C57BL/6 mice and the Signal Transducer and Activation Transcription (Stat)-1^-/- (B6.129S(Cg)-Stat1tm1Dlv/J, stock #012606) mice were obtained from The Jackson Laboratory. C57BL/6J MOG_35-55-specific TCR transgenic (2D2) mouse strain was kindly provided by Dr. R Pacheco (Fundación Ciencia & Vida, Chile). All mice were maintained under specific pathogen-free (SPF) conditions. All experimental procedures complied with the Helsinki Declaration of animal experiments and were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Chile and Northwestern University.

EAE was induced in 8- to 12-week-old mice with a s.c. injection of 150 µg myelin oligodendrocyte glycoprotein-derived 35-55 peptide (MOG_35-55, MEVGWYRSPFSRVVHLYRNGK, CPC Scientific, California, US), emulsified in Incomplete Freund’s adjuvant containing 500 µg Mycobacterium tuberculosis (BD Difco, Detroit, Michigan, US) followed by an i.p. injection with 500 ng Bordetella pertussis toxin (List Biological, Campbell, California, US) on the day of immunization and 48 h later. Body weight and clinical symptoms were monitored daily using a standard clinical score of 0-6 as previously described (36). For treatment with IFN-γ, 1 μg/mouse/day of recombinant murine IFN-γ (Biolegend, San Diego, California, US) was administered i.p. for 5 days starting at the peak of EAE. Non-immunized (NI) mice (without EAE) and EAE mice injected with phosphate-buffered saline (PBS) (Gibco, Grand Island, New York, US) were used as control groups ( Figure 1A ).

Mice were deeply anesthetized and intracardially perfused with PBS (0.1 M) followed by 4% paraformaldehyde (pH=7.4). Thoracic and lumbar spinal cord (SC) sections were removed, post-fixed in 4% paraformaldehyde, and embedded in paraffin. Serial sections with 6 μm thickness were cut, followed by hematoxylin and eosin (H&E) and luxol fast blue (LFB) staining. All reagents were purchased from Sigma (Saint Louis, Missouri, USA). LFB images were captured on an Olympus BX51 multichannel light/epifluorescence microscope (Olympus, Tokyo, Japan). H&E images were captured on a NanoZoomer XR slide scanner (Hamamatsu Photonics, Japan) employing the NanoZoomer Digital Pathology scan software v3.0 (Hamamatsu Photonics, Japan). Density of infiltrating inflammatory cells was determined as the number of cell nuclei per 10.000 µm². The extent of demyelination was evaluated by measuring the percentage of demyelinated area over the total white matter area for each SC section. Quantifications were performed using ImageJ software (NIH, USA).

For immunostaining, thoracic SC sections were deparaffinized and rehydrated. Heat-induced antigen retrieval was performed in citrate buffer (pH=6.0) for 30 min. Sections were washed in 1X Tris-buffered saline (TBS), permeabilized, and blocked for 1 h at room temperature (RT) in blocking buffer [5% bovine serum albumin (BSA), 0.5% Triton-X 100 in 1X TBS)]. Tissues were incubated with the primary antibody polyclonal rabbit anti-ionized-calcium binding adaptor protein (Iba) 1 diluted in blocking buffer (1:300, FujiFilm Wako, Osaka, Japan) overnight at 4°C. The next day, slides were washed with 1X TBS and incubated with Alexa Fluor 555-labeled secondary antibody anti-rabbit (1:200, Invitrogen, Waltham, MA, US) for 3 h at RT ( Table 1 ). Slides were washed with 1X TBS, and cell nuclei were stained with 4’,6-diamidino-2-phenylindole (DAPI). Finally, slides were rinsed with 1X TBS, mounted with an anti-fade mounting media, and visualized in a Leica DMI8 inverted fluorescence microscope. Image J-assisted analysis was used to evaluate density of Iba1⁺ cells, determined as the number of Iba1⁺ cells per 100.000 µm²; average Iba1⁺ cell size, determined as cell area of Iba1⁺ cells; and percentage Iba1 coverage, determined as the percentage of the total section area occupied by Iba1⁺ cells (µm²).

