Research Article

Single-cell profiling identifies myeloid cell subsets with distinct fates during neuroinflammation

See allHide authors and affiliations

Science  25 Jan 2019:
Vol. 363, Issue 6425, eaat7554
DOI: 10.1126/science.aat7554

A myeloid cell atlas of neuroinflammation

Myeloid cells, such as dendritic cells and macrophages, in the central nervous system (CNS) play critical roles in the initiation and exacerbation of multiple sclerosis (MS). Jordão et al. combined high-throughput single-cell RNA sequencing and intravital microscopy to compile a transcriptional atlas of myeloid subsets in experimental autoimmune encephalomyelitis (EAE), a mouse model of MS. Microglia and other CNS-associated macrophages expanded and transformed into various context-dependent subtypes during EAE. Furthermore, dendritic cells and monocyte-derived cells, but not resident macrophages, played a critical role by presenting antigen to pathogenic T cells. This exhaustive characterization may inform future therapeutic targeting strategies in MS.

Science, this issue p. eaat7554

Structured Abstract

INTRODUCTION

Under homeostasis, the central nervous system (CNS) hosts microglia (MG) and CNS-associated macrophages (CAMs). During experimental autoimmune encephalomyelitis (EAE), myeloid complexity drastically increases, with dendritic cells (DCs) and monocytes seeding the CNS. However, which disease-specific populations can be found during neuroinflammation remains largely unknown.

RATIONALE

An important step for the initiation of EAE and multiple sclerosis (MS) is the infiltration of the CNS by encephalitogenic T cells, which potentially become reactivated by encountering their self-cognate antigens presented at the brain interfaces. Myeloid cells have been shown to play a critical role in antigen presentation. Consequently, their transcriptomic profile and dynamics during neuroinflammation are crucial for understanding neuroinflammatory pathology.

RESULTS

High-throughput single-cell sequencing (scRNA-seq) of CD45+ cells isolated from several CNS compartments (including leptomeninges, perivascular space and parenchyma, and choroid plexus) allowed us to assemble a transcriptional atlas comprising 3461 immune cells, identified as homeostatic (“h”) or disease-associated (“da”) myeloid subsets. Profiling of all CAMs unraveled a core signature that consists of Mrc1, Pf4, Ms4a7, Stab1, and Cbr2. During disease, only Ms4a7 remained stably expressed, and a strong increase of antigen-presentation molecules (such as Cd74) was observed. Microglia expressed genes that included P2ry12, Tmem119, Sparc, and Olfml3. Although most of the core genes were down-regulated during disease, Sparc and Olfml3 expression remained unaltered and were accompanied by an up-regulation of Ly86. Several monocyte populations were observed during EAE, including monocyte-derived cells expressing Mertk and Mrc1 or expressing Zbtb46 and Cd209a. Although DCs were scarce in the homeostatic CNS, their density highly increased during disease, and diverse disease-associated DCs could be identified.

We next established the spatiotemporal relationship between infiltrating monocytes and resident macrophages using the Cx3cr1CreERT2 system. Local proliferation of resident macrophages occurred alongside continuous monocytic infiltration up to the peak of disease. Monocytes were transiently integrated into the CNS, and resident macrophages underwent apoptosis during the chronic phase. An evaluation of microglial expansion by using Cx3cr1CreER:R26Confetti mice revealed their clonal expansion during neuroinflammation.

We then investigated the capacity of resident and hematopoietic stem cell–derived myeloid cells for antigen presentation. Time-lapse imaging of Cx3cr1CreERT2:R26tdTomato:Cd2GFP and Ccr2RFP: Cd2GFP mice showed prolonged T cell interactions with circulating myeloid cells rather than tissue-resident macrophages during neuroinflammation. Accordingly, MOG35-55 immunization of Cx3cr1CreERT2:H2-Ab1flox mice showed no overt changes in disease development, indicating that resident macrophages are redundant for antigen presentation. By contrast, Cd11cCre:H2-Ab1flox mice were highly resistant to EAE, pointing to the potential role of DCs and monocyte-derived cells in EAE onset.

CONCLUSION

In this study, we unraveled the complexity of the CNS myeloid landscape and the dynamics of several myeloid populations during neuroinflammation. Although CNS-resident macrophages quickly generated context-dependent subsets during disease, their role as APCs was irrelevant for the initiation of pathology. DCs and monocyte-derived cells, highly diverse during EAE, remain the major players in antigen presentation. The comprehensive characterization presented here will provide a strong basis for their future targeting.

Myeloid cell diversity during neuroinflammation.

The homeostatic CNS includes microglia and different CAMs. During disease, microglia clonally expand, and the transcriptomic profile of microglia and CAMs drastically change. Diverse DC and monocyte subsets simultaneously populate the CNS. The role of resident macrophages for antigen presentation is redundant, whereas DCs and/or monocyte-derived populations show high antigen-presentation capacity, pointing to their crucial role in experimental autoimmune encephalomyelitis.

Abstract

The innate immune cell compartment is highly diverse in the healthy central nervous system (CNS), including parenchymal and non-parenchymal macrophages. However, this complexity is increased in inflammatory settings by the recruitment of circulating myeloid cells. It is unclear which disease-specific myeloid subsets exist and what their transcriptional profiles and dynamics during CNS pathology are. Combining deep single-cell transcriptome analysis, fate mapping, in vivo imaging, clonal analysis, and transgenic mouse lines, we comprehensively characterized unappreciated myeloid subsets in several CNS compartments during neuroinflammation. During inflammation, CNS macrophage subsets undergo self-renewal, and random proliferation shifts toward clonal expansion. Last, functional studies demonstrated that endogenous CNS tissue macrophages are redundant for antigen presentation. Our results highlight myeloid cell diversity and provide insights into the brain’s innate immune system.

Under steady-state conditions, the central nervous system (CNS) hosts a heterogeneous population of myeloid cells, including parenchymal microglia and non–parenchymal perivascular (pvMΦ), meningeal (mMΦ), and choroid plexus macrophages (cpMΦ), which are collectively called CNS-associated macrophages (CAMs) (13). The various anatomical localizations of endogenous myeloid cells within the CNS (parenchyma versus brain interfaces) have been associated with functions such as antigen presentation to encephalitogenic T cells (4) and the drainage of protein aggregates from the CNS (5) during pathology. However, clear-cut experimental evidence for myeloid-cell-subtype–specific functions in vivo is scarce. Whether the myeloid milieu differs across the CNS and how it changes under pathological conditions such as CNS autoimmunity remains largely unexplored.

Various immune cells have been described to initiate the inflammatory processes that underlie demyelinating inflammatory diseases, such as multiple sclerosis (MS) and experimental autoimmune encephalomyelitis (EAE), a mouse model of autoimmune demyelination (6, 7). The immune subsets found in the inflamed CNS include macrophages, several types of monocytes [Ly6Chi and Ly6Clo monocytes and monocyte-derived cells (MCs)], classical dendritic cells (cDCs), plasmacytoid DCs (pDCs), B cells, T cells, and natural killer (NK) cells (8, 9). During MS, immune cells are repeatedly recruited from the periphery, reinforcing the local inflammatory reaction within the CNS. These newly engrafted immune cells engage in a dynamic interplay with local endogenous macrophages in the CNS, which is still incompletely understood. Until now, a clear distinction between tissue-resident macrophages and invading myeloid cells has been mostly accomplished solely on the basis of their localization and morphology by using immunohistochemistry. Only recently, comprehensive profiling of bulk populations of CNS macrophages, including of entire transcriptomes (10) and proteomes (11), has helped to uncover several myeloid cell states during homeostasis (12) and disease (13). Although these approaches provided important insights, they all suffered from some limitations because they were restricted to probing a few selected proteins or RNAs, impeding the possibility of studying comprehensive landscapes and of discovering previously unrecognized cell subsets owing to a bias toward precharacterized molecules (14). The profiling of potentially heterogeneous cell populations led to an average signature, which obscured the putative diversity of CNS macrophages (1517).

Additionally, the antigen presentation required for T cell activation (18) has been attributed to either major histocompatibility complex (MHC) class II+ CNS intrinsic cells (19) or circulating DCs (20). In order to address these contradictory results and understand myeloid cell diversity during neuroinflammation, we combined single-cell RNA sequencing (scRNA-seq) with fate mapping, transgenic mouse lines, and in vivo imaging. We challenge the prevailing view that only a few myeloid cell subsets exist. Furthermore, we provide insights into the transcriptional networks, ancestry, and turnover of macrophages at CNS boundaries and peripheral myeloid cells during autoimmune neuroinflammation. We uncovered signature molecules that distinguish myeloid-cell populations involved in demyelinating neuroinflammatory conditions by highlighting context-dependent myelomonocytic subsets and their distinct signals. These data provide potential therapeutic targets and a complementary resource for the study of disease mechanisms in the CNS.

Results

scRNA-seq identifies new myeloid subsets in distinct CNS compartments during autoimmune inflammation

In order to analyze the diversity of hematopoietic cells during neuroinflammation on a single-cell level, we dissected several CNS compartments. Cells from the leptomeninges, choroid plexus, perivascular space, parenchyma, and blood from naïve and MOG35-55–immunized mice were isolated and subjected to high-throughput scRNA-seq accompanied with unbiased clustering (Fig. 1).

Fig. 1 Molecular census of hematopoietic cells during neuroinflammation in different CNS compartments.

(A) t-SNE representation of 3461 individual hematopoietic cells from all CNS compartments measured with scRNA-seq. Each dot represents an individual cell. Dashed lines indicate different hematopoietic populations. CAMs, CNS-associated macrophages; DCs, dendritic cells. (B) t-SNE plot depicting the expression levels of known core signature genes for microglia, CAMs, MCs, and DCs among all hematopoietic cells that underwent scRNA-seq. (C) t-SNE representation of 1052 individual meningeal cells, 1324 perivascular and parenchymal cells, and 701 choroid plexus cells measured with scRNA-seq and RaceID3 clustering. Yellow dots highlight the analyzed cells from homeostasis and different stages of EAE. (D) Identification of the main cell populations in the leptomeninges, perivascular space plus parenchyma, and choroid plexus. mMΦ, meningeal macrophages; pvMΦ, perivascular macrophages; cpMΦ, choroid plexus macrophages; MCs, monocyte-derived cells; Granulo, granulocytes; Lympho, lymphocytes. (E) Unbiased cluster analysis of subpopulations of cells found in the leptomeninges, perivascular space, parenchyma, and choroid plexus during EAE measured with scRNA-seq. hmMΦ, homeostatic meningeal macrophages; damMΦ, disease-associated meningeal macrophages; hpvMΦ, homeostatic perivascular macrophages; dapvMΦ, disease-associated perivascular macrophages; hMG, homeostatic microglia; daMG, disease-associated microglia; hcpMΦ, homeostatic choroid plexus macrophages; dacpMΦ, disease-associated choroid plexus macrophages; mDCs, meningeal dendritic cells; cpDCs, choroid plexus dendritic cells; mMCs, meningeal monocyte-derived cells; pMCs, perivascular and parenchymal monocyte-derived cells; cpMCs, choroid plexus monocyte-derived cells.

