Immunosuppressive Immature Myeloid Cell Generation Is Controlled by Glutamine Metabolism in Human Cancer

Tumor-associated myeloid cells are a group of heterogeneous myeloid cells that have emerged as one of the predominant regulators of immune response in cancer (1–3). These myeloid cells can enhance cancer cell stemness (4), support angiogenesis and metastasis (5), and promote resistance to chemotherapy, radiotherapy, and immune-checkpoint blocking (ICB) therapy (6–8). Immunosuppressive myeloid cells are generated in the bone marrow from myeloid progenitor cells (9), and then recruited by tumor tissues. Hematopoietic progenitor cells (HPC) can be directly recruited and accumulated in peripheral tissues, where they serve as an important source of descendant cells (10, 11). However, we know little about how these infiltrating HPCs differentiate into functional myeloid cells in the tumor microenvironment (TME).

Growing evidence indicates that the metabolism of a cell can have a marked impact on its function (12). For example, high blood glucose levels in type 2 diabetes cause endothelial cells to increase fatty acid oxidation (FAO), subsequently promoting atherosclerosis and microangiopathy (12, 13). Activated T cells require upregulation of glucose and amino acid catabolism for rapid proliferation and cytokine production (14). On the other hand, polarization of macrophages from M1 to M2 is regulated by the switch from glycolysis to lipid metabolism and oxidative phosphorylation (OxPhos; refs. 15, 16). These observations suggest that the metabolism of cells determines their differentiation and function, potentially offering therapeutic insight.

Nutrient-limiting TME can metabolically reprogram infiltrating immune cells and propagate immune escape (17). Glucose and nutrient deficiency in the TME impairs T-cell receptor signaling in activated T cells, resulting in anergy and exhaustion (18, 19). Notably, some immunosuppressive cells, such as regulatory T cells (Treg) and macrophages, can alter their metabolic pathway to adapt to TME (20, 21). They have low glycolysis rates, but sustain high mitochondrial biogenesis, relying on FAO to exert their immunosuppressive effect (22, 23). However, little is known about how the metabolism regulates the generation of suppressive myeloid cells in the TME.

Here, we identified that hematopoietic precursor cells were enriched in several human cancers. These hematopoietic precursor cells can further differentiate into immature myeloid cells (IMC) with strong immunosuppressive capacity. To explore how metabolism regulates the generation of IMCs in the TME, we used an in vitro model to induce IMCs from CD34⁺ precursors with GM-CSF and G-CSF. Our work revealed that generation of IMCs relied on glutaminolysis, regardless of glucose availability. Glutamine-derived α-ketoglutarate (αKG) supported the expansion of IMCs, and glutamate, another intermediate, conferred the suppressive capacity through the NMDA receptor axis. Moreover, inhibition of glutaminase GLS1 improved the therapeutic efficacy of anti–PD-L1 treatment in an ICB-resistant mouse model. Thus, the glutamine metabolism of suppressive myeloid cells is a promising target to relieve the resistance of ICB therapy.

The detail of materials used in this article is summarized in Supplementary Table S1.

Peripheral blood, cord blood, non-tumor, and tumor tissue samples were obtained from the Cancer Center or the First Affiliated Hospital of Sun Yat-Sen University. All patients with cancer were pathologically confirmed without previous anticancer therapy, and individuals with concurrent autoimmune disease, HIV, or syphilis were excluded. Before processing, samples were shortly stored at 4°C refrigerator. All samples were coded anonymously in accordance with local ethical guidelines (as stipulated by the Declaration of Helsinki), and written informed consent was obtained. The protocol was approved by the Review Board of Sun Yat-Sen University.

Paired fresh tumor and non-tumor tissues from patients with hepatocellular (n = 20), breast (n = 17), and colorectal (n = 40) carcinomas were used for the isolation of tissue-infiltrating leukocytes, and for IHC and immunofluorescence staining. Peripheral blood samples from 27 patients with hepatocellular (n = 11), breast (n = 5), and colorectal (n = 11) carcinomas were taken before treatment. Control blood samples were obtained from 15 healthy donors, all negative for HCV, HBV, HIV, and syphilis, who had routine physical examinations at our cancer center.

Peripheral leukocytes were isolated by Ficoll density gradient centrifugation. Infiltrated leukocytes from fresh tumor and non-tumor tissues were obtained as described in our previous studies (24). Liver, breast, and colon tissue biopsy specimens were cut into small pieces and digested in RPMI-1640 supplemented with 0.05% collagenase type IV (Sigma-Aldrich), 0.002% DNase I (Roche), and 20% fetal calf serum (FCS, HyClone Laboratories). The dissociated cells were then filtered through a 150-μm mesh and the mononuclear cells (MNC) were obtained by Ficoll density gradient centrifugation. The MNCs were washed and resuspended in media supplemented with 1% heat-inactivated FCS for FACS analysis.

