Introduction
The nerve/glial antigen 2 (NG2; also known as chondroitin sulfate proteoglycan 4, CSPG4) is a chondroitin sulfate proteoglycan expressed by a number of cell types within and outside the central nervous system (CNS). Within the CNS, NG2, together with platelet-derived growth factor receptor alpha (PDGFRα), is a marker of oligodendrocyte precursor cells (OPCs), which downregulate both markers as they differentiate into mature oligodendrocytes [
16,
28,
39,
47]. The OPCs respond with increased proliferation to CNS injury and large numbers of NG2-expressing cells are found at sites of inflammation in the CNS [
10]. In particular, OPC levels markedly increase at the acute phase of experimental autoimmune encephalomyelitis (EAE), the purported model for multiple sclerosis [
16,
19], but are severely decreased in chronic EAE [
16]. These cells have been shown to respond to demyelination by generating new oligodendrocytes [
19,
27,
50,
55]. At the CNS microvessels, pericytes are involved in ensheathment of endothelial cells, blood–brain barrier (BBB)-endothelial cell differentiation, and maintenance of adult endothelial barrier function [
2,
23]. Although NG2 is not expressed in adult CNS pericytes [
53], it is an early marker of pericyte activation during CNS development and pathological conditions [
40,
53]. Thus, upon CNS injury, NG2-reactive pericytes are found along microvessels, where they act as sensors for inflammation and support the immunosurveillance and effector function of extravasated neutrophils and macrophages [
48]. Non-CNS resident infiltrating macrophages that express NG2 have also been observed in CNS inflammation [
5,
33], but, to date, the expression of NG2 on other immune cell subsets has not been reported.
These observations suggest a role for NG2 in CNS injury associated with inflammation, a possibility corroborated by the accumulation of NG2 at the glial scar, due not only to the recruitment of OPCs, but also to the shedding of the extracellular domain of NG2 from OPCs and to its incorporation into the extracellular matrix [
3,
9]. Moreover, there is controversial evidence for a role of NG2 in axon regeneration, remyelination, and CNS recovery [
10]. Thus, in vitro studies have shown that NG2 inhibits neurite growth, suggesting that it could inhibit axonal regeneration [
49,
52]. In contrast, other studies both in vitro [
57] and in vivo [
9,
24] suggest that NG2 is unlikely to be a major inhibitor of axonal regeneration after CNS injury.
On the basis of the above data, it appears highly likely that NG2 could play a role in EAE, a disease associated with increased BBB permeability, inflammatory infiltrates, and CNS damage. Moransard et al. studied EAE in NG2 knock-out (KO) mice to assess this possibility and did not observe any difference at clinical, immune response, and pathological levels, between NG2KO and wild-type mice [
35]. In this study, we investigated EAE in NG2KO mice of the same origin [
17] and found, in contrast, that NG2 plays an important role in EAE, with NG2 KO resulting in a significantly milder EAE disease, at clinical and neuropathological levels. Unexpectedly, we observed that NG2 significantly contributes to EAE pathogenesis and progression, not only through its expression on CNS-resident OPCs and pericytes, but also through its expression on immune cells, which had not been reported and studied previously, in particular at the encephalitogenic T cells/dendritic cells axis.
Materials and methods
Mice
Wild-type C57Bl/6J mice were purchased from Harlan (Bresso, Italy). The homozygous NG2
−/− mouse colony was generated as follows: NG2 knock-out (NG2KO) mice, developed by Stallcup and colleagues [
17] on the C57BL/6J background and kindly provided by Prof. Malatesta (University of Genoa, Italy), were cross-bred with WT C57Bl/6J mice and the F1 heterozygous NG2
+/− mice were themselves crossed to obtain homozygous NG2
−/− F2 progeny. The NG2
−/− (NG2KO) mice were then interbred to maintain the homozygous colony and used for all experiments. NG2KO mice were regularly genotyped by PCR to ensure the maintenance of the insertion leading to gene disruption (Supplementary Fig. S1). The F2 NG2
+/+ mice were themselves interbred to generate “wild-type” (WT) control mice. NG2KO mice were viable, healthy, fertile, and indistinguishable from WT mice. All mice were housed in pathogen-free conditions with food and water ad libitum. All applicable international, national, and/or institutional guidelines for the care and use of animals were followed (Decreto Legislativo 4 marzo 2014, n. 26, legislative transposition of Directive 2010/63/EU of the European Parliament and of the Council of 22 September 2010 on the protection of animals used for scientific purposes). The research protocol was approved by the Ethical Committee for Animal Experimentation of the University of Genoa, Italy.
