components of mural cells were macrophages (Fig. S4 D), a small population of dendritic cells was tdTomato⁺. Exposing bone marrow–derived dendritic cells to mural cells also suppressed their ability to present MOG antigens to T cells in vitro (Fig. S5 G). These data suggest that, although mural cells have a greater capacity to interact with macrophages than dendritic cells in the dura, the ability for mural cells to suppress a T_H cell response is not cell type specific.

To test the functional effect of transferring cytoplasmic components of mural cells to macrophages, we cocultured labeled CNS mural cells with isolated splenic macrophages and sorted macrophages based on their ability to take up cytoplasmic components of labeled mural cells with fluorescence-activated cell sorting (FACS; Fig. 4 F and Fig. S5 H). Preconditioned macrophages that took up cytoplasmic components of mural cells suppressed the proliferation of CD4⁺ T_H cells, whereas preconditioned macrophages without cytoplasmic components of mural cells failed to suppress T_H cells (Fig. 4 G). In addition to suppressing the proliferation of activated T_H cells, preconditioned macrophages suppressed CD4⁺ T cells skewing to T_H17 cells (Fig. 4, H and I; and Fig. S5 I). No effects were observed for skewing of T_H1 cells (Fig. 4, J and K). Therefore, we concluded that mural cells could program macrophages to block the proliferation and differentiation of T_H17 cells.

CNS-associated mural cells can physically connect to and communicate with neighboring cells to control local functions. They can form tunneling nanotubes with endothelial cells, other mural cells, and macrophages and have been shown to pass cytoplasmic components such as vesicles, and even larger organelles (e.g., mitochondria) to target cells (Errede et al., 2018; Alarcon-Martinez et al., 2020). We previously identified a novel form of communication between MSCs and lung macrophages involving the cytoplasmic transfer of processing bodies (PBs) that reprogram inflammatory macrophages to suppress a T_H response (Min et al., 2021). This was unlike efferocytosis (i.e., where the targets are apoptotic cells) and involved transferring PBs between two living cells. PBs are membrane-less organelles that contain RNA-binding proteins, microRNAs, and mRNAs enriched for regulatory functions (Standart and Weil, 2018) and have the potential to extensively reprogram a cell’s transcriptome (Loll-Krippleber and Brown, 2017). MSCs have a perivascular origin; express canonical markers PDGFRβ, NG2, CD90, CD146, and CD44; and have many overlapping functions with pericytes (Smyth et al., 2018; Crisan et al., 2008). Like with MSCs, we detected an abundance of PBs in mural cells in the dura (Fig. 5 A). Macrophages in the dura extended projections that contacted mural cells and the highest levels of PBs within dural macrophages were detected at these contact sites (Fig. 5, B and C), suggesting a potential location to transfer PBs. To further assess the ability of mural cells to transfer PBs, we labeled PBs in cultured brain mural cells using GFP-DCP1A (Fig. 5 D) and incubated them with splenic macrophages (Fig. 5 E). After 2 h, we detected GFP-DCP1A within macrophages (Fig. 5, F and G). To investigate if PBs in mural cells contributed to reprograming macrophages, we knocked down DDX6 (which is critical for the formation of PBs) using siRNA to deplete PBs in mural cells (Fig. 5, H and I). After coculturing DDX6-knocked-down mural cells with macrophages, the number of DCP1A puncta transferred to macrophages decreased (Fig. 5 J). Additionally, the ability of macrophages to suppress a T_H response was blocked when preconditioned with mural cells lacking PBs (Fig. 5 K).

In summary, our data suggest a novel mechanism for regulating the trafficking and skewing of CNS antigen-specific T cells in the dura. Mural cells, including pericytes, maintain dural macrophages in a surveillance phenotype through direct cell-to-cell contact and transfer of cytoplasmic material. In presymptomatic EAE, mural cells in the brain, leptomeninges, and dura were transcriptionally altered, and mural cell coverage along blood vessels decreased in the dura. In the dura and leptomeninges, mural cells upregulated cell projection organization and leukocyte migration pathways, and specifically in the dura, macrophages were enriched for antigen processing and presentation pathways in presymptomatic EAE. Partially depleting mural cells in mice demonstrated that interactions between mural cells and macrophages were important to suppress antigen-dependent T cell trafficking to the dura. At this time, we can only speculate that T cells are homeostatically circulating through tissues, and CNS antigen-specific T cells are being retained in the dura via cognate antigen engagement with local APCs (Radjavi et al., 2014). Mural cells transfer cytoplasmic components, including PBs, to dural macrophages to suppress the pathological skewing of T_H17s. Susceptibility variants in surrounding genomic DNA of DDX6 have been identified by genome-wide association studies for MS (Disanto et al., 2014; International Multiple Sclerosis Genetics Consortium, 2013), and this mechanism could contribute to the initial steps in the dura, breaking immune tolerance in CNS autoimmune diseases.

