Metabolic determinants of the immune modulatory function of neural stem cells

NSCs lines were prepared from the subventricular zone (SVZ) of 7–12-week-old female SJL mice (18–20 gr. Charles River). NSCs were grown in a serum-free basal medium, NeuroCult basal medium plus mouse NeuroCult proliferation supplements (Complete Growth Medium, CGM; Stem Cell Technologies) supplemented with 2 μg/ml heparin (Sigma-Aldrich), 20 ng/ml purified human recombinant epidermal growth factor (EGF; Provitro) and 10 ng/ml human recombinant fibroblast growth factor (FGF)-2 (Provitro) [9]. At time of passaging (every 3–5 days), neurospheres were dissociated by enzymatic digestion with Accumax (Sigma) at 37 °C for 15 min, the number of viable cells determined by trypan blue exclusion and viable cells re-plated at 8000 cells/cm² (1.3 × 10⁶ cells with 15-ml media in T162 flask) [33]. NSCs at passage number ≤15 were used in all experiments.

NSCs were plated in CGM with or without either Th1-like (200 U/ml recombinant mouse TNF-α, Pepro Tech Inc.; 500 U/ml recombinant mouse IFN-γ, Pepro Tech Inc; 100 U/ml recombinant mouse IL-1β, Pepro Tech Inc) or Th2-like (10 ng/ml recombinant murine IL-4, R&D; 10 ng/ml recombinant mouse IL-5, R&D; 10 ng/ml recombinant mouse IL-13, R&D) cytokine cocktails for 16 h in vitro [33]. At the end of the conditioning, NSCs were washed three times with phosphate-buffered saline (PBS) to remove cytokine contamination before cell harvesting. Finally, NSCs and NSC-conditioned media were processed according to the analysis to be performed.

NSCs were plated at 348,000 cells/ml, collected after 16 h of incubation in CGM without or with Th1 or Th2 cytokines and washed with PBS. Conditioned media were collected and immediately frozen at −80 °C before the protein quantification. Protein quantification was performed using Direct Detect Spectrometer (Millipore). One hundred microgram of total protein for NSC-conditioned media were separated by SDS-PAGE using 4–12 % precast NuPAGE Bis-Tris gels (1.5-mm thickness) under reducing (Reducing Buffer, 10×) conditions and MES running buffer and then transferred onto nitrocellulose membranes (0.45-μm pore size, Hybond ECL) using XCell II Blot Module and NuPAGE transfer buffer (all from Invitrogen). Molecular weight marker: SeeBluePlus2PrestainedStandard (Invitrogen). For immunoblot analysis, the membranes were blocked for 1 h at room temperature with Tris-buffered saline/Tween (TBST; 10 mM Tris-HCl, 150 mM NaCl, 0.1 % Tween 20 pH 7.6) containing 5 % nonfat dry milk and then incubated overnight at 4 °C with rabbit anti-arginase I (H-52, Santa Cruz Biotech) or anti-arginase II antibody (H-64, Santa Cruz Biotech) by using antibody dilution 1:200. Mouse liver (sc-2256, Santa Cruz) and rat kidney (sc-2394, Santa Cruz Biotech) extracts were used as positive control for arginase I and arginase II, respectively. After washing with TBST, the nitrocellulose membranes were incubated with the appropriate donkey anti rabbit IgG HRP antibody (Amersham NA9340V; dilution used: 1:3000) for 1 h at room temperature. Immunoreactivity was revealed by using an ECL detection kit (Pierce).

1 ×10⁶ cells were plated in T25 flasks and cultured in CGM supplemented with U-¹³C₆ L-arginine (Cambridge Isotopes Laboratories) for 2 h and 16 h in the absence and in the presence of Th1 and Th2 cytokines cocktails. The cells were then harvested, washed with PBS (2 × 5 min), and lysed for 15 min at 4 °C under agitation with 1 ml of 50 % methanol and 30 % acetonitrile in water kept in dry ice/methanol bath (−80 °C). The insoluble material was pelleted in a cooled centrifuge at 14,000g for 10 min at 4 °C and the supernatant were collected for LC-MS analysis. XBridge Amide column (3.5 μm, 150 × 2.1 mm) was used for LC separation and the detection of metabolites was performed using a Thermo Scientific Q-Exactive high-resolution mass spectrometer with electrospray (ESI) ionization, examining metabolites in both positive and negative ion modes, over the mass range of 75–1000 m/z. The mobile phase for elution was a gradient established between water acidified with 0.1 % formic acid (A) and acetonitrile acidified with 0.1 % formic acid (B) at a flow rate of 200 μl/min. The gradient used was preceded by an isocratic step at 20 % A/80 % B. This step was followed by a linear decrease of B at 20 % in 25 min and re-equilibration at 20 % A/ 80 % B for 11 min. The injected volume was 5 μl.

