Inhibition of Fatty Acid Oxidation Modulates Immunosuppressive Functions of Myeloid-Derived Suppressor Cells and Enhances Cancer Therapies

Stromal cells in the tumor microenvironment promote tumor growth and metastatic spread, limit the antitumor response to immunotherapy, and protect tumors from the effect of chemotherapy and radiotherapy (1–3). Prominent in the tumor microenvironment are tumor-infiltrating myeloid-derived suppressor cells (T-MDSC) that, in addition to blocking T-cell function and protecting tumors from the effect of chemotherapy and radiotherapy, support the expansion of regulatory T cells (Treg; refs. 4, 5), further enhancing this highly immunosuppressive microenvironment. T-MDSC use several mechanisms to block T-cell function, including the depletion of l-arginine by arginase I, the induction of T-cell apoptosis by nitric oxide (NO), and the synthesis of peroxynitrite (PNT; refs. 6–8). The plasticity of MDSC and the redundancy of these mechanisms have been shown by the fact that blocking one specific immunosuppressive mechanism induces the upregulation of the remaining pathways and only results in a partial recovery of T-cell function. Thus, therapies aimed at inhibiting MDSC have been limited to myelosuppressive chemotherapeutic agents (gemcitabine and 5-fluoruracil) and multi-tyrosine kinase inhibitors (sunitinib; refs. 9, 10). Therefore, better approaches to inhibit MDSC and enhance cancer therapies, in particular cancer immunotherapy, are needed.

The last decade has seen major progress in understanding the energy metabolic pathways used by different immune cell subpopulations (11–14). Effector T cells are highly glycolytic, whereas Tregs and memory T cells use fatty acid oxidation (FAO; refs. 15–17). Similarly, M1 macrophages and granulocytes preferentially use glycolysis (18), whereas M2 macrophages rely on FAO (11, 19, 20). In the present study, we aimed to characterize the energy metabolic pathway(s) used by T-MDSC, establish its impact on the immunosuppressive mechanisms, and test whether inhibition of this pathway would block MDSC and enhance antitumor therapies. The results showed that, upon infiltrating the tumor, MDSC increased the incorporation of fatty acids and activated FAO. This was accompanied by an increased mitochondrial biogenesis, upregulation of key FAO enzymes, and increased oxygen consumption rate (OCR). Inhibition of FAO blocked the tolerogenic function and immunosuppressive mechanisms of T-MDSC and resulted in a T-cell–dependent inhibition of tumor growth. More importantly, FAO inhibition enhanced the antitumor effect of low-dose chemotherapy and adoptive cellular therapy (ACT). Therefore, targeting FAO represents a novel approach to globally inhibiting the function of T-MDSC and enhancing the antitumor effect of various cancer therapies.

Samples were obtained from consented patients and donors under approved Institutional Review Board protocols. Peripheral blood mononuclear cells were separated on Ficoll-Paque Plus (GE Healthcare Life Sciences). MDSC (CD14^neg CD33⁺ HLA-DR^neg CD66b⁺) were sorted using a BD FACSAria (BD Biosciences). Polymorphonuclear granulocytes (PMN) were isolated by suspension over 3% dextran, as described (21).

C57BL/6 mice (8–10-week-old females) were purchased from Harlan laboratories, and OT-1 T cell antigen receptor (TCR) transgenic mice were from The Jackson Laboratory. Lewis lung carcinoma (3LL) and MCA-38 colon adenocarcinoma (American Type Culture Collection) were cultured in RPMI 1640 (Lonza-Biowhittaker) supplemented with 10% FBS (Hyclone), 25 mmol/L HEPES, 4 mmol/L l-glutamine, and 100 U/mL penicillin, streptomycin (all from Life Technologies). 3LL cells expressing ovalbumin (3LL-OVA) were generated, as previously described (22). 3LL and MCA-38 cells were periodically tested and validated to be Mycoplasma free.

