Despite considerable progress in understanding MDSC maturation both in vivo and in vitro, mechanisms of immunosuppression at the metabolic level are still unclear. Specific immune functions, including antigen processing and presentation, cytokinesis and activation, are known to be energetically-costly [4]. Moreover, immune cells, such as lymphocytes and macrophages, modulate their metabolism and respiration to fulfill their energy requirements. The energy metabolism of immune effector cells was thus considered a potential target for immunotherapy. Indeed, drugs that affect cell bioenergetics can reduce ATP production, for the treatment of psoriasis, rheumatoid arthritis, and cardiac arrhythmia. These drugs act either by inhibiting reactions related to substrate oxidation or by increasing proton permeability through the mitochondria, in turn resulting in the uncoupling of oxidative phosphorylation [4].

Therefore, to better understand the nutritional and energy requirements of MDSCs, two complementary experimental models were used. The first consisted of in vitro maturation of bone marrow (BM)-derived MDSCs using a combination of granulocyte-macrophage colony-stimulating factor (GM-CSF) and interleukin (IL)-6, as previously reported [5], which generated a CD11b⁺/Gr-1^low population, the most tolerogenic and immunosuppressive sub-population among CD11b⁺/Gr-1⁺ cells [6]. This cell population expresses iNOS and ARG1, and has a similar genetic signature to tumor-infiltrating MDSCs [5]. To discern whether the metabolic changes that occur during maturation are related to the direct effects of GM-CSF and IL-6 on metabolic pathways and nutrient transporters, or are associated with the activation of iNOS and ARG1, we used a second model that inhibited iNOS and ARG1 activity in MSC-1 cells, an immortalized cell line derived from mouse MDSCs. Being phenotypically similar to primary MDSCs, the MSC-1 cell line represents a relevant model system. MSC-1 cells constitutively expressed iNOS and ARG1 and inhibited antigen-specific proliferation and functions of cytotoxic T-lymphocytes without any additional treatment with specific cytokines or endotoxins, as previously demonstrated [7,8]. Comparing the results, we further confirmed the MSC-1 cell line as a model system.

iNOS and ARG1 activities were respectively inhibited by N-[3-(aminomethyl)-benzyl]-acetamidine (1400W) and [(S)-(2-boronoethyl)-L-cysteine] (BEC). 1400W was selected as it is a tight-binding and highly selective inhibitor of iNOS exhibiting the highest potency of reported iNOS inhibitors [9]. BEC is a potent slow-binding competitive and selective inhibitor of ARG1 [10]. Both 1400W and BEC inhibitors act directly on enzymatic activity, without affecting enzyme mRNA or protein levels [9,10].

Exposure of BM cells to GM-CSF and IL-6 induced a continuous up-regulation of iNOS and ARG1 activities after 24 and 16h, respectively (Figure ​(Figure11A​A1B).1B). The delayed effects can be attributed to cytokine internalization and to activation of the CCAAT-enhancer-binding protein (C/EBP)β transcription factor [5] and other signaling pathways (JAK/STAT3, MAPK and PI3-K) that were shown to regulate the expression and activation of L-Arg metabolizing enzymes [25]. However, BM-derived MDSCs only exhibited their immunosuppressive potential after 48 to 72h exposure to cytokines, which corresponded to the initiation of Jurkat cell growth inhibition and loss of viability, respectively (Figure ​(Figure1C,1C, D). The immunosuppressive activity of BM-derived MDSCs was not observed until sufficient L-Arg was sufficiently removed from the culture medium allowing NO derivatives to accumulate.

The activation of iNOS and ARG1 was accompanied by the up-regulation of glucose uptake (Figure ​(Figure2A)2A) and glycolysis, as shown by the AMP-to-ATP ratio, which was up to 5-fold higher than in the control culture (Figure ​(Figure4F),4F), and by the accumulation of G-6-P and F-6-P (Figure ​(Figure2B).2B). This accumulation may suggest that cells continue producing these intermediates without consuming them, a behavior previously associated with the initiation of cell death [17]. However, cells remained viable and grew throughout the culture. L-Arg and lactate were continuously produced, suggesting that cells consumed these intermediates to support both the catabolic processes and the sparse synthesis of anabolism-related macromolecules. Thus, the accumulation is caused by a higher rate of production compared with consumption. The cell specific concentration of F-6-P, which was approximately 60% to 70% that of G-6-P, suggests that glucose is mainly processed by glycolysis (Figure ​(Figure2B).2B). This may be a consequence of the low specific growth rate exhibited by BM-derived MDSCs, which may have limited fluxes of the non-oxidative reactions of the PPP. These results are in agreement with previous studies by Ando et al., where IL-6 enhanced the expression of the glycolytic enzymes hexokinase 2 and 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase-3 in mouse embryonic fibroblasts via the IL-6/STAT3 pathway [26]. Similarly, GM-CSF promoted rapid glucose transport in Xenopus oocytes via the PI3-K pathway [27]. L-Gln uptake was also increased in the presence of GM-CSF and IL-6 (Figure ​(Figure3A).3A). In addition to the possible direct effect of cytokines on nutrient uptake, the enzymatic activities of iNOS and ARG1 indirectly regulated glucose and L-Gln consumption since the inhibition of iNOS and ARG1 activities in MSC-1 cells down-regulated glucose and L-Gln uptake. Therefore, the abolition of MSC-1 cells immunosuppressive potential in the presence of 1400W and BEC may have decreased cellular demand of intermediates and energy, and so regulated nutrient uptake to adjust to the lower requirements. The decrease of nutrient uptake was not caused by the direct effects of 1400W and/or BEC on the expression or activity of glucose or glutamine transporters. The presence of these inhibitors in Jurkat cell cultures did not induce noticeable effects at the nutritional level (data not shown), although Jurkat cells and macrophages, as principal sub-populations of MSC-1 cells, express the same nutrient transporters [28,29]. This hypothesis requires further confirmation using specific siRNA to iNOS and ARG1 to avoid possible unknown effects associated with inhibitors that may have caused the down-regulation of nutrient uptake.