1.74 × 10⁶ cells (cellular density 348,000 cells/ml) were seeded onto T25 flask (5 ml medium) in the absence or in the presence of either Th1 or Th2 cytokines cocktails. NSCs at different experimental conditions were homogenized in 100 μl Assay Buffer (Urea Assay Colorimetric Kit, Biovision), centrifuged at 15,000g for 10 min to remove insoluble materials. Fifteen-microliter samples were directly added to a 96-well plate for the urea determination according to the manufacturer’s instructions.

The enzymatic activity of arginase within NSCs (with or without Th1 or Th2 cytokines cocktails) was determined by arginase activity colorimetric assay kit (BioVision). Briefly, 1 × 10⁶ cells at the same cellular density used for the metabolomic studies (348,000 cells/ml) were dissolved with 100 μl ice-cold arginase buffer and centrifuged at 10,000g for 5 min. The supernatants were collected, and 40 μl were used to perform the assay for each samples.

Axillary, brachial, cervical, inguinal, and mesenteric lymph nodes were surgically excised from SJL mice and reduced to single cell suspension by mechanical disruption.

In selected experiments, CD4⁺ T cells were purified by negative selection using anti-CD8 (clone KT1.5) and anti-I-Ab-d/I-E (clone B21-22) rat Abs and sheep anti-rat-coated magnetic beads (Dynal Biotech, UK) to a purity of 95 %. Purified CD4 lymphocytes or unfractionated lymph node cells (LNC) were seeded in 96-well plates (1 × 10⁵ per well) in a final volume of 200 μl of CGM and were stimulated with anti-mouse-CD3/CD28 beads (Dynabeads mouse T-activator CD3/CD28, Invitrogen) at a ratio of 1:1 or left untreated. NSCs were pre-conditioned for 16 h with Th1 or Th2 cytokines cocktails at 348,000 cells/ml and then co-cultured with LNCs (NSC/LNC ratio 1:2) in the same well. NSCs were washed three times with PBS to avoid any cytokine carry-over in the co-culture system.

In selected experiments, LNCs were labelled with the vital dye, 5-(and-6)-carboxyfluorescein diacetate succinimidyl ester (CFSE; Molecular Probes, Invitrogen), final concentration of 1 μmol/L. Briefly, the cells were washed twice with PBS and suspended at 20 × 10⁶/ml in PBS. An equal volume of CFSE at 2 μmol/L was then added. After 8-min incubation under gentle shaking at room temperature, an equal volume of FCS was added to quench the reaction. LNCs were then washed twice in RPMI with 10 % FCS and plated with NSCs in co-cultures for 48 and 72 h, as above. LNCs were surface stained with rat anti-mouse CD4 Pacific Blue conjugated (clone RM4-5, BD Pharmingen) and CD44PerCP-Cy5.5 (BioLegend). The nuclear counterstain and dead cell indicator TO-PRO3 Iodide (Molecular Probes) was added before the flow cytometry acquisition. In a parallel co-culture experiment, the cells were labelled for the last 5 h of a 72-h-long culture with 10 μM 5-ethynyl-2′deoxyuridine (EdU). EdU-Click iT Flow Cytometry Assay (Alexa 647, Invitrogen, Molecular Probes) was used to detect EdU incorporation according to the manufacturer’s instructions (reaction volumes were adjusted to 50 μl). LNCs were surface stained with rat anti-mouse CD4 Pacific Blue conjugated (clone RM4-5) and CD44PerCP-Cy5.5.

N^ω-hydroxy-L-arginine (nor-NOHA; 30 μM) was added to cultures to inhibit arginase activity. Flow cytometry was performed with FACSCalibur (BD Biosciences), and a minimum of 15,000 events were acquired. Analysis of FACS data was performed using the FlowJo_V10 software.