C57BL/6, Pdgfr-β-CreER^T2, ROSA-iDTR, 2D2 TCR, OT-II TCR, SOD1-G93A, and Cx3cr1^GFP transgenic mice were originally purchased from Jackson Laboratories and bred in house. Purchased mice were given at least 1 wk to habituate prior to breeding or starting an experiment. Mice were randomly assigned to groups and the experimenter was always blinded to genotypes and conditions. All experiments were performed in accordance with Duke University Institutional Animal Care and Use Committee’s policies.

To induce active EAE, 8–10-wk-old male C57BL/6 mice were immunized by subcutaneous injection of MOG peptide (MOG_35–55) emulsified in Complete Freund’s adjuvant (CFA; Hooke Laboratories, LLC). On the day of immunization and 24 h later, mice received 200 ng of pertussis toxin (intraperitoneally [IP]; Hooke Laboratories, LLC). Mice were observed daily and euthanized at appropriate time points (i.e., presymptomatic on day 7 and disease onset around day 13–15).

To induce passive EAE, mice received in vitro–differentiated T_H17 (5 × 10⁶ cells per mouse) by intravenous injection, and typical or atypical EAE symptoms were monitored using the following scoring scale. Typical EAE: 0, no detectable symptoms; 1, complete tail paralysis; 2, wobbly gait; 3, complete hind limb paralysis; 4, complete hind and fore limb paralysis or moribund; and 5, dead. Atypical EAE symptom: 0, no detectable signs; 1, tail paralysis, hunched appearance, unsteady walk; 2, ataxia, head tilt, hypersensitivity; 3, severe ataxia, spasticity or knuckling, severe proprioception defects; 4, moribund; 5, dead. Daily scores for an individual mouse reflect the more severe readout of either set of symptoms.

To isolate dura as previously described (Louveau et al., 2018a), mice were euthanized and transcardially perfused with PBS containing heparin and, for some experiments, 2% PFA. If necessary, skull caps and brains were further submerged in 4% PFA for 24 h at 4°C. Skull caps were washed with PBS and dura carefully removed from the skull cap under a dissection microscope for downstream procedures. To collect leptomeninges, brains were put in 30-mm dish with PBS and leptomeninges were carefully peeled from the brain surface using fine forceps under a dissection microscope. The PBS in the dish was collected in a tube and leptomeninges were collected by centrifugation. For flow, tissues from two mice were pooled for downstream digestion. Cortex was collected from brains after the leptomeninges were removed. To detach dura from spinal cord, tissues were isolated from spine and fixed using 4% PFA overnight at 4°C. Outer membranes collected from spinal cord were processed for further analysis.

To determine the mural cell coverage on blood vessels, the dura was stained using tomato lectin (TL) with anti-CD13 or anti-NG2 antibodies and imaged using confocal microscopy (SP8 Upright Confocal; Leica). CD13⁺ or NG2⁺ area was measured on TL⁺ blood microvessels (defined as vessels <50 μm in diameter).

8-wk-old male or female C57BL/6 mice were immunized by subcutaneous injection of MOG peptide (MOG_35–55) emulsified in CFA. On the day of immunization and 24 h later, mice received 200 ng of pertussis toxin (IP). 7 or 14 d after MOG injection, MOG-treated mice and naïve control mice were sacrificed for collecting dura, leptomeninges, and brains. Tissue from n = 3 mice was pooled and processed for RNA-sequencing. Tissues were enzymatically digested in EBSS solution (GIBCO) containing 30 U/ml papain (Worthington Biochemical) and DNase I (Sigma-Aldrich) for 30 min at 37°C, triturated using a Dounce homogenizer, and passed through a 50-μm cell strainer. Brain samples were resuspended in 30% Percoll (Sigma-Aldrich) and centrifuged for 10 min at 700 g. The supernatant was discarded to avoid myelin debris. The pelleted cells were resuspended in PBS containing 0.5% FBS.

