GM-CSF Promotes the Survival of Peripheral-Derived Myeloid Cells in the Central Nervous System for Pain-Induced Relapse of Neuroinflammation

It has been genetically shown that multiple sclerosis (MS) is a CD4⁺ T cell–mediated autoimmune disease of the CNS and can be divided into at least four types: (1) relapse-remitting MS (>80% of patients), (2) secondary progressive MS (the late phase of relapse-remitting type), (3) primary progressive MS (∼10% of patients), and (4) relapse-progressing MS (∼5% patients) (1–4). MS inflammatory lesions are composed of various types of cell, including peripheral-derived CD4⁺ T cells and MHC class II⁺ (MHC II⁺) CD11b⁺ myeloid cells, and result in the impairment of the blood-brain barrier (BBB) followed by the dysregulation of neurologic functions caused by a loss of myelin and disruption of neural circuits (5–9).

It has been reported that GM-CSF is required for the induction of active experimental autoimmune encephalomyelitis (EAE) (10), and that the transfer of GM-CSF–deficient CD4⁺ T cells fails to induce a transfer model of EAE (tEAE) (11, 12), suggesting that GM-CSF from activated CD4⁺ T cells is critical for the development of tEAE. In contrast, it was reported that GM-CSF is produced by nonhematopoietic cells, including blood endothelial cells (BECs), in a manner dependent on IL-1β during the development of active EAE (13, 14).

We found that the activation of specific neural circuits triggered by environmental stimulations, such as pain, gravity, stress, light, and inflammation, or artificial neural stimulations establish gateways for immune cells, including autoreactive CD4⁺ T cells at specific blood vessel sites in organs, to develop tissue-specific inflammatory diseases (15). Pain is often associated with MS and is sometimes a determinant of MS activity (16, 17). In tEAE relapse, pain activates neural circuits to express the chemokine CX3CL1 from endothelial cells and MHC II^HiCD11b⁺ cells in a manner dependent on norepinephrine-mediated IL-6 amplifier activation to breach the BBB, particularly in the presence of myelin oligodendrocyte glycoprotein (MOG)-specific autoreactive CD4⁺ T cells (18). During the first clinical sign of tEAE, immune cells, including MOG-specific autoreactive CD4⁺ T cells and MHC II^HiCD11b⁺ cells, invade the fifth lumbar (L5) spinal cord from the bloodstream via the gravity gateway reflex (15). These MHC II^HiCD11b⁺ cells express CX3CR1, a receptor for CX3CL1, and survive a long time during the remission phase, in which there are no clinical signs (18). Because they are not activated microglia, but activated monocytes with a peripheral origin (18), we named these cells “peripheral-derived myeloid cells.” Pain-mediated sensory–sympathetic cross-talk recruits these cells around the ventral vessels of the L5 spinal cord in a manner dependent on the norepinephrine-CX3CL1 axis. This recruitment compromises the BBB and presents MOG peptides, allowing for the invasion and activation of autoreactive CD4⁺ T cells into the CNS and tEAE relapse (18). Because the L5 cord is the first inflammation site (15) and peripheral-derived myeloid cells are sustained in high numbers there, tEAE relapse begins from the L5 cord even though pain-mediated neural circuits induced a weak norepinephrine-CX3CL1 axis around ventral vessels of all spinal cord levels (18).

In this study, we found that GM-CSF molecules constitutively expressed from BECs in the CNS are important for the survival of peripheral-derived myeloid cells, suggesting the GM-CSF axis in the CNS could be a promising therapeutic target for relapse-associated CNS diseases, including MS.

C57BL/6 mice 6–8 wk old were obtained from Japan SLC (Tokyo, Japan). All mice were maintained under specific pathogen-free conditions according to the protocols of Hokkaido University. The animal experiments used in this study were approved by the Institutional Animal Care and Use Committees of Hokkaido University.

EAE was induced as previously described (15, 18, 19). In short, C57BL/6 mice were immunized s.c. with MOG 35–55 peptide (Sigma-Aldrich) emulsified in CFA (Sigma-Aldrich) on day 0. Pertussis toxin (Sigma-Aldrich) was injected i.v. on days 0, 2, and 7. On day 9, CD4⁺ T cells were isolated from immunized mice with anti-CD4 microbeads (Miltenyi Biotec). CD4⁺ T cells were cocultured with rIL-23 (1 ng/ml; R&D Systems), irradiated splenocytes, and MOG peptide (25 μg/ml; Sigma-Aldrich). After 2 d, CD4⁺ T cells were isolated and injected i.v. into normal mice (1.5 × 10⁷ cells). The severity of EAE was monitored and scored daily: 0, normal; 1, limp tail; 2, mild paralysis of the hind limbs with uneven gait; 3, rear limbs paralysis; 4, front and rear limbs paralysis; and 5, moribund (15, 18, 19).

