Neural precursor cell–secreted TGF-β2 redirects inflammatory monocyte-derived cells in CNS autoimmunity

Multiple sclerosis (MS) is an autoimmune disease of the central nervous system (CNS), representing the most common CNS inflammatory disorder and the second leading cause of disability in young adults. In experimental autoimmune encephalomyelitis (EAE), the animal model of MS, myelin-reactive pathogenic T helper (Th) cells enter the CNS at the level of pial vessels and are reactivated in the subarachnoid space by meningeal perivascular macrophages (1) that allow pathogenic Th cells tissue entry and initiation of inflammation (2). Pathogenic Th cells induce, through granulocyte-macrophage colony-stimulating factor (GM-CSF), an inflammatory program in monocytes and their progeny, monocyte-derived dendritic cells (moDCs) (2, 3). In the CNS, moDCs function as antigen-presenting cells (APCs) to pathogenic Th cells, secreting polarizing cytokines such as IL-1, IL-23, IL-6, and TNF-α (4–7), and directly contribute to myelin and axonal loss via phagocytosis and the production of ROS and other neurotoxic mediators (3, 8).

Although available therapies for MS effectively target the recruitment of T cells from the periphery into the CNS (9), there is no conclusive evidence regarding their ability to target either the self-maintaining inflammatory cycle that occurs locally within the CNS or the pathogenic phenotype of CNS-infiltrating monocyte-derived cells (MCs) (10).

Neural precursor cells (NPCs) have been previously studied as a therapeutic strategy for MS for their regenerative potential, and, more recently, for their neuroprotective and immunomodulatory properties (11–14). NPCs express chemokine receptors and adhesion molecules that allow them, upon transplantation, to reach inflamed areas in the CNS (11, 15–17) and to release a variety of cytokines and neurotrophins that can inhibit the adaptive and innate immunity (18–20). While previous studies found intrathecal transplantation of NPCs to ameliorate EAE disease severity (14, 21, 22), the immunological mechanism underlying the therapeutic efficacy of NPCs delivered in the inflamed CNS still remains unknown.

In this work we investigated whether intrathecal NPC treatment could directly interfere with the CNS-restricted crosstalk between pathogenic Th cells and inflammatory MCs. We found that NPC treatment impaired the accumulation of infiltrating MCs in the CNS and reduced EAE disability. In fact, NPCs limited the expression of cytokines required for the polarization of encephalitogenic GM-CSF–producing Th cells. NPCs inhibited in vivo and in vitro DC differentiation and maturation, favoring a fate switch toward an antiinflammatory phenotype. We identified TGF-β2 as the primary factor secreted by NPCs for the observed therapeutic effect in the chronic phase of EAE.

Intrathecal NPC transplantation ameliorates EAE by reducing inflammatory MCs in the CNS. To examine the capacity of NPC transplantation to inhibit the CNS-restricted inflammation occurring in EAE, we first validated the therapeutic potential of intrathecal NPC injection in C57BL/6 mice immunized with myelin oligodendrocyte glycoprotein (MOG)_35–55 peptide. One million NPCs in PBS (NPC-treated) or PBS alone (PBS-treated) was intrathecally injected into the cisterna magna at the peak of disease severity, 4 days after clinical onset. We observed that NPC-treated mice displayed faster recovery and lower disability compared with EAE mice treated only with PBS (Figure 1A). At 1 month after treatment (60 days postimmunization [dpi]), NPC-treated mice showed significantly lower neurological impairment, and at the end of the follow-up (80 dpi), NPC-treated mice showed either complete functional recovery or minor tail paralysis, while PBS-treated mice maintained an overt paresis of the hind limbs. Accordingly, NPC-treated mice displayed a lower lesion burden measured as demyelination and axonal loss in the spinal cord at the end of follow-up (Figure 1, B–D). Interestingly, we found 2 weeks after treatment (40 dpi) a significant reduction of the inflammatory infiltrate areas (Figure 1E) and in particular of CD3⁺ T cells (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/JCI92387DS1) and of IB4⁺ phagocytes (Supplemental Figure 1B) in the spinal cord of NPC-treated compared with PBS-treated mice. We further observed that NPC-treated mice, compared with PBS-treated controls, already at this early time point, displayed a parallel reduction of both inflammation (IBA1⁺ areas) and demyelination (loss of MBP⁺ areas) in the spinal cord (Supplemental Figure 1, C and D). At this same time point, we observed that most transplanted GFP⁺ NPCs localized in the subarachnoid space of cerebellum, brainstem, and cervical spinal cord, with similar spatial distribution of infiltrating extraparenchymal CD11b⁺ myeloid cells (Figure 2, A and B, and Supplemental Figure 2A). Transplanted NPCs were found in proximity to meningeal vessels and CD45⁺ perivascular inflammatory cells, including CD4⁺ T cells, CD11b⁺ myeloid cells, MHC class II⁺ cells, and CD11c⁺ DCs (Figure 2, C–F). Transplanted NPCs retained an undifferentiated phenotype up to the end of follow-up, at 80 dpi, displaying only in low percentage markers for neuronal (β-tubulin III, β-tub III⁺), oligodendroglial (oligodendrocyte lineage transcription factor 2, Olig2⁺), or astroglial lineage (glial fibrillary acidic protein, GFAP⁺) (Supplemental Figure 2, B–D and G). Further, transplanted NPCs expressed the stem cell marker nestin and the proliferation marker Ki-67 (Supplemental Figure 2, E and F). Notably, intrathecally transplanted GFP⁺ NPCs were detected only in the CNS of NPC-transplanted mice (Supplemental Figure 3, A and B), while NPCs were found neither in peripheral organs (Supplemental Figure 3, C–E) nor in secondary lymphoid organs (Supplemental Figure 3, F–H).

Figure 1

Intrathecally administered NPCs ameliorate chronic EAE. (A) Top: Clinical scores of EAE mice treated at the peak of the disease (4 days after the clinical onset, arrow) with intrathecal injection (into cisterna magna) of either NPCs (red dots) or PBS (black dots), n = 20 mice per group. Each point represents the mean ± SEM; 70 to 82 dpi, 2-way ANOVA with Bonferroni’s post-test, *P ≤ 0.05. Bottom: Linear regression curves of clinical score from the day of treatment with either NPCs or PBS until 60 or 80 dpi. Dashed lines indicate 95% CI. **P ≤ 0.01; ***P ≤ 0.001. (B–D) Quantification and representative images of spinal cord demyelination (Luxol fast blue staining) (B) and axonal loss quantified on Bielschowsky silver staining (C; at bottom right, high-magnification insets, ×100) and immunohistochemistry for neurofilaments (NF) (D) at 80 dpi in PBS-treated (black bars) versus NPC-treated (white bars) EAE mice (n = 15 sections per mouse, 3–6 mice per group). (E) Quantification and representative images of inflammatory infiltrates (H&E staining) at 30 dpi in PBS-treated (black bars) versus NPC-treated (white bars) EAE mice (n = 12–20 sections per mouse, 3 mice per group). Data are mean ± SEM and represent the percentage area of damage over total section area. *P ≤ 0.05, **P ≤ 0.01, unpaired t test. Scale bars: B–E, 100 μm.