Dissociated cells were loaded into a 10X Genomics Chromium Controller, and sequencing libraries were prepared according to the manufacturer’s instructions (10X Genomics). The resulting libraries were sequenced using the Illumina NovaSeq 6000 platform. Raw sequence data were processed using Cell Ranger software (10X Genomics) to align sequences to the mm10 reference genome supplemented with the Pan troglodytes heparin-binding EGF-like growth factor (HBEGF) gene, annotate genes, and count reads.

Raw counts were normalized and cells were clustered by expression patterns using the Seurat package in R (Stuart et al., 2019). Identification of cell types clustered together using Uniform Manifold Approximation and Projection and t-distributed stochastic neighbor embedding was established by using cell type–specific markers (Fig. 1 D). Differential gene expression was determined using variance-stabilizing transformation followed by analysis with the Wilcoxon Rank Sum test with Benjamini–Hochberg correction for false discovery rate. Enrichment of biological processes was determined using Gene Ontology (GO) Enrichment Analysis (Ashburner et al., 2000; Gene Ontology Consortium et al., 2023; Thomas et al., 2022).

To deplete mural cells, 4–6-wk-old Pdgfr-β-CreER^T2::iDTR mice received 75 mg/kg of tamoxifen (Sigma-Aldrich) once a day (IP) for 3 d. 2 wk later, mice received a single injection (IP) of DTx (0.2 μg/mouse; Cayman Chemical Company).

To test blood vessel integrity, control or mural cell–depleted mice received fluorochrome-conjugated dextran (10, 40, or 70 kD; Thermo Fisher Scientific) by IV injection. 15 min after injection, mice were sacrificed and dura, brain, and spinal cord were isolated. Tissues were briefly fixed in 4% PFA, and brains and spinal cords were frozen in OCT and sectioned on a Leica cryostat (15–18-μm sections). Tissues were stained using DyLight 649-conjugated TL, and images were acquired using a Leica SP8 upright confocal microscope (Leica Microsystems, Inc). For controls, cryogenic lesions were induced by placing a liquid nitrogen–cooled metal rod on the surface of the exposed skull. To measure blood vessel permeability, dextran⁺ area and TL⁺ blood vessel area were quantified from each image and a permeability index was calculated as (dextran⁺ area − TL⁺ area)/TL⁺ area.

Mouse dura was detached from skull cap, fixed in 4% PFA, and incubated in blocking buffer (5% normal serum, 2% BSA, and 0.1% Triton X-100 in PBS). Other tissues, including brains, spinal cords, and diaphragms, were similarly fixed and washed in PBS prior to submerging in 30% sucrose. Tissues were frozen in OCT and sectioned on a Leica cryostat (15–18-μm sections). Free-floating tissues were labeled with antibodies raised against CD13, IBA1, DCP1A, NG2, or αSMA in a blocking buffer. After incubating with fluorochrome-conjugated secondary antibodies (1:200) and fluorochrome-conjugated TL, tissues were mounted on a slide and covered with VectaShield medium (Vector Labs) and coverslips. Images were acquired using a Leica SP8 upright confocal microscope or Zeiss LSM780 Inverted Confocal with AiryScan. Z-stacks were analyzed using Imaris Bitplane software 9.1.2 (Oxford Instrument) for 3D reconstruction.

Mural cells or MSCs were plated on collagen-coated coverslips and, 2 d later, fixed in 4% PFA and incubated in blocking buffer. Cells were labeled using anti-collagen I or DCP1A antibodies and incubated with fluorochrome-conjugated secondary antibodies in blocking buffer. Macrophages that were cocultured with mural cells were removed by gentle pipetting, plated on slide glass using cytospin (1,200 rpm for 5 min), and fixed with 4% PFA. Cells were incubated with blocking buffer and labeled using anti-IBA1 antibody. After incubating with fluorochrome-conjugated secondary antibodies, cells were mounted with VectaShield and imaged using a Leica SP8 upright confocal microscope.

Lymph nodes and spleens were isolated from donor mice and pooled. Red blood cells were removed with ACK lysing buffer, and the remaining cells were washed with PBS prior to isolating CD4⁺ T cells via MACS columns (CD4⁺ T Cell Isolation kit; Miltenyi). T cell purity was confirmed by flow cytometry, and cells (5 × 10⁶ cells of each genotype) were transferred to recipient mice via tail vein.