Fresh tumor tissues from patients with breast (n = 8) and colorectal (n = 11) carcinomas were used for the detection of tissue-infiltrating CD133⁺ cells by flow cytometry, and the clinical information of these patients was summarized in Supplementary Table S2.

CD133⁺ cells were purified from tumor-infiltrating MNCs in colon and breast cancer patients with CD133 Isolation Kit (Miltenyi Biotec, cat. #130-097-049). The isolated cells were plated and cultured in Methocult H4034 Optimum media (STEMCELL Technologies, cat. #04034). Colonies were assigned scores after 14 to 16 days of culture on the basis of morphology according to the Atlas of Human Hematopoietic Colonies (STEMCELL Technologies, cat. #28700). The types and frequency of colonies were then recognized and counted manually under inverted microscope.

CD133⁺ HPCs were isolated from tumor tissues of colon and breast cancer patients. After that, cells were seeded at 1–2 × 10⁵/well in 24-well plates and cultured in complete medium containing 10% heat-inactivated fetal bovine serum (FBS) supplemented with GM-CSF (40 ng/mL, R&D Systems, cat. #215-GM-010) and G-CSF (40 ng/mL, R&D Systems, cat. #214-CS-025) for 3 days. Differentiated cells were further cocultured 1:1 with allogenic T cells at 37°C in a 5% CO₂-humidified atmosphere for another 6 days.

The CD34⁺ cells were expanded as described in our previous work (25). In brief, CD34⁺ cells were purified from fresh human cord blood or peripheral blood of healthy donors using a direct CD34 progenitor cell isolation kit, according to the manufacturer's instructions (Miltenyi Biotec, cat. #130-046-702). These CD34⁺ cells were plated at 5 × 10⁴ cells/mL in 6-well plates (Corning) with 3 mL/well of HSC expansion media (StemSpan SFEM, Stem Cell Technologies) supplemented with 100 ng/mL SCF, 100 ng/mL FLT-3L, 100 ng/mL TPO, and 20 ng/mL IL3 (R&D). The cells were cultured at 37°C in 5% CO₂ for 7 days. Fresh expansion media were changed on days 3 and 5. The cell density was kept at about 3–5 × 10⁵ cells/mL. The information of media and cytokines was summarized in Supplementary Table S1.

The suppressive IMCs were induced by GM-CSF (40 ng/mL) and G-CSF (40 ng/mL) as described in our previous work (25). To obtain CD34⁺ precursor–derived IMCs, the expanded CD34⁺ cells were plated at 2.5 × 10⁵/well in 24-well plates in complete DMEM supplemented with GM-CSF (40 ng/mL) and G-CSF (40 ng/mL), and cultured at 37°C in 5% CO₂-humidified atmosphere for 3 to 4 days. The information of cytokines was summarized in Supplementary Table S1.

For experiments performed in the absence of glucose or glutamine, cells were differentiated in DMEM without glucose or glutamine (GIBCO) supplemented with BSA (1.5 g/L, GIBCO), GM-CSF (40 ng/mL), and G-CSF (40 ng/mL). Glucose utilization was blocked by addition of 2-DG (1 mmol/L, Sigma-Aldrich). Glutamine utilization was blocked, as indicated, by addition of BPTES (20 μmol/L; APExBIO), DON (50 μmol/L; Sigma-Aldrich), or aminooxyacetic acid (AOA) (1 mmol/L, Sigma-Aldrich), and rescue experiments were performed in the presence of DMK (7 mmol/L, Sigma-Aldrich), Embryomax nucleosides mix (Merck-Millipore), or desaturated lysophospholipid (C18:1; 20 μmol/L, Avanti polar lipids). These drugs were dissolved in DMSO or phosphate-buffered saline (PBS) as appropriate. The chemicals used in cell culture were summarized in Supplementary Table S1.

CD34⁺ cells were cultured in expansion media for 5 days. Thereafter, they were left untreated or transfected with 40 nmol/L of a negative control siRNA or an ASCT2-specific siRNA (si-ASCT2#1: 5′-CCT GGG CTT GGT AGT GTT T-3′ or si-ASCT2#2: 5′-CCT GGA TCT TGC GAG AAA T-3′) using Lipofectamine-RNAi MAX (Invitrogen). After transfection for 48 hours, cells were collected, washed, and cultured in complete DMEM supplemented with GM-CSF and G-CSF for 3 days.

IMCs were resuspended in PBS and centrifuged twice to discard chemical in the culture media. Washed IMCs were cocultured with carboxyfluorescein diacetate succinimidyl ester (CFSE)-labeled T cells in the presence of coated anti-CD3 and soluble anti-CD28 antibody.