EAE induction
Chronic EAE was induced in female WT and NG2KO mice (6–8 weeks of age, weighing 18.5 ± 1.5 g) by subcutaneous injection at two different sites in the right and left flanks and one site at the tail base with an emulsion (300 µl total) containing 200 µg myelin oligodendrocyte glycoprotein peptide spanning amino acids 35–55 (MOG35–55) (Espikem) in incomplete Freund’s adjuvant (Sigma-Aldrich) supplemented with 300 µg
Mycobacterium tuberculosis (strain H37RA; Difco). Mice were injected in the tail vein with 400 ng pertussis toxin (Sigma-Aldrich) in 100 µl of phosphate buffer saline solution (PBS, pH 7.6) immediately, and 48 h after the immunization. The mice were scored daily for clinical manifestations of EAE on a scale of 0–5 [
59].
Immunohistochemistry
Mice under deep anesthesia with ketamine/xylazine cocktail (90 mg and 4.5 mg/kg, respectively; intraperitoneal injection) were transcardially perfused with 100–150 ml of 2 % paraformaldehyde (PFA) and 0.2 % glutaraldehyde (Sigma-Aldrich) solution. Whole brains and spinal cords were removed and post-fixed by immersion in the same fixative at 4 °C for 4 h, then washed in PBS overnight at 4 °C, and the samples were stored in 0.2 % PFA in PBS at 4 °C. Spinal cord demyelination and inflammatory activity were evaluated as previously described [
14,
59]. Briefly, demyelination was measured as the percent of demyelinated areas in the area of entire section of the spinal cord as follows: for each mouse (
n = 5 mice per group), three sections per each of cervical, thoracic, and lumbar regions were stained with Luxol Fast Blue and demyelination was scored as follows: 1 = traces of subpial demyelination; 2 = marked subpial and perivascular demyelination; 3 = confluent perivascular or subpial demyelination; 4 = massive perivascular and subpial demyelination involving one half of the spinal cord with presence of cellular infiltrates in the CNS parenchyma; and 5 = extensive perivascular and subpial demyelination involving the whole cord section with presence of cellular infiltrates in the CNS parenchyma. In chimeric mice (see below), demyelination was assessed by staining for myelin basic protein (MBP) with anti-MBP antibody (diluted 1:90, Abcam). The inflammatory activity was analyzed by immunohistochemistry with anti–mouse CD3 (diluted 1:200, Santa Cruz) and anti–mouse IBA-1 (diluted 1:500, Abcam) antibodies, which identify T cells and microglia/macrophages, respectively. We used antibodies against glial fibrillary acidic protein (GFAP, diluted 1:500, Agilent Technologies) for the detection of astrocytes, and against inducible nitric oxide synthase (iNOS, diluted 1:100, Abcam) as inflammatory activity marker. All primary antibodies were detected with a biotin- or rhodamine-labeled secondary antibody (Bio-Rad) followed by diaminobenzidine-tetrahydrochloride detection according to standard protocol. Quantification was performed for the same areas as for demyelination in sequential sections of the same spinal cord regions, i.e., nine sections per animal, from approximately three segments apart. Images were captured at 40× magnification with an Axio Imager M1 microscope and analyzed with AxioCam software (Zeiss). The captured images were coded and quantified in a blinded manner. Data are expressed as mean ± standard error of the mean (SEM) of the total number of positive elements exclusively in the white matter and reported as number of positive elements per mm
2.
To assess NG2 expression on immune cells, splenocytes were seeded onto poly-lysine-coated glass coverslips and then placed in 24-well plates in RPMI1640 (Thermo Fisher Scientific), 1 mM glutamine (Thermo Fisher Scientific), 2-mercaptoethanol (5 × 10−5 M, Sigma-Aldrich), penicillin/streptomycin (50 U/ml, Thermo Fisher Scientific), supplemented with 10 % fetal bovine serum (FBS, Thermo Fisher Scientific). After overnight incubation, cells were fixed with 4 % PFA and processed for immunocytofluorescence with rabbit anti-NG2 (diluted 1:150, Merck), anti-CD11b (diluted 1:100, Merck), anti-CD11c (diluted 1:200, Merck), and anti-CD3 (diluted 1:50, Santa Cruz) antibodies. Primary antibody binding was detected with the appropriate secondary antibody: IgG anti-rabbit FITC-conjugated for anti-NG2 and IgG anti-rabbit PE-conjugated for anti-CD11b, anti-CD11c, and anti-CD3 (diluted 1:100, Thermo Fisher Scientific). The reactivities were detected under an Axio Imager.M1 microscope and analyzed with AxioCam software (Zeiss).