We assembled a transcriptional atlas comprising a total of 3461 immune cells recovered from all CNS compartments and blood (table S1). We represented these data using dimensionality reduction by use of t-distributed stochastic neighbor embedding (t-SNE) (Fig. 1A and fig. S1A). RaceID3 analysis (21) predicted 26 cell clusters, which mainly contained innate immune cells such as myeloid cells but also lymphocytes (fig. S1A and table S2). In order to define the transcriptional changes that allowed us to distinguish cell types in the inflamed CNS, we generated maps for the major myeloid cell populations based on previously described key signature genes (Fig. 1B) (2225).

Projections of individual cells by using t-SNE analysis from naïve and EAE-diseased mice across the CNS revealed the differential involvement of distinct CNS immune compartments over time (Fig. 1C). The main hematopoietic populations found were distinguishable on the basis of their transcriptomic signature (Fig. 1D).

We then identified molecularly distinct classes and subclasses of cells from different CNS compartments. Across the disease course, we identified 10 hematopoietic cell populations in the leptomeninges, 15 populations in the parenchyma and perivascular space, and 13 subsets in the choroid plexus (Fig. 1E). Clustering showed that the CNS myeloid cell compartments were separated into opposing states during the course of neuroinflammation: homeostatic (“h”) and disease-associated (“da”) clusters. Consequently, CNS tissue macrophages were found as homeostatic mMΦ (hmMΦ), pvMΦ (hpvMΦ), cpMΦ (hcpMΦ), or homeostatic parenchymal microglia (hMG) subsets in the healthy CNS. EAE-associated populations were designated as disease-associated mMΦ (damMΦ), pvMΦ (dapvMΦ), cpMΦ (dacpMΦ), and disease-associated microglia (daMG) (Fig. 1E). DC and MC subclasses were enriched during disease stages. Identified myeloid subpopulations showed both spatially specific and disease stage–specific kinetics when quantitatively analyzed (fig. S1B). Thus, the innate immune compartment in the adult CNS comprises transcriptionally distinct myeloid cell populations, which exhibit different disease stage–related subclasses with variable distribution across the CNS.

Uncovering tissue-resident macrophage subsets during neuroinflammation

Profiling of all CAM populations—including individual mMΦ, pvMΦ, and cpMΦ from healthy mice—identified the presence of three subsets of CAMs, designated hCAM1, hCAM2, and hCAM3 (fig. S2, A to D), which expressed not only Mrc1 but also Ms4a7, Pf4, Stab1, Cbr2, and Cd163 (fig. S2E). Fcrls, which was previously described to be microglia-specific (25, 26), was also detectable in all hCAM subsets. A comparison between hCAM1 and hCAM3 revealed an activated phenotype for the hCAM3 cluster with higher levels of the MHC class II–related molecules H2-Aa, H2-Ab1, H2-Eb1, and Cd74 (fig. S2F). hCAM3 also showed high levels of Ccr2 due to the contribution of cpMΦ, which are known to be partially derived from CCR2+ monocytes (24).

Disease-associated clusters of CNS endogenous macrophages were transcriptionally distinct from their counterparts during homeostasis (Fig. 2). However, CNS myeloid cells are generally characterized by their inherent ability to dynamically change their transcriptional profile and accompanying surface marker landscape upon activation. This makes distinguishing yolk sac (YS)–derived CNS endogenous tissue macrophages from hematopoietic stem cell (HSC)–derived circulating myeloid cells during neuroinflammation very challenging (8, 27, 28). In order to unequivocally characterize the lineage and fate of the myeloid subsets identified with scRNA-seq, we took advantage of a tamoxifen-inducible Cx3cr1CreERT2:R26tdTomato line, which distinguishes long-lived YS-derived CX3CR1+ MG, mMΦ, pvMΦ, and cpMΦ from HSC-borne short-lived CX3CR1+ myeloid cells, such as monocytes and DCs (24, 29, 30).

Fig. 2 Identification of disease-induced populations of CNS-endogenous tissue macrophages.

(A) mMΦ, (F) pvMΦ, and (J) cpMΦ subsets during disease identified by means of unbiased clustering. The arrow indicates the direction of cell cluster differentiation during disease. (B, G, and K) Log-transformed of expression levels of differentially expressed genes, given as significantly increased based on the negative binomial distributions in (B) mMΦ, (G) pvMΦ, and (K) cpMΦ subsets. Data are shown as whisker plots with means ± SEM of expression value. Data are representative from three independent experiments (n = 6 mice per experiment for naïve stage, n = 5 mice from onset, and n = 5 mice from peak phase). (C, H, and L) Top differentially regulated genes in (C) hmMΦ compared with damMΦ1, (H) hpvMΦ compared with dapvMΦ1, and (L) hcpMΦ (hcpMΦ1, hcpMΦ2, and hcpMΦ3) compared with dacpMΦ (dacpMΦ1 and dacpMΦ2). Data are presented as log2-fold changes. Arrows highlight genes that were verified on the protein level. (D, E and I) Representative immunofluorescence picture of Cx3cr1CreERT2:R26tdTomato mice depicting (D) LYVE-1, CD74, and CCL5 expression on resident mMΦ (tdTomato+) and (I) expression of LYVE-1, CTSD, CD74, and CCL5 on resident pvMΦ (tdTomato+) at naïve stage and peak of EAE. The dashed line indicates the barrier between the meninges (Men) and parenchyma (PC), or the vessel lumen and perivascular space (PV). Asterisks or arrows indicate resident macrophages either expressing or not expressing the specified proteins, respectively. Scale bars, 50 μm (overview) and 10 μm (inset). Representative pictures of four mice from two independent experiments are depicted. Quantification of (E) resident mMΦ expressing LYVE-1 and CCL5 and (I) resident pvMΦ expressing LYVE-1, CTSD, and CCL5 during the naïve stage and the peak of EAE. Data are from four mice from two independent experiments and are presented as mean ± SEM. An unpaired two-tailed Mann-Whitney U test revealed significant differences between the groups. (M) Representative immunofluorescence for CD74 and IL-1β in EAE-diseased Cx3cr1CreERT2:R26tdTomato mice at naïve stage and peak of disease. Asterisks show CD74 and IL-1β expression by resident cpMΦ (tdTomato+). Scale bars, 50 μm (overview) and 10 μm (inset). A representative picture of three mice from two independent experiments is displayed. (N) Top highly expressed genes in dacpMΦ1 and dacpMΦ2 subsets. Specified subsets are highlighted in the t-SNE plots. (O) t-SNE plots showing the mRNA expression of Ms4a7 during homeostasis and neuroinflammation in all CNS hematopoietic cells. The t-SNE plot reflects the map shown in Fig. 1A. The dotted line limits the cells belonging to CAM subsets, and arrows indicate the dynamics of cell populations during disease. hCAM, homeostatic CNS-associated macrophages; daCAM, disease-associated CNS-associated macrophages.

Profiling of single CAMs in different CNS immune compartments identified one hmMΦ population in the leptomeninges that was distinct from the disease-associated damMΦ1 (Fig. 2A). Both mMΦ subsets expressed Mrc1, Pf4, Cbr2, Ms4a7, Stab1, Fcrls, Cd163, and Siglec1 (Fig. 2B, and fig. S2G). Individual hmMΦ expressed higher levels of Cxcl2, Lyve1, and Nfkbiz, whereas damMΦ1 cells exhibited increased levels of the inflammatory chemokine Ccl5 and H2-Ab1, H2-Aa, H2-Eb1, and Cd74, suggesting a functional antigen presentation role for mMΦ in the CNS (Fig. 2C). LYVE-1 expression could be confirmed at the protein level in tdTomato+ mMΦ during both the naïve stage and peak of disease (Fig. 2D). In agreement with the transcriptional profile of damMΦ, we observed a significant down-regulation of the LYVE-1 expression levels in tdTomato+ damMΦ at peak of disease (Fig. 2E). CCL5 immunoreactivity was also up-regulated on resident damMΦ1 (Fig. 2E). pvMΦ, present during homeostasis as hpvMΦ, also showed a previously unidentified subset in the context of disease (dapvMΦ1) (Fig. 2F), with similarities to mMΦ (Fig. 2, G and H, and fig. S2H). In general, dapvMΦ did not show a strong transcriptional differentiation during neuroinflammation. LYVE-1 expression in tdTomato+ resident pvMΦ was significantly down-regulated at the peak of disease (fig. S2M). However, its down-regulation did not represent a complete loss of LYVE-1, and thus, the number of LYVE-1–expressing pvMΦ was not significantly altered (Fig. 2I). By contrast, the number of CTSD-expressing pvMΦ was down-regulated during disease (Fig. 2I). CD74 and CCL5 were observed to be highly expressed by tdTomato+ pvMΦ at peak of disease (Fig. 2I). In the choroid plexus, macrophage heterogeneity was higher during homeostasis with three hcpMΦ subsets. Two dacpMΦ subsets were also found during disease (Fig. 2J). cpMΦ shared core genes with mMΦ and pvMΦ, such as Mrc1, Ms4a7, Pf4, Stab1, Cbr2, and Fcrls (Fig. 2K). dacpMΦ populations showed an activated phenotype with increased levels of Il1b and MHC class II–related molecules (Fig. 2L). Accordingly, CD74 and interleukin-1β (IL-1β) were found to be highly expressed in the choroid plexus at peak of disease, but specificity for resident macrophages is challenging to address because of the dual ontogeny of cpMΦ (Fig. 2M). dacpMΦ1 expressed higher levels of S100a9, S100a8, and Lcn2, whereas MHC class II–related molecules and Ctss were highly expressed in dacpMΦ2 (Fig. 2N). Both populations showed an enrichment for the chemotaxis pathway, and dacpMΦ1 also up-regulated pathways related to cell motility as well as antigen processing and presentation of exogenous peptide antigen (fig. S5). We also identified a core signature for all CAMs independent of their localization comprising Mrc1, Pf4, Ms4a7, Stab1, Cbr2, and Fcrls (Fig. 2O and fig. S3, A and B). Fcrls was expressed by all tissue-resident macrophages, including MG and CAMs (fig. S3A), whereas Cbr2, Mrc1, Stab1, and Pf4 were highly enriched for CAMs but down-regulated during disease (fig. S3B). By contrast, Ms4a7 was distinctly and constantly expressed by individual CAMs throughout health and disease (Fig. 2O). Only a few genes were found to be different when CAMs were compared among the different CNS localizations during both homeostasis and inflammation (fig. S2K).