Figure 2

Transplanted NPCs distribute in the subarachnoid space within the inflammatory infiltrates. (A and B) Representative 3D reconstruction of the relative distribution of transplanted NPCs (GFP⁺ in green) and CD11b⁺ inflammatory cells (in red) within the hindbrain (A, rostral and caudal view) and the cervical spinal cord (B) (outlined in gray), the cerebral aqueduct and IV ventricle (outlined in yellow), in an NPC-treated mouse at 7 days after transplantation, obtained from a total of n = 33 sequential stereology sections. (C–F) Representative immunofluorescence images of intrathecally transplanted NPCs (GFP⁺ in green, arrowheads in the magnified insets, shown below, of the above dashed boxes) that localize in the meningeal perivascular spaces of the spinal cord and brainstem, often detected in proximity to von Willebrand factor–positive (vWF⁺) endothelial cells (white in C), to CD45⁺ blood-derived leukocytes (red in C), to perivascular infiltrating CD11b⁺ myeloid cells (red in D), to CD4⁺ T cells (white in D), and to MHC-II⁺ as well as CD11c⁺ local APCs (red in E and F, respectively). Nuclei are stained by DAPI (blue). Scale bars: C–F, 25 μm.

To better investigate the mechanism regulating the clinical and pathological recovery induced by transplanted NPCs, we performed flow cytometry analysis of CNS-resident and infiltrating inflammatory cells of EAE mice 7 days after NPC or PBS treatment. We observed a significant reduction in frequency and cell number of inflammatory MCs (CD45^hiCD11b⁺Ly6G^–Ly6C⁺CD11c⁺MHC-II⁺) (23) as well as a trend of reduction of infiltrating lymphoid cells (CD45^hiCD11b^–) in the CNS of NPC-treated compared with PBS-treated EAE mice (Figure 3, A–C). Microglia (CD45^loCD11b⁺Ly6G^–Ly6C^–) were not quantitatively altered after NPC treatment (Figure 3, A and D), but displayed lower expression of MHC-II (Figure 3, E and F), suggestive of a lower activation state.

Figure 3

NPC transplantation reduces the accumulation of inflammatory MCs and the activation of microglia during the effector phase of EAE. (A–D) Flow cytometry of CNS leukocytes from PBS- and NPC-treated EAE mice, assessed at day 7 after transplantation. Frequencies and total cell number of CNS-infiltrating lymphoid cells (LCs; CD45^hiCD11b^–) and inflammatory MCs (CD45^hiCD11b⁺Ly6G^–Ly6C^hiCD11c⁺MHC-II⁺) are shown in the plots (A) and quantified in the graphs (n = 4–5 mice per group) (B and C). Frequencies and total cell number of CNS-resident microglia (CD45^intCD11b⁺Ly6G^–Ly6C^–) are shown in the plots (A) and quantified in the graphs (n = 4–5 mice per group) (D). (E and F) Level of MHC-II on microglia in PBS-treated (black line) and NPC-treated EAE mice (red line) at day 7 after transplantation is shown in the histogram (mean fluorescence intensity, MFI) (E) and quantified in the graph (n = 4 mice per group) (F). Black bars, PBS-treated mice; white bars, NPC-treated mice. Data are mean ± SEM and are representative of 2 independent experiments. *P ≤ 0.05, unpaired t test.

NPC treatment hampers the effector phase of EAE. The accumulation of myeloid cells in the CNS during the effector phase of EAE relies on the ability of CNS-invading Th cells to secrete GM-CSF (2, 3). NPC-transplanted mice had significantly lower frequencies of GM-CSF⁺ cells in CNS-infiltrating Th1 and Th17 cells (Figure 4A) as well as in total CD4⁺ T cells (Figure 4B). No significant difference between NPC- and PBS-treated mice, neither in the frequency of FoxP3⁺ Tregs nor in the level of FoxP3 expression by CNS-infiltrating CD4⁺ T cells, was observed (Figure 4, C and D). Consistently, CD4⁺ T cells in the CNS showed significantly lower expression of transcripts related to Th17 pathogenicity (24), such as Il23r, Rorc, Tbx21, Il22, Cxcl3, and Csf2, in NPC-treated mice (Figure 4E). Furthermore, GM-CSF levels were significantly reduced in the hindbrain and cervical spinal cord of NPC-treated mice (Figure 4F).

Figure 4

NPC transplantation dampens the pathogenic signature of the CNS-infiltrating Th cells. (A and B) Production of GM-CSF by CNS-infiltrating Th1 (CD45^hiCD3⁺CD4⁺IFN-γ⁺) and Th17 cells (CD45^hiCD3⁺CD4⁺IL-17A⁺) in EAE mice 7 days after PBS or NPC treatment as analyzed by flow cytometry. Frequency of GM-CSF⁺ Th1, GM-CSF⁺ Th17, and total GM-CSF⁺ Th cells are shown in the plots (A) and quantified in the graph (n = 4 mice per group) (B). (C) Frequency of FoxP3⁺ CNS-infiltrating CD4⁺ T cells (CD45^hiCD11b^–CD4⁺) in PBS- and NPC-treated EAE mice at the same time point as A is shown in plots and graph (n = 4 mice per group). (D and E) Expression of FoxP3 (D) and Rorc, Il23r, Tbet, Il22, Csf2, and Cxcl3 (E) in CD45^hiCD11b^–CD4⁺ Th cells sorted from the CNS of PBS- and NPC-treated EAE mice at the same time point, relative value to PBS-treated mice (n = 4 mice per group). (F) GM-CSF quantification by ELISA in hindbrain and cervical and dorsal spinal cord (SC) tissues of PBS- and NPC-treated EAE mice at 7 days after transplantation (n = 6 mice per group). Black bars, PBS-treated mice; white bars, NPC-treated mice. Data are mean ± SEM and are representative of 2 independent experiments. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, unpaired t test.

The capacity of autoreactive Th cells to secrete GM-CSF depends on the production of proinflammatory cytokines by moDCs (2, 3, 7, 25). Seven days after transplantation, we found significantly lower expression of Il23a and Tnfa transcripts in the CNS inflammatory infiltrate of NPC-treated mice (Figure 5A). Moreover, IL23p19 protein levels were significantly reduced in the CNS of NPC-transplanted mice (Figure 5B). Accordingly, cytokine profiling of CNS tissues revealed also decreased levels of IL-1β and TNF-α and of IL-23–related factors such as G-CSF, CXCL1, and IL-17A (Figure 5C). These findings support the idea that NPC treatment dampens the production of polarizing cytokines by inflammatory MCs, which in turn reduce GM-CSF production and the pathogenic potential of CNS-invading Th cells.

Figure 5

NPC transplantation impairs the production of polarizing cytokines by CNS-infiltrating APCs. (A) Expression of Il23 and Tnf in CNS leukocytes isolated from EAE mice at 7 days after PBS or NPC treatment, relative value to PBS-treated mice (n = 3 mice per group). Data are representative of 2 independent experiments. (B) ELISA for IL23p19 in hindbrain and cervical and dorsal spinal cord tissues from PBS- and NPC-treated EAE mice obtained at the same time point as A (n = 3 mice per group). Data are representative of 2 independent experiments. (C) Cytokine profile in hindbrain tissues of PBS- and NPC-treated mice obtained at the same time point as A. Hindbrain tissue lysates were analyzed for expression of various cytokines by Luminex-based assay (n = 10 mice per group). Black bars, PBS-treated; white bars, NPC-treated. Data are mean ± SEM. *P ≤ 0.05, **P ≤ 0.01, unpaired t test.