Mononuclear cells were isolated from SC of EAE mice treated with PBS or IFN-γ at day 19 post-immunization as previously described (37). Briefly, spinal cord homogenates were obtained and incubated with 0.5 mg/ml collagenase (Roche, Manheim, Germany) and 10 units/ml DNAse I (New England Biolabs, Ipswich, Massachusetts, US) at 37°C for 1 h. Mononuclear cells were purified using 40%/70% discontinuous Percoll gradients (Amersham, Piscataway, New Jersey, US), and total cell numbers were determined using a hemocytometer with viability assessed by trypan blue exclusion. In some assays, fresh cells were immediately analyzed by flow cytometry ( Figure 1B ). In other experiments, primary cell cultures were established; seeding cells in 24-well plates at a density of 2x10⁵ cells per well in 1 ml Iscove´s Modified Dulbecco´s Medium (IMDM) supplemented with 10% fetal calf serum (FCS), 100U/ml penicillin plus 100 µg/ml streptomycin, 1 mM sodium pyruvate, 50 µM beta-mercaptoethanol, 2 mM glutamine, and non-essential amino acids (all from Gibco, Carlsbad, California, US) ( Figure 1C ). Primary cell cultures usually contained 80-90% of adherent CD11b⁺ myeloid/microglia (MC/MG) ( Supplementary Figure 1 ). For nitric oxide (NO) determination, MC/MG were re-stimulated with 1 ng/ml IFN-γ for 24 h and then pulsed with 5 μg/ml MOG_35-55 for an additional 24 h. Then, re-stimulated MC/MG cultures were challenged with 1 µg/ml bacterial lipopolysaccharide (LPS) for 72 h in IMDM culture media. Thereafter, the medium was collected, and nitrite, a stoichiometric and stable metabolite of NO, was determined from supernatants by Griess reaction (Promega, Madison, Wisconsin, US). For cell co-culture experiments, MC/MG were re-stimulated with varying amounts of IFN-γ (0.1-1-10 ng/ml) for 24 h, washed, and then pulsed with MOG_35-55 (5 μg/ml) for an additional 24 h. After washing, cells were co-cultured with 1x10⁶ CD4⁺ T cells previously purified by negative selection from spleens of 2D2 mice using a CD4⁺ T cell isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany). After 4 days of co-culture, cell culture supernatant was collected for further cytokine analysis, and the cells were analyzed for Treg cell frequency (CD4⁺CD25^highFoxp3⁺) by flow cytometry ( Figure 1C ).

For cell surface staining, isolated cells were immediately fixed with fixation buffer (eBioscience, San Diego California, US) overnight at 4°C. For blocking non-specific Fc receptor-mediated antibody binding, cells were incubated with anti-FcγR III/II antibody for 15 min at 4°C in 2% fetal calf serum (FCS) PBS. Then, cells were stained with antibodies against CD11b, lymphocyte antigen 6 complex locus G (LY6G), CD45, CX3C chemokine Receptor 1 (CX3CR1), programmed death ligand 1 (PD-L1), CD86, CD80, CD40, MHC class II (MHC-II) molecules, and transmembrane protein (TMEM) 119 for 30 minutes at 4°C ( Table 1 ). For intracellular staining, cells were permeabilized with Permeabilization kit (eBioscience, San Diego California, US) for 30 minutes at RT and then incubated with antibodies against forkhead box P3 (FoxP3) (Biolegend, USA) for 30 minutes at RT. Next, cells were resuspended in PBS and analyzed using a Fortessa Flow Cytometer (BD Biosciences, USA) and FlowJo software (Tree Star, USA). Flow cytometry gating strategies are described in Supplementary Figure 2.