To analyze T cell subsets in the dura, mice were sacrificed and perfused as described above. Dura were isolated from skull caps and digested with 1 mg/ml Collagenase D (Sigma-Aldrich) and 0.5 mg/ml DNase at 37°C for 15 min. Cells were filtered through a 70-μm cell strainer and fixed using 2% PFA. Cells were permeabilized using FACS permeabilizing buffer (2% FBS, 0.1% saponin in PBS) and labeled with antibodies raised against CD4, T-bet, GATA3, and RORγt. For restimulation, cells were resuspended in a T cell culture medium containing PMA (100 nM; Sigma-Aldrich), ionomycin (1 μg/ml; Sigma-Aldrich), and monensin (2 μM; BioLegend) for 5 h. Cells were then washed with FACS buffer, fixed using 2% PFA, washed, and permeabilized before labeling with antibodies raised against CD45.1, CD4, IFN-γ, and IL-17A. Cells were analyzed using a FACSLyric (BD Bioscience), and data were analyzed using the BD FACSuite software (BD Bioscience).

4-wk-old Pdgfr-β-CreER^T2::ROSA-tdTomato::Cx3cr1^GFP mice received 75 mg/kg of tamoxifen and, 2 wk later, were anesthetized and imaged using two-photon microscopy. Mice were anesthetized with ketamine (85 mg/kg), xylazine (13 mg/kg), and acepromazine (2 mg/kg) in PBS and maintained at a core temperature of 37°C. Two-photon imaging through the thinned skull was adapted from methods previously described for cortical imaging (Yang et al., 2010). Briefly, a metal bracket containing an imaging window was glued to the skull. The skull bone above the cortex was thinned to ∼30 μm within the imaging window. Imaging was performed with a water-dipping 25×/1.0 IRAPO W Immersion Objective with a motorized correction collar on a Leica DIVE resonance scanning system. An Insight 3 laser tuned to 910 nm was used to acquire seven optical slices 2.5 μm apart. Two HyD detectors were used to collect light between 470 and 513 nm (GFP) and 637–705 nm (tdTomato).

To deplete resident macrophages, mural cell–depleted mice (tamoxifen-treated Pdgfr-β-CreER^T2::iDTR mice) received 1.5 mg of PLX3397 (Selleck chemical) twice per day via IP injection. PLX3397 stock in DMSO was dissolved in buffer containing 5% DMSO, 45% PEG300, and 5% TWEEN 80 and administered to mice promptly after mixing. 2 d after mice started receiving PLX3397, mice received DTx. 2 d after DTx treatment, 5 × 10⁶ 2D2 T cells were intravenously administered. 3 d after 2D2 T cells were administered, mice were sacrificed and tissues collected for further analysis. For flow cytometry, the dura was digested in PBS containing 1 mg/ml Collagenase D and 0.5 mg/ml DNase at 37°C for 15 min with shaking. Tissues were ground using frosted slide glass and filtered through a 70-μm cell strainer. Cells were treated with Mouse BD Fc Block (BD Biosciences) and labeled with anti-CD11b, -Ly6G, -CD206, -MHC class II, -OX40L, and -ICOS (BioLegend). Cells were analyzed using a FACSLyric and data were analyzed using FACSuite and FlowJo (BD Biosciences) software.

Mouse brain mural cells were isolated and cultured as previously described (Boroujerdi et al., 2014). Briefly, brains or dura were collected from 6–8-wk-old mice and minced thoroughly with a sterilized razor blade. Minced tissues were enzymatically digested in EBSS solution containing 30 U/ml papain and 40 μg/ml DNase I for 70 min at 37°C. Tissues were then triturated by passing through an 18-gauge needle. 1.7 vol of 22% BSA (Sigma-Aldrich) was added and mixed. After centrifugation at 4,000 rpm for 10 min, cells were washed in endothelial cell growth medium (ECGM: 10% FBS, heparin, ascorbic acid, L-glutamine, penicillin/streptomycin, and endothelial cell growth supplement [Sigma-Aldrich] in Hams F12 [GIBCO]), suspended in ECGM, and plated on a 0.02% collagen I-coated 6-well plate. During the first two passages, cells were cultured in ECGM. Following the third passage, cells were maintained in pericyte medium (ScienCell Research Laboratories).