Pan T cells were purified from peripheral blood of healthy donors using a Pan T-cell Isolation Kit (Miltenyi Biotec, cat. #130-096-535). Pan T cells and splenocytes were stained with 2.5 μmol/L CFSE for 15 minutes before stimulation, according to the manufacturer's instructions (Invitrogen Molecular Probe, cat. #C34554). After CFSE staining, pan T cells or splenocytes were cultured only or were cocultured with washed IMCs (at a 1:1 ratio) in the presence of 2.5 μg/mL coated anti-CD3 (human: eBioscience, cat. #16-0037-85; mouse: BioLegend, cat. #100314) and 5 μg/mL soluble anti-CD28 (human: eBioscience, cat. #16-0289-85; mouse: BioLegend, cat. #102112) at 37°C in a 5% CO₂-humidified atmosphere for 6 days. Subsequently, the cells were collected, stained with surface markers, and analyzed by flow cytometry.

4T1 (ATCC, CRL-2539) cells were obtained in 2014. 4T1 cells were cultured with RPMI-1640 medium (Gibco) supplemented with 10% FBS, penicillin (100 U/mL), streptomycin (100 mg/mL) at 37°C in 5% CO₂-humidified atmosphere, and were used for experiments within 10 passages. The cell line was tested to ensure negativity for Mycoplasma. This cell line was not reauthenticated within the past year.

Female BALB/c mice (6–8 weeks of age) were purchased from Guangdong Medical Laboratory Animal Center (Guangzhou, China). All mice were maintained under specific pathogen-free conditions in the animal facilities of Sun Yat-sen University Cancer Center (Guangzhou, China), and all animal experiments were performed according to state guidelines and approved by the IACUC of Sun Yat-sen University.

In the 4T1 tumor model, 2 × 10⁵ cells were injected subcutaneously into the flank of BALB/c mice, and tumor growth was monitored for up to 25 days. Tumor dimensions were measured with caliper once when tumors were palpable. Tumor volumes were calculated using the equation (l² × w)/2. Mice cage and treatment allowance were randomized 7 days after tumor implantation. For therapeutic treatment, anti–PD-L1 antibody (200 μg per mouse, clone 10F.9G2, Bio X Cell, cat. # BE0101) or the corresponding IgG2b isotype control (Bio X Cell, cat. # BE0090) in 100 μL PBS was administered intraperitoneally to mice, twice a week for 2 weeks. For glutamine metabolism studies, mice were treated with the selective glutaminase inhibitor BPTES (12.5 mg/kg, ApexBio, cat. # B6008) or vehicle intraperitoneally twice a week.

The flow cytometry was performed as previously described (25). For surface marker staining, MNCs from peripheral blood, fresh tissue biopsy specimens or in vitro cultured cells were prepared and suspended in PBS supplemented with 1% heat-inactivated FCS. In some experiments, the cells were stained with surface markers, fixed and permeabilized with the Intracellular (IC) Fixation Buffer (00-8222, eBioscience) or Foxp3/Transcription Factor Fixation/Permeabilization Concentrate and Diluent (00-5521, eBioscience) reagent, and finally stained with related Abs. Data were acquired on a Gallios flow cytometer (Beckman Coulter) and analyzed with Kaluza 1.2 software (Beckman Coulter).

For cell sorting, cells from in vitro culture systems were resuspended in sorting buffer (1 × PBS; 3% FBS [v/v]; 3 mmol/L EDTA [v/v]) before sorting with a MoFlo XDP instrument (Beckman Coulter). DAPI was used for exclusion of dead cells (eBioscience). The fluorochrome-conjugated Abs used were summarized in Supplementary Table S1.

Paraffin-embedded cancer samples were cut into 4-μm sections and processed for IHC or immunofluorescence staining, as previously described (24). In brief, the sections were sequentially deparaffinized and rehydrated with xylene and a decreasing ethanol series. Then, the slides were soaked in 0.3% H₂O₂ for 10 minutes to quench endogenous peroxidase activity and boiled in 10 mmol/L citrate buffer (pH 6.0) for 10 minutes for heat-induced epitope retrieval. The tissues were stained with rabbit anti-human CD11b (dilution 1:500, Invitrogen, cat. # PA5-29633) and rabbit anti-human Glut1 (dilution 1:500, Millipore, cat. # 07-1401) or rabbit anti-human ASCT2 (dilution 1:500, Novus, cat. # NBP1-89327), or rabbit anti-human xCT (dilution 1:500, Novus, cat. # NB300-318). Signal in chromogenically labeled slides was visualized with DAB by horseradish peroxidase–conjugated anti-rabbit/mouse Dako REAL EnVision detection systems (Dako, cat. # K5007) and counterstained with Mayer's hematoxylin. Signal in immunofluorescently labeled slides was visualized with AlexaFluor 488, AlexaFluor 594, and AlexaFluor Cy5 TSA Kits (Invitrogen), and nuclei were visualized with Prolong Gold Antifade Reagent with 4′,6-diamidino-2-phenylindole (DAPI; Invitrogen). The images were viewed by optical microscope or scanning confocal microscope (Leica) and analyzed with LAS AF Lite software. All stained slides were evaluated in a blinded fashion by 2 pathologists. The antibodies were summarized in Supplementary Table S1.