NPC-secreted factors inhibit maturation of DCs and their capacity to trigger GM-CSF production in myelin-reactive Th cells. We hypothesized that an NPC-secreted factor could inhibit the antigen-presenting and proinflammatory properties of myeloid cells. To investigate this hypothesis, we exposed in vitro bone marrow–derived DCs (BMDCs) to NPC-conditioned medium (NCM) or control DC culture medium (DCM) during stimulation with CD40L, mimicking T cell interaction with APCs. BMDCs exposed to NCM expressed lower levels of MHC-II and of the costimulatory molecules CD80 and CD86 compared with BMDCs in DCM (Figure 6, A and D). Cytokine profiling showed that BMDCs exposed to NCM produced a significantly lower amount of polarizing cytokines, including TNF-α, IL23p19, IL-6, and IL12p70 and chemokines such as CXCL10, CXCL1, and G-CSF (Figure 6B). We further confirmed the inhibitory effect of NCM on IL-23 and TNF-α production by ELISA (Figure 6C) and quantitative PCR (Figure 6D). Finally, we assessed the ability of BMDCs exposed to NCM during CD40L stimulation to reactivate CD4⁺ T cells from MOG_35–55 peptide–immunized 2D2 mice. Conditioning of BMDCs with NCM during CD40L stimulation decreased their ability to induce the production of GM-CSF in Th1 and Th17 cells (Figure 7).

Figure 6

NPC-secreted factors impair the maturation of BMDCs. (A) Levels of MHC-II, CD86, and CD80 on CD40L-stimulated BMDCs in the absence or presence of NCM for 48 hours. Gray lines represent immature unstimulated BMDCs (iDCs), black lines BMDCs matured in control medium (mDC), red lines CD40L-matured BMDCs in the presence of NCM (mDC-NCM), and filled gray histograms fluorescence minus one controls (n = 3 samples per group). *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, 1-way ANOVA with Bonferroni’s post-test. (B) Unsupervised clustering of cytokine profile in supernatants of CD40L-matured BMDCs in the absence (mDC) or presence of NCM (mDC-NCM). Supernatants were collected after 48 hours in culture, 0.2-μm-filtered, and analyzed for expression of various cytokines by Luminex-based assay (individual samples are shown). (C) Protein levels of indicated cytokines in iDCs, mDC, and mDC-NCM, validated by ELISA (n = 3 samples per group). *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, 1-way ANOVA with Bonferroni’s post-test. (D) Expression of Cd80, Cd86, Il23a, Il12b, Tnf, Il1b, and Il6 in BMDCs treated as in B for 18 hours, relative value to BMDCs matured in control medium (n = 3 samples per group). *P ≤ 0.05, **P ≤ 0.01, unpaired t test. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, 1-way ANOVA with Bonferroni’s post-test. Gray bars, iDCs; black bars, mDC; white bars, mDC-NCM. Data are mean ± SEM and are representative of 2 independent experiments.

Figure 7

BMDCs matured in the presence of NPC-secreted factors fail to polarize MOG_35–55-specific CD4⁺ T cells toward GM-CSF production. (A and B) Frequency of Th1 (GM-CSF⁺IFN-γ⁺) and Th17 (GM-CSF⁺IL-17⁺) among 2D2 CD4⁺ T cells stimulated with MOG_35–55 in the presence of BMDCs cultured as described in Figure 6 are shown in the plots (A) and graphs (B) (n = 3–4 samples per group). *P ≤ 0.05, ***P ≤ 0.001, 1-way ANOVA with Bonferroni’s post-test. Gray bars, iDCs; black bars, mDC; white bars, mDC-NCM. Data are mean ± SEM and are representative of 2 independent experiments.

In conclusion, NPC-secreted factors inhibit in vitro the capacity of DCs to induce GM-CSF–producing pathogenic Th cells.

NPC-secreted factors guide transcriptional reprogramming of inflammatory DCs toward an alternative antiinflammatory activation. We next sought to dissect the sequence of molecular events triggered by NPCs that induces immunomodulation in inflammatory myeloid cells. To follow over time the dynamic changes induced by NPCs in BMDCs, we performed gene expression profiling on BMDCs after 6 and 18 hours of CD40L stimulation in the absence (mDC) or presence of NCM (mDC-NCM). WT unstimulated immature DCs (iDCs) were used as reference group to identify maturation-related transcripts (Figure 8A).

Figure 8

NPC-secreted factors redirect the transcriptional program of BMDC maturation toward an alternative activation. (A) Workflow of microarray gene expression analysis at 6 and 18 hours showing the differential expression between control medium (mDC) and NCM-cultured BMDCs (mDC-NCM) during CD40L-induced maturation. (B) Unsupervised hierarchical clustering of expression profiles of 130 and 175 differentially modulated genes with ≥1.3-fold change and a P ≤ 0.01 in mDC-NCM and mDC at 6 and 18 hours in culture, respectively, compared with the expression profiles of the same transcripts in unstimulated BMDCs (iDCs). Dashed lines indicate subclusters of genes that in mDC-NCM return to the expression levels observed in iDCs. (C) Gene ontology enrichment of differentially modulated genes between mDC-NCM and mDC at 6 and 18 hours. The gray scale indicates the statistical significance (FDR ≤0.05). (D) Color-coded ratio of mDC-NCM over mDC showing NCM-regulated genes at 18 hours, clustered to the indicated pathways (blue means downregulation, red upregulation). Significance is shown alongside. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, unpaired t test. (E and F) Quantitative PCR validation of indicated differentially expressed genes in mDC-NCM compared with mDC at 6 (E) and 18 hours (F); the expression is relative to mDC (n = 3 samples per group). *P ≤ 0.05, **P ≤ 0.01, unpaired t test.

Bioinformatics analysis identified 2 sets of unique genes that were differentially regulated between mDC and mDC-NCM after 6 or 18 hours in culture with a significance threshold of P ≤ 0.01 (Figure 8A and Supplemental Table 1).

Unsupervised hierarchical clustering clearly separated NCM-treated mDC from control mDC, both at the early (6 hours) and at the late (18 hours) time point (Figure 8B). Subclustering analysis at each time point highlighted subsets of genes shared only between mDC-NCM and iDCs, suggesting an impairment of the CD40L-induced maturation by NPC-secreted factors (Figure 8B). However, the exposure to NCM during CD40L stimulation instructed in DCs a specific gene expression signature, demonstrating the induction of an alternative program of activation/differentiation.