In vivo tumor models were generated by injecting C57BL/6 mice s.c. with 1 × 10⁶ 3LL cells or 2.5 × 10⁵ MCA-38 cells, followed by daily i.p. injections of 50 mg/kg of the FAO inhibitors etomoxir or ranolazine (Sigma-Aldrich). For depletion of T-cell subsets, mice were injected i.p. with 500 μg/mouse anti-CD4 (GK1.5) or anti-CD8 (2.43) (BioXCell) 1 day before and 2 days after tumor injection, followed by injection of 250 μg/mouse every 5 days throughout the experiment. To test the synergy between FAO inhibition and low-dose chemotherapy, 3LL or MCA-38 tumor–bearing mice were treated daily for 6 days with etomoxir, starting on day 1 after tumor injection, followed by a single i.p. dose of cyclophosphamide (CTX; Sigma-Aldrich) at 200 mg/kg on day 7. To test the effect on established tumors, tumor cells were allowed to grow for 5 days, followed by etomoxir injections on days 6 through 10 (5 days) and a single injection of CTX (200 mg/kg) on day 11. To determine the effect of FAO inhibition on ACT, mice were injected s.c. with 1 × 10⁶ 3LL-OVA cells and treated with etomoxir daily throughout the experiment. OT-1 T cells (2.5 × 10⁶) were adoptively transferred on day 14, and mice were vaccinated with 100 μg/mouse SIINFEKL peptide (American Peptide Company) on day 15 after tumor injection. Ten days after the transfer, spleens were recovered and challenged with SIINFEKL for 24 hours, after which they were monitored for IFNγ production by Elispot (R & D systems). Activated OT-1 T cells were generated by stimulating splenocytes from OT-1 TCR transgenic mice in complete media with SIINFEKL peptide (1 μg/mL) and IL2 (100 U/mL) for 3 days. Tumor volume was measured using calipers and calculated using the formula: [(small diameter)² × (large diameter) × 0.5]. Experiments using animals were approved by the LSU-Institutional animal care and usage committee.

CD3⁺ T cells were isolated from spleens of C57BL/6 mice using a T-cell–negative isolation kit (Life Technologies). Purity exceeded 95%. For T-MDSC, tumors were digested with DNAse and Liberase (Roche USA) at 37°C for 1 hour, and T-MDSC were isolated from tumor single-cell suspensions, as described (23). Similarly, splenic MDSC were isolated from spleens of tumor-bearing mice, and normal myeloid cells (nMC) were isolated from spleens of control mice. The purity of different MDSC preparations ranged from 90% to 99%. Ly6C⁺ and Ly6G⁺ MDSC subsets were isolated by flow cytometric cell sorting. For the generation of bone marrow–derived MDSC (BM-MDSC), bone marrow cells were harvested from femurs and tibias of control mice and cultured with G-CSF (100 ng/mL), GM-CSF (20 ng/mL), and IL13 (80 ng/mL), as previously described (24). Cytokines were purchased from R&D Systems. When indicated, etomoxir (100 μmol/L) was added on day 2 of culture.

CD3⁺ T cells were labeled with 1 μmol/L carboxyfluorescein diacetate succinimidyl ester (CFSE; Molecular Probes; Life Technologies) and were cocultured with BM-MDSC or T-MDSC at a 4:1 T-cell:MDSC ratio, in the presence of plate-bound anti-CD3 (145-2C11) and anti-CD28 (37.51; 1 μg/mL each; BD Biosciences). T-cell proliferation was measured after 72 hours by CFSE dilution. When indicated, IFNγ production was assessed by ELISA (Biolegend).

Anti-human antibodies used to characterize cell subpopulations were: anti-CD33 (VM53), anti-HLA-DR (G46-6), anti-CD66b (80H3; Beckman Coulter), and anti-CD14 (61D3; eBioscience). Mouse antibodies specific for CD11b (M1/70), Gr1 (RB6-8C5), Ly6C (AL-21), Ly6G (1A8), CD8 (53-6.7), and CD45.1 (A20) were obtained from BD Biosciences. A Live/Dead stain kit was from Molecular Probes (Life Technologies). A Gallios flow cytometer (Beckman Coulter) was used for flow cytometry acquisition. Samples were analyzed with FlowJo software (TreeStar).

OCR and extracellular acidification rate (ECAR) were measured using XF-24 and XFe-24 Extracellular Flux Analyzers, respectively (Seahorse Bioscience) following the manufacturer's instructions. OCR was measured in XF media containing 11 mmol/L glucose and 1 mmol/L sodium pyruvate under basal conditions and in response to 1 μmol/L oligomycin, 1 μmol/L carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP), and 0.1 μmol/L rotenone plus 0.1 μmol/L antimycin A. ECAR was measured in XF media containing 2 mmol/L l-glutamine under basal conditions and in response to 10 mmol/L glucose, 1 μmol/L oligomycin, and 100 mmol/L 2-Deoxy-D-glucose (2DG).