Single-cell suspensions from in vitro–induced IMCs were plated in XF24 Cell Culture Microplates (Seahorse Bioscience) previously coated with Cell-Tak (BD Biosciences, cat. # 354240) at a cellular density of 1 × 10⁴ cells/well. For oxygen consumption rates (OCR) determination, cells were incubated in XF Base Medium (Seahorse Bioscience) supplemented with 2 mmol/L glutamine, 10 mmol/L glucose, and 1 mmol/L pyruvate for 1 hour, prior to the measurements using the XF Cell Mito Stress Kit (Seahorse Bioscience). Concentrations of oligomycin and FCCP were both 1 μmol/L. For glycolytic metabolism measurements, cells were incubated in XF Base Medium supplemented with 2 mmol/L glutamine prior to injections using the Glycolytic Test kit (Seahorse Bioscience). Experiments were run in an XFe24 analyzer (Seahorse Bioscience). Lactate production was measured by colorimetric assay (BioVision). The information of these assay kit was summarized in Supplementary Table S1.

Total RNA was isolated using TRIzol reagent (Life Technologies, cat. #15596-018). Aliquots containing 2 μg of total RNA were reverse transcribed using MMLV reverse transcriptase (Promega, cat. #M1701). PCR was performed in triplicate using SYBR Green Real-time PCR MasterMix (TOYOBO, cat. #QPS-201) in a Roche LightCycler 480 System. Expression levels were determined using threshold cycle (Ct) values normalized to ACTB (ΔCt) and expressed with 2^–ΔCt. All results are presented in arbitrary units relative to human ACTB mRNA expression. The specific primers used in RT-PCR were summarized in Supplementary Table S1.

The proteins from IMCs were extracted with RIPA buffer (Thermo Fisher Scientific, cat. #89900). The protein concentrations of cell lysates were determined by BCA protein assay kit (Thermo Scientific, cat. #23225). Equal amounts of cellular proteins were separated by 8% or 10% SDS-PAGE, immunoblotted with anti-Arginase1 Ab and anti-β-actin, and visualized by ChemiDoc Imaging Systems (Bio-Rad) with ECL kit (Millipore, cat. #WBKL S0500). The antibodies used in immunoblotting were summarized in Supplementary Table S1.

All microarray data are publicly available in the GEO database under accession number GEO GSE92367.

All experiments were performed using at least 3 different samples. Two-tailed Student t test and one-way ANOVA with Bonferroni adjustment were performed, and P < 0.05 was considered statistically significant. All analyses were performed using SPSS 19.0 (SPSS).

We previously found that circulating and splenic HPCs are myeloid biased in cancer and could be recruited to tumor sites (11, 25). To investigate the differentiation potential of tumor-infiltrating HPCs, CD133⁺ cells were isolated from tissues of colon and breast cancer patients. Their differentiation capacity was tested by the methylcellulose assay. Apart from erythroid differentiation potential, HPCs enriched in tumor tissues were also capable of forming granulocyte-macrophage colonies (CFU-GM; Fig. 1A), similar to those from peripheral blood (25). The HPCs from tumor tissues can differentiate into IMCs in response to tumor-secreting cytokines, such as GM-CSF and G-CSF (26, 27). When cocultured with T cells activated by anti-CD3 and anti-CD28 antibodies, these IMCs exhibited strong suppressive activity against both Th and Tc cells (Fig. 1B).

To dissect the composition of suppressive myeloid cells, we used flow cytometry to characterize different subsets of myeloid cells (CD45⁺CD33⁺HLA-DR^low/−CD3⁻CD19⁻CD56⁻7AAD⁻; ref. 28) in paired tumor and adjacent non-tumor tissues from colon, breast, and liver cancer patients. In addition to classic myeloid derived suppressor cells (MDSC), a CD33⁺CD11b^low/− myeloid subset was enriched by tumor tissues (21.3% ± 3%) compared with non-tumor tissues (5.2% ± 0.9%; Fig. 1C and D, and Supplementary Fig. S1). Most of these CD33⁺CD11b^low/− cells had low expression of mature myeloid markers, including CD14 and CD15 (Fig. 1C; Supplementary Fig. S1), but maintained the expression of the stem/progenitor cell marker CD133 (Supplementary Fig. S2A). Thus, these cells could be characterized as immature myeloid cells or early-stage MDSCs (28). Compared with monocytic MDSC (M-MDSC) and polymorphonuclear MDSC (PMN-MDSC) subsets, these IMCs expressed more of the proliferation marker Ki67 (7.2% ± 1.0% IMC, 0.17% ± 0.06% M-MDSC, 0.17% ± 0.05% PMN-MDSC) and macrophage colony-stimulating factor receptor (M-CSFR; MFI: 25.9 ± 3.1 IMC, 7.6 ± 1.1 M-MDSC, 7.9 ± 1.1 PMN-MDSC), a key marker of suppressive myeloid cells (refs. 29, 30; Fig. 1E–H; Supplementary Fig. S2B). These results indicate that tumor-infiltrating HPCs are sources of IMCs in human cancers.