Laser confocal microscopy immunofluorescence
These methods are essentially similar to what we previously described [
11]. Briefly, the perfused brains and spinal cords were serially sectioned using a vibrating microtome (Leica) in about 120 serial sections (20 µm-thick) per sample and organized in a multiwell archive stored at 4 °C. Single and multiple immunostainings were carried out with the following primary antibodies: anti-CD45 (diluted 1:50; Novus Biologicals), anti-collagen type IV (Coll IV; diluted 1:50; Acris Antibodies GmbH), anti-PDGFRα (diluted 1:70; Millipore), anti-G-protein coupled receptor 17 (GPR17, diluted 1:350; Cayman Chemical), anti-Alzheimer precursor protein A4 (APP; diluted 1:100; Millipore), anti-MBP (diluted 1:90, Abcam), anti-CD13 (diluted 1:50; Aminopeptidase N; BD Pharmigen), anti-Claudin-5 (diluted 1:25; Invitrogen), anti-occludin (diluted 1:50; Invitrogen), anti-NG2 (diluted 1:200; Merck), and anti IBA-1 (diluted 1:50; Abcam) antibodies, then detected by appropriate fluorophore-conjugated (Alexa Fluor
® 488, 555 and 568; Invitrogen) or biotinylated (Vector Laboratories) secondary antibodies and fluorophore-conjugated streptavidin (Invitrogen) diluted in PBS containing 1 % bovine albumin and 2 % FBS. After immmunolabeling, the sections were counterstained with TO-PRO-3 (Alexa Fluor
® 633; diluted 1:10 in PBS; Invitrogen). Negative controls were prepared by omitting the primary antibodies, pre-adsorbing the primary antibodies with an excess of antigen when available, and mismatching the secondary antibodies. The sections were examined under a Leica TCS SP5 confocal laser-scanning microscope (Leica Microsystems) using a sequential scan procedure. Confocal images were taken at 250–500 nm intervals through the z-axis of the sections with 40× and 63× oil lenses.
Laser confocal microscopy morphometry
Quantitative assessment of PDGFRα+ OPCs, GPR17+ pre-myelinating oligodendrocytes, and MBP+APP+ axons was carried out through computer-aided morphometric analysis with Leica Confocal Multicolor Package (Leica Microsystems) and ImageJ software. Randomly chosen fields across the cerebral cortex and spinal cord white matter were analyzed for each mouse on 4 sections per brain and 3 sections per each of cervical, thoracic, and lumbar spinal cord regions. The number of positive cells and APP+ transversely cut axons was interactively counted (Cell counter ImageJ) on serial 0.5 μm optical sections across the entire stack of the projection images (20 μm thick). The results for the cells were normalized to the same volume (1 mm3), whereas the results for axons were normalized to a bi-dimensional value (reference area = 1 mm2).
Mice were injected with a solution of heparin (100 Units/kg) containing 71-kDa FITC-labeled dextran (5 mg/ml, Sigma-Aldrich) into the tail vein. Two minutes later, the mice under deep anesthesia with ketamine/xylazine cocktail (90 mg and 4.5 mg/kg, respectively; intraperitoneal injection) were killed, and brain and spinal cord were collected immediately. The tissues were processed by post-fixation as described above, and 20-µm-thick slices were counterstained with TO-PRO3 (Alexa Fluor® 633, diluted 1:10 in PBS; Invitrogen) and analyzed by confocal microscopy (Leica Microsystems) with 40× oil lens.