Glucose uptake was measured using a flow cytometry–based assay, in which single-cell suspensions were incubated with 100 μmol/L fluorescent 2-(N-[7-nitrobenz-2-oxa-1,3-diazol-4-yl]amino)-2-deoxyglucose (2NBDG) for 2 hours, followed by staining with different cell surface markers (16). Fatty acid uptake in MDSC was determined by a fluorometric fatty acid uptake kit (Abcam). Cells were serum-deprived for 1 hour at 37°C and incubated with fatty acid mixture for 30 minutes; the fluorescence signal was measured in a microplate reader.

Staining with Mitotracker green was used to determine mitochondrial mass, Mitosox red to determine mitochondrial superoxide, and 2′,7′-Dichlorofluorescein Diacetate (DCFDA) to determine cellular reactive oxygen species (ROS; Molecular Probes; Life Technologies). Nitrite production was evaluated using standard Griess reagent (Molecular Probes; Life Technologies). PNT production was assessed by quantifying nitrotyrosine residues from cell lysates by ELISA (Millipore). The levels of ATP were quantified using a bioluminescence assay kit (Molecular Probes; Life Technologies).

Total RNA was isolated using the RNAeasy Mini Kit (Qiagen). cDNA was generated using the iScript cDNA Synthesis Kit (Bio-Rad). Real-time PCR was performed for carnitine palmitoyltransferase 1 (CPT1), Acyl CoA dehydrogenase (ACADM), peroxisome proliferator-activated receptor gamma coactivator 1-beta (PGC1β), and 3-hydroxyacyl-CoA dehydrogenase (HADHA) with Taqman primers from Applied Biosystems. Gene expression was calculated relative to 18s rRNA using 2^–ΔΔCT method.

Western blot analysis was performed using standard protocols. Proteins were electrophoresed in 8% TrisGlycine gels, transferred to polyvinylidene difluoride membranes, and immunoblotted with antibodies against arginase I (19; BD Biosciences) and β-actin (AC-74; Sigma).

Cytokines and chemokines were assessed in cell lysates using a Bio-Plex immunoassay (Bio-Rad Laboratories). CPT1 activity was measured in cell lysates based on the release of CoA-SH from palmitoyl CoA using thiol reagent 5,5-dithio-bis 2-nitrobenzoic acid (DTNB; Sigma; ref. 25).

Mouse cancer cells (3LL and MC38) were cultured for 7 days in ultra-low attachment polystyrene 6-well plates (Corning; #3471). The cells from nearly confluent regular cell culture were washed twice in PBS and resuspended in X-Vivo 20 serum-free medium (Lonza; #04-448Q). The cells were seeded in 3 mL of X-Vivo 20 medium (2.0 × 10³ MC38 cells or 1 × 10³ 3LL cells) with or without etomoxir (ET) (100 μmol/L). After 7 days, spheres were counted with phase contrast microscope, with exclusion of spheres below 50 μm (which can be cell aggregates).

Spheres were collected in 5 mL tubes, centrifuged, and resuspended in 4 mL of 0.25% Trypsin, 2.21 mmol/L EDTA in Hanks Balanced Salt Solution (Cellgro; #25-053-Cl). After 30-minute incubation at 37°C, cells were centrifuged and resuspended in Trizol reagent (Life Technologies; #15596-026). Samples were processed according to the manufacturer's instructions. Reverse transcription was performed with Cloned AMV First-Strand cDNA Synthesis Kit (Life Technologies; #12328-040). qPCR was performed with Fast SYBR Green Master Mix (Life Technologies; #4385612), using the following primers: Mouse pouf51 (OCT3/4), forward primer 5′-GGAGGGATGGCATACTGTGG-3′ and reverse primer 5′-ACCTTTCCAAAGAGAACGCC-3′; Mouse nanog, forward primer 5′-TCGAATTCTGGGAACGCCTC-3′ and reverse primer 5′-CAGGTCTTAACCTGCTTATAGCTCA-3′; Mouse Sox2, forward primer 5′-GGAGGAGAGCGCCTGTTTTT-3′ and reverse primer 5′-CTGGCGGAGAATAGTTGGGG-3′; mouse GAPDH, forward primer 5′-ATGACATCAAGAAGGTGGTG-3′ and reverse primer 5′-CATACCAGGAAATGAGCTTG-3′.