For FACS ( Figure 1D ), freshly isolated SC cells of 5-6 mice were incubated with anti-FcγR III/II antibody for 15 min at 4°C in 2% FCS PBS and immune stained with antibodies against CD11b, LY6G, CD45, CX3CR1, PD-L1, for 30 minutes at 4°C. Cells were immediately sorted on a FACSAria™ III (BD Biosciences, US), collected in tubes containing 2.4 ml of RLT lysis buffer (Qiagen, Hilden, Germany), and frozen at -80°C until RNA-Seq analysis. Usually, 6x10⁴ to 1x10⁵ cells were obtained, and cell viability was higher than 97% (not shown). Post sorting analysis confirmed that >97.5% of sorted cells were CD11b⁺LY6G^-CX3CR1^highPD-L1^low.

The concentration of IFN-γ, IL-1β, tumor necrosis factor (TNF)-α, granulocyte macrophage colony-stimulating factor (GM-CSF), IL-2, IL-4, and IL-10 in cell co-culture supernatants was determined by a multiplex assay using Luminex technology (Merck Millipore, Darmstadt, Germany) according to manufacturer´s instructions. Total and active transforming growth factor (TGF)-β production was determined by ELISA (Invitrogen, Vienna, Austria). The standard curve was diluted in medium containing 10% fetal calf serum, so TGF-β production was calculated over the basal level contained in the fetal calf serum ( Figure 1C ).

Results were analyzed using a Mann–Whitney nonparametric test or one-way ANOVA using GraphPad Prism v.5.0 (GraphPad Software). P values <0.05 were considered statistically significant.

First, we determined the effect of systemic administration of IFN-γ for 5 days starting at the peak of EAE. The results showed that IFN-γ significantly decreased the severity of clinical symptoms and body weight loss compared to PBS-treated mice ( Figure 2A ). After cessation of treatment, disease severity returned to levels similar to PBS-treated mice ( Supplementary Figure 3 ). Histological analyses showed that thoracic and lumbar SC sections from IFN-γ-treated EAE mice had significantly less infiltration of inflammatory cells and fewer demyelinated areas compared to PBS-treated-EAE mice ( Figures 2B, C ). Interestingly, flow cytometry analysis revealed that in vivo IFN-γ-treatment resulted in a significantly lower absolute number of mononuclear cells and lower frequency and absolute numbers of CD11b⁺ cells and non-neutrophil MC/MG (CD11b⁺Ly6G^-) compared to SC from PBS-treated EAE mice; however, levels were still higher than the non-immunized (NI) group ( Figure 3A–C ). There was no significant difference in the frequency and absolute number of neutrophils (CD11b⁺Ly6G⁺) between the PBS- and IFN-γ-treated mice ( Figure 3D ). Similar total number of cells was determined in draining lymph nodes (dLN) and spleen from IFN-γ-treated EAE mice and control mice. Frequency and absolute number of CD11b⁺ cells and neutrophils and macrophages in dLN was not affected by IFN-γ treatment. In contrast, a significantly lower frequency and absolute number of CD11b⁺ cells and neutrophils were found in spleen from IFN-γ-treated EAE mice compared to those from PBS-treated EAE mice; whilst macrophages were not significantly altered ( Supplementary Figure 4 ).

Several studies have reported that in a neuroinflammatory or tumor microenvironment MG are induced to upregulate CD45 expression (5, 41–49). Consequently, an increase in the frequency of CD45^hi cells would reflect the activation of CD45^low MG into CD45^hi cells resembling peripheral infiltrating MC (44–46). Our results showed that IFN-γ-treatment resulted in a significantly lower frequency and absolute number of activated MC/MG cells (CD11b⁺Ly6G^-CD45^high) compared to PBS-treatment. In addition, IFN-γ-treatment induced a significant increase in the percentage of resting MG compared to PBS-treated EAE mice; but was still lower than NI mice ( Figure 3E, F ). Consistent with the activation status of these cell populations, activated MC/MG (CD11b⁺Ly6G^-CD45^high) obtained from both IFN-γ- and PBS-treated EAE mice showed a significantly higher expression of MHC-II molecules, CD80, CD40, and PD-L1 than resting MG (CD11b⁺Ly6G^-CD45^low) ( Supplementary Figure 5 ).