T-cell proliferation assay
For the measurement of WT and NG2KO ex vivo proliferative responses, mice were immunized with MOG35-55 using the same protocol as for EAE induction, except that they did not receive pertussis toxin, and lymph node cells were isolated from the draining inguinal lymph nodes 9 days after immunization. MOG35–55-primed lymph node cells were cultured (2 × 105 cells/well) in 96-well flat-bottom plates in the presence of 10 μg/ml MOG35–55, 1 μg/ml Concanavalin A (ConA, Sigma-Aldrich) as positive control, or 10 µg/ml proteolipid protein peptide spanning amino acids 139–151 (PLP), as negative control, for 3 days. Irradiated (25 Gy) splenocytes, isolated from the spleen of the same MOG-immunized mice and depleted of T cells through negative selection (Pan T cell Isolation Kit II, Miltenyi), were used as supplementary antigen-presenting cells (1 × 106 cells/well). Cell proliferation was measured as DNA incorporation of 5-ethynyl-2′-deoxyuridine (EdU) with Click-iT® EdU Microplate Assay (Thermo Fisher Scientific).
Elisa
Cytokines were evaluated in supernatants from MOG35-55-primed WT and NG2KO T-cell cultures by ELISA Standard Set kits (IL-4, IL-17 and IL-10 ELISA MAX, Biolegend; IFN-γ, Mabtech). Briefly, 96-well ELISA plates coated overnight with appropriate capture antibodies at 4 °C were washed with PBS supplemented with 0.1 % Tween 20 and blocked with PBS containing 5 % FBS for 2 h at room temperature. Supernatants of cell cultures (100 μl) were added and the plates were incubated for 2 h at room temperature. Plates were washed and incubated with the relevant horseradish peroxidase-coupled detection antibodies for 1 h at room temperature. The plates were washed, substrate (3,3′,5,5′tetramethylbenzidine, Sigma-Aldrich) was added, and the plates were developed for 20–30 min. The reaction was stopped with 2 N H2SO4 and plates read at OD 450 nm on a Multilabel Victor3 reader (Perkin Elmer).
Generation and activation of bone-marrow-derived dendritic cells
Primary bone-marrow-derived dendritic cells (BMDDCs) were obtained as described previously [
6]. Briefly, at day 0, bone-marrow cells from WT and NG2KO mice were flushed from the femur and tibia and passed through a 70-µm nylon cell strainer (Becton–Dickinson, New Jersey, USA). The cell suspension was seeded in the presence of granulocyte macrophage colony-stimulating factor (20 ng/ml, Miltenyi) and after seven days, the cells were analyzed by FACS for surface marker CD11c. Purity of CD11c
+ cells assessed by FACS with APC-conjugated anti-CD11c antibody (Becton–Dickinson) was at least 65 %. Activation of CD11c
+ BMDDCs was then carried out by stimulation for 24 h with 10 μg/ml lipopolysaccharide (LPS, Sigma-Aldrich).
Generation and polarization of bone-marrow-derived macrophages
To generate primary bone-marrow-derived macrophages (BMDMs), bone-marrow cells were flushed from the femur and tibia and passed through a 70-µm nylon cell strainer (Becton–Dickinson). The cell suspension was seeded in 6-well plates (6 × 106/well) in the presence of macrophage colony-stimulating factor (100 ng/ml, Miltenyi) in Minimum Essential Medium alpha (Thermo Fisher Scientific) with 1 % penicillin/streptomycin (50 U/ml) and 10 % FBS. After 7 days, polarization of WT and NG2KO BMDMs towards M1 and M2 phenotypes was performed for 72 h by stimulation with IFN-γ (20 ng/ml, R&D Systems) or with IL-4 (20 ng/ml, R&D Systems), respectively. Polarization to the appropriate phenotype was confirmed by FACS analysis for the expression of M1- and M2-specific surface maturation markers (see below).
FACS analysis
Surface marker staining and intracellular cytokine staining were performed using the Cytofix/Cytoperm, Fixation/Permeabilization Solution Kit (Becton–Dickinson), according to the manufacturer’s protocol. Briefly, cells (5 × 105) were resuspended in 100 µl of FACS buffer (PBS, pH 7.2, containing 0.5 % bovine serum albumin) and stained with appropriate conjugated antibodies for 30 min at 4 °C.