Both constitutive and disease-induced microglial heterogeneity was also observed in the parenchyma, with two populations (hMG1 and hMG2) present during homeostasis and four subsets (daMG1, daMG2, daMG3, and daMG4) during disease (Fig. 3A). All MG populations expressed Bhlhe41, Gpr34, Hexb, Olfml3, P2ry12, P2ry13, Sall1, Serpine2, Siglech, and Sparc (Fig. 3B and fig. S2J), but daMG clusters showed lower expression of P2ry12, Maf, and Slc2a5 and higher expression of Ccl2, Cxcl10, Ly86, and Mki67, indicating a proliferative capacity of daMG alongside the production of chemokines (Fig. 3C). Additionally, these daMG subpopulations showed reduced levels of Gpr34, Bhlhe41, Serpine2, Siglech, and Sall1 (fig. S3C) during disease. The most inflammatory disease-associated microglial subsets (daMG2, daMG3, and daMG4) strongly down-regulated the core microglial genes P2ry12, Tmem119, and Selplg and up-regulated Ly86 (Fig. 3D). Both P2RY12 and TMEM119 immunoreactivities were clearly down-regulated within the core of spinal cord lesions, whereas CD162 (encoded by Selplg) was only weakly reduced (Fig. 3E). By contrast, microglial MD-1 (encoded by Ly86) was strongly up-regulated in the lesions (Fig. 3E). Because of their transcriptional profile and their P2RY12loTMEM119loMD-1hi phenotype (Fig. 3, F and G, and fig. S2L), we determined that only daMG2, daMG3, and daMG4 localized within the lesion sites. A comparison of the transcriptomic profile between these three subsets revealed distinct signatures, which could be confirmed in situ with daMG2 as CD74hiCXCL10lo tdTomato+ microglia, daMG3 as CXCL10hiCD74lo tdTomato+ microglia, and daMG4 as CCL5hiCD74lo tdTomato+ microglia (Fig. 3, F and G). Microglial cells were also observed to be able to express high levels of CXCL10, CCL5, and CD74, which is an indication of the interplay between daMG2, daMG3, and daMG4 described subsets, likely at different stages of activation. Among the core microglial genes, only Sparc and Olfml3 were shown to be stable during pathology (Fig. 3H). Accordingly, SPARC protein was expressed by all tdTomato+ microglia during homeostasis and disease (Fig. 3, E and I).

Fig. 3 Identification of microglia subsets during health and disease.

(A) Microglia (MG) subsets found during homeostasis and EAE. The arrow indicates the direction of cell cluster differentiation during disease. (B) Log-transformed expression of enriched genes in MG subsets, given as significantly increased based on the negative binomial distributions, related to other myeloid cell populations. Data are presented as whisker plots with means ± SEM of expression value. Data are representative of 16 mice pooled from three independent experiments. (C) Most differentially regulated genes in hMG subsets (hMG1 and hMG2) in comparison with both extreme daMG populations (daMG3 and daMG4). Data are presented as log2-fold changes. The arrow indicates the gene that was verified on protein level. (D) t-SNE plot displaying the expression of genes down- or up-regulated in daMG populations (daMG2, daMG3, and daMG4). t-SNE at the left illustrates daMG (daMG2, daMG3, and daMG4) with red dots, whereas the same population is highlighted by a red dotted line in the right t-SNE plots. (E) Representative immunofluorescence pictures of Cx3cr1CreERT2:R26tdTomato mice during peak of EAE showing the expression of CD162, P2RY12, TMEM119, MD-1, and SPARC on resident (tdTomato+) microglia within the lesion compared with outside of the lesion. Scale bars, 50 μm (overview) and 10 μm (inset). Asterisks indicate resident tdTomato+ microglia. Representative pictures of five mice from three independent experiments are depicted. (F) Transcriptomic profile of daMG2, daMG3, and daMG4 subsets. Specified subsets are highlighted in the t-SNE plots, and arrows indicate the up- or down-regulation of specific genes. (G) (Left) Representative immunofluorescence pictures of Cx3cr1CreERT2:R26tdTomato mice during peak of EAE showing the down-regulation of P2RY12 and up-regulation of MD-1 in the lesion and the expression of P2RY12 and low expression of MD-1 in the nonlesion site of the spinal cord. Asterisks indicate resident tdTomato+ microglia. Scale bars, 500 μm (overview) and 10 μm (inset). (Right) Representative immunofluorescence pictures depicting daMG2 (CXCL10loCD74hi), daMG3 (CXCL10hiCD74lo), and daMG4 (CCL5+CD74lo) subsets within the core of lesions of the spinal cord. Scale bars, 10 μm. Pictures are representative of four mice from two independent experiments. (H) t-SNE plots showing the expression of Olfml3 and Sparc during homeostasis and neuroinflammation in all CNS hematopoietic cells. The dotted line limits the cells belonging to MG subsets. Arrows indicate the dynamics of cell populations during disease. hMG, homeostatic microglia; daMG, disease-associated microglia. (I) Quantification of resident microglia expressing SPARC at naïve stage and peak of disease. Data are from five mice from two independent experiments and are presented as mean ± SEM. An unpaired two-tailed Mann-Whitney U test revealed no significant difference between the groups.

Thus, our single-cell profiling identified previously unknown mMΦ, pvMΦ, cpMΦ, and MG subclasses associated with neuroinflammation, suggesting that homeostatic subsets of CNS endogenous tissue macrophages are able to quickly change their phenotypes and generate context- and time-dependent subsets. Moreover, CAMs strongly up-regulate MHC class II molecules during neuroinflammation, suggesting a prominent role for antigen presentation as previously postulated (31, 32).

Heterogeneity of HSC-derived myeloid cells during CNS autoimmunity revealed with single-cell analysis

We then used a combination of high-throughput scRNA-seq profiling and Cx3cr1CreERT2-based fate mapping to identify disease-induced subsets of MCs and DCs that could contribute to EAE pathogenesis (20, 3335). We identified several neuroinflammation-associated MC subsets in the leptomeninges (Fig. 4A). One subset, termed meningeal monocyte-derived cell subset 1 (mMC1), was reminiscent of Ly6Chi monocytes (fig. S4, A, B, and C) and showed high levels of Ly6c2, Ccr2, Fn1, and Cd44. This combination of genes was previously described (12) and confirmed by us as being highly expressed by peripheral immune cells, including monocytes and MCs (Fig. 4B and figs. S4C and S6). We further recognized MCs subsets that, in addition to their monocytic signature, also expressed macrophage- or DC-associated genes. In the leptomeninges, mMC1, mMC2, and mMC3 expressed the macrophage marker Mertk, whereas high levels of Mrc1 were observed in mMC4. Moreover, mMC2 and mMC3 both expressed high levels of Arg1 and Spp1 mRNA (Fig. 4B and fig. S4D). Given the proximity of mMC4 to the resident damMΦ1 subset, we then compared their transcriptomic profiles (Fig. S4F). We could observe high expression of monocyte-associated genes, such as Ccr2 and Vim in mMC4. However, we were unable to exclude the possibility of resident macrophages up-regulating these markers during neuroinflammation. CD206 (Mrc1) expression was confirmed on tdTomato mMCs alongside with resident tdTomato+CD206+ mMΦ (Fig. 4C). By contrast, mMC5 appear to resemble monocyte-derived DCs (moDCs) as described by others (36, 37), which not only expressed the monocytic markers Ly6c2 and Ccr2 but additionally expressed Cd209a, Kmo, Zbtb46, and Clec9a (Fig. 4B). DC-SIGN (CD209a) was confirmed on CD11c+ cells in the leptomeninges (Fig. 4D). All individual monocytes and MCs expressed Fcgr1, which encodes CD64 and was previously described to be present in all macrophages and monocytes (22). Thus, we contend that Ly6c2+Ccr2+CD44+Fcgr1+ cells are likely cells of monocytic origin that may acquire macrophage or DC markers.

Fig. 4 Single-cell profiling of monocyte-derived populations during EAE.

(A, E, and H) Visualization of (A) meningeal monocyte-derived cell subsets (mMCs), (E) perivascular and parenchymal monocyte-derived cell subsets (pMCs), and (H) choroid plexus monocyte-derived cell subsets (cpMCs) on a t-SNE 2D map. Arrows indicate the direction of cluster differentiation during disease. (B, F, and I) Expression levels of enriched genes in different monocyte-derived populations in (B) the leptomeninges, (F) perivascular space and parenchyma, and (I) choroid plexus. Data are depicted as whisker plots, where mean ± SEM of expression value is presented on a logarithmic scale. Data are representative of 16 mice from three independent experiments. (C) Representative immunofluorescence picture from Cx3cr1CreERT2:R26tdTomato mice showing infiltrating tdTomato MCs (arrows) and resident tdTomato+ mMΦ (asterisks) expressing CD206 in the leptomeninges. The dashed line delimits the barrier between the meninges (Men) and parenchyma (PC). Scale bars, 10 μm. A representative picture of three mice from two independent experiments is displayed. (D) Representative immunofluorescence image of DC-SIGN (encoded by Cd209a) on CD11c+ cells in the leptomeninges. Scale bars, 10 μm. The dashed line delimits the barrier between the meninges (Men) and parenchyma (PC). Representative picture from five investigated mice from two independent experiments is shown. (G) Representative immunofluorescence picture from Ccr2RFP mice showing CD206+CCR2+ MCs (arrows) and CD206+CCR2 resident macrophages (asterisks) in the leptomeninges and perivascular space. The dashed line delimits the barrier between the meninges (Men) and parenchyma (PC) or the perivascular space (PV). Scale bars, 50 μm (overview) and 10 μm (inset). A representative picture of four mice from two independent experiments is displayed.

Within the perivascular space and parenchyma, we identified five monocyte-derived subsets (pMCs). The pMC1 population was observed at an earlier inflammatory stage, whereas pMC2, pMC3, pMC4, and pMC5 subsets were present only during fully developed CNS inflammation (Fig. 4E). All five populations were characterized by a Ly6c2, Ccr2, Fn1, and Cd44 signature, whereas Mertk was present only in the pMC2, pMC3, and pMC4 subsets (Fig. 4F). Reporter CCR2-RFP mice allowed us to confirm the presence of CCR2+CD206+ MCs beside CCR2CD206+ resident mMΦ and pvMΦ in the leptomeninges and perivascular space, respectively (Fig. 4G). Thus, these cells have the capacity to gain phenotypic markers often associated with their resident counterparts. Similar to the leptomeninges, pMC5 in the perivascular space and parenchyma expressed Kmo and Zbtb46. However, CD209a was absent in this subset (Fig. 4F). We were further able to identify three disease-associated monocyte-derived subsets in the choroid plexus (cpMCs), with one subset presenting an undifferentiated monocytic profile (cpMC1) and the others presenting a transcriptional signature, including DC (cpMC2) or macrophage (cpMC3) markers (Fig. 4, H and I).