Analysis of basal and cytokine-induced expression and localization (cytosol vs. mitochondria) of arginase I and II was done on mouse NSCs transduced with a 3rd generation Mito-DsRed lentiviral vector. NSCs were plated in CGM (laminin 1:100) at a density of 80,000 cells/cm². After 12 h, the cells were treated for additional 16 h in vitro with Th1 or Th2 cytokine cocktails as above. Untreated NSCs were used as controls. NSCs were fixed with paraformaldehyde (PFA) 4 % and then stained using a chicken anti-nestin (1:500; Abcam) and a goat anti-arginase I (1:100; Santa Cruz) or a rabbit anti-arginase II (1:100; Santa Cruz) primary antibodies. Appropriate Alexa Fluor conjugated secondary antibodies were then used. Nuclei were counterstained with 4, 6-diamidino-2-phenylindole (DAPI). For the assessment of arginase I and II expression, immunofluorescent staining were evaluated adopting a CCD camera/fluorescent microscope; n = 5 equally distributed regions of interest (ROIs) were chosen and acquired via ×20 objective lens for analysis (n = 4 cover slides were used for each group). Fluorescence intensity was calculated with ImageJ Software (NIH) and normalized on the number of DAPI⁺ cells/ROI. Data were expressed as arginase I or II mean normalized intensities over micron² ± SEM.

Ex vivo histopathological fluorescent images of ArgI/ArgII expression at level of SVZ were obtained from 4–8-week-old C57BL/6 female mice with myelin oligodendrocyte glycoprotein (MOG) peptide 35–55 induced experimental autoimmune encephalomyelitis (EAE) at 25 and 50 days post immunization (dpi), as described [33]. Strain-, age-, sex- and weight-matched healthy mice were used as controls. The mice were transcardially perfused with saline EDTA and 4 % paraformaldehyde. Brains were removed and post-fixed in PFA 4 % for 24 h at 4 °C and then washed with PBS. The brains were cryoprotected with 30 % sucrose (Sigma) and then embedded in optimal cutting temperature (OCT) Tissue Tek compound (EM sciences) for freezing. Twenty five-micrometer-thick coronal brain sections were cut and collected onto SuperfrostPlus slides (Menzel-Glaser, Thermo Fisher Scientific) and processed for histopathology. The following primary antibodies were used: goat anti-arginase I (1:100 Santa Cruz), rabbit anti-arginase II (1:100 Santa Cruz), chicken anti-nestin (1:500 Abcam) and a rat anti-CD45 (BD). Appropriate Alexa Fluor-conjugated secondary antibodies were used. Nuclei were counterstained with DAPI. Confocal microscopy images of SVZ healthy controls and MOG-induced EAE mice were acquired at ×40 and ×63.

Statistically significant differences between experimental groups were determined by Kruskal-Wallis followed by unpaired two-tailed t test (for comparisons between groups). */°/^p ≤ 0.05; **/°°p ≤ 0.01; ***p ≤ 0.001 and n.s. = not significant. Statistical analyses were performed using Prism software (v6.0f, GraphPad Software, San Diego, CA).

Priming with environmental stimuli that include cytokines, growth factors, morphogens and toll-like receptor (TLR) ligands modulates several stem cell functions [16, 34]. Here, we asked whether exposing NSCs to inflammatory cytokines, that are known to play a central role in triggering and perpetuating chronic inflammation, induces a licensing switch in their intrinsic immune regulatory functions [35].

To evaluate the effects of NSCs on acute T cell responses, we measured CD3/CD28-induced lymph node cell (LNC) proliferation by flow cytometry using a carboxyfluorescein diacetate succinimidyl ester (CFSE) dilution assay at both 48 and 72 h in vitro. These time points were selected based on previous data on proliferation and activation of rodent LNCs in vitro [36]. LNCs were also stained with anti-CD44 antibodies to detect their activation. NSCs were first primed for 16 h in vitro with a Th1-like cytokine cocktail composed of IFN-γ, TNF-α and IL-1β or a Th2-like cytokine cocktail composed of IL-4, IL-5 and IL-13 [33], and then tested for their immune modulatory activity in same well co-cultures with LNCs (Th1 NSC/LNC and Th2 NSC/LNC, respectively). Co-cultures of basal (e.g. not exposed to cytokines) NSCs and LNCs (NSC/LNC) or LNC preparations only (LNC) were used as controls.