Resting MG are also characterized by a ramified cell morphology with numerous thin processes that upon activation are drawn back into the soma, resulting in a rounded amoeboid-like appearance. In order to evaluate morphological changes associated with microglial activation, thoracic SC sections from NI mice and IFN-γ- and PBS-treated EAE mice were immunostained for Iba1, a known cell marker used to evaluate MC/MG morphology and activation (50–54). SC from PBS-treated mice were characterized by extensive Iba1 staining and amoeboid-shaped Iba1⁺ cells. Instead, IFN-γ-treated mice had predominantly ramified-shaped, Iba1-stained cells similar to NI mice ( Figure 4A ). In addition, SC from IFN-γ-treated EAE mice showed the density of Iba1⁺ cells, average Iba1⁺ cell area (indicative of cell size), and percentage Iba1 coverage (indicative of simultaneous alterations in cell density and morphology) significantly reduced in comparison to PBS-treated mice, but similar to NI mice ( Figure 4B–D ).

IFN-γ also plays a protective role in preventing hindbrain neuroinflammation, an effect dependent on the interaction between IFN-γ and CNS cells (55–57). Therefore, changes in MG and myelination in the cerebellum obtained from EAE mice treated with IFN-γ or PBS by immunofluorescence were examined. A significantly higher expression of TMEM119, a recently described homeostatic marker for MG (11, 58), and of MBP staining was observed in response to IFN-γ compared to control treatment ( Supplementary Figure 6 ; Method S1). Taken together, these results indicate that in vivo IFN-γ-treatment starting at the peak of EAE induces amelioration of clinical symptoms, reduction of body weight loss, attenuation of neuroinflammation associated with significantly less infiltration of inflammatory cells and demyelination, reduced activation of MC/MG, and enhanced expression of homeostatic microglial markers.

It has been reported that low concentrations of IFN-γ induce a tolerogenic phenotype in MG from neonatal mice capable of inducing regulatory T (Treg) cells (59). Thus, we were interested in determining the tolerogenic activity and phenotype of primary MC/MG cultures established from EAE-induced mice treated with IFN-γ or PBS. First, cell cultures were ex vivo pre-conditioned with low doses (0.1, 1, and 10 ng/ml) of IFN-γ and MOG_35-55, and then co-cultured with CD4⁺ T cells obtained from transgenic 2D2 mice for 96 h ( Figure 1C ). The results showed that MC/MG obtained from IFN-γ-treated EAE mice and ex vivo stimulated with 0.1 and 1 ng/ml IFN-γ induced a significantly higher frequency of Treg cells compared to untreated cells. Interestingly, ex vivo re-stimulation with 1 ng/ml IFN-γ and MOG_35-55 induced a significantly higher frequency and absolute number of Treg cells in co-cultures containing MC/MG obtained from IFN-γ-treated EAE mice than in co-cultures containing MC/MG from PBS-treated EAE mice ( Figure 5A ). No differences in the expression of MHC-II, CD86, CD80, and CD40 in primary MC/MG obtained from either IFN-γ-treated EAE mice or PBS-treated mice and stimulated with low doses of IFN-γ and MOG_35-55 were detected prior to culture with 2D2 CD4⁺ T cells ( Figure 5B ). Cell culture supernatants obtained at the end of the co-cultures of pre-conditioned MC/MG and 2D2 CD4⁺ T cells were analyzed by immunoassays. Cell co-cultures containing MC/MG obtained from IFN-γ-treated EAE mice had lower production of IFN-γ, TNF-α, and GM-CSF than conditioned cell co-cultures containing MC/MG obtained from PBS-treated EAE mice. However, only the secretion of IFN-γ in co-cultures containing MC/MG from IFN-γ-treated EAE mice and pre-conditioned with 10 ng/ml IFN-γ and MOG_35-55 was statistically lower than in those from control mice. Interestingly, cell co-cultures containing MC/MG isolated from IFN-γ-treated EAE mice and ex vivo pre-conditioned with 1 ng/ml IFN-γ and MOG_35-55 showed a significantly higher production of total TGF-β than control co-cultures ( Figure 5C ). Furthermore, active TGF-β production was higher in co-cultures containing MC/MG isolated from IFN-γ-treated EAE mice and ex vivo conditioned with 1 and 10 ng/ml IFN-γ and MOG_35-55 than control co-cultures ( Figure 5C ). In contrast, cell co-cultures containing MC/MG obtained from PBS-treated EAE mice and stimulated ex vivo with low doses of MOG_35-55 alone or in combination with IFN-γ produced significantly lower levels of active TGF-β than unstimulated cell co-cultures. There was no difference in the production of IL-4, IL-10, IL-2, and IL-1β between both groups of co-cultures. These results suggest that ex vivo re-stimulation with low doses of IFN-γ and MOG_35-55 endow MC/MG with the capacity to induce conversion of CD4⁺ T cells into Treg cells in association with high secretion of TGF-β.