To culture dura mural cells, the dura was isolated from the skull cap and digested with 1 mg/ml Collagenase D and 0.5 mg/ml DNase at 37°C for 15 min and then ground using a Dounce homogenizer. Cells were filtered through a 70-μm cell strainer, washed, suspended in ECGM, and plated on a 0.02% collagen I–coated 6-well plate. Cells isolated from dura were cultured in ECGM until the second passage, when the culture medium switched over to pericyte medium.

For T cell suppression assays, T cells were isolated using MACS, as described above, from spleens and lymph nodes and labeled with 2.5 μM of carboxyfluorescein succinimidyl ester (CFSE; Thermo Fisher Scientific). T cells were activated with Dynabeads Mouse T-Activator CD3/CD28 (Thermo Fisher Scientific) for 3 d. Upon completion, cells were treated with Mouse BD Fc Block (BD Bioscience), washed, labeled with anti-CD4 antibody (BD Bioscience), and analyzed using a FACSLyric flow cytometer. Data were analyzed using BD FACSuite software.

To precondition splenocytes for suppression assays, T cells were depleted from splenocytes using the CD3ε MicroBead Kit (Miltenyi). T cell–depleted splenocytes were then cocultured with CFSE-labeled T cells, and these cells were stimulated using Dynabeads Mouse T-Activator CD3/CD28. For isolating T cells, the Pan T cell Isolation Kit II (Miltenyi) and a magnetic column were used to select T cells from mouse spleen and lymph nodes. Isolated T cells were stained using CFSE and cocultured with mural cell preconditioned macrophages for 3 d. Macrophages were isolated from either spleen or meninges (dura and arachnoid layers combined) with a CD11b MACS MicroBead kit (Miltenyi). Cells were labeled with anti-CD4 antibody and analyzed using a FACSLyric flow cytometer. Data were analyzed using BD FACSuite software. To isolate mural cells that had interacted with macrophages, Qtracker-labeled brain mural cells were plated on collagen-coated 6-well plates. The next day, splenocytes were added, centrifuged briefly (1,000 rpm for 5 min), and then cultured for 2 d. Splenocytes were then collected and labeled with anti-CD11b and -Ly6G antibodies. Qtracker positive or negative CD11b⁺Ly6C⁻ cells were sorted using a MoFlo Astrios Cell Sorter (Beckman Coulter).

For T_H1/T_H17 differentiation, CD4⁺ T cells were isolated from the mouse spleen and lymph nodes using a CD4⁺ T cell Isolation Kit and magnetic column (Miltenyi). Isolated CD4⁺ T cells were labeled using anti-CD44 and -CD62L antibodies and CD44⁻/CD62L⁺ naïve helper T cells were isolated using a Sony MA900 cell sorter. Cells were plated on a hamster IgG-precoated plate with anti-CD3 and anti-CD28 antibodies under T_H1 differentiation conditions or T_H17 differentiation conditions. 3 d later, cells were restimulated using PMA (100 nM) and ionomycin (1 μg/ml) with monensin (BioLegend) for 4 h. Cells were labeled using anti-CD4, -IL17A, and -IFNγ antibodies in permeabilization buffer (BioLegend). Cells were analyzed using a FACSLyric flow cytometer, and data were analyzed using the BD FACSuite software.

For pathogenic T_H17 differentiation for passive EAE induction, CD4⁺ T cells isolated from 2D2 mice were cultured on anti-hamster IgG (MP Biomedicals)-precoated plates with anti-CD3 (0.1 μg/ml; BD Biosciences), anti-CD28 (0.1 μg/ml; eBioscience), anti-IFNγ (2 μg/ml; BioLegend), anti-IL-4 (2 μg/ml; BioLegend) antibodies, IL-1β (20 ng/ml; BioLegend), IL-6 (10 ng/ml; BioLegend), and IL-23 (25 ng/ml; BioLegend) for 5 d. 2D2 T_H17 cells were restimulated on hamster IgG-precoated plates with anti-CD3 and anti-CD28 antibodies for 24 h and collected, washed using PBS, and resuspended in PBS for adaptive transfer to C57BL/6 mice.