Indeed, gene ontology analysis identified enriched functional groups associated with metabolic processes (Arg1, Arg2, ferroportin-1), macromolecule biosynthesis, regulation of transcription (Cbx5 and Itch), and programmed cell death in mDC-NCM at the early time point (Figure 8C). Genes related to oxidative stress and apoptosis resistance, such as Ak2, Tgm2, and Smarca2 (26, 27), were upregulated whereas proapoptotic genes such as Tnfrsf6 (Fas) were downregulated in mDC-NCM. In addition, NCM induced in DCs transcripts related to the alternative or antiinflammatory phenotype, such as Arg1, Maf, Itch, Stab1, and Tgm2 (28–32). Accordingly, mDC-NCM also upregulated the expression of Ddx21, which is required for the induction of S100A9 (33), an immaturity marker of myeloid-derived suppressor cells (MDSCs) (ref. 34; Figure 8, B and E; and Supplemental Table 1). At 18 hours, gene ontology analysis of NCM-regulated genes confirmed the NCM-induced shift of mDC toward an antiinflammatory activation. Indeed, NCM-treated mDC had a significantly lower expression of genes encoding proinflammatory chemokines (Ccl1, Ccl2, Ccl3, Ccl7, Ccl8, Cxcl2) (3, 35), positive regulators of cytokine production (CD36, Il33, Clec9a, and Clec5a) (36–39), and phagocytosis-related genes (Ptx3, Elmo2, Clec7a) (ref. 3 and Figure 8, B, D, and F). Conversely, NCM induced the antiinflammatory molecules Hspa1a, which impairs the stimulatory capacity of moDCs (40), and Il1r2, the IL-1 decoy receptor that blocks IL-1 signaling in inflammatory macrophages (41) and the NF-κB inhibitor Ascc1 (ref. 42; Figure 8, B, D, and F; and Supplemental Table 1).

Transcription factor analysis of early (6 hours) NCM-regulated genes showed an enrichment in positively regulated transcripts controlled by c-Myc, a master regulator in macrophage alternative activation (43, 44), and c/EBPβ, which is the main regulator of MDSC differentiation (ref. 45 and Supplemental Figure 4A). At the later time point (18 hours), many negatively NCM-regulated transcripts were enriched in genes controlled by PU.1 (46) and NF-κB1, both pivotal in DC differentiation and maturation (Supplemental Figure 4B, Supplemental Table 2, and refs. 46, 47).

When we compared the differentially expressed genes of CD40L-induced maturation in the absence (iDCs vs. mDC) or presence of NCM (iDCs vs. mDC-NCM), we identified pathways only modulated during the maturation induced in the presence of NPC-secreted factors. The pathways included at 6 hours the antiapoptotic TNFs/NF-κB/BCL-2 pathway and NFAT in immune response (Supplemental Figure 5A), and at 18 hours IL-4–induced regulators, p53 signaling, NO synthesis and signaling, Fas signaling, and TGF-β–dependent activation, and were associated with an alternative activation of DCs (Supplemental Figure 5B).

Specifically, among early NCM-regulated genes, Rad51 (48) and Arhgap15 (49), known to be inhibited by TGF-β, were downregulated, while Arg1, which is induced by TGF-β (ref. 50; Figure 8, A and E; and Supplemental Table 1), was increased. Similarly, CD36 and Il33, both inhibited by TGF-β, were significantly downregulated in mDC-NCM at 18 hours (ref. 51; Figure 8, B, D, and F; and Supplemental Table 1).

We therefore investigated whether NPCs could influence myeloid cell differentiation and activation also in vivo and possibly through TGF-β signaling during EAE. To this end, MOG_35–55-immunized EAE mice were injected, at the peak of the disease, with 10⁶ NPCs or vehicle. At 7 days after transplantation, CNS-infiltrating MCs were FACS-sorted (Figure 9A) and mRNA was extracted. Next-generation sequencing was performed, and genome-wide profiles were generated and compared between MCs isolated from the CNS of NPC-treated animals and PBS-treated animals in EAE. We identified 610 genes that were differentially expressed with the significance threshold of P ≤ 0.01 (Figure 9B and Supplemental Table 3). Hallmark gene set enrichment analysis confirmed the modulation of metabolic processes and of several immunological pathways, including TGF-β signaling, the IFN response, and genes upregulated by ROS, in MCs from NPC-treated mice (Figure 9, C and D). In line with in vitro data, transcription factor analysis found c-Myc and CREB1 (Supplemental Figure 6 and Supplemental Table 4), master transcription factors for the alternative activation of macrophages (43, 52), as main regulators of NPC-induced gene signature in CNS-infiltrating MCs.

Figure 9

NPC treatment alters the gene expression signature of CNS-infiltrating inflammatory MCs in EAE. (A) Cohorts of 4–7 MOG_35–55-immunized C57BL/6 mice intrathecally treated with either PBS or NPCs at the peak of the disease (2–4 days after clinical onset). At 7 days after transplantation, CNS tissues were pooled and CNS-infiltrating MCs were FACS-sorted according to the phenotype CD45^hiLy6G^–CD11b⁺Ly6C^hiMHC-II⁺. Sorting strategy used for 3 independent FACS sorting experiments is shown. (B) Next-generation sequencing was performed on RNA extracted from sorted cells of 3 independent experiments and respective CNS harvests. Six hundred ten genes that are significantly altered in 3 different statistical tests to a minimum significance threshold of P ≤ 0.01 are shown in the heatmap. (C) Volcano plot showing the fold change and significance of genes in MCs from PBS- versus NPC-treated EAE mice. (D) Bipartite graph of functional enrichment analysis using hallmark data set for NPC treatment obtained by Cytoscape software (http://www.cytoscape.org/). Black and gray nodes represent enriched pathways with sizes corresponding to FDR-adjusted enrichment P value (P ≤ 0.05). Red dots represent upregulated genes and blue dots downregulated genes, whereas the dot size indicates significance (P ≤ 0.01).

Overall these findings suggest that NPC-secreted factors skew the inflammatory differentiation program of MCs toward an alternative/antiinflammatory activation (53), possibly through TGF-β signaling, both in vitro, during DC activation, and in vivo, in the inflamed CNS.

TGF-β2 mediates the inhibitory effect of NPCs on DC maturation. We next focused on identifying the secreted factor(s) responsible for the observed immunomodulatory effects of NPCs on inflammatory DCs.

First, we observed that NCM inhibits the production of IL-23 by CD40L-stimulated BMDCs in a dose-dependent manner (Supplemental Figure 7A). To investigate the nature of the secreted factor, NCM frozen at –20°C was thawed and then filtered through a 0.1-μm filter (Supplemental Figure 7B). The process did not alter NCM inhibitory activity, indicating that the factor is neither released as exosome nor as microvesicle. Further, size fractionation demonstrated that the inhibitory activity of NCM on DC maturation was absent when the secretome of ≤3 kDa was used, excluding small molecules as possible mediators of NCM effect (Supplemental Figure 7C). However, proteinase K treatment or heat inactivation (80°C) abolished NCM activity (Supplemental Figure 7D), indicating that the active secreted factor was a protein.