A total of 23 archival biopsy samples from patients with colon adenocarcinoma, clear cell kidney carcinoma, and breast ductal carcinoma were obtained from the Louisiana Cancer Research Center (LCRC) tissue biorepository. All samples were de-identified. Immunohistochemistry was performed using avidin–biotin–peroxidase (Vector Laboratories). Briefly, our protocol includes deparaffinization in xylene, re-hydration through descending grades of alcohol up to water, nonenzymatic antigen retrieval in Citrate buffer, pH 6.0 for 30 minutes at 95°C, and endogenous peroxidase quenching with H₂O₂ in methanol for 20 minutes. After PBS wash, samples were blocked with 5% normal goat serum in 0.1% PBS/BSA. Primary antibodies included mouse monoclonal antibodies against HADHA (1:500 dilution; Abcam), CD66b (1:100 dilution; LS-B7134; Lifespan Biosciences), CPT1 (1:200 dilution; 8F6AE9; Abcam), and HLA-DR (1:100 dilution; L243; Santa Cruz; Biotechnology). After PBS washing, sections were incubated with a biotinylated anti-rabbit secondary IgG for 20 minutes, incubated with avidin–biotin–peroxidase (ABC) complexes, and developed with diaminobenzidine (Sigma). All sections were counterstained with hematoxylin, dehydrated in alcohol, cleared in xylene, and mounted with Permount (Fisher).

For immunofluorescence, double labeling of paraffin-embedded sections, deparaffinization, antigen retrieval, and blocking were performed as described above (endogenous peroxidase was omitted). Sections were then incubated with a first primary antibody overnight at room temperature and rinsed in PBS. A fluorescein-conjugated secondary antibody was incubated for 1 hour in the dark. After washing with PBS, a second primary antibody was incubated overnight in the dark, followed by rinsing with PBS and incubation with a second rhodamine-tagged secondary antibody for 1 hour. Sections were washed in PBS, mounted in aqueous mounting media with DAPI (Vector Laboratories), and visualized in a confocal microscope (Olympus FV1000).

Data were analyzed by either Student t test or one-way ANOVA followed by Tukey posttest using Graph-Pad Prism analysis software. Results were expressed as mean ± SEM, and P values are presented in the figures as *, P < 0.05; **, P < 0.01 or ***, P < 0.001).

The harsh tumor microenvironment created in part by the increased metabolic rate of tumor cells can result in a significant metabolic stress on other cells including T-MDSC. We aimed to determine the metabolic characteristics of T-MDSC and establish whether they could regulate the immunosuppressive functions of these cells. To identify the metabolic characteristics of T-MDSC, a single-cell suspension of s.c. 3LL tumor was simultaneously labeled with the fluorescent glucose analogue 2NBDG, and cell-surface markers for MDSC, defined as CD11b⁺ Gr1⁺ CD11c⁻ F480⁻. Control cells included nMCs from spleens of naïve mice and MDSC from spleens of tumor-bearing mice (splenic MDSC; Fig. 1A; Supplementary Fig. S1A). The nonmyeloid fraction containing mostly tumor cells had a significantly increased 2NBDG uptake (Fig. 1A, last panel; Supplementary Fig. S1A), whereas that in nMCs, splenic MDSC, and T-MDSC was all lower. Instead T-MDSC (but not nMCs or splenic MDSC) had a high fatty acid uptake, which was similar in the Ly6C⁺ monocytic MDSC (M-MDSC) and the Ly6G⁺ granulocytic MDSC (G-MDSC) subpopulations (Fig. 1B). T-MDSC had an increased metabolic activity overall as shown by the increased OCR and ECAR, reflecting an increased FAO and glycolysis (Fig. 1C). However, the ratio of OCR/ECAR clearly demonstrated a preferential increase in OCR, which confirmed the metabolic reprograming toward fatty acid oxidation (Fig 1D). In addition, the increased OCR in T-MDSC was accompanied by an increase in mitochondrial biogenesis as shown by higher staining with Mitotracker and enhanced production of superoxide (Mitosox) and ROS (DCFDA; Fig. 1E; Supplementary Fig. S1B–S1D). Moreover, T-MDSC displayed a significantly elevated expression of genes associated with FAO, including CPT1, ACADM, PGC1β, and HADHA (Fig. 1F).