Surgical procedures were performed under anesthesia using isoflurane. The left middle branch of the trigeminal nerve was exposed and ligated loosely with a polyglycolic acid suture (Akiyama Medical), as previously reported (18). In the sham surgery, the trigeminal neuron was exposed, but not ligated with the suture.

The spinal cord and brain were collected after perfusion with PBS and embedded in SCEM compound (20) (SECTION-Lab), and frozen sections were made using the microtome device CM3050 (Leica Microsystems), as reported previously (15, 18). Nonspecific staining was blocked with PBS or TBS-T containing 2% BSA (Sigma-Aldrich), and sections were incubated overnight at 4°C with primary Ab in the presence of anti-CD16/CD32 (2.4G2) for Fc receptor blocking. The primary Abs used were anti–MHC II (M5/114.15.2; eBioscience) and antiphosphorylated c-Fos (Cell Signaling Technology). Alexa Fluor 488–labeled secondary Abs (Life Technologies) were used to detect the primary Abs, followed by staining of the nucleus using Hoechst 33342 (Life Technologies). Sections were observed with a BZ-9000 microscope, and images were analyzed using HSALL software (KEYENCE).

Mice deeply anesthetized with pentobarbital were perfused with saline, followed with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). The vertebral columns with the spinal cords were dissected and immersed in 5% EDTA for 2–3 wk for decalcification. The fifth lumbar vertebrae were dissected and sagittally cut. They were then rinsed with PBS and put into 0.03% Triton X–containing PBS (Triton-PBS) for 2 h at 4°C. The tissues were blocked with Block Ace (DS Pharma Biomedical, Osaka, Japan) diluted 1:2 in Triton-PBS for 2 h and incubated with rat anti–MHC II Ab (M5/114.15.2; eBioscience) and rabbit anti-collagen type IV Abs (ab19808; Abcam) diluted in Triton-PBS for 3–5 d at 4°C. The tissues were incubated with Cy3-labeled anti-rat IgG and Alexa Fluor 488–labeled anti-rabbit IgG diluted in Triton-PBS in the dark for 4 h at room temperature. The stained spinal cords were put on a slide glass with the meninges side up and observed under a confocal laser microscope (FV300; Olympus, Tokyo, Japan).

The following experiments were performed under anesthesia using pentobarbital. Brains and spinal cords were harvested after perfusion with PBS. Samples were dissociated with RPMI 1640 (Thermo Fisher Scientific) containing 10% FCS (Thermo Fisher Scientific) and Dispase I (1500 U/ml; Wako). 30% Percoll (GE Healthcare) was then added, and the mixture was centrifuged at 2500 rpm for 15 min to remove myelin. The cells were seeded in a 96-well plate with RPMI 1640 containing 10% FCS and cultured for 36 h. GolgiPlug (BD Biosciences) was added, and the cells were cultured for another 12 h. To stimulate CD4⁺ T cells, we added PMA (1 μg/ml) and ionomycin (1 μg/ml) 6 h before the culture finished.