The TME is known to induce metabolic changes in immune cells, regulating their phenotype and function. To examine the metabolic status of IMC subsets, we assessed the expression of glucose transporter Glut1 (SLC2A1). The results showed that Glut1 was only expressed in a minor proportion of M-MDSC and PMN-MDSC subsets from tumor (2.9% ± 0.5%) or non-tumor (0.7% ± 0.1%) tissues (Fig. 1C and I; and Supplementary Fig. S1), and was rarely detectable on circulating mature myeloid cells (Supplementary Fig. S2C). In contrast, the expression of Glut1 was selectively upregulated on IMCs derived from tumor tissues (37.2% ± 3.5% tumor vs. 10.8% ± 1.4% non-tumor; Fig. 1C and I; Supplementary Fig. S1). By using confocal microscopic analysis, we further confirmed the Glut1 expression on myeloid cells that were localized in the invasive tumor margin in situ (Supplementary Fig. S2D). The selective upregulation of Glut1 on the IMCs indicates that a glycolytic mode of energy metabolism might be critical for the generation of IMCs.

We and others have shown that G-CSF and GM-CSF are the dominant cytokines that regulate myelopoiesis of HSPCs and induce subsequent accumulation of suppressive myeloid cells in several cancers (25, 31). There are limited available models that might reveal the differentiation process of human IMCs. To elucidate the potential metabolic mechanism that regulates the phenotype and function of IMC subsets, we used human hematopoietic progenitors and established a short-term culture system to reliably induce suppressive IMCs in vitro (25). During in vitro differentiation of IMCs, the expression of Glut1 was significantly upregulated (Fig. 2A), mainly on CD11b^low/− cells (Supplementary Fig. S3A). A profound elevation of M-CSFR was observed on these IMC (Supplementary Fig. S3B). Importantly, the inhibition of M-CSFR by pretreatment with neutralizing antibody significantly attenuated the acquisition of suppressive function after the differentiation process, indicating that M-CSFR was a key component of IMC activation program in vitro (Supplementary Fig. S3C and S3D).

Next, we examined the bioenergetic profiling of HPC-derived IMCs using the Seahorse XF-analyzer platform. IMCs displayed an approximately 3-fold increase in both the extracellular acidification rate (ECAR) and lactate production (Fig. 2B and C) compared with untreated cells. Accordingly, real-time PCR analysis revealed that these cytokine-induced IMCs upregulated expression of multiple proteins and enzymes involved in glycolysis. These include the transporter Glut1 (SLC2A1), glycolytic enzymes hexokinase 3 (HK3), glucose-6-phosphate isomerase (GPI), triosephosphate isomerase 1 (TPI1), enolase 1 (ENO1), pyruvate kinase muscle (PKM), and lactate dehydrogenase α (LDHA; Fig. 2D). These results indicated that human IMCs differentiated from HPCs exhibited high glycolytic metabolism.

We examined whether glucose affected IMCs function by inducing IMCs in normal or glucose-deprived medium. To our surprise, induction of IMCs in glucose-insufficient conditions did not impair the expression of suppressive molecule arginase 1 (ARG1) and the suppressive activity of IMCs, but slightly elevated the proportion of the CD15⁺M-CSFR⁺ subset in IMCs (Fig. 2E–I). Consistent with this finding, IMCs induced in glucose-deprived medium maintained high levels of catabolism (as indicated by lactate production and ECAR) with increased expression of Glut1 (Fig. 2B and C; Supplementary Fig. S4A). Although these cells reduced their basal and maximal OCR, their spare respiratory capacity (SRC) was unchanged in glucose-deprived conditions (Fig. 2J and K). Thus, our data indicated that IMCs could maintain normal catabolism to support their functional features under glucose-deprived conditions.

We also used 2-DG, a prototypical inhibitor of glycolysis via blocking hexokinase, to test the importance of glycolysis in IMC differentiation. Distinct from glucose deprivation, 2-DG treatment abolished the glucose uptake and lactate production of IMCs (Supplementary Fig. S4A–S4C). Moreover, 2-DG treatment inhibited the proliferation of CD11b^low/− IMCs (Supplementary Fig. S4D and S4E) and markedly reduced the expression of M-CSFR and ARG1 of IMCs under both normal and glucose-deprived conditions (Supplementary Fig. S3B; Supplementary Fig. S4A, S4B, and S4F). CFSE assays showed that 2-DG treatment also significantly attenuated the suppressive activity of IMCs on T cells (Supplementary Fig. S4G and S4H).