Stimulation of glycolysis and glutaminolysis provides the TCA cycle with specific intermediates, thus ensuring its enhanced activity (Figure ​(Figure3B,3B, C, D). Interestingly, malate accumulation was initiated around 8 to 16h before fumarate and α-KG accumulation. This may appear illogical since fumarate and α-KG are the upstream metabolites of malate. The study of metabolic fluxes and enzymatic kinetics is therefore crucial for understanding the differences in the trends of metabolite concentrations. The use of labeled nutrients may also offer interesting insight to the origin and fate of each metabolite. Particularly, the accumulation of fumarate suggests that L-Arg is continuously synthesized and supports a permanent immunosuppressive activity (Figures ​(Figures11​13C).3C). Furthermore, the enhanced activity of the TCA cycle resulted in the accumulation of high levels of malate compared to PEP (Figures ​(Figures22D​D3D),3D), both of which are precursors of pyruvate. This may suggest that the TCA cycle contributes to lactate accumulation since the lactate production-to-glucose consumption rates reached values higher than 2 (Figure ​(Figure2D).2D). BM-derived MDSCs undergo anaerobic glycolysis despite non-limiting oxygen conditions, a typical behavior for tumor cells that produce lactate rather than obtaining high energy yields from respiration. This phenomenon results in the acidification of the tumoral microenvironment, a condition that promotes tumor progression and metastasis [30]. Moreover, the accumulation of fumarate was previously associated with the inhibition of HIF hydroxylases and the stable expression of HIF [31]. On that note, our results agree with recent findings demonstrating that MDSCs express hypoxia-induction factor-1 α to adapt to the quasi-hypoxic conditions encountered in tumors [32]. Similarly, previous work by Wada et al. revealed that GM-CSF induces a rapid glucose-dependent extracellular acidification that is regulated by protein kinase C and the sodium/proton antiporter [33]. Moreover, iNOS activity is associated with an increased glucose consumption rate, increased glycolysis and PPP, and the inhibition of oxidative phosphorylation in zymosan-treated macrophages [34]. Likewise, a study by Irace et al., on LPS-treated macrophages reported a bi-directional regulation between NO and TCA cycle, that supports our findings. NO was shown to regulate aconitase activity and α-ketoglutarate production and alterations of the TCA cycle correlated with the inhibition of NO biosynthesis [35].

The two principal NADPH producing pathways in mammalian cells, the TCA cycle and PPP, were also stimulated in the presence of GM-CSF and IL-6. As the non-oxidative phase of PPP is devoted to anabolic processes and BM-derived MDSCs exhibit a low growth rate, the conversion of G-6-P into F-6-P was probably dominant to the production of 6-phosphogluconolactone, the first PPP intermediate. However, NADPH is mostly derived from the TCA cycle, particularly via malate dehydrogenase and isocitrate dehydrogenase, since fluxes through these pathways were considerably higher than those through the PPP (Figures ​(Figures2C,2C, ​C,3B–D),3B–D), although the oxidative phase of the latter was shown to be stimulated. As BM-derived MDSCs exhibited low specific growth rates in vitro, the oxidative phase, which is responsible for NADPH production, was probably more active than the non-oxidative phase, which is related to anabolic processes. This resulted in the recycling of PPP intermediates to the glycolysis pathway.

Although BM-derived MDSCs undergo glycolytic metabolism, with a low energy yield, the specific cell concentration of ATP increased gradually during maturation (Figure ​(Figure4B).4B). However, the decreasing trend of the ATP-to-ADP ratio suggests that ATP was continuously depleted from the intracellular pool (Figure ​(Figure4E).4E). This may be due to ATP consumption, forming ADP, or ADP production from AMP. Nevertheless, the production of ADP from AMP, via the enzymatic activity of adenylate kinase, requires ATP. Thus, the decrease of the ATP-to-ADP ratio was probably caused by ATP consumption.