Mice were anesthetized with pentobarbital and perfused with PBS. Brains and spinal cords were harvested and digested enzymatically using RPMI 1640 containing FCS and Dispase I. Cells were stained with the following Abs: anti–CD4-allophycocyanin (RM4-5; BioLegend), anti–CD4-PE (RM4-5; BioLegend), anti–CD4-PacificBlue (RM4-5; eBioscience), anti–CD8a-PE/Cyanine7 (53-6.7; BioLegend), anti–CD11b-FITC (M1/70; BioLegend), anti–CD11b-PE (M1/70; BioLegend), anti–CD11b-PerCP (M1/70; BioLegend), anti–CD19-BV421 (6D5; BioLegend), anti–CD19-allophycocyanin (6D5; BioLegend), anti–Ly-6C-allophycocyanin (HK1.4; BioLegend), anti–GM-CSFRα-PE (698423; R&D Systems), anti-common β-chain-biotin (REA193; Miltenyi Biotec), anti–CD45-PE (30-F11; BD Pharmingen), anti–MHC II-allophycocyanin (M5/114.15.2; BioLegend), anti–MHC II-FITC (M5/114.15.2; eBioscience), anti–CD31-allophycocyanin (390; BioLegend), anti-TCR β-FITC (H57-597; eBioscience), anti-TCR β-PE (H57-597; BioLegend), anti-TCR β-biotin (H57-597; eBioscience), anti-γ δ TCR-biotin (eBioGL3; eBioscience), anti–IgD-biotin (11-26c.2a; BioLegend), anti–Podoplanin-biotin (NC-08; BioLegend), streptavidin-BV421 (BioLegend), Rat IgG2a-PE (eBR2a; eBioscience), and Human IgG1-biotin (Ancel). For intracellular staining, cells were surface stained and fixed with Cytofix/Cytoperm (BD Bioscience) for 10 min on ice. After fixation, anti–Bcl-xL-PE (54H6; CST), anti–Bcl-2-PE (BCL/10C4; BioLegend), and anti–GM-CSF–FITC (MP1-22E9; BioLegend) Abs diluted in Perm/Wash buffer (BD Biosciences) were stained for 1 h on ice. For the analysis, cyan flow cytometry (Beckman Coulter) was used. To purify the cells, we used Moflo cell sorters (Beckman Coulter). Cell purity was routinely >98%. The data were analyzed with FlowJo software (Tree Star).

Mice were anesthetized with isoflurane before the surgery. Approximately 20 μl of CSF was collected from the cisterna magna using a glass capillary. GM-CSF was measured using the Milliplex cytokine/chemokine magnetic bead panel (Merck Millipore) according to the manufacturer’s instructions. In some experiments, the GM-CSF concentration in the cell-culture supernatant was determined with an ELISA kit (BD Biosciences) according to the manufacturer’s protocol. The detection range was 15.625–1000 pg/ml. Concentrations were calculated using a standard curve generated with specific standards provided by the manufacturer. Each sample was measured in duplicate.

Frozen sections were made and fixed with a Tissue FIX container (PAXgene; Qiagen) for a few minutes. After washing with saline, the parenchymal and submeningeal areas of the L5 cord were collected using the microdissection device DM6000B (Leica Microsystems). Harvested tissues were immediately lysed in lysis buffer from an RNeasy Micro Kit (Qiagen) for RNA extraction.

Total RNA was extracted using the RNeasy Micro Kit. cDNA was synthesized with M-MLV reverse transcriptase (Promega) after DNase I (Qiagen) treatment. Quantitative PCR (qPCR) was performed using KAPA PROBE FAST qPCR Master Mix (KAPA Biosystems) or KAPA SYBR FAST qPCR Master Mix (KAPA Biosystems). The primers used were as follows: Hprt probe, 5′-FAM-ATCCAACAAAGTCTGGCCTGTATCCAACAC-TAMRA-3′, 5′-AGCCCCAAAATGGTTAAGGTTG-3′, and 5′-CAAGGGCATATCCAACAACAAAC-3′; Gm-csf probe, 5′-FAM-CGCCCTTGAGTTTGGTGAAATTGCCCC-TAMRA-3′, 5′-CCTTCAAGAAGCTAACATGTGTG-3′, and 5′-GGGCAGTATGTCTGGTAGTAGC-3′; Hprt sybr, 5′-GATTAGCGATGATGAACCAGGTT-3′ and 5′-CCTCCCATCTCCTTCATGACA-3′; GM-CSFRα sybr, 5′- CTGCTCTTCTCCACGCTACTG-3′ and 5′-GAGACTCGCCGGTGTATCC-3′; and common β-chain sybr, 5′-AAAAACAGCCAGTGTCCTGTG-3′ and 5′-GATGCTGACGTTCTTGGGAAG-3′. The condition for real-time PCR was as follows: 40 cycles at 95°C for 3 s followed by 60°C for 30 s (probe) or 40 cycles at 95°C for 5 s followed by 60°C for 1 min (sybr).

Mice were anesthetized with isoflurane before the surgery. A total of 20 μg of anti–GM-CSF Ab (MP122E9; R&D Systems), 20 μg of Rat IgG2a (RTK2758; BioLegend), or 0.2 μg GM-CSF (PeproTech) was diluted with 10 μl saline. The injection was performed using a microsyringe with a 30G needle. The site of the injection was between the L6 cord and the first sacral (S1) cord. The infusion of the reagent into the intradural space was performed for >5 min. In some experiments, osmotic pumps (Alzet) were used for the intrathecal administration of anti–GM-CSF Ab or rat IgG2a (infusion rate: 0.5 μl/h, 20 μg/d) (21).