Given that hematopoietic precursors from peripheral blood could be directly recruited to tumor sites to differentiate into IMCs, we also used circulating CD34⁺ cells from peripheral blood and obtained similar results (Supplementary Fig. S5). Thus, these data suggest that an alternative energy resource might be necessary to support the generation and suppressive activity of IMCs in glucose-insufficient conditions.

We next performed microarray analysis to compare the transcriptional profiling from IMCs induced in normal and glucose-deprived conditions. Using the significance analysis of the microarray algorithm, we identified 575 glucose deprivation-induced genes, and 47 of them were metabolism-related genes. Using Gene Ontology enrichment analysis, we found that these 47 genes were enriched for the cellular amino acid metabolic process (Fig. 3A). Interestingly, glucose deprivation upregulated a set of genes in IMCs that have been implicated in glutaminolysis, including GLS, GLUL, BCAT1, SLC1A3, and SLC7A11, which was further confirmed by quantitative RT-PCR (Fig. 3B and C). In addition, ASS1, a key enzyme in the urea cycle that is metabolically linked with amino acid metabolism, was also upregulated in IMCs induced under glucose-deprived condition (Fig. 3B–E). These data strongly indicate that glutaminolysis might be an alternative metabolic pathway to support IMC generation in glucose-deprived conditions. To test this, cytokine-treated precursors were cultured with complete medium (CM), glucose-deprived medium (Non-glc), glutamine-deprived medium (Non-gln), or glucose and glutamine double-deprived medium (Non-glc-gln). Glutamine deprivation markedly reduced the proportion of the CD15⁺M-CSFR⁺ subset and the expansion capacity and suppressive activity of IMCs both in the presence and absence of glucose. The expression of Glut1 and ARG1 on IMCs was also markedly downregulated in IMCs after glutamine deprivation (Fig. 3F–I). Altogether, these data point to a crucial role of glutamine metabolism in the differentiation of IMCs in both normal and glucose-deprived conditions.

The uptake of glutamine is primarily mediated by the amino acid transporter ASCT2 (SLC1A5; ref. 32). To investigate the role of glutamine metabolism in the differentiation of IMCs, we first examined ASCT2 protein expression on induced IMCs and tumors in situ. ASCT2 is highly expressed on cytokine-induced IMCs and myeloid cells at the invasion margin of tumor tissues (Supplementary Fig. S6A and S6B). Next, we pursued an siRNA-mediated interference approach to determine whether ASCT2 was directly responsible for the observed glutamine-dependent effects on IMCs. The results showed that activation of IMCs, indicated by M-CSFR expression, was significantly attenuated by treatment with siASCT2 compared with cells treated with scrambled sequence (Supplementary Fig. S6C and S6D). In addition, 2-DG treatment and glucose deprivation had obviously different effects on IMCs metabolic reprogramming. IMCs can acquire normal levels of ASCT2 expression in glucose-deprived conditions, while completely lost ASCT2 expression with 2-DG treatment (Supplementary Fig. S6B).

ASCT2 also transports other neutral amino acids, including alanine, serine, and cysteine. To determine whether the role of ASCT2 in IMCs is due to its capacity to transport glutamine, we blocked glutaminolysis with several inhibitors that act through different mechanisms (Fig. 4A). Mitochondrial utilization of glutamine occurs via the deamination of glutamate, a reaction catalyzed typically by either glutamate dehydrogenase (GDH) or transaminases. We detected significant upregulation of the expression of transaminases PSAT1 and GPT2 (Fig. 4B), but not GDH in cytokine-induced IMCs. This result pointed to the importance of transaminases in IMC differentiation. Indeed AOA, a pan transaminase inhibitor (33), markedly reduced the generation of functional IMCs (Fig. 4C–E). We also tested bis-2-(5-phenylacetamido-1,3,4-thiadiazol-2-yl)ethyl sulfide (BPTES) and 6-diazo-5-oxo-L-norleucine (DON), 2 mitochondrial glutaminase inhibitors that target glutamine metabolism (34), and observed that treatment significantly abrogated IMCs commitment, as shown by the reduction of functional IMCs and ARG1 expression (Fig. 4C–E).