The presence of MerTK+ and CD209+ MCs may represent an ongoing differentiation into monocyte-derived macrophages or monocyte-derived DCs, but their function remains to be investigated.

DC subsets in the leptomeninges were barely detectable in the healthy CNS. However, during disease, one subset (mDC1) was clearly distinguishable (Fig. 5A). This population expressed Ly75, P2ry10, Ccr7, and Flt3 (Fig. 5B). No DCs could be detected in the perivascular space. By contrast, DCs in the choroid plexus fell into four disease-associated populations (cpDC1, cpDC2, cpDC3, and cpDC4) (Fig. 5C), with differential levels of Ly75, Ccr7, Ccl22, Cadm1, Cd81, and Ccl5 expression (Fig. 5D). Ccr7 expression by mDC1 and cpDC2 was indicative of their migratory capacity (Fig. 5D). Immunoreactivity for Ly75, which is enriched in specific DCs subsets (fig. S4H), was found in tdTomato cells in all different CNS compartments (fig. S4I).

Fig. 5 Profiling of DC subsets at different CNS compartments during neuroinflammation.

(A and C) DC subsets in the (A) leptomeninges (mDCs) and (C) choroid plexus (cpDCs) on a t-SNE 2D map. (B and D) mRNA expression levels of enriched genes in DCs from (B) the leptomeninges and (D) choroid plexus compared with other myeloid cells. Whisker plots are displayed as mean ± SEM of expression value. Logarithmic scales are used. Data are representative of 16 mice pooled from three independent experiments. (E) Representative fluorescence-activated cell sorting (FACS) gating strategy for cDC1 (MerTKCD64CD11c+MHC-II+CD11bloCD24hi), cDC2 (MerTKCD64CD11c+MHC-II+CD11bhiCD24lo), pDCs (MerTKCD64Ly6C+B220+Siglec-H+), and CD209+ MCs (MerTKCD64+Ly6C+CD44+CD209a+) in the CNS. Gating strategy is representative of five mice from two independent experiments. (F) Quantification of the absolute cell number (per milliliter) of cDC1, cDC2, pDCs, and CD209+ MCs present at naïve stage and during the course of EAE in different CNS compartments. Data are representative of five to seven mice per time point from two independent experiments for onset and three independent experiments for naïve, peak, and chronic phases and are presented as mean ± SEM. Two-way ANOVA followed by Tukey’s multiple comparisons test revealed significant differences between the groups.

Extensive work on DCs has been performed by others (37), where cDC1 are defined as CD11bCD103+CD24hiIrf8hiIrf4lo, cDC2 as CD11b+CD24loCD11chiCD172hiIrf8loIrf4hi, and pDCs as CD11bCD11cintB220+Ly6C+Siglec-H+ (37, 38). Our single-cell analysis failed to identify any pDCs within the populations examined. Furthermore, cDCs were only present as rare populations. By comparing the transcriptional profile of CNS DC subsets and CD209+ MCs, we observed shared gene signatures between some of the subsets across CNS compartments (fig. S4G). CD209+ MCs could be identified in all compartments as expressing Irf8, Itgam, Cd24, and Sirpa. The cDCs found in the leptomeninges and choroid plexus showed differential expression of these markers (fig. S4G), and their relationship to cDC1 and cDC2 requires further assessment.

We next performed flow cytometric analysis to evaluate DC population kinetics in different CNS compartments during the course of disease (Fig. 5E). We confirmed that homeostatic CNS possessed few DCs, mostly cDCs (Fig. 5F). Although DCs are scarce in the homeostatic CNS, their density increases dramatically during disease. Although pDCs and CD209+ MCs showed substantial increase in the leptomeninges, perivascular space, and parenchyma during the peak of disease, they were reduced in the choroid plexus.

The distinction between cDC2, CD209+ MCs, and activated macrophages remains challenging because of overlapping phenotypic markers. Although CD64/Fcgr1 is often solely expressed by macrophages and monocytes (22), cDC2 can also express CD64 in certain tissues (38) or under specific inflammatory conditions (13). We confirmed by means of flow cytometry that typical cDC2 can express Ly6C, CD44, and CD209a (Fig. 5E). Consequently, it is possible that CD209+ MCs may resemble a subset of cDC2 that are more closely related to the remaining MCs in the scRNA-seq analysis because of the expression of some monocytic markers. Using a fate-mapping approach, CD209+ MCs were verified to differ from CX3CR1+ long-lived resident macrophages because no tdTomato expression was observed within the former population (fig. S4J). Instead, MerTK+ MCs showed a surface marker profile similar to that of the CD209+ MCs (fig. S4K). Thus, these results suggest the emergence of disease-related monocyte and DC populations in distinct CNS compartments.

CNS-resident macrophages accumulate and expand clonally during neuroinflammation

The accumulation of myeloid cells during inflammation can occur either through local proliferation of tissue macrophages or the recruitment of peripheral monocytes from the blood. Although engrafted Ly6Chi monocytes rapidly die during autoimmune inflammation, the microglial pool quickly expands owing to local self-renewal (39). However, macrophage kinetics at CNS interfaces during inflammation are only poorly understood.

To establish the spatiotemporal relationship between microgliosis and the expansion of infiltrating myeloid cells, pvMΦ, and mMΦ, we used Cx3cr1CreERT2:R26tdTomato mice in which tissue-resident macrophages are efficiently labeled in contrast to monocytes (fig. S7, A and B). Although leakiness of tdTomato expression was observed in CAMs and microglia in Cre+ mice without tamoxifen treatment, blood monocytes showed no tdTomato expression, validating the use of these mice to discriminate between resident macrophages and peripheral immune cells (fig. S7C). We immunized Cx3cr1CreERT2:R26tdTomato mice and analyzed spinal cord sections at different phases of disease (fig. S7D). The recruitment of peripheral myelomonocytic cells (IBA-1+tdTomato) was first observed when animals reached the onset phase (Fig. 6A) and continuously increased up to the peak of disease. Their numbers then declined during the chronic stage. This suggests that circulating blood cells are not permanently integrated into the CNS, as previously hypothesized (39). IBA-1+tdTomato cells were initially found in proximity to the leptomeninges and perivascular spaces, suggesting that the entry of circulating myeloid cells into the CNS occurs through these compartments. The contribution of IBA-1+tdTomato MCs was shown to be more prominent when compared with the resident macrophage pool during full blown inflammation. In addition to infiltrating IBA-1+tdTomato cells, the number of IBA-1+tdTomato+ microglia dramatically increased during the peak of disease, whereas pvMΦ and mMΦ expanded more modestly (Fig. 6A). Ki67+ proliferating mMΦ were already evident during the onset phase, suggesting that the expansion of mMΦ is an early event in EAE pathogenesis (Fig. 6B). At peak of the disease, all IBA-1+tdTomato+ resident macrophages significantly expanded. A robust decrease in the frequency of proliferating MG and mMΦ was observed in the chronic phase, whereas pvMΦ proliferation remained stable. This drop in cell proliferation was accompanied by the presence of TUNEL+ apoptotic CNS macrophages in the chronic phase (Fig. 6B and fig. S6E). Resident macrophages and infiltrating monocytes at the leptomeninges were indistinguishable by morphology, whereas infiltrating monocytes in the perivascular space were significantly smaller compared with resident pvMΦ (fig. S7, F and G). The discrimination of different microglial subsets within the spinal cord allowed us to recognize the robust proliferation of daMG (CD45lotdTomato+MD-1+) in comparison with hMG (CD45lotdTomato+MD-1) at the peak of disease (Fig. 6C). daMG from EAE mice showed a significantly higher rate of proliferation when compared with hMG from naïve mice, demonstrating the differential impact of neuroinflammation on microglia. An evaluation of individual lesion-associated daMG subsets revealed higher proliferative capacity for daMG3 and daMG4 (Fig. 6D).

Fig. 6 Fate mapping of resident tissue macrophages in the CNS and infiltrating monocytes during EAE.

(A) Resident macrophages (asterisks) and infiltrating MCs (arrows) in diseased Cx3cr1CreERT2:R26tdTomato mice (representative pictures of at least four mice are shown) and quantification thereof. Data represent mean ± SEM. of four to six mice per group from two independent experiments for naïve and onset phases and three independent experiments for peak and chronic phases. Dotted lines indicate the leptomeningeal-parenchymal barrier or the perivascular space (PV). (B) Quantification of proliferating (percent Ki67+) and apoptotic (percent TUNEL+) mMΦ, pvMΦ, and MG in Cx3cr1CreERT2:R26tdTomato mice. Data show mean ± SEM of four mice per group from two independent experiments. n.d., nondetectable. (C) (Left) Representative flow cytometric gating strategy for the proliferation analysis of hMG (tdTomato+MD-1) and daMG (tdTomato+MD-1+) at both the naïve stage and the peak of disease. (Bottom right) Proliferation presented as mean ± SEM. Data are representative of six mice from two independent experiments. (D) Quantification from immunohistochemistry of the proliferation of daMG2, daMG3, and daMG4 at peak of the disease. Data show mean ± SEM of four mice per group from two independent experiments. (E) Immunofluorescence images of naïve and diseased Cx3cr1CreERT2:R26Confetti mice showing Confetti+ MΦ (asterisks). Pictures are representative of four mice from two independent experiments. Scale bars, 50 μm (overview) and 10 μm (inset). (F) Density of Confetti+ MΦ. Bars represent mean ± SEM of three mice per group from two independent experiments. (G) (Left) Representative confocal image from Cx3cr1CreERT2:R26Confetti mice at peak of EAE. Scale bars, 30 μm (overview) and 50 μm (inset). Nine mice were investigated from three independent experiments. (Right) Representation of the confocal image shown in the left demonstrating the analysis of Confetti+ microglial density. Microglia labeled by YFP, RFP, and CFP are represented by yellow, red, and cyan spheres, respectively. Magenta spheres show IBA-1+Confetti microglia. The density of same-colored cells is quantified within rings (gray) of increasing radius (arrows) from the reference cell (RC). (H) Densities of Confetti+ microglia according to (G). Microglial clonal expansion is assumed if the EAE distribution (blue line) lies outside of the random distribution area (gray, 98th percentile of Monte Carlo simulation). Data were obtained from nine mice at the peak of EAE from three independent experiments. (I) Analysis of Confetti+ microglia clone sizes for naïve (n = 3) and diseased (n = 9) mice for up to a 200-μm radius from each RC. Data are representative from two (naïve) or three (EAE) independent experiments. In (A), (C), and (D), Kruskal-Wallis test followed by Dunn’s multiple comparisons; in (B), two-way ANOVA followed by Tukey’s multiple comparisons test; and in (F), two-way ANOVA followed by Sidak’s multiple comparisons test were used to calculate significant differences between groups.