The metabolic reprograming was paralleled by an upregulation of arginase I, elevated production of NO and PNT (Fig. 1G–I; Supplementary Fig. S1E), and increased ability to inhibit T-cell proliferation (Fig. 1J). In addition, T-MDSC produced higher levels of G-CSF, GM-CSF, IL1β, IL6, and IL10, cytokines known to promote and sustain MDSC development; however, they also produced higher levels of IL12 (Supplementary Fig. S1F).

We then explored the effect of FAO inhibition on the induction and function of MDSC using BM-MDSC from normal C57BL/6 mice. Bone marrow precursors were activated with G-CSF, GM-CSF, and IL13 and incubated with or without etomoxir, a specific inhibitor of CPT1, which is the first and rate-limiting enzyme in the FAO cycle. Incubation of BM-MDSC with etomoxir lowered CPT1 enzymatic activity (Supplementary Fig. S2A), but did not alter the proportion of G-MDSC or M-MDSC, and did not induce apoptosis or block proliferation of BM-MDSC (Supplementary Fig. S2B–S2D). Etomoxir, however, decreased the basal and maximal OCR in BM-MDSC (Fig. 2A), diminished fatty acid uptake (Fig. 2B), and decreased ATP levels by approximately 40% to 50% (Fig. 2C; Supplementary Fig. S2E). Furthermore, etomoxir-treated BM-MDSC had a significantly decreased ability to block T-cell proliferation (Fig. 2D) and had a lower expression and activity of arginase I (Fig. 2E and F). NO levels were not detected in BM-MDSC generated in vitro, and PNT levels remained unchanged (data not shown).

We next tested the effect of FAO inhibition in vivo on the accumulation, metabolism, and function of T-MDSC. The initial experiments tested the effect of daily i.p. injections of etomoxir (50 mg/kg) into C57BL/6 mice bearing s.c. 3LL tumors, starting on day 1 after tumor injection and up to day 20, at which time tumors were harvested. Mice did not show any overt toxicity at this dose. Etomoxir treatment decreased the enzymatic activity of CPT1 in T-MDSC in vivo (Supplementary Fig. S3A), but did not alter the percentage of total T-MDSC or that of G-MDSC and M-MDSC subsets infiltrating the 3LL tumors (Fig. 3A) or significantly increase their apoptosis. Similar data were found in MCA-38 colon carcinoma (data not shown). However, treatment with etomoxir significantly inhibited fatty acid uptake (Fig. 3B) and ATP production (Fig. 3C) and decreased OCR and ECAR (Fig. 3D). This suggested that inhibition of FAO decreased the overall metabolic activity of T-MDSC and that T-MDSC appeared to be unable to compensate by increasing glycolytic stress response. More importantly, FAO inhibition in vivo decreased the immunosuppressive function of T-MDSC, as demonstrated by their inability to block T-cell proliferation and IFNγ production (Fig. 3E). This correlated with a lower expression and production of arginase I, ROS, NO, and PNT (Fig. 3F–I; Supplementary Fig. S3B). In addition, T-MDSC from etomoxir-treated mice produced significantly lower levels of cytokines critical to the induction and differentiation of MDSC, such as G-CSF, GM-CSF, IL6, and IL10 (Fig. 3J). In contrast, the levels of IL1β and IL12 remained unchanged. Treatment with etomoxir also diminished the accumulation of CD4⁺ FoxP3⁺ Tregs in the spleens of mice (Supplementary Fig. S3C); minimal numbers of Tregs were found infiltrating 3LL tumors. Cumulatively, these results suggest that inhibiting FAO blocks the immunosuppressive mechanisms and function of T-MDSC. Genetic confirmation of the effect of CPT1 inhibition could not be done at present because CPT1 knockout mice are embryonic lethal and conditional CPT1 knockouts are not yet available for testing.