By 48 h, significantly higher proportions of CD4⁺ T cells started to proliferate (CSFE^dim) in NSC/LNC [43.7 % (±4.2)], Th1 NSC/LNC [48.2 % (±1.5)] and Th2 NSC/LNC [39.2 % (±1.1)] vs. LNC [2.64 % (±0.18); p ≤ 0.0001] (Fig. 1a, b). By 72 h, the fraction of CFSE^dim CD4⁺ T cells further increased in all conditions, still being slightly more represented in NSC/LNC [83.9 % (±0.9)], Th1 NSC/LNC [86.8 % (±1.7)] and Th2 NSC/LNC [87.8 % (±0.1)], vs. LNC [77.3 % (±2.72); p ≤ 0.05 and p ≤ 0.01] (Fig. 1a, b). Nevertheless, we noticed a significantly higher proportion of CFSE^dim CD4⁺ T cells expressing low levels of the activation marker CD44 (CFSE^dimCD44^lowCD4⁺) within NSC/LNC, Th1 NSC/LNC and Th2 NSC/LNC at both 48 and 72 h [p ≤ 0.001 vs. LNC] (Fig. 1a, c). Comparable results were obtained when NSCs were co-cultured with purified CD4⁺ T cells for 72 h in vitro (Additional file 1: Figure S1), thus suggesting a direct effect of NSCs on CD4⁺ T cells.

Significantly higher proportions of CSFE^dimCD44^high CD4⁺ T cells were also observed by 48 h in NSC/LNC [36.8 % (±3.3)], Th1 NSC/LNC [39.3 % (±1.6)], and Th2 NSC/LNC [31.4 % (±0.8)] compared to LNC [1.6 % (±0.1); p ≤ 0.0001]. By contrast, no differences in the proportion of CD4⁺CSFE^dimCD44^high T cells were observed by 72 h in any of the conditions tested (Fig. 1a, d).

LNCs in co-cultures at 72 h also showed a higher CFSE content (CFSE^high) vs. control LNC (Fig. 1e). This CFSE^high late LNC behaviour did not account for any significant differences in the proportion of TO-PRO3⁺ apoptotic CD4⁺ T cells; yet, it was associated to significant 4–8-fold higher survival of LNCs in co-cultures [p ≤ 0.01 and p ≤ 0.001] vs. control LNCs (Additional file 2: Figure S2).

We therefore reasoned that these results might suggest that only in the presence of NSCs, T cells survive better and undergo transient early activation and proliferation, while prematurely exiting from the cell cycle [37]. To address this hypothesis, we analysed the ability of T cells in co-cultures to enter S-phase (synthesis phase) of the cell cycle at 72 h upon exposure to a 5-h-long pulse with 5-ethynyl-2′-deoxyuridine (EdU).

We found a significant 38 % reduction of the fraction of CD4⁺ T cells incorporating EdU in NSC/LNC co-cultures [22.50 % (±0.4)] vs. control LNC [35.80 % (±2.9); p ≤ 0.01] (Fig. 2). This effect became even more striking in Th1 NSC/LNC and Th2 NSC/LNC, which showed 56–61 % inhibition of the EdU uptake [Th1 NSC/LNC: 14.17 % (±3.3); Th2 NSC/LNC: 15.9 % (±1.8); all p ≤ 0.001 vs. controls], thus confirming that cytokine-induced NSC-reactive signals may hinder optimal cell cycle progression in target T cells [8].

To investigate whether the exposure to a Th1-like cytokine cocktail induced a metabolic alteration in NSCs, which could interfere with the NSCs-T cell crosstalk, we performed untargeted intracellular (endometabolome) and extracellular (exometabolome) metabolomic profiling of basal and Th1 NSCs. NSCs primed with Th2-like cytokine cocktails (Th2 NSCs) were used as additional controls. GC/MS and LC/MS in positive ion mode and negative ion mode were performed according to the different metabolite ionization properties. In total, 215 and 86 metabolites were identified for the NSCs endo and exo, when comparing basal, Th1 or Th2 NSCs. These analytes spanned many classes of metabolites, including amino acids, carbohydrates, peptides, lipids and nucleotides (Additional file 3: Table S1).