Generally, glutamine is a nonessential amino acid that can be converted to αKG and utilized in the tricarboxylic acid (TCA) cycle by cells. Glutamine is also converted to glutamate by glutaminase (GLS) in nitrogen-donating reactions for nucleotide synthesis (12, 35). To elucidate the mechanisms by which glutamine supports the differentiation and function of IMCs, we carried out rescue experiments with supplementation of nucleotides and αKG, glutamine-derived metabolic products, in glutamine-deprived conditions. Supplementing cells with nucleosides did not rescue the commitment of glutamine-deprived IMCs. In contrast, supplementation with dimethyl-αKG (DMK), a cell-permeable form of αKG, could largely restore the Glut1 expression and cell proliferation capacity of IMCs (Fig. 5A and B). DMK only partially rescued the production of CD15⁺M-CSFR⁺ cells (Fig. 5A and B). Combination of nucleosides with DMK did not have synergistic effects (Fig. 5A and B).

αKG can be either oxidized to succinate (standard TCA cycle reaction) or reductively carboxylated to citrate, which can be further synthesized to lipids (36, 37). We next investigated which of these pathways glutamine-derived αKG contributed to. Lipids incorporated from glutamine can be rescued by the desaturated lysophospholipid (C18:1; ref. 38). However, we found that this could not rescue the glutamine-dependent commitment of IMCs (Supplementary Fig. S7A). We then analyzed OCR, an estimation of OxPhos, to determine whether αKG could rescue the TCA cycle of IMCs in glutamine-depleted conditions. The results showed that addition of DMK could effectively restore both the basal and maximal OCR of IMCs in glutamine-deprived condition (Fig. 5C). These data demonstrate a critical role of glutamine-derived αKG in supporting the proliferation and differentiation of IMCs via restoring the TCA cycle.

The generation of functional IMCs from hematopoietic precursors is thought to be a two-stage process, myeloid differentiation followed by activation (39). We next examined the kinetic effects of glutamine on myeloid differentiation and cellular expression of M-CSFR. The results showed that depletion of glutamine in culture media did not affect myeloid differentiation, but almost completely blocked expression of M-CSFR, the key component for activation of IMC-suppressive function (Fig. 5D).

The above results collectively indicate a critical role of glutamine metabolism in the activation of IMCs, a process that is independent of glutamine-derived nucleotide and lipid biosynthesis. Moreover, α-KG can only partially rescue the expression of M-CSFR. These results indicate that other component(s) in the glutamine metabolic pathway might contribute to IMC activation. Glutamate was originally found as a major excitatory neurotransmitter in the CNS and can also act as a signaling molecule in other tissues (40). Recent studies have shown that glutamate could regulate the activation and proliferation of T cells through the glutamate receptor NMDAR (41). Therefore, we examined the effect of the NMDAR inhibitor MK-801 on IMC activation. The results showed that treatment with MK-801 significantly reduced the expression of M-CSFR and immune-suppressive function of IMCs (Fig. 5E and F). We subsequently examined the effect of exogenous glutamate on IMC activation. In glutamine-insufficient culture conditions, glutamate at a low dose (100 μmol/L) could effectively restore M-CSFR expression. Interestingly, exogenous glutamate could also increase M-CSFR expression on IMCs under normal culture conditions (Fig. 5G). These data indicate a critical role of exogenous glutamate in IMC activation.

Although cancer cells utilize a large amount of glutamine for tumor growth, they also secrete the metabolite glutamate out to the TME. Glutamate secretion by cancer cells could be mediated by xCT (SLC7A11) cysteine/glutamate antiporter (42). Indeed, expression of xCT antiporter was markedly elevated in tumor tissues, including tumor stroma cells (Supplementary Fig. S7B and S7C). We subsequently examined the in situ distribution of xCT in sections of colon cancer tissues by confocal microscopy. xCT expression is closely linked with IMCs, and they are well colocalized at invasion margin of tumor tissues (Supplementary Fig. S7D). These data indicate that glutamate, as a metabolite from the TME, can be an important strategy for tumors to promote the activation of IMCs.

Suppressive myeloid cells contribute to the resistance of ICB therapy by suppressing T-cell function (8). To find out the role of glutamine utilization in the generation of suppressive myeloid cells in vivo, we tested the effect of BPTES, a selective inhibitor of glutaminase GLS1, on the function of tumor-infiltrating myeloid suppressors in an ICB-resistant 4T1 tumor model. The myeloid cells were purified and cocultured with CFSE-labeled splenocytes from untreated tumor-free mice for 3 days. The suppressive function of tumor-infiltrating myeloid suppressors was significantly attenuated in the BPTES-treated group (Fig. 6A). Although glutamine metabolism is critical for many cancer cells, treatment with BPTES alone only had slight/negligible effect on tumor growth (Fig. 6B). However, BPTES enhanced the therapeutic efficacy of anti–PD-L1 treatment and delayed tumor growth (Fig. 6B). We examined the infiltration of tumor-infiltrating myeloid suppressors, as well as T cells in the 4T1 model. Although BPTES or anti–PD-L1 alone had a slight/negligible effect on the total number of Ly6G⁺ myeloid suppressor cells in tumor tissues, BPTES treatment significantly decreased the expression of ARG1 on the myeloid cells (Fig. 6C and D). These data collectively indicated that targeting glutamine metabolism could abrogate the differentiation and suppressive function of tumor-infiltrating myeloid cells. Neither BPTES nor anti–PD-L1 alone could efficiently rescue the cytotoxic potential of tumor-infiltrating-T cells. The combination of both BPTES and anti–PD-L1 treatment markedly increased the percentage of CD8⁺, IFNγ⁺, and granzyme B⁺ T cells in tumors (Fig. 6E–H). These data indicate that targeting the generation of suppressive myeloid cells from tumor-infiltrating HPCs by BPTES relieves the resistance to ICB therapy.