For transient expression of fluorescent-labeled DCP1A in in vitro cultured brain mural cells, cells were plated on 24- or 48-well plates or collagen-coated coverslips. The following day, plated cell density was confirmed at 70–80% confluency, and transfection was performed using TransIT-2020 transfection reagent (Mirus Bio) following the manufacturer’s protocol. DNA complexes were prepared by combining DCP1A-GFP in the pACGFP-C1 vector (Addgene) with TransIT-2020 reagent (Mirus Bio) in Opti-MEM I reduced-serum medium (Thermo Fisher Scientific) at a DNA:reagent ratio of 1:3. The cells were cultured in Opti-MEM I reduced-serum media with the transfection reagent and DNA complexes. Images were taken 24 h after the transfection.

Mural cells isolated from mouse brain were plated on collagen-coated plates and, 1 d later, transfected with DDX6 Dicer-substrate or non-targeting negative control dicer-substrate short interfering RNAs (TriFECTa RNAi kit; IDT) using TransIT-2020 transfection reagent. DDX6 and PB depletion was confirmed using western blot analysis and imaging. 2 d after transfection, CD11b⁺ cells isolated from mouse spleen and lymph nodes using CD11b MicroBeads (Miltenyi) were plated on the mural cells and cocultured for 2 d. CD11b⁺ cells were isolated and mixed with CFSE-labeled mouse helper T cells isolated from spleen and lymph nodes and stimulated using Dynabeads Mouse T-Activator CD3/CD28. 3 d later, cells were labeled using anti-CD4 antibody, and CFSE intensity on helper T cells was measured using a FACSLyric flow cytometer.

Brain mural cells were labeled using Qtracker (Thermofisher) and plated on a collagen-coated 24-well plate. 1 d later, cells were washed twice and CD11b⁺ macrophages isolated by MACS separation from spleen using CD11b MicroBeads (Miltenyi) in RPMI containing 10% FBS. 3 d later, cells were lysed and RNA collected using the RNeasy kit (Qiagen). Dura, leptomeninges, and brain were collected from tdTomato⁺ mice and digested. Cells were labeled using anti-CD11b and -PDGFRβ antibodies. TdTomato⁺ or tdTomato⁻ CD11b⁺ cells and PDGFRβ⁺ cells were sorted using a Sony MA900 cell sorter. RNA was collected using the RNeasy kit. Complementary DNA was prepared using SuperScript VILO cDNA Synthesis Kit (Invitrogen). RNA expression was determined using TaqMan Fast Advanced Master Mix (Thermo Fisher Scientific). The following TaqMan probes were used: Ciita (assay ID: Mm00482914_m1), Rfx5 (assay ID: Mm01263513_g1), H2-DMb2 (assay ID: Mm00783707_s1), Ox-40l (assay ID: Mm00437214_m1), Tdtomato (assay ID: Mr07319439_mr), β-actin (Mm04394036_g1), and Gapdh (assay ID: Mm99999915_g1).

Control or DDX6^KD cells were lysed in radioimmunoprecipitation assay lysis buffer containing protease inhibitor (Sigma-Aldrich) and separated using SDS polyacrylamide gel. The protein was transferred to a polyvinylidene fluoride membrane and blocked with 5% skim milk in PBS containing 0.05% Tween 20 and blotted using anti-DDX6 and -GAPDH antibodies (Cell Signaling Technology). Subsequently, the blots were developed using the ECL Detection Kit (Bio-Rad Laboratories) and protein bands were visualized using a C-digit blot scanner (LI-COR Biosciences).

To generate bone marrow–derived dendritic cells, cells obtained from the mouse tibia were cultured in RPMI 1640 (GIBCO) containing L-glutamine, 25 mM HEPES, 10% FBS, penicillin/streptomycin, 55 μM β-ME (Sigma-Aldrich), and 100 ng/ml of Flt-3L (Peprotech). After 7 d, 2 × 10⁵ dendritic cells and 4 × 10⁵ CFSE-labeled CD4⁺ T cells were cocultured with 1 mg/ml MOG for 3 d. Intensity of CFSE on CD4⁺ T cells was measured using a FACSLyric flow cytometer.