Since transcriptome analyses suggested the involvement of TGF-β signaling in the modulation of myeloid inflammatory cells by NPCs (Figures 8 and 9), we focused our investigation on elements of the TGF-β family, morphogenic factors with known immunomodulatory and tolerogenic effects on both T cells and DCs (54–58). NPCs release in vitro TGF-β2 and are equipped to promote activation of latent TGF-β2, via matrix metalloproteinases and thrombospondins (Figure 10, A and B). We further showed that BMDCs express receptors to transduce TGF-β signaling (Figure 10A). Supplementation of BMDC culture with TGF-β2 resulted in marked and dose-dependent inhibition of IL-23 secretion by CD40L-stimulated BMDCs (Figure 10C). Consistently, pharmacological inhibition of TGF-β signaling with SB431542, the tyrosine kinase inhibitor of TGF-β receptor 1 (TGFBRI), or with ITD1, which induces the proteosomal degradation of TGF-β receptor 2 (TGFBRII) (59), reverted almost completely the effect of NCM on DCs (Figure 10, D and E).

Figure 10

NPCs inhibit BMDC maturation in vitro via TGF-β2. (A) Color-coded expression heatmap of selected TGF-β family growth factors and TGF-β activating enzymes in NPCs (left) and of their relative receptors on mDC (CD40L-stimulated BMDCs; right). Data are represented as ΔCt over Gapdh (note that a higher ΔCt denotes a lower expression level). (B) ELISA for TGF-β2 performed on NCM from NPCs cultured in DC medium for 2, 6, 24, and 48 hours. (C) Quantification by ELISA of IL23p19 secreted in the supernatant of mDC cultured either in control medium (black bar) or in the presence of recombinant TGF-β2 at increasing concentrations (gray bars). (D and E) Quantification by ELISA of IL23p19 secreted by mDC cultured either in NCM alone (white bar) or in the presence of the TGFBRI tyrosine kinase inhibitor SB431542 (gray bars) (D) or with the TGFBRII inhibitor ITD1 (gray bar) (E). (F) Quantification by ELISA of the IL23p19 secreted by mDC from control littermates CD11c-Cre⁺Tgfbr2^WT/WT (white bar) and CD11c-Cre^–Tgfbr2^fl/fl (gray bar) versus CD11c-Cre⁺Tgfbr2^fl/fl (gray striped bar) mDC lacking TGFBRII in CD11c⁺ cells. Data in D-F are fold induction (FI) relative to mDC matured in control medium subjected to the same treatment. (G) Expression of Tgfb2 mRNA in WT (Tgfb2^+/+), Tgfb2^+/–, and Tgfb2^–/– eNPCs cultured for 48 hours in DCM relative to WT NPCs. (H) TGF-β2 protein quantification by ELISA secreted by Tgfb2^+/+, Tgfb2^+/–, and Tgfb2^–/– eNPCs cultured for 48 hours in DC medium. (I) IL23p19 production (measured by ELISA) by mDC cultured in NCM from Tgfb2^+/+ (white bar) and Tgfb2^–/– (gray bar) eNPCs. Data are FI relative to mDC matured in control medium. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, 1-way ANOVA with Bonferroni’s post-test (C, D, F, H, and I); *P ≤ 0.05, **P ≤ 0.01, ****P ≤ 0.0001, unpaired t test (E, G). Data are mean ± SEM and are representative of at least 2 independent experiments.

Furthermore, we confirmed that TGF-β2 signaling on BMDCs was required for NPC immunomodulation, since the genetic ablation of TGFBRII on CD11c⁺ BMDCs, obtained from CD11c-Cre⁺Tgfbr2^fl/fl mice (60, 61), made DCs insensitive to the NCM effect (Figure 10F).

To investigate the role of TGF-β2, we generated NPCs from Tgfb2tm1Doe mice (Tgfb2^–/–), which constitutively lack functional TGF-β2 (62). Given the perinatal lethality of homozygous mutant mice (62), NPCs were derived from E15.5 Tgfb2^–/– mouse embryos (embryonic NPCs [eNPCs]). Tgfb2^–/– eNPCs were not affected in their viability and proliferative capacity in vitro (Supplemental Figure 8A); they efficiently differentiated in vitro into astrocytes and oligodendrocytes, producing slightly more neurons compared with WT (Tgfb2^+/+) eNPCs (Supplemental Figure 8, B–E). Importantly, Tgfb2^–/– eNPCs maintained the expression of CD44, integrin α₄, and CXCR4 (Supplemental Figure 8F), adhesion molecules crucial for the homing and persistence of transplanted NPCs in the CNS during EAE (11).

The expression of TGF-β2 in eNPCs derived from heterozygous and homozygous Tgfb2tm1Doe mice was reduced in an allele-dosage manner at both the mRNA and the protein level (Figure 10, G and H). In line with this, NCM obtained from Tgfb2^–/– eNPCs had significantly lower inhibitory activity on IL-23 production by BMDCs (Figure 10I).

Overall, these data pointed to TGF-β2 as the main driver of the immunomodulatory effect of NPC on BMDC differentiation and maturation.

TGF-β2 is nonredundant for the therapeutic effect of NPCs in chronic EAE. Upon intrathecal transplantation of NPCs in EAE mice, we observed that NPCs secrete TGF-β2 in vivo (Supplemental Figure 9, A and B) and that adjacent CNS-infiltrating CD11b⁺ cells were positive for phosphorylated SMAD2 (p-SMAD2), suggestive of an activation of the TGF-β pathway in the MCs induced by NPC treatment (Supplemental Figure 9C).

To understand the therapeutic relevance of TGF-β2 produced by NPCs, we intrathecally transplanted Tgfb2^–/– eNPCs in EAE mice. Tgfb2^–/– eNPCs distributed within the perivascular space of the leptomeningeal vessels nearby inflammatory CD11b⁺ cells, similarly to WT eNPCs (Tgfb2^+/+ eNPCs) (Figure 11, A and B). However, Tgfb2^–/– eNPCs were not effective in inducing the clinical amelioration observed after WT eNPC transplantation (Figure 11C). Moreover, Tgfb2^–/– eNPCs did not hamper demyelination, axonal damage, and inflammatory infiltration due to EAE (Figure 11, D–F). Consistently, Tgfb2^–/– eNPCs failed to prevent accumulation of the inflammatory MCs in the CNS of EAE mice (Figure 11G) and upregulation of MHC-II on microglia (Figure 11H). Accordingly, WT NPC treatment, unlike Tgfb2^–/– eNPC or PBS treatment, significantly increased the frequency of SMAD2 phosphorylation in CNS-infiltrating CD11b⁺ myeloid cells, thus supporting that the therapeutic effect of NPCs acts through the activation of the TGF-β pathway in CNS-infiltrating MCs (Figure 11I).