Microglial self-renewal under steady-state conditions constitutes a stochastic process that during facial nerve axotomy shifts from random to selected clonal expansion (40). To investigate whether the EAE-driven expansion of microglia and CAMs was random or clonal, we immunized Cx3cr1CreERT2:R26Confetti mice. These animals feature the stochastic recombination and expression of nuclear green fluorescent protein (nGFP), cytoplasmic yellow fluorescent protein (YFP), cytoplasmic red fluorescent protein (RFP), and membrane-tagged cyan fluorescent protein (mCFP) in resident macrophages (fig. S7H). At the peak of disease, Confetti-labeled cells could be observed in the CNS interfaces and parenchyma (Fig. 6E). Because the increase of Confetti+IBA-1+ cell density was only significant within the parenchyma (Fig. 6F), we focused on these macrophages for clonal analysis. Computational analysis of confocal images from the spinal cords of Cx3cr1CreERT2:R26Confetti mice allowed us to evaluate the distribution of all Confetti+ microglia within the tissue (Fig. 6G), revealing clonal expansion of microglia during EAE (Fig. 6H). Microglia only developed clones of same-colored cells under disease conditions (Fig. 6I). Thus, during autoimmune inflammation, macrophages appear to expand through local self-renewal alongside the recruitment of peripheral myeloid cells.

Prolonged T cell interactions occur with circulating myeloid cells during neuroinflammation

In the context of EAE, CD4+ T cells primed in the periphery migrate to the CNS and potentially become reactivated by encountering their self-cognate antigens at the site of brain interfaces (41). Antigen presentation to T cells has been shown in vivo for leptomeningeal myeloid cells (4, 19) and in vitro for several myelomonocytic populations (42). However, the precise nature of the myeloid cell type involved in T cell activation during neuroinflammation remains enigmatic with tissue-resident and circulating myeloid cells as potential candidates.

With scRNA-seq, we first determined the myeloid subsets with the most prominent antigen-presenting cell (APC)–related genetic profile across the different CNS immune compartments. In the leptomeninges, damMΦ1, and Cd209+ (mMC5) and Mertk+ (mMC2 and mMC4), MC populations showed the highest expression of the core antigen presentation signature (Fig. 7A). Increased levels of CD74 during disease could be confirmed in situ on MCs (IBA-1+tdTomato) compared with lower levels on tdTomato+ resident mMΦ (Fig. 7B). Within the perivascular space and parenchyma, daMG3, resident pvmMΦ1, and pMCs induced the highest levels of APC genes (Fig. 7C). These cells could be detected in spinal cord sections, where the highest CD74 levels were observed in monocytes from both compartments (Fig. 7, D and E). In the choroid plexus, inflammation-induced dacpMΦ, and Cd209+ (cpMC2) and Mertk+ (cpMC3), MCs showed the highest induction of APC-related molecules (Fig. 7F). Because of the dual origin of cpMΦ from pre- and postnatal hematopoietic sources (24), we were unable to apply our Cx3cr1CreERT2-based fate mapping to discriminate tissue-resident cells from monocyte-derived subsets. However, high levels of CD74 on myeloid cells in the choroid plexus were confirmed by using the pan-macrophage marker IBA-1 (Fig. 7G).

Fig. 7 In vivo imaging of T cells and CNS-resident and circulating myeloid cells.

(A, C, and F) t-SNE plot showing genes associated with antigen-presentation capacity (APC) across cells found at the (A) meninges, (C) perivascular space and parenchyma, and (F) choroid plexus. Populations with highest levels of the APC-associated genes are shown. (B) Images of CD74+ mMΦ and MCs (asterisks) in the meninges of diseased Cx3cr1CreERT2:R26tdTomato mice (representative of four mice from two independent experiments) and quantification thereof (data are presented as mean ± SEM of at least four mice per group pooled from three independent experiments). Scale bars, 10 μm. Dotted lines reveal the leptomeningeal-parenchymal barrier. (D) Images of CD74+ pvMΦ, MG, and MCs (asterisks) at peak of the disease (representative of five mice from three independent experiments). Scale bars, 10 μm. Dotted lines indicate the vessels. (E) Quantification from (D). Data are presented as mean ± SEM of five mice per group from three independent experiments. (G) (Left) CD74+IBA-1+ cells (including both cpMΦ and MCs) in the choroid plexus. Pictures are representative of four mice from two independent experiments. Scale bars, 10 μm. (Right) Quantification thereof. Data are presented as mean ± SEM of four mice per group from three independent experiments. (H and I) Confocal picture from (H) Cx3cr1CreERT2:R26tdTomato and (I) Ccr2RFP mice crossed with Cd2GFP mice at onset of EAE. Scale bar, 50 μm. Representative picture of five to seven mice from three independent experiments. (J) Percentage of cells labeled in Ccr2RFP mice as determined by means of flow cytometric analysis at the onset of EAE. Data are representative of five to seven mice from two independent experiments. (K) Contacts between macrophages/microglia or CCR2+ cells with T cells. Data are presented as mean ± SEM and are representative of five to seven mice from three independent experiments. (L) Quantification thereof. Each dot indicates a single cell, and the dotted line indicates the threshold limit past which a contact is considered long-lasting. Data are representative of four to seven mice from three independent experiments. (M) Quantification. Data are presented as the mean percentage ± SEM and are representative of four to seven mice from three independent experiments. (N) 3D reconstruction from Cx3cr1CreERT2:R26tdTomato mice (representative of four mice from two independent experiments). Asterisks indicate the contact points. Scale bars, 10 μm (overview) and 2 μm (inset), and quantification [representative of four mice (26 hMG and 34 daMG cells) from two independent experiments]. (O) Contacts of daMG2 subsets with T cells at onset of the disease. Data are representative of four mice (19 daMG2 and 15 daMG3/daMG4 cells) from two independent experiments. In (B), (E), (N), and (O), a two-tailed Mann–Whitney U test; in (J), Kruskal-Wallis test followed by Dunn’s multiple comparisons test; and in (K) and (M), one-way ANOVA followed by Bonferroni’s multiple comparison tests revealed significant differences between the groups. n.s., not significant.

To analyze which of these APC-competent myeloid subsets physically interacted with infiltrating T cells, we tracked the contact dynamics of fluorescently labeled CD2+ T cells with CNS-resident myeloid cells in spinal EAE lesions of Cx3cr1CreERT2:R26tdTomato:Cd2GFP mice by use of intravital microscopy (Fig. 7H). Time-lapse imaging revealed that mMΦ, pvMΦ, and microglia had similar T cell contacts (Fig. 7K). Likewise, the average duration of these contacts did not significantly differ between CNS-resident macrophages and microglia, with most lasting only a few minutes (Fig. 7L). Because CCR2 controls EAE susceptibility (34, 43), we next determined Ccr2 RNA expression levels on all immune-cell subsets during disease. In all CNS immune compartments, Ccr2 expression was highest in blood-derived populations such as Cd209+ and Mertk+ mMCs. By contrast, it was absent in pvMΦ and mMΦ populations (fig. S8D). To directly monitor interactions between HSC-derived myelomonocytic cells and T cells, we next crossed Ccr2RFP/WT mice to Cd2GFP T cell reporter mice (Fig. 7I). Although pDCs expressed almost no RFP, ~30% of cDC1 and cDC2 were targeted in this line, which is in agreement with recent findings (13). Similar labeling was observed in CD209+ MCs. Among all subsets within the CNS, MerTK+ MCs expressed the highest levels of RFP. As expected, circulating monocytes were also efficiently labeled in this model (Fig. 7J). In vivo imaging of acute spinal EAE lesions revealed that CCR2+ myelomonocytic cells were equally likely as their CNS-resident counterparts to be in contact with T cells (Fig. 7K). However, the average duration of such contacts was substantially longer (Fig. 7L), and the proportion of long-lasting contacts (>20 min) was significantly increased (Fig. 7M). Evaluation of the interaction between T cells and hMG (tdTomato+P2RY12hi) or daMG (tdTomato+P2RY12lo/−) showed preferential contact with the daMG subset (Fig. 7N). Discrimination between the different daMG subsets revealed that daMG2 were more likely to interact with T cells as compared with daMG3/daMG4 (Fig. 7O).

Antigen recognition by encephalitogenic T cells is associated with long-lasting T cell–APC contacts (19). Our results suggest that preferentially HSC-derived myeloid cells such as cDCs, CD209+, or MerTK+ MCs play a crucial role in antigen presentation during CNS autoimmune disease.

MHC II on circulating myeloid cells plays a pivotal role in neuroinflammation

The prevailing concept that macrophages at CNS borders present antigens and subsequently reactivate T cells to induce full encephalitogenicity during neuroinflammation is largely based on previous work with irradiated bone-marrow chimeras (31, 32). However, irradiation of the host primes the tissue and induces the artificial engraftment of donor-derived cells into the CNS (44, 45).

To test whether antigen presentation on CAMs contributes to the pathogenesis of neuroinflammation, we again used the Cx3cr1CreERT2 system to conditionally ablate MHC class II on pvMΦ, mMΦ, cpMΦ, and MG (Fig. 8A). This led to the robust deletion of this molecule in all resident CNS macrophages and microglia (Fig. 8A). In agreement with a recent study (46), Cx3cr1CreERT2:H2-Ab1flox mice surprisingly showed no overt changes in disease development (Fig. 8B). This indicates that MHC class II on CX3CR1+ macrophages, including resident macrophages at brain interfaces, is redundant for disease pathogenesis. Accordingly, accompanying histological analysis of Cx3cr1CreERT2:H2-Ab1flox mice revealed no changes in demyelination, myeloid-cell infiltration, T cell or B cell density, or amyloid precursor protein (APP) deposits (Fig. 8C).

Fig. 8 Absence of MHC II on circulating CD11c+ myeloid cells prevents autoimmune inflammation in the CNS.