Since nitric oxide (NO) is recognized as an important effector molecule produced by macrophages and microglia in response to inflammation (60), we evaluated NO production. Pre-conditioned MC/MG cultures obtained from IFN-γ-treated EAE mice produced significantly lower nitrite in response to LPS stimulation than those from PBS-treated EAE mice ( Figure 5D ). Taken together, these results show that ex vivo re-stimulation with low doses of IFN-γ and MOG_35-55 induces tolerogenic and anti-inflammatory activity in MC/MG.

Next, we examined the expression of MHC class II molecules, costimulatory molecules (CD80, CD86, CD40), coinhibitory molecules (PD-L1), and a microglial marker (CX3CR1) in MC/MG obtained from IFN-γ and PBS-treated EAE mice by flow cytometry ( Supplementary Figure 7A ). Interestingly, we found that in vivo IFN-γ treatment induced a significantly higher expression of CX3CR1 in MC/MG cells than in control cells ( Figure 6A ). In turn, an analysis of CX3CR1^highCD11b⁺Ly6G^- cells showed that IFN-γ treatment induced a significantly higher frequency of these cells expressing low PD-L1 ( Figure 6B ; Supplementary Figure 7B ). There was no difference in the absolute number of CX3CR1^highPD-L1^lowCD11b⁺Ly6G^- cells from mice treated with IFN-γ or PBS. However, the total number of live mononuclear cells isolated from the SC of IFN-γ-treated EAE mice was almost half of PBS-treated EAE mice ( Figure 3A ). These CX3CR1^highPD-L1^lowCD11b⁺Ly6G^- cells were 75-85% TMEM119⁺ ( Supplementary Figure 8A ). In addition, CX3CR1^highPD-L1^lowCD11b⁺Ly6G^- cells showed a strong expression of gene markers for MG and weak expression for MC, oligodendrocytes, and astrocytes ( Supplementary Figure 8B ). Therefore, these results strongly suggest that this cell subpopulation is an enriched subset of CX3CR1^high MG expressing low PD-L1 (CX3CR1^highPD-L1^low MG). The remaining cells were 47-52% TMEM119⁺ ( Supplementary Figure 8A ) and were considered MC/MG. A significantly lower frequency and absolute number of MC/MG was observed in SC from IFN-γ-treated EAE mice than in SC from PBS-treated EAE mice ( Figure 6B ). Interestingly, NI mice exhibited a significantly higher frequency and absolute number of CX3CR1^highPD-L1^low MG (60.2% ± 5.4%) than MC/MG (39.5% ± 5.6%) ( Supplementary Figure 9A ).

To determine whether the increase in CX3CR1^highPD-L1^low MG induced by IFN-γ is due to proliferation, mice treated with IFN-γ or PBS received a simultaneous i.p. injection of 5-bromo-2´-deoxyuridine (BrdU) starting at the peak of EAE. After 5 days, mononuclear cells were isolated from SC, and the frequency of proliferating CX3CR1^highPD-L1^low MG was analyzed by flow cytometry. The results showed that there was no difference in the frequency of proliferating CX3CR1^highPD-L1^low MG between EAE mice treated with IFN-γ or PBS, suggesting that enhanced CX3CR1^highPD-L1^low MG induced by IFN-γ might be explained by microglial plasticity ( Supplementary Figure 9 ; Method S2).