Figure 11

NPC-secreted TGF-β2 is nonredundant for the therapeutic effect. (A and B) Representative immunostainings showing similar localization of GFP⁺ Tgfb2^+/+ (A) and Tgfb2^–/– (B) eNPCs (green) in meningeal perivascular areas in proximity to CD11b⁺ cells (red). Nuclei stained with DAPI (blue). Scale bars: 25 μm. (C) Left: Clinical scores of EAE mice treated with intrathecal injection of PBS (black dots), Tgfb2^+/+ eNPCs (red dots), or Tgfb2^–/– eNPCs (gray dots) at the peak of the disease (arrow) (n = 10 mice per group). Each point represents the mean ± SEM. *P ≤ 0.05, 2-way ANOVA with Bonferroni’s post-test. Right: Linear regression curves from day of treatment (18 dpi); dashed lines indicate 95% CI. ***P ≤ 0.001. (D–F) Quantification of spinal cord demyelination (Luxol fast blue staining) (D), axonal loss (Bielschowsky silver staining) (E), and inflammatory infiltration (H&E staining) (F) at 60 dpi in EAE mice treated with Tgfb2^–/– eNPCs (gray bars), Tgfb2^+/+ eNPCs (white bars), and PBS (black bars) (n = 20 sections per mouse, 4 mice per group). Data are mean ± SEM and represent the percentage area of damage per total section area. *P ≤ 0.05, unpaired t test. (G) Flow cytometry of CNS leukocytes from EAE mice treated as in C, assessed at day 7 after transplantation (n = 4 mice per group). Frequencies of inflammatory MCs (CD45^hiCD11b⁺Ly6G^–Ly6C^hiCD11c⁺MHC-II⁺) are shown in the plots and quantified in the graph. Black bars, PBS; white bars, Tgfb2^+/+ eNPCs; gray bars, Tgfb2^–/– eNPCs. (H) Levels of MHC-II on microglia in EAE mice treated with PBS (black line), Tgfb2^+/+ eNPCs (red line), and Tgfb2^–/– eNPCs (gray line) at the same time point as I are shown in the histogram and quantified in the graph (n = 4 per group). *P ≤ 0.05, 1-way ANOVA followed by Bonferroni’s post-test. (I) Confocal images and frequency of phospho-SMAD2⁺ cells over CD11b⁺ cells in the spinal cord of EAE mice treated with PBS, Tgfb2^+/+ eNPCs, and Tgfb2^–/– eNPCs, 7 days after treatment (n = 8–12 inflammatory infiltrates per mouse; n = 3 mice per group). ***P ≤ 0.001, 1-way ANOVA followed by Bonferroni’s post-test. Data are mean ± SEM and are representative of 2 independent experiments.

We tried to recapitulate the therapeutic effect of NPCs via intrathecal injection in the cisterna magna of a lentiviral vector expressing TGF-β2. The resulting increased expression of TGF-β2 at the peak of EAE severity within the CNS was sufficient to induce clinical amelioration in EAE mice, when compared with control mice receiving GFP-expressing control lentiviral construct (Supplemental Figure 10).

Altogether our data indicate that TGF-β2 production by transplanted NPCs is necessary for NPC immunomodulatory effect on pathogenic properties of myeloid cells in vitro and in vivo and is nonredundant for NPC therapeutic effect in EAE.

Furthermore, increasing the intrathecal levels of TGF-β2 at the peak of EAE disease severity resulted in accelerated clinical recovery.

In this study we provide evidence that intrathecally injected NPCs interfere with the pathogenic properties of CNS-infiltrating MCs via the secretion of the immunomodulatory cytokine TGF-β2.

We initially observed that intrathecally transplanted NPCs distribute within the leptomeningeal perivascular space of EAE mice, close to myeloid infiltrating cells and pathogenic Th cells, and maintain mainly an undifferentiated phenotype. This localization within the leptomeninges, an important site for the reactivation of autoaggressive Th cells by meningeal perivascular macrophages (1), allows NPCs to impact on CNS-restricted immunopathogenic events.

Indeed, intrathecal transplantation of NPCs reduces during overt EAE the accumulation within the CNS of inflammatory infiltrating cells and in particular of monocytes and their progeny, MCs, and also restrains microglia activation, leading to a long-term amelioration of disease-related disability. Inflammatory Ly6C^hi-monocyte-derived DCs are indeed a key pathogenic myeloid cell population in the CNS of EAE mice (63, 64). EAE severity correlates with the number of monocytes in the spinal cord, and mice in which monocytes are inhibited from entering the CNS are protected against EAE development (35, 65, 66). The reduced infiltration of MCs in the CNS of NPC-treated mice and the evidence of lower microglia activation have been similarly observed in GM-CSF–depleted EAE mice (2, 67, 68). In fact, mice lacking GM-CSF do not develop EAE, since the accumulation of leukocytes is abolished, despite the initial CNS invasion by CD4⁺ T cells (2, 69, 70).

Interestingly, we found that, early after treatment, NPC-injected EAE mice have reduced GM-CSF–producing pathogenic Th cells infiltrating the CNS and lower expression of genes related to Th17 pathogenic signature, including Il23r, essential for the maturation of effector Th17 cells (24, 71–74). The acquisition of Th17 pathogenic potential is a multistep process that starts in secondary lymphoid organs and can be completed only in the target tissue (75). MCs act as blood-borne APCs, sustaining encephalitogenic Th cell reactivation and full differentiation within the inflamed CNS through the secretion of proinflammatory cytokines (4, 5, 7, 63). Our data reveal that NPC treatment reduces in the CNS the burden of polarizing cytokines, including IL-1 and IL-23, required for the generation of pathogenic GM-CSF–producing Th cells.

Considering that intrathecally transplanted NPCs maintain an undifferentiated phenotype, we hypothesized a paracrine effect of NPCs, acting via the secretion of cytokines, growth factors, or morphogens to modulate infiltrating myeloid inflammatory cells. Accordingly, NPC-secreted factors recapitulated in vitro the NPC inhibitory effect on the ability of DCs to secrete polarizing cytokines, including IL-23, IL12p70, TNF-α, and IL-1β, and to drive the differentiation of GM-CSF–producing pathogenic Th cells.

Transcriptome analyses revealed that NPC-secreted factors induce a profound rearrangement of chromatin organization, metabolism, and biosynthetic processes in myeloid cells. These modifications lead to the reprogramming of inflammatory DCs toward an alternative activation state, defined by the downregulation of proinflammatory chemokines (Ccl1, Ccl2, Ccl3, Ccl7, Ccl8, Cxcl2) (35) and of positive regulators of cytokine production (Cd36, refs. 36, 37; Il33, ref. 38; Clec9a, ref. 39; and Clec5a) and by the upregulation of tolerogenic genes, such as hepcidin (76), Il1r2 (41), and Ascc1 (42).

Interestingly, transcription factor analyses suggested that the effect of NPCs on MCs might involve the activation of c-Myc, c/EBPβ, and CREB1, known to control M2 polarization and MDSC differentiation (45, 77), as well as the inhibition of PU.1, known to maintain the identity and inflammatory phenotype of DCs (46, 78). Gene set enrichment analysis confirmed that NPC treatment of EAE promoted in CNS-invading MCs a program of alternative activation via c-Myc activation and indicated TGF-β signaling as a possible upstream coordinator of the NPC-dependent antiinflammatory polarization of CNS-infiltrating MCs.