Enhanced AMPK activity, as suggested by the continuous increase of the AMP-to-ATP ratio (Figure ​(Figure4D)4D) and confirmed by p-AMPK blotting (Figure ​(Figure5A),5A), is most likely responsible for the up-regulation of ATP-producing processes. Indeed, AMPK is considered an energy sensor in several metabolic disorders, such as cancer and diabetes, where enzymes switch cellular metabolism from anabolic to catabolic in reaction to deficits in cellular energy [36]. IL-6 up-regulated AMPK activity in rat skeletal muscle cells, and the IL-6-induced STAT3 enzyme was localized in mitochondria, resulting in enhanced oxidative phosphorylation and consequently increased cell ATP levels [37,38]. However, we observed that BM-derived MDSC respiration decreased in the presence of GM-CSF and IL-6, probably caused by HIF-1α expression [32]. The role of AMPK in HIF-1α expression is cell type dependent. Hypoxia induces HIF-1α expression in an AMPK-independent manner in mouse embryonic fibroblasts, whereas enhanced AMPK activity is important for HIF-1α transcriptional activity under hypoxic conditions in prostate cancer cell lines [39]. Therefore, the enhanced activity of AMPK in BM-derived MDSCs may have had multiple effects, by switching cells from oxidative phosphorylation to glycolysis, supported by Cidad et al., where inhibition of mitochondrial respiration by NO rapidly stimulated glucose uptake through AMPK [19]. Alternatively, to compensate for a deficient energy yield from glycolytic metabolism, BM-derived MDSCs further stimulated TCA cycle activity through glutaminolysis to produce ATP at a higher rate than required by maturation and other biochemical reactions. Nevertheless, the relative low ATP-to-ADP ratio in the iNOS and 1400W-inhibited MSC-1 cell culture indicated that the ATP consumption rate was higher than that of its net production, when compared to the control culture. However, this low ATP production rate did not stimulate AMPK in response to the energy deficit as revealed by the low AMP-to-ATP ratio when compared to the control culture. This suggests that the enzymatic activities of iNOS and ARG1 probably regulate AMPK activity, and that their inhibition may render AMPK unresponsive to energy deficit.

To confirm the implications of AMPK in the maturation process of BM cells to MDSCs, we incubated freshly isolated BM cells with Comp-C, a potent selective and ATP-competitive inhibitor of AMPK [40]. Since Comp-C is a reversible inhibitor, the expression levels of p-AMPK were monitored throughout the duration of culture to confirm the inhibition of AMPK activity (Figure ​(Figure5B).5B). The inhibition of AMPK in Comp-C pre-treated BM cells cultured in the presence of GM-CSF and IL-6 caused a net decrease in the uptake rate of major nutrients (glucose and L-Gln). This may have resulted in the down-regulation of TCA cycle activity and of the related metabolic pathways, such as L-Arg endogenous synthesis and energy production. Furthermore, AMPK inhibition down-regulated the GM-CSF and IL-6–induced activation of iNOS and ARG1 (Figure ​(Figure5C,5C, D). NADPH, a co-factor of iNOS, is principally derived from the glutaminolysis/TCA cycle axis and the decrease of cell specific levels of NADPH caused the down-regulation of iNOS activity. This finding is consistent with a previous report where the inhibition of AMPK activity by Comp-C, or dominant negative AMPK, down-regulated the activity of PGE2-induced eNOS (endothelial isoform of NOS), in human epithelial progenitor cells [20]. Interestingly, GM-CSF and IL-6 failed to activate ARG1 in the AMPK-inhibited BM cells (Figure ​(Figure5D),5D), although we recently showed that ARG1 activity was not associated with any specific energy requirements [41]. The activation of the mitogen-activated protein kinase (MAPK) signaling pathways including p38MAPK, ERK1/2 and SAPK/JNK, which are implicated in BM-derived MDSC maturation [5], was strictly linked to the induction of iNOS expression in macrophages [42]. Moreover, the AMPK inhibition-induced down-regulation of L-Arg endogenous synthesis and the low affinity of ARG1 for L-Arg (approximately 10mM) may have contributed to the inactivation of ARG1. The inhibition of AMPK activity in macrophages was also shown to inhibit cyclooxygenase-2 activity, which is crucial for MDSC accumulation in the tumor microenvironment [43,44]. Therefore, AMPK, via downstream-activated signaling pathways, is implicated in the maturation of BM cells to MDSCs. Although the phenotype of GM-CSF and IL-6-treated AMPK-inhibited BM cells was not investigated, iNOS and ARG1 activities were suppressed and led to a non-immunosuppressive cell population, as revealed by cytotoxicity assay, where the density and viability of Jurkat cells cultured in the presence of cytokine-treated Comp-C-pre-treated BM-cells were similar to those observed in the control culture (Jurkat cells only) (Figure ​(Figure5E,5E, F). As for all pharmacological inhibitors, Comp-C may have off-target effects that affect the reliability of the results observed here. Although no Comp-C-related side effects on cell metabolism were reported in the literature, Comp-C could potentially interact with other cell compartments, proteins or other molecules resulting in experimental bias. Studying the maturation of BM cells to MDSCs when the AMPK gene is silenced, using either AMPK knockout mice or specific siRNA to AMPK, is thus crucial before initiating in vivo studies.