To assess the NG2 expression on immune cells, splenocytes were isolated, passed through a 70-µm nylon cell strainer (Becton–Dickinson) to prepare a single-cell suspension, and used to optimize staining conditions. To establish background staining and to check the positive signal using the appropriate amount of anti-NG2 Alexa Fluor®488 antibody (diluted 1:100, Merck), we performed a limited set of pilot experiments where quadrants were set based on staining with isotype-matched control antibody and NG2KO splenocytes were tested as negative control. WT splenocytes were double-stained with anti-NG2 antibody and antibodies against relevant surface markers, PE-conjugated anti-CD3 (diluted 1:100, Biolegend), APC-conjugated anti-CD11c, and PE-conjugated anti-CD11b (diluted 1:100, Becton–Dickinson) antibodies, for T cells, DCs, and macrophages, respectively.
BMDMs from WT and NG2KO mice were assessed by FACS for surface-specific markers of macrophage maturation using PercP-conjugated anti-CD86, PE-conjugated anti-CD206 (diluted 1:100, both from Biolegend) and PE-conjugated anti-MHCII (diluted 1:100, Becton–Dickinson) antibodies. WT and NG2KO LPS-primed BMDDCs and MOG35–55-stimulated lymph node cells were incubated with APC-conjugated anti-CD11c antibody (diluted 1:100, Becton–Dickinson) for 30 min at 4 °C, fixed/permeabilized for 15 min at 4 °C, and processed for intracellular cytokine staining using PE-conjugated anti-IL-12 (p70) antibody (diluted 1:100, BD Pharmingen) for 30 min at 4 °C. Data were acquired on a FACS Canto II (Becton–Dickinson) and analyzed using DIVA 6.1 software.
Real-time PCR
Total RNA was isolated using TRIzol
® Reagent (Thermo Fisher Scientific). First-strand cDNAs were generated starting from 1 μg RNA using Transcriptor First Strand cDNA Synthesis Kit (Roche). Real-time PCR was performed with 20 ng cDNA using a Roche Light Cycler 480 (Roche). Experiments were performed in triplicates, with glyceraldehyde 3-phosphate dehydrogenase (Gapdh) analyzed as housekeeping gene to normalize the expression data. The relative gene expression of target genes in reference to Gapdh expression was analyzed using the comparative CT threshold method [
41], and the normalized expression was expressed as fold induction over the control sample. The following forward and reverse primers (TIB MOLBIOL) were used: Il12, forward primer: 5′-GGAGTCCAGTCCACCTCTACA-3′, reverse primer: 5′-ATCGTTTTGCTGGTGTCTCC-3′; Tnf, forward primer: 5′-GGTCTGGGCCATAGAACTGA-3′, reverse primer: 5′-TCTTCTCATTCCTGCTTGTGG-3′; Nos2, forward primer: 5′-TTCATGATAACGTTTCTGGCTCT-3′, reverse primer: 5′-TGAACTTGAGCGAGGAGCA-3′; Chil3, forward primer: 5′-GAGACCATGGCACTGAACG-3′, reverse primer: 5′-AAGAACACTGAGCTAAAAACTCTCCT-3′; Arg1, forward primer: 5′-GAATCCTGGTACATCTGGGAAC-3′, reverse primer: 5′-GAATCTGCATGGGCAACC-3′; Ifng, forward primer: 5′-TTCAAGACTTCAAAGAGTCTGAGG-3′, reverse primer: 5′-GGAGGAACTGGCAAAAGGAT-3′; Il1b, forward primer: 5′-TTTGAAGCTGGATGCTCTCAT-3′, reverse primer: 5′-AGTTGACGGACCCCAAAAG-3′; Sox, forward primer: 5′-AGCCTCATCGGAGGGCTA-3′, reverse primer: 5′-AGGCAACTGCAGGCTGTAAAA-3′; Gapdh: forward primer: 5′-AATCTCCACTTTGCCACTGC-3′, reverse primer: 5′-ATGGTGAAGGTCGGTGTGA-3′.
Generation of chimeric mice
To create bone marrow chimera, female WT and NG2KO recipient mice (8 weeks of age) were irradiated with a lethal dose (9.5 Gy) and immediately transplanted with bone-marrow cells (5 × 10
6 per mouse in 100 µl of PBS injected in the tail vein) isolated from hind legs of syngeneic 8-week-old donor male mice. Reconstitution in the recipient was evaluated 8 weeks later by real-time PCR for sex-determining region Y (Sox) gene [
37] in circulating leucocytes. To avoid host infections, 50 mg/ml Gentamicin (Sigma-Aldrich) was added to the drinking water, 1 week before irradiation and for 2 weeks thereafter.