In line with our findings, a recent report has demonstrated that the upregulation of inhibitor of differentiation 1 (ID1), in response to tumor-derived factors, including TGF-β, is responsible for the switch from DC differentiation to MDSC expansion during tumor progression (78). In fact, the generation of the 2 tolerogenic myeloid cell populations, MDSCs and tumor-associated macrophages, in tumor-bearing hosts requires the integration of at least 2 types of signals: factors that expand myeloid precursors such as M-CSF, G-CSF, and GM-CSF (79), followed by factors that activate immune-regulatory programs. Besides classic Th2 cytokines (e.g., IL-4 and IL-13), chemokines (e.g., CCL5, CXCL12, and CX3CL1) as well as growth factors and noncanonical chemotactic peptides (e.g., VEGF-α, TGF-β, basic FGF, and antimicrobial peptides) (80) can contribute as “second signal” to monocyte recruitment and tolerogenic macrophage differentiation (45). In our study of intrathecal NPC treatment during the effector phase of EAE, we hypothesized that the inflammatory milieu provides abundant proliferative stimuli to the CNS-infiltrating myeloid compartment, while NPC-secreted TGF-β2 could impart the antiinflammatory signal. TGF-β signaling is necessary to prevent autoimmunity (81), and the signaling through the TGFBRs in DCs is a prerequisite to control EAE autoimmune response (82). Moreover, TGF-β limits at the site of inflammation the differentiation of highly mature DCs as a means of restricting Th17 cell differentiation and controlling autoimmunity (83).

In our experimental setting, TGF-β2 was found to be the most expressed among TGF-β family members in NPCs, and gain- and loss-of-function experiments demonstrated that TGF-β2 recapitulates the immunomodulatory effect of NPCs in vitro. In support of the role of TGF-β2, the transplantation of TGF-β2–deficient NPCs, during the EAE effector phase, failed to induce the expected clinical and neuropathological improvement, since it did not prevent the accumulation of MCs in the CNS. This finding indicates that TGF-β2 is required for efficient immunomodulation by transplanted NPCs, since a different source of NPCs (eNPCs), lacking only TGF-β2, did not display a therapeutic effect in our experimental setting. Moreover, increasing TGF-β2 levels in the CNS with intrathecally targeted gene therapy significantly ameliorated the chronic phase of EAE. TGF-β2 is constitutively expressed by astrocytes in the adult CNS, and little is known about its function during neuroinflammation. It has been suggested that TGF-β2 plays a central role in the maintenance of CNS immune privilege, and downregulation of astrocyte-derived TGF-β2 by T cell– and myeloid-secreted inflammatory cytokines appears to be a critical step for the induction and maintenance of neuroinflammation (84–86).

Altogether our data prove that increasing TGF-β signaling in the CNS is sufficient to interfere with differentiation and activation of MCs, which are recognized as the primary cell population sustaining neuroinflammation and the final effectors of tissue damage in CNS autoimmunity. In line with our findings, a recent work demonstrated that the selective deletion of TGFBRII in moDCs leads to the retention of moDCs in the CNS and to the development of more severe chronic EAE with increased irreversible tissue damage and no remission (87).

Compelling evidence demonstrates how ablating monocytes in the effector phases of EAE ameliorates disease progression and disability (35, 88). However, a phase II clinical trial with MLN1202, a human antibody blocking CCR2, failed to induce clinical amelioration in MS patients (3). A possible explanation is that this antibody could mainly act via peripheral ablation of monocytes, many of which, at this time point, might have already migrated into the CNS and undergone local differentiation and activation (3). Instead, a successful myeloid-directed therapeutic strategy in established CNS autoimmunity might focus on modulating in the CNS the differentiation and activation of MCs, rather than ablating their entry from the periphery (3).

In conclusion, we here propose, for the first time to our knowledge, a treatment able to target CNS-restricted pathogenic mechanisms of EAE effector phase. We identify TGF-β2 as the nonredundant secreted mediator of NPC immunomodulation in the inflamed CNS and as a potent immunomodulatory cytokine able to reprogram proinflammatory MCs to antiinflammatory/suppressive-like myeloid cells. The reduced amount of MC-derived polarizing cytokines and proinflammatory mediators in the CNS limits tissue damage and inhibits the terminal differentiation of pathogenic GM-CSF–producing Th cells. The reduction of GM-CSF levels acts synergistically with TGF-β2 in the CNS to impair the effector properties of MCs, disrupting the positive-feedback loop sustaining leukocyte recruitment and tissue damage in EAE. This allows accelerated recovery and significantly less accumulation of irreversible axonal loss in the chronic phase of the disease.

Mice. C57BL/6 female mice 6–8 weeks old were purchased from Charles River Italy. Mice with a transgene expressing T cell receptor specific for MOG_35–55 (2D2 mice; The Jackson Laboratory) (89) were provided by R. Furlan from the San Raffaele Scientific Institute. B6.Cg-Tg(Itgax-cre)1-1Reiz/J mice (The Jackson Laboratory) (61) expressing Cre recombinase under the CD11c promoter were obtained from M. Falcone from the San Raffaele Scientific Institute. B6;129-Tgfbr2tm1Karl/J mice (The Jackson Laboratory) (60) were provided by M. Greter (University of Zürich, Zürich, Switzerland). Heterozygous Tgfb2tm1Doe/J mice were kindly provided by T. Doetschman (University of Arizona, Tucson, Arizona, USA) (62).

Adult and embryonic mouse NPC culture. Mouse NPCs were obtained as previously described from dissected subventricular zone of adult (6- to 8-week-old) C57BL/6 female mice (NPCs) (14) and from dissected cortex of E15.5 C57BL/6 embryos (embryonic NPCs [eNPCs]) (90) from Tgfb2^–/– and Tgfb2^+/+ mice. Details are provided in Supplemental Methods.

In vitro differentiation and activation of BMDCs. BMDCs were prepared from bone marrow cell suspension from flushed tibias and femurs of naive C57BL/6 mice as previously described (91, 92). At day 6 of in vitro culture, BMDCs were used to investigate CD40L-induced DC maturation: BMDCs were plated at the concentration of 2 × 10⁶ per milliliter in the presence of recombinant CD40L (10 μg/ml; R&D Systems) in DCM or NCM. In most of the experiments, BMDCs and culture supernatants were collected at 48 hours for FACS analysis and cytokine measurements, respectively. For gene expression analysis, BMDCs were harvested at 6 and 18 hours.

In indicated experiments the following cytokines or pharmacological inhibitors were added at the indicated concentrations: TGF-β2 (R&D Systems), SB431542 (Tocris), and ITD1 (Tocris). In a set of experiments, BMDCs were derived from transgenic CD11c-Cre⁺Tgfbr2^fl/fl, CD11c-Cre⁺Tgfbr2^WT/WT, and CD11c-Cre^–Tgfbr2^fl/fl mice obtained by in-house breeding. Since CD11c^–Cre⁺Tgfbr2^fl/fl animals develop a severe, often fatal, autoimmune disease after 8 weeks of age, these mice were used before 8 weeks of age for ethical issues (60). To obtain a pure population of CD11c⁺ cells in which TGFBRII had been deleted, BMDCs derived from these latter 3 genotypes were FACS-sorted (Moflo XDP, Beckman Coulter) (achieving a 97% of purity) before being seeded for CD40L stimulation.