(A) Quantification of MHC II expression in different immune cells from Cx3cr1CreERT2:H2-Ab1flox mice as determined with flow cytometric analysis. (Left) MHC II expression in resident mmΦ, pvMΦ, and cpMΦ at naïve stage, 8 weeks after tamoxifen application. (Right) MHC class II expression in microglia (MG), cDC1, cDC2, pDCs, CD209+, and MerTK+ MCs, and B cells at the peak phase of EAE. Bars represent means ± SEM of 9 mice in total (four Cre and five Cre+) from two independent experiments. A two-tailed Mann-Whitney U test revealed a significant difference between the groups. n.s., not significant. (B) Course of EAE in Cx3cr1CreERT2:H2-Ab1flox mice. Each data point represents the mean ± SEM of 33 mice in total (14 Cre and 19 Cre+) from four independent experiments. A two-tailed Mann-Whitney U statistical test revealed no significant difference between the groups. (C) Histology of spinal cord sections from Cx3cr1CreERT2:H2-Ab1flox mice by using Luxol fast blue (LFB) for demyelination (blue), MAC-3 for macrophages (brown), CD3 for T lymphocytes (brown), B220 for B cells (brown), and amyloid precursor protein for APP deposits (brown). Scale bars, 10 μm. In total, eight mice (four Cre and four Cre+) from three independent experiments were used. Graphs show quantification of infiltrates. A two-tailed Mann-Whitney U statistical test revealed a significant difference between the groups. (D) Quantification of MHC II expression in CAMs/MerTK+ MCs, microglia (MG), cDC1, cDC2, pDCs, CD209+ MCs, and B cells from Cd11cCre:H2-Ab1flox mice at peak phase of EAE as determined with flow cytometric analysis. Bars represent means ± SEM of 11 mice in total (four Cre and seven Cre+) from two independent experiments. A two-tailed Mann-Whitney U test revealed a significant difference between the groups. (E) Clinical EAE course in Cd11cCre:H2-Ab1flox mice. Each data point represents the mean ± SEM of 27 mice in total (11 Cre and 16 Cre+) pooled from four independent experiments. Significant differences as obtained from two-tailed Mann-Whitney statistical test are depicted in the graph. (F) Spinal cord histopathology from Cd11cCre:H2-Ab1flox mice by using LFB (blue), MAC-3 (brown), CD3 (brown), B220 (brown), and APP (brown). Scale bar, 10 μm. Graphs show quantification of infiltrates. Data are presented as mean ± SEM and were collected from eight mice (four Cre and four Cre+) pooled from three independent experiments. A two-tailed Mann-Whitney statistical test revealed significant differences between the groups. (G) The course of passive EAE in Cd11cCre:H2-Ab1flox mice until day 25 after adoptive T cell transfer. Data are shown as the mean ± SEM of 28 mice in total (13 Cre and 15 Cre+ mice) from three independent experiments. Significant differences as obtained from two-tailed Mann-Whitney statistical test revealed significant differences between the groups.

To determine whether the expression of MHC class II on circulating myeloid cells was essential, we examined the development of EAE in mice lacking MHC class II on CD11c+ cells (Fig. 8, D to I). Because resident macrophages also up-regulate CD11c during neuroinflammation, the Cd11cCre transgene–mediated excision of MHC class II was detectable in both tissue-resident cells and blood myelomonocytic cells that engrafted in the diseased CNS (Fig. 8D). Cd11cCre:H2-Ab1flox mice were greatly resistant to MOG35-55 immunization (Fig. 8E), and neuropathological changes were not observed. Cd11cCre:H2-Ab1flox mice did not exhibit CNS-infiltrating immune cells (such as B and T lymphocytes), and their myelin remained undamaged. By contrast, Cre controls showed typical EAE inflammatory responses (Fig. 8F). The adoptive transfer of encephalitogenic T cells into Cd11cCre:H2-Ab1flox mice confirmed the essential role of MHC class II on CD11c+ APCs (Fig. 8I).

Thus, CD11c-expressing peripheral immune cells show a more critical role for T cell priming and initiation of the pathology.

Discussion

This study provides an unbiased view of the transcriptional landscapes of CNS-resident and circulating myeloid cells during homeostasis and CNS autoimmunity. We found that homeostatic cell-specific profiles are rather uniform throughout the various analyzed CNS regions, whereas a considerable compartment- and disease stage–specific myeloid subtype specification with high plasticity emerges during the development and maintenance of neuroinflammatory pathology. This phenomenon is mediated by the specific occurrence of distinct ratios of tissue-macrophage populations.

Unlike most peripheral tissue macrophages, MG and most CAMs such as pvMΦ and mMΦ arise from Linc-Kit+ erythromyeloid progenitors in the YS during embryonic development and are maintained throughout life by self-renewal (40, 47) independent of bone-marrow precursor cells (45, 48). By using the Cx3cr1CreER:R26tdTomato fate-mapping system, which tracks long-lived YS cells, we showed that CAMs have considerable longevity and remain stable even during CNS autoimmunity. By contrast, blood-derived IBA-1+tdTomato myeloid cells, presumably monocytes, infiltrated the spinal cord during disease but were only transiently present at the lesion site. This confirms previous studies, which lacked the proper fate-mapping tools (39, 49). Furthermore, we used a recently established tissue macrophage–focused multicolor-reporter mouse model (40), which revealed the organization of spinal cord MG and CAM networks during health and disease. We discovered selective microglial clonal expansion in response to inflammatory damage in the CNS, which may support the self-renewal and longevity of these tissue macrophages. Within the different CAM populations, different numbers of TUNEL+ apoptotic CNS endogenous macrophages could also be detected. This indicates the differential macrophage response to inflammation, which is suggestive of the functional diversity of these cells during disease.

We provide in vivo evidence of single-cell MG and CAM heterogeneity in the normal mouse CNS. Cellular diversity among CAMs in the healthy CNS was especially apparent in cpMΦ in comparison with mMΦ or pvMΦ. This finding may reflect the dual origin of cpMΦ from prenatal (YS/fetal liver) and postnatal (bone marrow) sources compared with the purely prenatal origin of mMΦ and pvMΦ (24).

MG and CAM populations were detectable in the steady-state CNS as distinct clusters. Indeed, during EAE, the ability of CNS-tissue macrophages to swiftly adapt to environmental changes was observed with several unappreciated cell subsets appearing within the MG population as well as for each CAM subtype. MG within the inflammatory lesions of the spinal cord down-regulated several markers designated to their core signature repertoire (25), such as P2ry12, Tmem119, and Selplg. By contrast, MD-1 (encoded by Ly86 and a member of the Toll-like receptor family) was up-regulated during CNS autoimmune disease. Moreover, the expression of different chemokines was increased in disease-linked microglial subsets. Three distinct microglial subsets were identified that were linked with the lesion sites during neuroinflammation: daMG2, daMG3, and daMG4. Although they all shared the P2RY12loTMEM119loMD-1hi profile, differences were found in the expression of specific chemokines, cytokines, and cysteine proteases. Moreover, differential cellular dynamics could be allocated to specific microglial subsets. daMG2 expressed high levels of Cd74, Ctsb, and Apoe. This population also proliferated less but made more contacts with encephalitogenic T cells compared with other daMG subsets. By contrast, daMG3 expressed high levels of Cxcl10, Tnf, and Ccl4, whereas daMG4 expressed high levels of Ccl5, Ctss, and Itm2b. Both populations showed high proliferative capacity and rather low frequency and duration of T cell contacts. Despite the distinct daMGs profiles, Olfml3 and Sparc remained unaltered in the daMG cluster, indicating that these genes may serve as robust microglial markers in health and disease.

This thorough microdissection of the several CNS compartments together with scRNA-seq enabled us to acquire an independent and unbiased overview of the inflammation-driven responses across the CNS. The individual transcriptional profiling of CAMs highlighted their close relationships during both homeostasis and inflammation irrespective of their anatomical localization within the CNS. A core gene signature comprising Mrc1, Pf4, Ms4a7, and Cbr2 was found to distinctly characterize CNS-associated macrophages and distinguish them from other myeloid populations. Furthermore, we identified several daCAM subsets during inflammatory disease. These daCAM subsets showed surprising transcriptional similarity among the various CNS immune compartments, including the shared reduced expression of key core genes such as Lyve1. By contrast, immune-related molecules were especially strongly induced. CD74, for example, was most highly expressed in CAMs during disease. CD74 associates with MHC class II and is an important chaperone regulating antigen presentation during immune responses. It also serves as a cell-surface receptor for the cytokine macrophage migration inhibitory factor, which initiates survival pathways and myeloid-cell proliferation (50). Moreover, CAMs did not completely lose their homeostatic core signature and maintained levels of molecules such as Ms4a7 unchanged. The functions of Ms4a7 are not completely understood but are thought to be associated with mature cellular function within the monocytic lineage. It may also be a component of a receptor complex involved in signal transduction (51). Recent studies showed Ms4a7 expression by embryonic microglia by embryonic day 14.5 (52) and suggested that they are ontologically diverse from the YS-derived macrophages (53) and potential progenitors of CAMs (52).

Strategically positioned at the CNS barriers, mMΦ, pvMΦ, and cpMΦ are thought to modulate immune-cell entry and phenotype possibly by presenting antigens to circulating lymphocytes (5456). In order to unravel mMΦ, pvMΦ, and cpMΦ APC function, we used the Cx3cr1CreERT2:H2-Ab1flox mice, which show specific MHC class II deficiency on long-living tissue macrophages. Surprisingly, we found no changes, either in the clinical course or in the histopathology of mice afflicted with EAE. Similarly, we observed that infiltrating T cells preferentially show long-lasting interactions with CCR2+ myeloid cells of peripheral origin, whereas most in vivo interactions with MG and CAMs were transient. Because T cells arrest in response to antigen recognition (19), these observations support an important role for infiltrating myeloid cells in the activation of T cells during EAE. When we subsequently deleted MHC class II expression on both peripheral and CNS-resident myeloid cells in Cd11cCre:H2-Ab1flox mice, EAE induction was effectively prevented.

The precise nature of the CD11c+ myeloid cells that drive CNS pathology by amplifying T cell responses via MHC class II in EAE remains unclear. One possible candidate subset comprises DCs that represent the intersection of the innate and adaptive immune systems (57). This idea is supported by the fact that CD11c-driven transgenic overexpression of MHC class II facilitates EAE pathogenesis (20). However, this previous approach did not exclude the potential contribution of MHC class II on CD11c+ MG or CAMs during disease. We observed the presence of disease-specific CD209+ and MerTK+ MCs and disease-associated DCs mostly in the choroid plexus, leptomeninges, and, to a lesser extent, the perivascular space. Thus, we identified these locations as putative entry sites for MHC class II–dependent T cells.

The role of microglia during MS/EAE remains controversial. For several decades, the activation of microglia has been described during CNS inflammation and considered to be an initial event in MS pathology (58). Even in early stages of MS, activated microglial clusters (so-called microglial nodules) are found in preactive lesions in the white matter of MS patients (59). To gain further insights into the involvement of activated microglia, several transgenic mouse models have been developed. These suggested major roles of microglia, including the expression of proinflammatory mediators and effector molecules at the peak of disease (30, 60) and the removal of debris, which allows proper remyelination during the recovery phase (49, 61). Furthermore, microglia are thought to contribute to antigen-dependent T cell activation (58). However, selective gene-targeting experiments now demonstrate that MHC class II–mediated T cell priming by microglia is not critical for the induction or progression of EAE.