To obtain mechanistic insight into the induction of CX3CR1^highPD-L1^low MG by IFN-γ, mice lacking STAT-1, the major STAT activated in response to engagement of IFN-γ receptor, were induced with EAE and then treated with IFN-γ for 5 days starting at the peak of disease. IFN-γ treatment had no effect on disease progression in EAE-induced Stat-1^-/^- mice ( Figure 6C ). Furthermore, the lack of STAT-1 inhibited the IFN-γ-induced expression of CX3CR1 in MC/MG cells ( Figure 6A ), suppressed the increase of CX3CR1^highPD-L1^low MG, and reversed the decreased frequency of MC/MG induced by IFN-γ in WT mice ( Figure 6B ). We thus conclude that the IFN-γ/STAT-1 signaling axis is involved in symptom amelioration and induction of CX3CR1^highPD-L1^low MG in EAE.

Next, we analyzed the impact of IFN-γ-treatment on the activation state and the expression of MHC-II molecules and costimulatory molecules in CX3CR1^highPD-L1^low MG from EAE-induced WT mice treated with IFN-γ or PBS. A similar frequency of activated (51.5 ± 6%) and resting (44.2 ± 12%) CX3CR1^highPD-L1^low MG was found in SC from IFN-γ- and PBS-treated EAE mice ( Figure 6D ). However, activated CX3CR1^highPD-L1^low MG from IFN-γ-treated EAE mice exhibited a significantly lower frequency and absolute number of cells expressing MHC-II molecules than cells from PBS-treated EAE mice ( Figure 6E ). There was no significant difference in the frequency or absolute number of cells expressing CD80, CD86, or CD40 between CX3CR1^highPD-L1^low MG from IFN-γ- and PBS-treated EAE mice. Interestingly, in NI mice most of CX3CR1^highPD-L1^low MG (91.9% ± 4.2%) were in a resting state and had significantly lower frequency of cells expressing CD40 compared to activated cells, whereas the expression of MHC-II molecules, CD86, and CD80 were not statistically significant ( Supplementary Figure 9B ).

We next wanted to identify genes and signaling pathways in which IFN-γ may be differentially regulating the activity of CX3CR1^highPD-L1^low MG in EAE by analyzing the transcriptional profile of these cells. CX3CR1^highPD-L1^low MG were isolated from IFN-γ- and PBS-treated EAE mice for FACS sorter (97.5% purity), and the transcriptional profile was analyzed by RNAseq. A total of 12,524 genes were detected above the threshold (see Methods), with 336 genes upregulated and 188 down-regulated after IFN-γ treatment ( Figure 7A ). Using more stringent criteria, lowering the p-value to 0.01, setting the Log2 Fold change to a minimum of 1.0, and raising the threshold of minimum expression to 40 read units, resulted in 25 upregulated and 22 down-regulated genes in response to IFN-γ ( Figures 7B, C ). Interestingly, the gene expression of ATPase phospholipid transporting 8A2 (Atp8a2), a selective microglial gene marker, and several genes associated with anti-inflammatory processes such as dual-specificity phosphatase 5 (Dusp5), MLX interacting protein-like (Mlxipl, ChREBP), dishevelled segment polarity protein 1 (Dvl1) and gamma-aminobutyric acid type B receptor subunit 1 (Gabbr1), were upregulated in IFN-γ-treated EAE CX3CR1^highPD-L1^low MG ( Figures 7B, C ; Table 2 ). Several genes associated with M1 inflammatory activity such as the cd38 molecule (Cd38), complement factor B (Cfb), inhibin subunit Beta A (Inhba), serum amyloid A3 (Saa3), and APN-like peptidase cytosolic alanyl-aminopeptidase (Anpep) were down-regulated in IFN-γ-treated EAE CX3CR1^highPD-L1^low MG ( Figures 7B, C ; Table 2 ). Unexpectedly, arginase 1 (Arg1), a M2 classic gene, was down-regulated in CX3CR1^highPD-L1^low MG from IFN-γ-treated EAE mice. However, the cationic amino acid transporter 2 (Slc7a2), a gene involved in the uptake of arginine, was also down-regulated, suggesting a decreased substrate availability for NO production in these cells ( Figures 7B, C ; Table 2 ).