The following antibodies were used: CD13 goat polyclonal antibody (1:200; R&D Systems), DCP1A rabbit polyclonal antibody (1:200; Abcam), IBA1 rabbit polyclonal antibody (1:200; FUJIFILM Wako Pure Chemical), DDX6 rabbit polyclonal antibody (1:5,000; Abcam), GAPDH rabbit monoclonal antibody (1:1,000, D16H11 clone; Cell Signaling), αSMA mouse monoclonal antibody (1A4 clone; R&D Systems), NG2 rat monoclonal antibody (1:200, 546930 clone; R&D Systems), B220 rat monoclonal antibody (1:100, RA3-6B2 clone; BioLegend), CD3ε mouse monoclonal antibody (1:100, 145-2C11 clone; BD Bioscience), CD4 rat monoclonal antibody (1:100, GK1.5 clone; BD Bioscience), CD11b rat monoclonal antibody (1:100, M1/70 clone; BD Bioscience), CD11c mouse monoclonal antibody (1:100, N418 clone; BioLegend), CD45 mouse monoclonal antibody (1:100, HI30 clone; BioLegend), CD45.1 mouse monoclonal antibody (1:100, A20 clone; BD Bioscience), CD206 rat monoclonal antibody (1:50, C068C2 clone; BioLegend), Ly6G rat monoclonal antibody (1:100, 1A8 clone; BioLegend), Ly6C mouse monoclonal antibody (1:200, HK1.4 clone; BioLegend), T-bet mouse monoclonal antibody (1:100, 4B10 clone; BioLegend), RORγt mouse monoclonal antibody (1:50, Q31-378 clone; BD Bioscience), GATA3 mouse monoclonal antibody (1:50, 16E10A23 clone; BioLegend), IL-17A mouse monoclonal antibody (1:100, TC11-18H10.1 clone; BioLegend), IFNγ rat monoclonal antibody (1:100, XMG1.2 clone; BD Bioscience), ICOSL rat monoclonal antibody (1:100, HK5.3 clone; BioLegend), and OX40L rat monoclonal antibody (1:100, RM134L clone; BioLegend).

The hanging wire test was performed by placing the mouse on the top of the conventional wire cage lid and then inverting it above the home cage. The inverted lid was held at a height of ∼30 cm and the latency of the mouse to fall was recorded (maximum of 300 s per trial).

Data were analyzed using the statistical methods stated in each figure legend using Prism (GraphPad). In general, comparison of multiple groups was performed using one-way ANOVA followed by a post hoc Tukey’s test. Comparisons of two data sets were performed using two-tailed unpaired t tests. EAE data were analyzed using a two-way repeated measures ANOVA. P values <0.05 were considered significant. Group sizes were determined via power analysis or prior experiments. Additional statistical information has been listed in the associated data sheet supplied with this manuscript (Table S1).

Fig. S1 shows mural cell counts and coverage in mice with EAE, G93A-SOD1 mutation, and in aged mice. Fig. S2 shows additional readouts characterizing the impact of partial mural cell depletion. Fig. S3 contains representative images tracking T cells, after adoptive transfer, in numerous tissues after partially depleting mural cells. Fig. S4 shows additional data assessing APCs that uptake cytoplasmic components of mural cells and which immune cells are depleted after PLX3397. Fig. S5 shows data characterizing mural cell cultures and additional data relating to in vitro T cell suppression assay presented in Fig. 5. Table S1 lists additional statistical information.

Single-cell sequencing data generated and presented in Figs. 1, 2, 3, and S2 are available from Gene Expression Omnibus (GSE242720). All other source data are included in the associated data sheet supplied with this manuscript (Table S1).

The authors would like to thank all members of the Filiano lab and members of MC3 for their thoughtful discussion of the work in this manuscript. We would also like to thank Drs. Saban and Valdivia at Duke University for their generous support and reagent sharing, and Dr. Jesse Troy for consulting on statistics. Additionally, we would like to acknowledge the Duke Molecular Physiology Institute Molecular Genomics core for the assistance generating sequencing libraries. Our illustrations were created with https://BioRender.com. Graphical abstract was created by Lauren Begg.

This work was supported by grants from the National Institutes of Health (NS123084) and grants from the Hartwell Foundation, the Cord Blood Association Foundation, and the Marcus Foundation.

Author contributions: H. Min, S.M. O’Neil, and A.J. Filiano conceived the study, designed and performed experiments, analyzed data, and wrote the manuscript. L. Xu and E.A. Moseman performed experiments and edited the manuscript. J. Kurtzberg provided intellectual guidance and edited the manuscript.