Student t tests (two-tailed) were used for the statistical analysis of differences between two groups. For more than two groups, one-way ANOVA tests were used. A p value <0.05 was considered statistically significant.

The EAE phenotype started around 1 wk after the MOG-specific autoreactive CD4⁺ T cell transfer (pathogenic T cell transfer), peaked at around 2 wk, and remitted after 3 wk (Fig. 1A) (18). Because a majority of peripheral-derived myeloid cells in tEAE mice under the remission phase (tEAE-recovered mice) express high levels of MHC II and CD11b compared with other myeloid cells, including microglia (18), we investigated whether their abundance ratio in the CNS increased after the development of tEAE. We found that there were low numbers of immune cells, including MHC II^HiCD11b⁺ cells, in the L5 cord before tEAE development. MHC II^HiCD11b⁺ cells together with other immune cells, such as CD4⁺ T cells, CD8⁺ T cells, and B cells, increased 2 wk after the tEAE induction. Interestingly, the MHC II^HiCD11b⁺ cell ratio still tended to be high under the remission phase in the L5 cord of tEAE-recovered mice (Fig. 1B). Notably, MHC II^HiCD11b⁺ cells were more abundant than other cells in the lumbar cord at the remission phase. Therefore, we used the surface markers MHC II and CD11b to isolate peripheral-derived myeloid cells.

We next investigated the cytokine receptor expression and found that GM-CSFRα and common β-chain molecules were highly expressed on peripheral-derived myeloid cells compared with microglia cells during the EAE remission phase (Fig. 1C, Supplemental Fig. 1). A qPCR analysis confirmed this finding (Fig. 1D). Because GM-CSF is a differentiation and survival factor for several myeloid cells (22) and because we found GM-CSF protein is present in cerebrospinal fluid not only in the remission phase of EAE mice but also in the steady state of naive mice (Fig. 1E), we focused on the role of the GM-CSF–GM-CSFR axis on the survival signal of peripheral-derived myeloid cells.

To investigate the functional significance of GM-CSF signaling on the survival of peripheral-derived myeloid cells, we administered rGM-CSF by intrathecal injections into tEAE-recovered mice. This treatment significantly increased the number of peripheral-derived myeloid cells in tEAE-recovered mice compared with naive mice (Fig. 2A), possibly because of the higher GM-CSFRα expression on peripheral-derived myeloid cells than residential CNS myeloid cells in steady state.

Antiapoptotic factor Bcl-xL, but not Bcl-2, was induced in peripheral-derived myeloid cells by the GM-CSF treatment in vivo (Fig. 2B, 2C). Consistently, intrathecal injections of a neutralizing Ab against GM-CSF significantly reduced the number of peripheral-derived myeloid cells (Fig. 2D) but hardly affected the number of microglia (Fig. 2E). The flow cytometry results for peripheral-derived myeloid cells and microglia cells are shown in Supplemental Fig. 1 (23). We hypothesize three potential reasons as to why the reduction of peripheral-derived myeloid cells was not complete after blockade of the GM-CSF pathway: (1) the Ab concentrations are not optimal; (2) the peripheral-derived myeloid cells are heterogenous, and only a fraction of the population is dependent on GM-CSF signaling; and (3) the survival of the peripheral-derived myeloid cells is dependent not only on GM-CSF but also other factors. In addition to GM-CSF, M-CSF is known to potentiate myeloid cell survival, proliferation, and differentiation (24). However, intrathecal injections of rM-CSF or anti–M-CSFR did not have significant effects on the number of peripheral-derived myeloid cells (data not shown). These results suggest that the GM-CSF signaling is important for the survival of peripheral-derived myeloid cells during the remission phase of tEAE.

It was reported that CD45⁺ bone marrow–derived cells, including activated CD4⁺ T cells, secrete GM-CSF (11, 12). We found CD45⁺ cells from EAE-recovered mice contained high mRNA levels of GM-CSF compared with CD45⁻ cells (Fig. 3A). We then investigated which CD45⁺ cell populations mainly express GM-CSF mRNA, finding they were CD4⁺ T cells and not CD45^MedCD11b⁺ microglia cells or peripheral-derived myeloid cells (Fig. 3B). However, we did not detect protein levels of GM-CSF in CD4⁺ T cells isolated from the CNS of tEAE-recovered mice. These cells produced GM-CSF proteins only after in vitro reactivation (Fig. 3C). Moreover, the depletion of CD4⁺ T cells in the CNS by the injection of anti-CD4 Ab did not decrease peripheral-derived myeloid cells during the remission phase of tEAE (Fig. 3D), suggesting that GM-CSF from CD4⁺ T cells had little effect on the survival of peripheral-derived myeloid cells.