FAO inhibitors are used to treat severe coronary disease. Therefore, we tested the effect of FAO inhibitors, administered in different regimens in mice bearing s.c. 3LL lung carcinoma or MCA-38 colon carcinoma. Mice that started etomoxir treatment alone 1 day after tumor implantation and up to day 20 had a significant delay in tumor growth, compared with controls (Fig. 4A and B). Histologic studies showed that tumors from control mice had areas of necrosis, whereas tumors from etomoxir-treated mice had minimal necrosis, but were infiltrated by mononuclear cells (Fig. 4C, top). Tunel assay showed that tumors from etomoxir-treated mice had a higher number of cells undergoing apoptosis (Fig. 4C, middle). Immunohistochemical staining showed no changes in CD31⁺ endothelial cells or in the number of blood vessels, suggesting that etomoxir did not affect angiogenesis (Fig. 4C, bottom). We also tested the effect of a second FAO inhibitor, ranolazine, that specifically blocks the trifunctional enzyme HADHA, which catalyzes the last three steps of the FAO cycle (26). Mice treated with ranolazine had a similar antitumor effect as did etomoxir (Fig. 4D). Clonogenic assays showed that FAO inhibition with etomoxir or ranolazine did not inhibit the in vitro growth of 3LL or MCA-38 cells (Supplementary Fig. S4A); etomoxir treatment did not change the number of cancer stem cells (stem cell spheres) or the expression of cancer stem cells markers Sox2, Nanog, or Oct3/4 (Supplementary Figs. S3D and S4B). However, the depletion of CD4⁺ or CD8⁺ T cells in etomoxir-treated mice virtually abrogated the antitumor effect of etomoxir (Fig. 4E; Supplementary Fig. S5A), suggesting that the antitumor effect was at least in part mediated by T cells. We also tested the possibility that etomoxir would augment T-cell function; however, in vitro cultures of T cells with etomoxir failed to show changes in proliferation, cytokine production, or cytotoxic function (or cytotoxic proteins; Supplementary Fig. S5B–S5D).

T-MDSC have been shown to decrease the antitumor effects of chemotherapy and radiotherapy (27) and impair the therapeutic effect of various forms of immunotherapy (1). Therefore, we tested whether FAO inhibition modulated the antitumor effects of low-dose chemotherapy or ACT. In the initial model, etomoxir was given only for 6 days after tumor implantation, followed by a single dose of CTX (200 mg/kg) on day 7. Results showed an increased antitumor effect on both 3LL and MCA-38 tumor–bearing mice treated with FAO inhibitors plus chemotherapy (Fig. 5A and B). This antitumor effect was also seen when treating mice with “established” 3LL tumors (6–7 days after tumor implantation), in which etomoxir was started on day 6 after tumor inoculation and given for 5 days (until day 10), followed by a single dose of CTX (Fig. 5C). A similar effect was seen with ranolazine (Fig. 5D).

Furthermore, the combination of etomoxir and ACT, using OT-1 T cells to treat OVA-expressing 3LL tumors, resulted in a significantly better antitumor effect (Fig. 5E). The increased efficacy of T-cell immunotherapy in etomoxir-treated mice correlated with a higher number of adoptively transferred OT-1 T cells (CD45.1⁺) infiltrating the tumors and increased number of cells producing IFNγ (Fig. 5F and G).

Peripheral blood MDSC from 23 patients with cancer (breast, renal cell carcinoma, bladder cancer, and colon cancer) were isolated and tested for fatty acid uptake. PMNs from normal donors and from the same patients were used as controls. In accordance with previous reports (21, 28), G-MDSC (CD14^neg CD33⁺ HLA-DR^neg CD66b⁺) were increased in the peripheral blood of these patients (Fig. 6A and B). Similar to murine T-MDSC, human G-MDSC had an increased fatty acid uptake (Fig. 6C), and although M-MDSC also incorporated fatty acids, they represented <0.9% of the total circulating MDSC (data not shown). We then examined the expression of CPT-1 and HADHA in biopsies from patients with colon adenocarcinoma, renal cell carcinoma, and breast ductal carcinoma. Immunohistochemistry for CPT-1 demonstrated its presence with a punctate cytoplasmic pattern (consistent with its expression in the mitochondria) in tumor cells and a markedly increased expression shown as a solid pattern in inflammatory cells (Fig. 6D, left). Double labeling with anti-CPT1 (fluorescein) and anti-CD66b (rhodamine) showed that the CD66b⁺ T-MDSC expressed significantly higher levels of CPT1, compared with tumor cells (Fig. 6D, right “merge” plots). Similarly, immunohistochemistry for HADHA showed the expression of the enzyme in both tumor cells and inflammatory cells in the stroma (Fig. 6E, left). Confocal microscopy following double labeling with anti-HADHA (fluorescein) and anti-CD66b (rhodamine) showed that all CD66b⁺ MDSC were also HADHA⁺ (Fig. 6E, right). These results suggest that human T-MDSC have similar metabolic characteristics as murine T-MDSC. The clinical and biologic significance of these results in patients with cancer is yet to be determined. However, the results support the possibility of testing FAO inhibitors as adjuvants to various cancer therapies.