Thus, the identification of disease- and CNS-compartment–specific myeloid subsets during EAE should provide the basis for implementation of therapeutic approaches, specific to previously unidentified subsets, for neuroinflammatory disorders. These should entail reduced risks compared with the somewhat global immune suppressive therapies currently administered to a large number of MS patients.

Material and methods

Mice

C57BL/6N mice were used as WT mice and all transgenic lines (Cx3cr1CreERT2, Cd11cCre, Cd2gfp, Ccr2RFP, and H2-Ab1flox) were on a C57BL/6 background. Mice were bred in-house under pathogen-free conditions. Cx3cr1CreERT2 were crossed to either R26tdTomato or R26Confetti mice. All mice are commercially available in The Jackson Laboratory. Littermate controls were used for the different experiments. All animal experiments were approved by the local administration and were performed in accordance to the respective national, federal and institutional regulations.

Tamoxifen treatment

For induction of the Cre recombinase in Cx3cr1CreERT2:R26tdTomato mice, 6-week-old animals were treated twice with 4 mg of tamoxifen (TAM, Sigma-Aldrich, Taufkirchen, Germany) dissolved in 200 μl of corn oil (Sigma-Aldrich, Taufkirchen, Germany), injected subcutaneously at two time points, 48 hours apart. For Cx3cr1CreERT2:R26Confetti mice, 6-week-old animals were treated once with 8 mg of tamoxifen dissolved in 200 μl of corn oil (subcutaneous injection).

Induction of experimental autoimmune encephalitis

For the induction of EAE, mice were immunized subcutaneously with 200 μg of MOG35-55 peptide emulsified in CFA containing 0.1 mg of Mycobacterium tuberculosis (H37RA; Difco Laboratories, Detroit, Michigan, USA). The mice received intraperitoneal injections with 250 ng of pertussis toxin (Sigma-Aldrich, Deisenhofen, Germany) at the time of immunization and 48 hours later. For experiments involving C57BL/6 mice, immunizations were performed in 6- to 8-week-old mice. For experiments using Cx3cr1CreERT2:R26tdTomato and Cx3cr1CreERT2:R26Confetti mice, immunizations were performed 8 weeks after TAM induction. Mice were scored daily according to their clinical symptoms (score 1: complete limp tail; score 1.5: limp tail and hindlimb weakness; score 2: hind limbs paresis; score 2.5: unilateral hind limb paralysis; score 3: bilateral hind limb paralysis). For the EAE experiments presented in this study, mice were analyzed either at the preclinical phase (day 8 after immunization), onset phase (score 1, typically observed between day 11-13), peak phase (score 3, typically observed between days 15 and 20), or at the chronic phase (day 30 post-immunization) of EAE.

Adoptive transfer EAE

Passive EAE was induced by intravenous injection of MOG-reactive lymphocytes (3 × 106/mouse) into recipient mice. Mice also received 200 ng of pertussis toxin intraperitoneally (i.p.) on the day of immunization and 2 days later.

Intravital imaging and image processing

Mice were imaged at the onset of EAE symptoms (only animals with an EAE score ≥ 1 were included). For this purpose, mice were anesthetized with medetomidin (0.5 mg/kg), midazolam (5.0 mg/kg), and fentanyl (0.05 mg/kg), placed on a heating pad, and then tracheotomized and intubated. The dorsal spinal cord was surgically exposed as previously described (62) and the opening constantly superfused with artificial cerebrospinal fluid (aCSF). For the imaging session, the vertebral column was fixed using a spinal clamping device (Narishige STS-a) and the spinal opening surrounded by a 4% agarose well. Mice were injected i.p. with 200 μg of dextran-AF647 (Life Technologies) to reveal the vasculature. In vivo imaging was performed using confocal or two-photon laser excitation on an Olympus FV1200 MPE setup equipped with a 25×/1.25 water immersion objective (Olympus) at a 1024 by 1024 pixel resolution. For confocal imaging, GFP, tdTomato, and dextran-AF647 were sequentially excited using 488 nm, 568 nm, and 647 nm lasers, respectively. For two-photon time-lapse imaging of interaction dynamics, the IR laser was tuned to 820 nm and fluorescence was collected using a standard green/red filter set (BA575-630). For image representation, maximum intensity projections of image stacks were gamma-adjusted and processed with a despeckling filter using Photoshop software (Adobe). The time and frequency of CD2+ T-cell contacts with different populations of CNS-resident and CCR2+ infiltrating myeloid cells were assessed manually in 60-min-long movies using Fiji software (63). Cellular interactions that began during the final 20 min of imaging were excluded from our analysis. To calculate the contact probability of different phagocyte subsets, the number of T-cell contacts on a given phagocyte per 60 min was adjusted for the T-cell infiltration density in the imaging area.

Histology

Mice were transcardially perfused with phosphate-buffered saline (PBS). Spinal cords were then dissected and fixed in 4% paraformaldehyde (PFA) overnight. Tissue was then embedded in paraffin and stained with Luxol fast blue to assess the degree of demyelination, rat anti-mouse MAC-3 (2.5 μg/ml, clone M3/84, BD Pharmingen) for macrophages and microglia, rat anti-human CD3 (3.5 μg/ml, clone CD3-12, Serotec, Düsseldorf, Germany) for T cells, rat anti-mouse B220 (2.5 μg/ml, clone RA3-6B2, BD Pharmingen) for B cells, and mouse anti-mouse APP (3 μg/ml, clone 22C11, Millipore) for indication of axonal damage. For MAC-3, CD3, B220, and APP immunohistochemistry, the primary antibodies were incubated overnight (4°C) followed by incubation with biotin-labeled goat anti-rat, goat anti-mouse, or goat anti-rabbit secondary antibodies (2.5 μg/ml, Southern Biotech) for 45 min at RT. Streptavidin (Southern Biotech) was then added for 45 min at RT. 3′ -Diaminobenzidine (DAB) brown chromogen (Dako) was used to resolve the aforementioned antibodies.

Immunofluorescence

After transcardial perfusion with phosphate-buffered saline (PBS), brains and spinal cords were fixed in 4% PFA for 6 hours, dehydrated in 30% sucrose and embedded in Tissue-Tek O.C.T.TM compound (Sakura Finetek Europ B.V., Netherlands). Cryosections were then blocked and permeabilized with PBS containing 5% normal donkey serum (NDS, Abcam) and 0.5% Triton-X 100 for 1 hour at RT. Primary antibodies were incubated overnight at a dilution of 1 μg/ml for rabbit anti-mouse IBA-1 (WACO, Japan), 1 μg/ml for goat anti-GFP (Rockland Immunochemicals Inc., Gilbertsville, USA), 1 μg/ml for rabbit anti-laminin (Sigma-Aldrich), 1 μg/ml for goat anti-collagen-IV (Millipore), 1 μg/ml for rabbit anti-Ki67 (Abcam), 3 μg/ml for rat anti-mouse CD206 (clone MR5D3, Bio Rad), 2 μg/ml for rat anti-mouse CD74 (clone In1/CD74, Biolegend), 1.5 μg/ml for rabbit anti- LYVE-1 (Abcam), 2 μg/ml for hamster anti-mouse CD11c (clone N418, eBioscience), 2 μg/ml for mouse anti-mouse CD209a (MMD3, Biolegend), 1 μg/ml for rabbit anti-mouse P2RY12 (AnaSpec), 1 μg/ml for rabbit anti-mouse TMEM119 (clone 106-6, Abcam), 2.5 μg/ml for rat anti-mouse MD-1 (clone MD113, Abcam), 3 μg/ml for mouse anti-SPARC (FITC-labeled, R&D Systems), 2.5 μg/ml for rat anti-mouse CD162 (clone 4RA10, BD Pharmingen), 1.5 μg/ml for rabbit anti-CCL5/RANTES (Novus Biological), and 2 μg/ml for goat anti-mouse CXCL10 (clone BAF466, R&D Systems) at 4°C. The following secondary antibodies were used: Alexa Flour 405-labeled donkey anti-goat 2 μg/ml, Alexa Flour 488-labeled donkey anti-rabbit 1 μg/ml, Alexa Flour 488-labeled donkey anti-rat 1 μg/ml, Alexa Flour 568-labeled donkey anti-rabbit 1 μg/ml, Alexa Flour 568-labeled donkey anti-rat 1 μg/ml, Alexa Fluor 647-labeled donkey anti-rabbit 1 μg/ml, and Alexa Fluor 647-labeled chicken anti-rat 1 μg/ml for 2 hours at RT (ThermoFisher Scientific). Nuclei were counterstained with DAPI (0.1 μg/ml) when necessary. Images were taken using conventional fluorescence microscopes (Olympus BX-61 with an Olympus XC10 camera and Keyence with a 2/3 inch, 1.5 million pixel monochrome CCD (colorised with LC filter) camera), and the confocal pictures were taken with Fluoview FV 1000 (Olympus). Images were processed with Photoshop (Adobe) or Fiji (63) software.

TUNEL assay

TUNEL assays were carried out using the In Situ Cell Death Detection Kit, TMR fluorescein (12156, Roche) according to the manufacturer’s instructions. Briefly, specimens were permeabilized with 5% NDS + 0.5% Triton-X 100 and subsequently incubated with the TUNEL reaction solution mixture in a humidified 37°C chamber for 1 hour. Cell nuclei were labeled with DAPI (0.1 μg/ml). Images were taken using Keyence fluorescence microscope. Images were processed with Photoshop (Adobe) and the percentage of TUNEL-labeled cells versus all tdTomato+ cells was calculated on Excel (Microsoft) and plotted with GraphPad Prism 6.

Three-dimensional (3D) reconstruction of macrophages

Free-floating 50-μm cryosections from spinal cords were blocked and permeabilized with 5% NDS + 0.5% Triton for 4 hours followed by incubation overnight with anti-IBA-1 and anti-collagen IV at 4°C. Secondary antibodies were incubated for 4 hours at RT. Imaging was performed on an Olympus Fluoview 1000 confocal laser scanning microscope using a 20×/0.95 NA objective with a 3× zoom. Z stacks assembled from 1.15-μm steps in the z plane, 1024 by 1024 pixel resolutions were recorded and analyzed using IMARIS software (Bitplane).