We further investigated the GM-CSF expression in the CD45⁻ cell fraction. The CD45⁻ cell fraction was divided into three populations: CD45⁻PDPN⁺CD31⁻ fibroblastic reticular cells (FRCs), CD45⁻PDPN⁻CD31⁺ BECs, and CD45⁻PDPN⁻CD31⁻ other neuronal cells (double-negative cells [DNs]) (25) (Supplemental Fig. 2). The BEC population, but neither the FRC nor DN population, expressed GM-CSF mRNA in EAE-recovered or naive mice (Fig. 4A). In contrast with CD4⁺ T cells, which express GM-CSF after activation (Fig. 3C), BECs expressed GM-CSF protein constitutively (Fig. 4B). The number of BECs in the CNS was also much higher than CD45⁺ cells (Fig. 4C). When GM-CSF mRNA levels were corrected by multiplying the expression quantity and cell number, we found the BEC fraction contained more mRNA for GM-CSF compared with other cell populations in naive and tEAE-recovered mice (Fig. 4D). Consistent with these observations, protein concentrations of GM-CSF in the cerebrospinal fluid were not significantly different between naive and EAE-recovered mice (Fig. 1E), although naive mice had significantly fewer CD45⁺ cells, including CD4⁺ T cells in the CNS (Fig. 4C). Finally, ELISA experiments using CD31⁺ and CD31⁻ cells in the CNS, including the spinal cord, showed only the former secreted GM-CSF after overnight culture (Fig. 4E). These results strongly suggest that BECs are the main cell population that supplies GM-CSF molecules for the survival of peripheral-derived myeloid cells under the tEAE remission phase.

We next investigated the localization of peripheral-derived myeloid cells in the L5 cord to investigate GM-CSF–expressing cells during the remission phase. Whole-mount immunohistochemical staining of the L5 cord revealed that peripheral-derived myeloid cells mainly localized in the subarachnoid space, but not in the parenchyma (Fig. 5A), and near type IV collagen⁺ blood vessels (Fig. 5B). In addition, some peripheral-derived myeloid cells were localized directly on the vessels (Fig. 5B). Consistent with this localization, GM-CSF mRNA was more abundant in submeningeal areas that included BECs, pial cells, and arachnoid membrane cells compared with the parenchymal area of the L5 cord (Fig. 5C). The submeningeal expression levels of GM-CSF mRNA were similar between naive and tEAE-recovered mice, suggesting the constitutive expression of GM-CSF even during steady state (Fig. 5C). These results suggest that peripheral-derived myeloid cells mainly localize in the subarachnoid space near blood vessels expressing GM-CSF during the tEAE remission phase.

We previously reported that pain sensation triggers tEAE relapse by accumulating peripheral-derived myeloid cells at the ventral vessels of the L5 spinal cord, and depletion of these cells by clodronate liposome injection into the CNS prevents the relapse (18). Because GM-CSF neutralization in the CNS significantly reduced the number of peripheral-derived myeloid cells (Fig. 2D), we investigated whether anti–GM-CSF treatment in the CNS suppresses tEAE relapse. Anti–GM-CSF neutralizing Ab or control IgG was injected intrathecally during the tEAE remission phase. Then, pain sensation was induced by ligation of the middle branch of the trigeminal nerves to induce tEAE relapse (18). We found that treatment with the anti–GM-CSF Ab significantly suppressed the relapse and inhibited the accumulation of peripheral-derived myeloid cells at the ventral vessels of the L5 cord (Fig. 6). Mice were further assessed for signs of neuroinflammation by paralysis of the tail, rear legs, and front legs, number of footsteps, and a moribund state, all of which are due to neural circuit dysregulation caused by pathogenic CD4⁺ T cell–mediated neuroinflammation (19). Anti–GM-CSF Ab treatment after pain induction suppressed EAE development and decreased the number of MHC II⁺ cells around the L5 cord, suggesting that the blockade of GM-CSF had therapeutic effects on EAE relapse (Supplemental Fig. 3). In this study, we investigated MHC II^Hi myeloid cells on day 40, because we hypothesized that MHC II^Hi myeloid cells were maintained around the L5 cord in the remission phase and not supplied from the blood. In other words, once removing MHC II^Hi myeloid cells from around the L5 cord prevents the accumulation of pain-mediated MHC II^Hi myeloid cells at the ventral vessels and relapse. In contrast, anti–GM-CSF treatment did not suppress sensory–sympathetic cross-talk via the anterior cingulate cortex (ACC), a brain somatosensory area important for tEAE relapse (18), as monitored by c-fos expression in the ACC (Supplemental Fig. 4). These results indicate that GM-CSF blockade in the CNS during the remission phase successfully suppresses tEAE relapse by decreasing the number of peripheral-derived myeloid cells.