Welch’s two-sample t tests were used to identify exo- and endo-metabolites that differed significantly between the experimental groups (Tables 1 and 2). On the one hand, we found a total of 14 and 7 exo-metabolites that were significantly upregulated in the conditioned media of Th2 NSCs and Th1 NSCs, respectively, compared to basal NSCs. On the other hand, we found a total of 2 and 1 exo-metabolites that were significantly downregulated in the conditioned media of Th2 NSCs and Th1 NSCs, respectively, compared to basal NSCs (Table 1).

When we instead analysed the endometabolome of NSCs, we found that Th1 NSCs had more pronounced alterations with respect to Th2 NSCs and basal NSCs. On the one hand, a total of 32 endo-metabolites were significantly upregulated in Th1 NSCs (p ≤ 0.05 vs. basal NSCs) compared to the only 9 endo-metabolites that were significantly upregulated in Th2 NSCs. On the other hand, 19 endo-metabolites were significantly downregulated in Th1 NSCs (p ≤ 0.05 vs. basal NSCs) compared to the only 4 endo-metabolites downregulated in Th2 NSCs (Table 2).

Most importantly, partial least squares discriminant analysis (PLS-DA) clearly segregated basal, Th1 or Th2 NSCs in separate clusters (Fig. 3a). Each condition clustered separately, thus indicating that each NSC metabolic profiles were highly specific. Out of the 215 identified endo-metabolites, 61 were significantly different between basal, Th1, and Th2 NSCs, as depicted in the heatmap of the hierarchical clustering of the MS peak intensities (Fig. 3b; p ≤ 0.05).

We then focused on the endometabolome, to investigate the possible metabolic determinants of the enhanced immune regulatory function of Th1 NSC on T cells. To identify the intracellular metabolic pathways that were mostly altered in Th1 NSCs (vs. basal NSCs), we performed a metabolic pathway analysis (MetPA) on differentially represented metabolites by using Metaboanalyst, a web-based tool for metabolomic data interpretations [38]. This analysis identified multiple intracellular metabolic pathways as being altered in Th1 NSCs (vs. basal NSCs). Metabolites belonging to the metabolism of arginine and proline, of phenylalanine and of tyrosine and tryptophan were differentially enriched in Th1 NSCs (vs. basal NSCs) (Fig. 4). In particular, when considering the log p value, differences in the arginine and proline metabolism of Th1 NSCs were the most significant when compared to basal NSCs (Additional file 4: Table S2). Accordingly, the arginine downstream metabolites agmatine, N-acetylputrescine and 4-guanidinobutanoate—which are all members of the arginine decarboxylase pathway—were significantly upregulated in Th1 NSCs (p ≤ 0.01 and p ≤ 0.05 vs. basal and Th2 NSCs) (Fig. 5).

Untargeted metabolomics can therefore be efficiently used to identify several differences in the endo- and exometabolome of primed NSCs. In particular, we found that arginine and proline are the most differentially regulated in the endometabolome of Th1 NSCs.

Arginine levels have been reported to be critical for T cell activation and proliferation [39]. Therefore, we decided to focus on the intracellular arginine metabolism to clarify which metabolic routes were specifically altered in Th1 NSCs. The cells were incubated with uniformly labelled U-¹³C₆ L-arginine in the absence or in the presence of Th1 and Th2 cytokines and isotopologues distribution of arginine and downstream metabolites was assessed by LC/MS (Fig. 6). Within 2 h, the amount of labelled intracellular arginine (m + 6) was significantly lower in Th1 NSCs (p ≤ 0.01), when compared to both basal and Th2 NSCs (Fig. 6a). In addition, labelled ornithine (m + 5) was undetectable in Th1 NSCs by 2 h (Fig. 6b), and overall levels of unlabelled ornithine (m + 0) were lower in these cells (p ≤ 0.05). By 16 h, differences in the arginine and ornithine intracellular levels between basal NSCs and Th1 NSCs were no longer statistically significant.