To determine if any molecular pathways were differentially regulated in CX3CR1^highPD-L1^low MG by IFN-γ, changes in the transcriptional profile were analyzed by gene set enrichment analysis (GSEA) using different reference databases. Using Gene Ontology (GO) biological processes and Reactome databases, we found that gene sets corresponding to RNA transcription and nucleus-cytoplasmatic transporter activity were highly and significantly upregulated in CX3CR1^highPD-L1^low MG from IFN-γ-treated EAE mice, compared to the control group (p< 0.001); indicating enhanced transcriptional and translational activities in response to IFN-γ stimulation ( Figure 8A ). The Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis showed that the top enriched gene set induced by IFN-γ were related to signaling transduction pathways (MAPK and RIG I like receptor), transcription factors (Basal transcription factors), and metabolism (Methionine and Glycerolipids metabolism), which might be related to a change in the activation pattern, oxidative status and lipid-related metabolism associated with a decreased inflammatory profile. Instead, downregulated enriched gene sets in CX3CR1^highPD-L1^low MG from IFN-γ-treated EAE mice were associated with oxidative metabolism (pyruvate metabolism), lipid metabolism, nitric oxide induction (peroxisome), and G-protein coupled receptor-associated to a neuroantigen response (neuroactive ligand-receptor interaction) ( Figure 8B ). Remarkably, over-representation analysis using the Reactome database showed that IFN-γ significantly induced an anti-inflammatory profile in EAE MG. Regarding immune cell function, genes in the TRAF3-dependent IRF activation pathway and Interleukin-10 signaling pathway were upregulated, while neutrophil degranulation, alternative complement activation, and activation of C3 and C5 pathways were down-regulated in CX3CR1^highPD-L1^low MG from IFN-γ-treated EAE mice ( Figure 8C ). Taken together, these results confirm that IFN-γ is a key inducer of anti-inflammatory pathways and suppressor of inflammatory mechanisms in EAE MG.

Because our results indicate the importance of IFN-γ/STAT-1 axis in the induction of CX3CR1^highPD-L1^low MG in EAE, STAT-1 target genes were analyzed in our RNAseq database using the Harmonizome database (p< 0.05) (79). The analysis revealed that 8 genes (Cfb, Dusp5, Anxa4, Neurl1b, C3, Naca, Anxa11, Slc15a2) were regulated by STAT-1. Using more stringent criteria (p< 0.01), Cfb and Dusp5 were functionally associated with STAT-1.

In order to search for molecular connections among genes regulated by IFN-γ-treatment, a molecular network of protein-protein interactions has been constructed based on the known interactions between the gene products of the 25 up- and 22 down-regulated genes by IFN-γ using the STRING database. Some regulated genes could not be considered because no interacting proteins were found in the STRING database. This resulted in 31 regulated gene products (primary nodes), 13 up-regulated, and 18 down-regulated by IFN-γ-treatment, with various numbers – up to 100 – of interacting proteins (secondary nodes) with each primary node. All secondary nodes were then searched for their occurrence in columns of at least two regulated gene products ( Figure 9A ). Inserting edges (connections) of all primary nodes with secondary nodes identified in their respective columns gave rise to the molecular network displayed in Figure 9 . As expected, Ifn-γ interacted with secondary nodes associated with inflammation, such as TNF-α, IL-6, IL-4, TGF-β1, or IL-10 ( Figure 9A ). Interestingly, we also found that Ifn-γ associated with secondary nodes interacting with three clusters of down-regulated inflammatory genes (cluster 1: Inhba, Cfb, and F7; cluster 2: Saa3, Cd38, and Anpep; cluster 3: acid phosphatase 5, tartrate resistant (Acp5), c-type lectin domain family 4 member d (Clec4d), and Arg1) and with secondary nodes interacting with up-regulated genes associated with anti-inflammatory roles (Dusp5, Mlxipl, and Gabbr1) ( Figures 9A, B ; Table 2 ).