MDSC are chronic inflammatory cells that inhibit T-cell function and play an important role in promoting tumor growth and facilitating the progression of chronic infections (4, 29). Thus, inhibiting MDSC has been pursued as a means of enhancing the effect of cancer therapies, in particular immunotherapies. However, MDSC have multiple immunosuppressive mechanisms, which they upregulate depending on the signals encountered in the microenvironment of different tumors. MDSC in renal cell carcinoma and colorectal carcinoma primarily produce arginase I, whereas MDSC in melanoma produce NO and PNT (21, 30, 31). Targeted inhibition of any of these pathways has not resulted in a significant therapeutic effect, limiting the anti-MDSC therapies to chemotherapeutic agents that suppress their production in the bone marrow. We therefore chose to study the metabolic characteristics of MDSC and test whether inhibition of these pathways could have a therapeutic application.

In recent years, targeting the energy metabolism pathways of tumors and immune cells has gained interest because of its potential to uncover novel prevention or therapeutic targets (32, 33). Thus, we aimed to characterize the energy metabolism pathway(s) of T-MDSC. These data presented here showed that highly immunosuppressive T-MDSC activate FAO. Inhibition of FAO using agents approved for the treatment of coronary disease further demonstrated the importance of this energy production pathway on the immunosuppressive functions of MDSC. The molecular mechanisms linking FAO with the regulatory mechanisms in MDSC are currently unknown; however, the antitumor effect in vivo supports the possibility of testing this therapeutic approach in patients. The mechanisms by which FAO inhibition can cause an antitumor effect are several. Our data strongly suggest that FAO inhibition blocks the immunosuppressive function of T-MDSC and thus allow T cells to kill tumor cells. However, it is also possible that FAO inhibition also decreases Treg function as proposed by Michalek and colleagues (34). In addition, recent reports have shown that some tumors, including pancreatic cancer stem cells and certain myeloid leukemias, rely on mitochondrial function for survival (35, 36). Although our data failed to demonstrate an inhibition of cancer stem cells in these two murine tumor models, it is still possible that FAO inhibition can affect multiple cells that support the tumor microenvironment. Our data also suggest that, in addition to blocking MDSC function, FAO inhibition could promote antitumor responses because it decreases the production of G-CSF, GM-CSF, and IL6, but does not alter the production of IL12. The net effect of these changes, therefore, appears to be a decrease in the overall immunosuppressive microenvironment in favor of the development of antitumor responses. This is especially significant given the additive effect of etomoxir with CTX, which has been used to modulate the immunosuppressive microenvironment. This combination appears to fully inhibit the immunosuppressive function of T-MDSC (Supplementary Fig. S6).

FAO is also important in other myeloid cell subsets. Elegant studies from several laboratories have shown that M2 macrophages primarily use FAO, whereas M1 macrophages use glycolysis (18–20). The pathophysiologic importance of this observation was recently demonstrated by Huang and colleagues, showing that inhibiting lipolysis by orlistat decreased the ability of M2 macrophages to control parasitic infection (37). In addition, Herber and colleagues recently showed that CD11c⁺ dendritic cells infiltrating tumors incorporate oxidized lipids that block antigen processing in lysosomes and their assembly on MHC class II, therefore preventing effective antigen presentation and T-cell stimulation (12). This process was inhibited by 5-(tetradecycloxy)-2-furoic acid (TOFA), an inhibitor of fatty acid synthesis. Another important myeloid subpopulation comprises granulocytes, which primarily use glycolysis as a source of ATP (38). Our data show that T-MDSC, but not splenic MDSC or nMCs, clearly upregulated FAO, suggesting that factors and/or signals in the tumor microenvironment are responsible for this metabolic reprogramming. In many cancer patients, G-MDSC are significantly increased in peripheral blood and tumors (28, 39), and as shown here, human G-MDSC have an increased fatty acid uptake and expression of FAO cycle enzymes CPT1 and HADHA.