Analysis of T cell–microglial interactions

Free-floating 20-μm cryosections from spinal cords were blocked and permeabilized with 5% NDS + 0.5% Triton for 4 hours followed by incubation overnight with rabbit anti-mouse P2RY12 (AnaSpec), 1.5 μg/ml for rabbit anti-CCL5/RANTES (Novus Biological), 2 μg/ml for rat anti-mouse CD74 (clone In1/CD74, Biolegend), 2 μg/ml for goat anti-mouse CXCL10 (clone BAF466, R&D Systems), and 2 μg/ml for rat anti-mouse CD4 (clone GK1.5, eBioscience) at 4°C. Secondary antibodies were incubated for 4 hours at RT. Imaging was performed on an Olympus Fluoview 1000 confocal laser scanning microscope using a 60×/0.95 NA objective with a 2× zoom. Confocal pictures were evaluated on Imaris software (Bitplane). The colocalization plugin was used to evaluate the interaction points between myeloid cells and T lymphocytes.

Dissection of CNS compartments and flow cytometry

To separate different CNS compartments, both brain and spinal cord were dissected from mice and placed in ice-cold PBS. Under binoculars, the leptomeninges were dissected from the spinal cord to obtain separate samples of the leptomeninges and of the parenchyma and perivascular space. All choroid plexuses were removed from the ventricles of the brain. Tissue samples were placed in HBSS containing 1 mg/ml of collagenase D (Sigma-Aldrich) and digested for 30 min at 37°C. After digestion, myeloid cells from the parenchyma and perivascular space were isolated using a 37% Percoll gradient (Sigma-Aldrich) from the homogenized tissue. Tissue from the choroid plexuses and leptomeninges was treated independently by mechanical dissociation through a 70-μm cell strainer to obtain cell suspensions. Monocytes were isolated from blood. Cells were acquired on FACSCanto II and LSRFortessa systems (BD Bioscience, Heidelberg, Germany) and analyzed with FlowJo software (TreeStar). Cell sorting was performed on a MoFlo Astrios (Beckman Coulter, Krefeld, Germany). The following antibodies were used for staining cells: anti-CD45 (BV786, 0.5 μg/ml, clone 30-F11, eBioscience, San Diego, USA), anti-CD11b (BV605, 0.5 μg/ml, clone M1/70, eBioscience, San Diego, USA), anti-CD3 (PE-Cy7, 0.8 μg/ml, clone eBio500A2, eBioscience, San Diego, USA), anti-CD19 (PE-Cy7, 0.8 μg/ml, clone eBio1D3, eBioscience, San Diego, USA), anti-NK1.1 (PE-Cy7, 0.8 μg/ml, clone PK136, eBioscience, San Diego, USA), and anti-MD-1 (Alexa 488, 1 μg/ml, Abcam). Before surface staining, dead cells were stained using the Fixable Viability Dye eFluor 780 or eFluor 506 (eBioscience, San Diego, USA) followed by incubation with Fc receptor blocking antibody CD16/CD32 (1 μg/ml, clone 2.4G2, BD Bioscience, Heidelberg, Germany).

scRNA amplification and library preparation

Myeloid cells from the different CNS compartments were obtained as detailed above. Samples containing cells from the different compartments (leptomeninges, parenchyma, and perivascular space, and choroid plexus) were subjected to single-cell sorting in the MoFlow Astrios machine. After lineage exclusion (Lin: CD3ɛ, CD19, NK1.1, and Ly6G), all CD45+ cells were sorted into 382-well plates. To enrich for rare myeloid populations, we separated the sorted cells using three strategies (CD45loCD11b+, CD45hiCD11b+, and CD45hiCD11blo/−). The same number of cells was sorted for each of these populations. Circulating Ly6Chi and Ly6Clo monocytes were sorted from blood. Single-cell RNA sequencing was performed using the mCEL-Seq2 protocol, an automated and miniaturized version of CEL-Seq2 on a mosquito nanoliter-scale liquid-handling robot (21, 64). Eight libraries with 192 cells each were sequenced per lane on Illumina HiSeq 2500 or 3000 sequencing system (pair-end multiplexing run) at a depth of ~130,000-200,000 reads per cell.

Quantification of transcript abundance

Paired-end reads were aligned to the transcriptome using bwa (version 0.6.2-r126) with default parameters (65). The transcriptome contained all gene models based on the mouse ENCODE VM9 release downloaded from the UCSC genome browser comprising 57,207 isoforms with 57,114 isoforms mapping to fully annotated chromosomes (1 to 19, X, Y, M). All isoforms of the same gene were merged to a single gene locus. Furthermore, gene loci overlapping by > 75% were merged to larger gene groups. This procedure resulted in 34,111 gene groups. The right mate of each read pair was mapped to the ensemble of all gene loci and to the set of 92 ERCC spike-ins in sense direction (66). Reads mapping to multiple loci were discarded. The left read contained barcode information: the first six bases corresponded by six bases representing the unique molecular identifier (UMI) followed by the cell-specific barcode. The remainder of the left read contained a polyT stretch. The left read was not used for quantification. For each cell barcode, the number of UMIs per transcript was counted and aggregated across all transcripts derived from the same gene locus. Based on binomial statistics, the number of observed UMIs was converted into transcript counts (67).

scRNA sequencing data analysis

Fifty-four libraries were sequenced and after quality control 3461 cells (PV: 1324, Men: 1052, CP: 701, Blood: 384) were analyzed. Data analysis and visualization were performed using the RaceID3 algorithm (21). Cells with a total number of transcripts < 1,500 were discarded and count data of the remaining cells were normalized by downscaling. Cells expressing > 2% of Kcnq1ot1, a potential marker for low-quality cells, were not considered for analysis. Additionally, transcript correlating to Kcnq1ot1 with a Pearson’s correlation coefficient > 0.65 were removed. The following parameters were used for RaceID3 analysis: mintotal = 1500, minexpr = 5, outminc = 5, FSelect = TRUE, probthr = 10-8. The top 20 principal components of the datasets were considered for clustering using the CCcorrect() function with the parameters nComp = 20 and mode = “pca”. We found no batch-associated variability in the dataset. Furthermore, to assess the batch effects associated with tissue dissociation-induced genes (68), RaceID3 was re-run where CGenes argument was initialized with the following genes: Dusp1, Jun, Fos, Hspa1a, Atf3 and Malat1. There was no visible dissociation-induced variability in the dataset. Differentially expressed genes between two subgroups of cells were identified similar to a previously published method (69). First, negative binomial distributions reflecting the gene expression variability within each subgroup were inferred based on the background model for the expected transcript count variability computed by RaceID3 (21). Using these distributions, a P value for the observed difference in transcript counts between the two subgroups was calculated and multiple testing corrected by the Benjamini-Hochberg method.

Confocal microscopy and image processing for microglia clonal expansion

Six-channel images were acquired using a gated SP8 STED-WS confocal microscope (Leica Microsystems) with a HCX PL HCL PL APO C 20×/0.75 NA glycerine objective lens and the LAS X software. Fluorophores, including Alexa Fluor 405, membrane-localized CFP, nuclear GFP, cytoplasmic YFP, cytoplasmic RFP, and Alexa Fluor 647, were detected using sequential and simultaneous acquisition mode with the HyD detectors in the gating mode. The excitation wavelengths used were: UV Diode Laser 405 nm, Argon Laser 458 nm, WL 480 nm, 525 nm, 565 nm, and 647 nm. The pinhole was set to 1 AU. The 30-μm stacks comprising 2048 pixel by 1500 pixel tiles were sampled at a 284-nm pixel size and 1-μm z-steps. Overlapping tiles were acquired and automatically stitched using XuvTools (70). An average of nine tiles per animal was imaged.

Computational analysis for clonal expansion

IBA-1+ and Confetti+IBA-1+ microglial cells were detected in three dimensions by using our custom Matlab-based program and assessment of microglial clonality as detailed previously (40). IBA-1+ meningeal and perivascular macrophages were excluded based on cell morphology and location (also indicated by collagen+ vessels). Briefly, we performed 10,000 Monte Carlo simulations to derive the baseline random distribution of Confetti+IBA-1+ microglial cells detected in each animal. The averages of measured Confetti+IBA-1+ microglial distributions were compared to the 98th percentile of simulated results to determine the probability of EAE-associated microglial clonal expansion. Clone sizes were estimated on the assumption that neighboring Confetti+IBA-1+ microglial cells of the same color belonged to the same clone given the extremely low number of Confetti+ cells in the homeostatic spinal cord.

Statistical analysis

Statistical analysis was performed using GraphPad Prism (GraphPad Software, Version 6.0, La Jolla, USA). Data were tested for normality applying the Shapiro–Wilk normality test. If normality was given, an unpaired t-test or one-way analysis of variance (ANOVA) was applied. If the data did not meet the criteria of normality, a Kruskall–Wallis test followed by Dunn’s multiple comparisons test was applied to datasets of more than two groups; otherwise, the Mann–Whitney U test was applied for data with two groups. Differences were considered significant when the P value was <0.05.

Supplementary Materials

References and Notes

Acknowledgments: The authors thank M. Oberle for excellent technical support. Funding: This work was supported by the the Sobek Foundation (M.P.); the Ernst-Jung Foundation (M.P.); the Deutsche Forschungsgemeinschaft (DFG) (SFB 992, SFB1160, SFB/TRR167, Reinhart-Koselleck-Grant) (M.P.); the Ministry of Science, Research and Arts, Baden-Wuerttemberg (Sonderlinie “Neuroinflammation”) (M.P.); the European Union’s Seventh Framework Program FP7 under grant agreement 607962 (nEUROinflammation) (M.P. and M.J.C.J.); the Bundesministerium für Bildung, Wissenschaft, Forschung und Technologie (BMBF)–funded competence network of multiple sclerosis (KKNMS) (M.P. and M.K.); DFG grant GR4980 (S.); and the Behrens-Weise-Foundation (S.), the Max Planck Society (S. and D.G.), the DFG grant KIDGEM (SFB 1140) (D.M.), the Centre for Biological Signalling Studies (BIOSS) (EXC294) (D.M.), and the European Research Council (ERC) Starting Grant (337689) (O.G). Author contributions: M.J.C.J., S.M.B., E.S., S.A., N.H., T.L.T., Ö.Ç., D.M., O.G., and T.F. conducted experiments, analyzed the data, or provided essential tools for the study. G.L., Y.-H.T., and M.K. contributed to the in vivo imaging studies. R.S. and S. analyzed the scRNA-seq data under the supervision of D.G.; M.J.C.J. and M.P. supervised the project and wrote the manuscript. Competing interests: The authors declare no competing interests. Data and materials availability: The scRNA-seq data are deposited in the Gene Expression Omnibus under the accession no. GSE118948. All other data needed to evaluate the conclusions in this paper are present either in the main text or the supplementary materials.
View Abstract

Navigate This Article