Statistical analysis
The results were analyzed with GraphPad Prism software (version 6.01; GraphPad Software). Intergroup differences were analyzed by one-way ANOVA followed by Bonferroni’s post hoc test. Comparison of two sets of data used the Mann–Whitney test. Differences between groups were considered to be significant when P ≤ 0.05 (*), P ≤ 0.01 (#) and P ≤ 0.0001 (**).
Discussion
In this study, we demonstrate that NG2, a marker of OPCs and pericytes, whose expression on various immune cell types we demonstrate here for the first time, plays an important role in the induction and progression of EAE. We further demonstrate that this role exerts itself at both the CNS/BBB and immune system levels. Thus, in EAE, the lack of NG2 resulted in reduced loss of OPCs at the chronic stage of EAE and in preserved TJ continuity, hence reduced BBB leakage, at the CNS level, and in a skewed Th2 response likely due to a shift of stimulating DCs to a less inflammatory profile. Our findings contrast markedly with those obtained by Moransard et al., who reported that NG2 expressed by CNS-resident cells or by infiltrating macrophages did not play a role in EAE [
35]. The reasons for this discrepancy are unclear as both studies use NG2KO mice derived from the original colony generated by Grako et al. [
17]. While the EAE induction protocols in the two studies differ, with the protocol used by Moransard et al. comprising less
Mycobacterium tuberculosis per injection but including a boost after 7 days [
35], the severity and course of the disease in the WT mice are equivalent in both studies. One can only speculate that the divergence in results might be attributable to differences in the mice themselves. The NG2 protein deficiency in these mice was obtained by homologous recombination to create a loss-of-function mutation through disruption of the NG2 coding sequence via insertion of a neomycin-resistance fragment in the targeting vector [
17]; it is, therefore, possible that mutation rescue might occur, with loss of the insertion [
20]. In this context, it is interesting to note that Moransard et al. did observe some NG2 reactivity of CNS resident cells in their NG2KO chimera reconstituted with WT bone marrow [
35]; unfortunately, they do not show the data for their NG2KO/NG2KO negative control chimera, nor do they show the confirmation of NG2 gene disruption in their NG2KO mice. We have verified the lack of NG2 in our NG2KO mouse colony through PCR analysis for the gene disruption (Supplementary Fig. S1) and through immunochemistry both at the CNS (OPCs and pericytes, Supplementary Fig. S3a and b) and immune cell (Supplementary Fig. S6) levels. Our data are supported by previous data obtained in the focal demyelination/remyelination model induced in mouse spinal cord by intrathecal injection of lysolecithin, where reduced inflammation was associated with less myelin damage and repair in NG2KO mice [
28]. Indeed, in that study, naïve WT and NG2KO mice did not display differences in the numbers of PDGFRα
+ OPCs, as we also observed; however, the acute demyelination stage (1 week after lysolecithin injection) was associated with lower numbers of OPCs and macrophages in NG2KO spinal cord, together with a significantly lower number of demyelinated axons, as compared to WT mice [
28]. We also observed a significant reduction in demyelinated areas in spinal cord of EAE-affected NG2KO mice at 40 dpi, which was associated with decreased inflammation. We have reported that at the acute phase of EAE, there is a significant increase in OPC numbers in WT mice [
16]; this does not occur in NG2KO mice, where the OPC numbers do not change throughout the disease. While this could relate to enhanced differentiation of NG2KO OPCs to pre-myelinating oligodendrocytes, it is more likely due to the reported decreased OPC proliferation in the absence of NG2 [
28,
29], as the levels of GPR17
+ PDGFRα
− cells paralleled those of OPCs in NG2KO mice throughout EAE, indicating that lack of NG2 is unlikely to alter OPC differentiation.