The factors responsible for inducing FAO in the tumor microenvironment are still unknown. Tumor-derived extracts trigger an increased synthesis of fatty acids in dendritic cells (12). In addition, signaling through Stat6 and IL4, which promote MDSC differentiation, can also induce PGC1β and activate FAO (20). Our data using in vitro–derived BM-MDSC suggest that G-CSF or GM-CSF may also induce FAO (data not shown). A previous study showed that GM-CSF and IL6, instead, upregulate AMPK and glycolysis in BM-MDSC (40). Recent reports suggest that lactic acid present in high concentrations in the tumor microenvironment can activate MDSC (29, 41); however, its role in inducing FAO is unknown. Thus, additional studies are needed to determine the mechanisms causing the metabolic shift toward FAO in T-MDSC.

How FAO inhibition blocks MDSC immunosuppressive functions is not fully understood, especially given the fact that etomoxir and ranolazine target different enzymes in this pathway. Etomoxir is a nonreversible inhibitor of CPT-1, the enzyme responsible for the initial step of internalization of fatty acids into the mitochondria. Ranolazine is a piperazine derivative that inhibits HADHA, a trifunctional enzyme that catalyzes the last three steps in FAO (26). Both drugs block fatty acid uptake, FAO, and ATP production. More importantly, however, the results demonstrated a novel and potentially important adjuvant effect when combining FAO inhibition with chemotherapy and/or immunotherapy. Phase I clinical trials with etomoxir for the treatment of coronary disease showed toxicities characterized by moderate increases in liver enzymes with chronic use (42). Ranolazine, however, is approved for the treatment of unstable angina. Other FAO inhibitors such as perhexiline (a CPT1 inhibitor) and trimetazidine (an HADHA inhibitor) or lipase inhibitors such as orlistat may, in the present context, inhibit MDSC. This could be an important development because the current therapeutic approaches aimed at blocking MDSC rely primarily on the use of chemotherapeutic agents, such as gemcitabine (43) and 5-fluorouracil (44), that suppress the bone marrow or the tyrosine kinase inhibitor sunitinib (45). In addition, certain chemotherapeutic agents increase rather than decrease the accumulation of MDSC and promote the expression of inhibitory pathways (46). Other approaches have included the use of all-trans-retinoic acid (ATRA) in an attempt to differentiate MDSC into mature granulocytes (47). Thus, targeting the energy metabolic pathways in T-MDSC may provide a broader range of effects by globally inhibiting several immunosuppressive mechanisms in T-MDSC, without causing bone marrow suppression, and allowing T cells to develop an antitumor function.

Results presented here support the possibility of testing approved FAO inhibitors in the context of chemotherapy or immunotherapy of cancer. It also highlights the importance of studying the metabolism of immune cells in diseases in which MDSC function may be modulated by metabolic manipulation (11, 12), including other malignancies (4), trauma and/or sepsis (48), and chronic infections, such as HIV, leishmaniasis, and tuberculosis (49).

No potential conflicts of interest were disclosed.

Conception and design: A.A. Al-Khami, P.C. Rodriguez, A.C. Ochoa

Development of methodology: F. Hossain, A.A. Al-Khami, D. Wyczechowska, L. Zheng, K. Reiss, W. Zou, A.C. Ochoa

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Hossain, A.A. Al-Khami, D. Wyczechowska, C. Hernandez, L.D. Valle, J. Trillo-Tinoco, T. Maj, W. Zou, A.C. Ochoa

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Hossain, A.A. Al-Khami, D. Wyczechowska, C. Hernandez, L.D. Valle, J. Trillo-Tinoco, P.C. Rodriguez, A.C. Ochoa

Writing, review, and/or revision of the manuscript: F. Hossain, A.A. Al-Khami, K. Reiss, L.D. Valle, P.C. Rodriguez, A.C. Ochoa

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.A. Al-Khami, C. Hernandez, L. Zheng, P.C. Rodriguez, A.C. Ochoa

Study supervision: A.A. Al-Khami, P.C. Rodriguez, A.C. Ochoa

This study was funded in part by R01 AI112402, R01CA082689, R01CA107974, and P20GM2013501 (to A.C. Ochoa) and partially supported from LA CaTS Center (U54GM104940; to A.A. Al-Khami and A.C. Ochoa) and the Al Copeland Foundation funds.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.