In this article, we describe the survival mechanism of MHC II^HiCD11b⁺ peripheral-derived myeloid cells in the CNS during the remission phase of tEAE. We found these cells express high levels of GM-CSFRα and common β chain compared with MHC II^LoCD11b⁺ microglia. Moreover, exogenous GM-CSF increased the cell population, whereas its blockade decreased it in tEAE-recovered mice. Importantly, the decrease by anti–GM-CSF treatment was sufficient to inhibit tEAE relapse induced by pain sensation. Furthermore, our data suggested that BECs are a major supplier of GM-CSF in the CNS. MS can be divided into at least four types of disease, and most patients experience disease relapse (1–4). This study highlights that GM-CSF from nonimmune cells could be a promising molecular target to prevent the relapse, but human studies are needed to confirm this possibility. Coincidently, a human mAb for GM-CSF, MOR103, has been tested in clinical trials and shown to have acceptable tolerance in MS patients (26). This Ab is expected to suppress the function of CNS-infiltrating pathogenic CD4⁺ T cells with GM-CSF expression and the mobilization and migration of peripheral myeloid cells from the bone marrow to the CNS. Therefore, we hypothesized that the systemic administration of anti–GM-CSF Ab suppresses both pathogenic CD4⁺ T cells and myeloid cells in the CNS, but its administration in the CNS suppresses mainly MHC II^Hi myeloid cells during MS development, because the function of pathogenic CD4⁺ T cells in the development of EAE is known to be dependent on GM-CSF expression (11). Thus, our current study may add another suppressive mode of action of anti–GM-CSF Ab in that it may prevent MS relapse by reducing the number of peripheral-derived myeloid cells in the CNS.

We found that peripheral-derived myeloid cells localized in the subarachnoid space, where GM-CSF–producing BECs also exist. These peripheral-derived myeloid cells are MHC II^Hi, which mainly persisted specifically in the L5 region because the L5 cord was the initial inflammatory site to accumulate blood immune cells, including peripheral-derived MHC II^Hi myeloid cells, via the gravity-gateway reflex (15). Indeed, L5 localization of the MHC II^Hi myeloid cells occurred during the initial phase of tEAE and the absence of region-specific expression of GM-CSF by L5 BECs. In other words, GM-CSF expression by L5 BECs may be necessary for MHC II^Hi myeloid cells to survive in this site, but it is not responsible for the initial localization, because GM-CSF in BECs is ubiquitously expressed in the entire CNS. We also hypothesized that BECs evolutionally secrete GM-CSF for microglia survival throughout the entire CNS. Consistently, single-cell RNA sequencing analysis showed that CD31⁺ cells expressed a low but similar level of GM-CSF mRNA expression in several CNS regions, such as the cerebellum, cortex, hippocampus, and striatum, by Tabula Muris Senis project (https://tabula-muris-senis.ds.czbiohub.org/).

Recently, it was reported that there is a subset of memory T cells, called resident memory T cells or tissue-resident memory T cells, that stay within an infection site long term and serve as a first-line defender on a reencounter with the same pathogen (27–29). Similarly, we hypothesize that our peripheral-derived myeloid cells stay within the CNS long term even during the remission phase and serve as a trigger for the relapse of neuroinflammation on specific neural signals (18), suggesting they are peripheral-derived resident myeloid cells. From this viewpoint, there is an important question whether peripheral-derived myeloid cells have physiological function in the CNS. Taken together, we propose that interactions between these peripheral-derived myeloid cells and BECs via the GM-CSF axis during the remission phase of tEAE promote the chronicity of inflammation by causing occasional relapse.

The authors have no financial conflicts of interest.

We appreciate the excellent technical assistance provided by M. Taru, K. Higuchi, and C. Nakayama, and we thank N. Yamamoto, S. Morita, and M. Ohsawa for excellent assistance. We thank Dr. P. Karagiannis (CiRA, Kyoto University, Kyoto, Japan) for carefully reading the manuscript.