On par with the findings in lysolecithin-induced demyelination in NG2KO mice [
28], we also observed a decreased expression of pro-inflammatory cytokines in CNS of EAE-affected NG2KO mice. These findings are further supported by our demonstration of the importance of NG2 in immune cell function, leading to a skewing of the T-cell response to the encephalitogenic peptide from a Th1 to a Th2 type. Kucharova et al. have suggested that NG2KO macrophages/microglia exhibit less inflammatory properties than their wild-type counterparts [
28], a possibility supported by the in vitro demonstration that NG2KO microglia have a reduced inflammatory phenotype [
13]. In vitro, Moransard et al. found that bone-marrow-derived macrophages, but not microglia, express low levels of NG2 that increase significantly upon exposure to TGF-β [
35]. In vivo, several studies have detected NG2 expression on macrophage-like cells in the injured CNS [
5,
26,
33], and Gao et al. observed colocalization of NG2 and OX42 (CD11b/c) immunoreactivity in the corpus callosum of rats injected with LPS which elicits the innate immune response [
13]. Similarly, Jones et al., using a combination of IBA-1 immunolabeling and morphological analysis, identified macrophages, but not microglia, as cells producing NG2 after spinal cord injury [
26]. In EAE, Moransard et al. suggest that there is overlap in macrophage distribution and NG2 immunoreactivity, albeit diffuse, in wild-type mice at the acute phase of the disease [
35]. We have assessed the expression of NG2 on IBA-1
+ macrophages/microglia in mice at 20 and 40 days after the encephalitogenic challenge and our results are in agreement with studies that showed the existence of a subpopulation of macrophage-like NG2
+ cells in the CNS after insult. Indeed, while we never detected NG2
+ IBA-1
+ cells in EAE CNS at 40 dpi, we observed the presence of a scarce NG2
+ macrophage-like population in the vicinity of EAE WT microvessels at 20 dpi (data not shown). In our hands, day 20 after immunization corresponds to the early chronic phase; it is, therefore, possible that we would have seen a greater number of NG2
+ IBA-1
+ cells at the acute phase, that is around days 10–15, as Moransard et al. report, albeit without defining the time frame of the acute phase in their EAE model [
35].
We show that not only macrophages but also T cells and dendritic cells express NG2 and that its absence results in a modified function of these cells. To our knowledge, expression of NG2 on peripheral blood and bone-marrow cells has not been reported before in mice, and NG2
+ cells were only found in blood of patients with special types of leukemia [
38,
42,
43], using a monoclonal antibody, MoAb 7.1, raised against an SV-40-transformed human marrow stromal cell line, which recognizes a 220–240 kDa protein with a high degree of homology to rat NG2 [
43]. It is not clear why NG2 should be expressed on normal immune cells in mice and not in humans, and we can only speculate that in humans NG2 expression on hematopoietic cells is aberrant and induced by the disease, as suggested [
38], or that the different observation might be due to the difference in antibodies used. Indeed, in the present study we used a rabbit polyclonal antibody raised against purified rat NG2 and our FACS analysis of human peripheral blood cells from three control individuals with this polyclonal antibody shows that, in fact, human T cells, monocytes, and NK cells do express NG2 on their surface (Supplementary Fig. S8); it is, therefore, possible that the unique epitope recognized by MoAb 7.1 is masked in normal human blood cells and revealed only in modified human leukemic cells.
The potential role of NG2 expression on immune cells is unknown. Bu et al. suggest that its transient expression on macrophages invading the CNS upon excitotoxic hipoccampal lesioning is associated with an activated state of these cells [
5]. However, we analyzed NG2 protein expression on non-activated cells freshly isolated from spleen or in vitro-derived from bone marrow and found that the majority of macrophages and T cells, and 50 % of dendritic cells express this molecule on their surface. As EAE is mediated by a T-cell response, mutations that affect T cells are likely to affect the T cell response and, therefore, the EAE course. In this context, we have observed a skewing of encephalitogen-reactive NG2KO T cells from a Th1 to a Th2 phenotype upon ex vivo recall response to MOG. Whether this is due to an intrinsic functional modification of NG2KO T cells or not is unclear. We have observed a strong upregulation of IL-4 mRNA expression in naïve NG2KO T cells as compared to WT T cells upon mitogenic stimulation, together with upregulation, albeit of a smaller extent, of IFN-γ and IL-17 mRNA (data not shown), suggesting that skewing towards a Th2 type could be an intrinsic feature of NG2KO T cells.