Myelin antigen presentation assay. Fifty thousand BMDCs, matured with CD40L in vitro (10 μg/ml; R&D Systems) for 48 hours in NCM or DCM, were cultured in IMDM medium with 10% FCS (Genenco), 2 mM l-glutamine (BioWhittaker), 100 U/ml penicillin, 100 mg/ml streptomycin (BioWhittaker), 50 mM 2-mercaptoethanol (Lonza) with 2 × 10⁵ CD4⁺ T cells isolated by CD4⁺ T cell MACS beads (Miltenyi Biotec), according to the manufacturer’s procedures, from splenocytes of MOG_35–55-immunized 2D2 mice at 7 dpi in the presence of MOG_35–55 peptide (20 μg/ml; Espikem) for 72 hours. DCs/CD4⁺ T cells were harvested and analyzed for cytokine production by flow cytometry.

Cytokine analysis. Multiplex quantification of cytokines and chemokines in BMDC supernatant and in CNS lysates from EAE mice was performed using a custom 16-plex Luminex plate (R&D Systems) according to the manufacturer’s recommendations. Unsupervised clustering was performed on normalized, median-centered data and reported as heatmap using GENE-E software (Broad Institute). For IL23p19, GM-CSF, TGF-β2, TNF-α, and IL-6, ELISA was performed (DuoSet ELISA kit, R&D Systems) according to the manufacturer’s instructions.

Gene expression microarray. Total RNA was isolated from at least 2 × 10⁶ BMDCs unstimulated (iDCs) or stimulated with CD40L in DCM (mDC) or NCM (mDC-NCM) at 6 and 18 hours of culture using the RNeasy Mini Kit (Qiagen). RNA quality was confirmed with a 2100 Bioanalyzer (Agilent). cRNA amplification was performed using the Illumina TotalPrep RNA Amplification Kit (Ambion), and 1.5 μg of biotin-labeled cRNA of each sample was hybridized on Illumina MouseWG-6 v2 Expression BeadChip (Illumina). Data were deposited at ArrayExpress with the accession number E-MTAB-5307. Details on bioinformatics analysis are provided in Supplemental Methods.

Next-generation sequencing. C57BL/6 mice were immunized and treated with NPC or PBS at the peak of disease as described above. Seven days after transplantation, CD45^hiCD11b⁺Ly6C^hiLy6G^–MHC-II⁺ MCs were FACS-sorted (BD FACSAria III) from hindbrain and spinal cord as described above. Total RNA from a minimum of 3 × 10⁵ up to 5 × 10⁵ cells was isolated with Qiagen RNeasy Plus Micro Kit according to the manufacturer’s instructions. RNA quality was confirmed with a 2100 Bioanalyzer (Agilent) resulting in RNA integrity number ≥8. Amplification of cDNA from total RNA was performed using the Ovation RNA-Seq system V2 (NuGEN). Then, cDNA was fragmented and ligated into a sequencing library using Ovation Ultralow Library System V2 (NuGEN). Libraries were PCR-amplified for 12 cycles and sequenced on an Illumina NextSeq 500 System with TruSeq reagent kits and software. Data were deposited at the NCBI’s Gene Expression Omnibus (GEO) database with the accession number GSE92395. Details on bioinformatics analysis are provided in Supplemental Methods.

EAE induction. Chronic EAE was induced in 6- to 8-week-old C57BL/6 female mice by s.c. immunization of MOG_35–55 peptide (200 μg per mouse) in incomplete Freund’s adjuvant containing 4 mg/ml of Mycobacterium tuberculosis. Pertussis toxin (500 ng) was injected i.v. at 0 and 2 dpi. The animals were assessed daily for body weight and clinical symptoms of EAE. Their disease stage was recorded by 3 researchers blinded to the treatment with the following scoring system (14): 0, healthy; 1, limp tail; 1.5, wobbling gait; 2, slight paresis of hind limbs/difficulty in righting; 2.5, severe paralysis of hind limbs; 3, complete paralysis of hind limbs; 3.5, initial paresis of forelimbs; 4, tetraparalysis; 5, death.

NPC intrathecal transplantation. At the treatment day, 4 days after clinical onset of EAE, animals were randomized into treatment groups and anesthetized with i.p. injection of 120–200 μl of 12.5% ketamine and 6.25% xylazine in 0.9% saline solution. Animals were fixed on a stereotactic device (David Kopf Instruments). A 10-μl Hamilton syringe was loaded either with 10⁶ NPCs, WT eNPCs, or Tgfb2^–/– eNPCs diluted in 10 μl of PBS or with control vehicle solution (PBS), according to experimental requirements. The Hamilton syringe was then stereotactically placed between the atlas and occipital bone at 35°, and advanced to puncture the cisterna magna. The following stereotactic coordinates were used: x = 0; y = –7.5 mm from λ; z = –4.5 mm from muscle plane. A total volume of 10 μl was injected in a 5-minute time period.

Isolation of CNS leukocytes. Extracted CNS tissue was cut into small pieces, incubated for 30 minutes with 0.2 mg/ml collagenase type IV (Sigma-Aldrich), and afterward passed through a 19G syringe to obtain a homogeneous cell suspension. CNS cell suspensions were further enriched by a Percoll gradient as previously described (93).

Flow cytometry, intracellular staining, and sorting. Flow cytometry was performed using a CantoII (Becton Dickinson) flow cytometer and analyzed with FlowJo software (Tree Star). FACS sorting was performed on BD FACSAria III (Becton Dickinson) or Moflo XDP (Beckman Coulter). Fluorochrome-conjugated mAbs specific for mouse MHC class II I-A/I-E (clone M5/114.15.2), CD11b (clone M1/70), CD11c (clone N418), CD45 (clone 30-F11), Ly6C (clone AL-21), Ly6G (clone 1A8), CD4 (clone GK1.5), CD3 (clone 17A2), CD80 (clone 16-10A1), and CD86 (clone GL1) were purchased from either BD Biosciences, eBioscience, or Biolegend.

For intracellular staining of T cell cytokines, cells were stimulated for 6 hours with phorbol 12-myristate 13-acetate (50 ng/ml) and ionomycin (500 ng/ml) in the presence of GolgiPlug (1:1,000; BD Pharmingen). Cells were first stained for surface molecules, fixed and permeabilized with a Cytofix/Cytoperm Plus kit (BD Biosciences), and stained with the following fluorochrome-conjugated mAbs: GM-CSF (clone MP1-22E9), IL-17A (clone TC11-18H10), and IFN-γ (clone XMG1.2) from BD Pharmingen. For analysis of intracellular FoxP3, cell preparations were stained for cell surface markers, then were fixed and made permeable with fixation-permeabilization buffers (eBioscience) and stained with fluorochrome-conjugated anti-FoxP3 mAb (clone FJK-16s) from eBioscience.

Dead cells were always excluded using a LIVE/DEAD stain kit (Invitrogen).

Histopathological analyses and immunofluorescence stainings. Details are provided in Supplemental Methods.

Statistics. Real-time PCR data, FACS data, EAE score, and histology results are expressed as mean ± SEM. Results were analyzed using the unpaired 2-tailed Student’s t test or 1-way ANOVA with Bonferroni’s post-test except for clinical score, which was assessed by 2-way ANOVA with Bonferroni’s post-test and linear regression. Analyses were performed using Prism v5.0a software (GraphPad). Statistical significance was accepted for P less than or equal to 0.05.

Study approval. All mouse experiments were conducted with approval from the Institutional Animal Committee of the San Raffaele Scientific Institute (IACUC number 568).

Further method descriptions are provided in the Supplemental Methods section.