The reduced amount of m + 6 arginine (Fig. 6a) and m + 5 ornithine (Fig. 6b), together with lower unlabelled m + 0 intracellular ornithine, supported a deregulation of arginine metabolism within the urea cycle in Th1 NSCs (Fig. 6e). Since arginase is the urea cycle enzyme involved in the conversion of arginine to ornithine and urea (Fig. 6e), we hypothesized that the observed metabolic changes were depending on the defects of arginase activity. To verify this hypothesis, we measured the arginase enzymatic activity in cell extracts and assessed urea production by a colorimetric assay. Th1 NSCs showed a significant decrease in the arginase activity (Fig. 6c) and in the rate of urea production, compared to basal and Th2 NSCs (p ≤ 0.05 and p ≤ 0.01) (Fig. 6d).

Altogether, these findings suggest that inflammatory cytokine signalling causes a selective defect of arginase activity in Th1 NSCs.

To further understand the regulation of arginase in cytokine-primed NSCs, we first investigated the expression of inducible (type I) and constitutive (type II) arginases in NSCs in vitro. We used immunofluorescence staining of Mito-DsRed-transduced Nestin⁺ NSCs to quantify the subcellular localization of cytosolic (type I) and mitochondrial (type II) arginase isoforms. We observed low basal expression of both arginases in NSCs and found that NSCs primed with Th1 cytokines had a significantly downregulated expression of arginase II, coupled with a significantly increased expression of arginase I (Fig. 7a).

As confirmation of these in vitro data, we interrogated a more physiological inflammatory context to examine the expression of both arginases by NSCs in vivo within brain germinal niches [40]. In mice with myelin oligodendrocyte glycoprotein (MOG)-induced experimental autoimmune encephalomyelitis (EAE), as model of Th1-like antigen-specific chronic inflammatory disease of the central nervous system [41], we found a downregulation of arginase II and concomitant upregulation of arginase I in Nestin⁺ NSCs of the SVZ at both 25 and 50 days after immunization vs. healthy controls (Fig. 7b).

These findings suggest that the previously observed decrease in the intracellular arginase activity in Th1 NSCs might well be ascribed to an imbalance between decreased arginase II expression (e.g. controlled at the post-translational level) and increased arginase I.

We then reasoned that (i) a shift in arginine metabolism and (ii) subsequent secretion of active arginases might be driving some of the immune modulatory properties of cytokine-primed NSCs.

To address the latter of these outstanding questions, we performed western blot analysis for both arginase I and II in NSC-conditioned media and found evidence of secreted arginase I, but not of arginase II (Fig. 7c). The secretion of arginase I might be an active secretion process by NSCs, instead of being a bystander effect of NSC death, as suggested by the comparable viability of basal, Th1 and Th2 NSCs (Additional file 5: Figure S3).

To address the role of extracellular arginases in the immune modulatory effects of Th1 NSCs, we co-cultured basal, Th1, and Th2 NSCs with LNCs in the absence or in the presence of the potent, reversible, pan-arginase inhibitor N^ω-hydroxy-nor-arginine (nor-NOHA) and measured EdU incorporation after 72 h in vitro.

Co-cultures with NSCs induced significant 40 % reduction of the fraction of CD44^highEdU⁺ CD4⁺ T cells vs. LNCs (p ≤ 0.01). This effect was more striking in co-cultures with Th1 and Th2 NSCs, respectively (58 and 52 % reduction; p ≤ 0.01 or p ≤ 0.05 vs. NSC/LNC and p ≤ 0.001 vs. LNC). Pan-arginase inhibition by nor-NOHA had no effects on NSC/LNC but induced a 13 % rescue of the anti-proliferative properties of Th1 NSCs (p ≤ 0.01 vs. NSC/LNC + nor-NOHA) and a further 1.8-fold lower rescue of the anti-proliferative effects in Th2 NSC/LNC (p ≤ 0.05 vs. Th1 NSC/LNC + nor-NOHA) (Fig. 7d, e).

The fact that cytokine-primed NSCs were still in part capable to inhibit T cells proliferation in the presence of pan-arginase inhibition suggests that mechanisms other than extracellular arginase I or II contribute to the broad NSCs immune regulatory activities, as described elsewhere [3].

Overall, our findings suggest that deregulation of the arginine pathway and the release of arginase I contribute to part of the NSCs immune regulatory activities, especially in the case of NSCs primed with extracellular inflammatory signals.