Nevertheless, it is likely that the skewing from Th1 to Th2 in MOG35-55-responsive NG2KO T cells is also related to a role of NG2 in DC activation. DCs are the main antigen-presenting cells in EAE [
18], co-infiltrating the CNS with T cells [
7]; their production of IL-12 is essential in the Th1-cell response implicated in the disease. IL-12 is a potent inducer of IFN-γ secretion by T cells; it drives the differentiation of naïve T cells to Th1 cells while impeding that to Th2 cells, and contributes to amplification and expansion of Th1 cells. In EAE, the role of IL-12 in disease expression is clear; thus, treatment with recombinant IL-12 increased disease severity [
31,
44] and in vitro treatment of encephalitogen-reactive T cells with recombinant IL-12 enhanced their encephalitogenicity upon adoptive transfer EAE [
31,
54]. We observed a reduction in IL-12 expression by stimulated NG2KO DCs at both mRNA and protein levels and the expression of IL-12-expressing CD11c
+ cells was significantly decreased in MOG35-55-primed NG2KO lymph node cells; while this could be due to a decrease in migratory capacity of the DCs in NG2KO mice, possibly resulting from a stronger CD40 signaling [
32], it could also indicate that IL-12 expression is governed by NG2 expression by the cells. Our results, which show that, in naïve WT mice, the proportion of DCs that express IL-12 is significantly lower in NG2
− than in NG2
+ DCs, support this hypothesis. However, how NG2 expression influences the expression of IL-12 is unclear, and further studies that include silencing the NG2 gene in WT DCs are necessary to sustain this possibility.
Our data on the role of NG2 in immune cell function suggest that NG2 KO results in less inflammatory T cells and DCs; while this could explain the reduced inflammatory infiltration we see in NG2KO mice during EAE, the latter is also likely due to the lack of NG2
+ pericytes and the maintained BBB in NG2KO mice during EAE. Thus, Stark et al. showed that NG2
+, but not NG2
−, pericytes interact dynamically with extravasated innate immune cells to promote their migration along capillaries and arterioles [
48]. Our observation of a maintained BBB in naïve NG2KO mice and its reinforced integrity during EAE was unexpected. Several studies on tumor vascularization, where NG2 expression is upregulated in activated pericytes, have indicated that lack of NG2 results in reduced pericyte/endothelial cell interaction and basal lamina assembly with increased vessel leakiness [
15,
25,
58]. However, CNS interendothelial TJs and BBB functionality had not been studied in healthy NG2KO mice until now. In our study, naïve NG2KO mice were viable, healthy, fertile, and indistinguishable from WT mice, indicating that NG2 was not essential in BBB-microvessel development. Nevertheless, the modified appearance of the junctional staining pattern described in naïve NG2KO mice implies that NG2 is involved in the TJ settling plan and that its lack during brain development could lead to subtler alterations of claudin-5 and occludin distribution. As revealed by freeze-fracture electron microscopy, TJ-associated integral membrane proteins form linear chains of particles, denoted TJ strands, which are arranged in regular meshes and correspond to focal attachments of the affronted junctional membranes [
45,
46,
51]. When considered in the context of strand organization, the aspect of naïve NG2KO TJs, upon claudin-5 and occludin immunostaining, may represent a reduced number of strands/meshes in the belt between two adjoining endothelial cells (bicellular TJs) and apparently normal points of intersections between three joining endothelial cells (tricellular TJs). The reason why vessel development is negatively impacted in tumors in the absence of NG2 may depend on a substantially different organization as compared to CNS microvessels. Thus, whereas tumor capillaries are built up from two cell types, pericytes and endothelial cells, the neurovascular unit typical of CNS microvessels is assembled from both cellular and non-cellular components, with the cellular component also including perivascular astrocytes that are thought to induce the barrier phenotype of endothelial cells [
1,
8,
21]. Unlike that observed for NG2-expressing macrophages in tumors, whose role prevails on NG2
+ pericyte activity in determining vascular abnormalities (Yotsumoto et al. 2015), the results obtained through the analysis of EAE-affected NG2KO mouse chimeras confirm that TJ organization in CNS microvessels is primarily controlled by perivascular cells, astrocyte and pericytes. Thus, chimera of NG2KO mice reconstituted with WT bone marrow and induced for EAE did not show increased leakiness as compared to their NG2KO counterparts reconstituted with NG2KO bone marrow.
In conclusion, we show that the lack of NG2 in NG2KO mice impacts EAE through its effect not only at the CNS level, more particularly at the BBB, but also most importantly at the immune response level. Our data further suggest that NG2, expressed on immune cells, might be involved in the activation of DCs and thereby of the encephalitogenic T-cell response, through controlling IL-12 expression.