Tumor-associated myeloid cells contain mixed groups of myeloid cells that are expanded and activated in tumor tissues (2). Here, we found that hematopoietic precursor cells are enriched in several types of human cancers. Tumor-infiltrating hematopoietic precursor cells can further differentiate into IMCs with strong immunosuppressive capacity. These IMC expressed high levels of the glucose transporter Glut1 and M-CSFR, a key marker of suppressive myeloid cells. In vitro, we demonstrated that the cytokines-induced immunosuppressive IMCs exhibited highly glycolytic metabolism. The generation of IMCs was not affected by glucose deprivation due to anaplerotic glutaminolysis. Glutaminolysis was an important catabolic pathway for the generation of myeloid suppressors. Glutamine supported the expansion of myeloid suppressors with glutamine-derived α-ketoglutarate (αKG) and regulated the activation and function of myeloid suppressors through the glutamate–NMDA receptor axis. These findings revealed a novel mechanism for the generation of functional IMCs in the TME.

The ability to access sufficient and appropriate nutrients is a fundamental requirement for cell activation and proliferation, particularly in the nutrition-limiting TME (18, 43, 44). Indeed, we found increased expression of Glut1 on IMCs isolated from human tumor tissues, indicating a strong need for glycolysis during IMC generation. Accordingly, cytokine-induced IMCs express high levels of Glut1 with upregulation of glycolytic enzymes and elevated glycolytic flux. However, the lactate production and differentiation of IMCs is not affected by glucose deprivation, which suggests the existence of an alternative metabolic switch in glucose-limiting conditions. Our data indicate that tumor-associated glycolysis may affect the IMC compartment, and glutamine is one critical element for the differentiation and suppressive function of IMC in the TME. However, it remains unclear how tumor-associated metabolism regulates the myeloid compartment at the genetic, transcriptional, and/or translational levels. In contrast, the glycolysis inhibitor 2-DG could block the metabolic reprogramming during IMC differentiation, probably due to its inhibiting effect on anaplerotic glutaminolysis, which was indicated by a reduced ASCT2 expression.

Glutamine is a primary nitrogen source and is second only to glucose as a carbon source for energy production and anabolic processes (45). The present study revealed that glutamine metabolism is both necessary and sufficient to support the generation of functional IMCs under normal and glucose-insufficient conditions. However, this glutamine-dependent process is neither due to the nitrogen-donating reactions of nucleotide synthesis, nor due to the reductive carboxylation of fatty acid production. Rather, it is glutamine-derived αKG that generates substantial respiration capacity for IMC differentiation by restoring the TCA cycle. The essential role of glutamine in fueling mitochondrial metabolism of rapidly dividing cells has been identified (46, 47). However, we found that although glutamine-derived αKG supports the expansion of IMCs, αKG only partially rescues the suppressive marker of IMCs in glutamine-deprived conditions. These data strongly suggest that other component(s) from glutamine metabolism are involved in regulating the activation stage of IMC generation.

Studies indicate that metabolites, such as kynurenine, succinate, and lactate, can act as signaling molecules to induce trained immunity or suppressive function (48–50). Here, we found that glutamate, a metabolite from glutamine metabolism, was essential for regulating the activation of IMCs. Exogenous glutamate could either promote or restore M-CSFR expression on IMCs in normal or glutamine-deprived culture conditions. Accordingly, antagonism of the glutamate–NMDAR axis effectively blocked the activation of IMCs, but did not alter the cell number of IMCs. Importantly, the expression of an xCT antiporter, which transports glutamate out of the cells (42, 51), is significantly upregulated in tumors. The xCT antiporter is present primarily in the invasion margin of tumor tissues and is colocalized with suppressive myeloid cells.

Growing evidence has shown that metabolism can directly regulate immune cell function and responses to immune-checkpoint blockade treatment (19). Our results suggest that there are fine-tuned collaborative actions that regulate IMC differentiation in the TME. In glucose-limiting TME, the expansion of newly recruited myeloid progenitors is supported by glutamine-derived αKG. Glutamate, exported by tumor tissues, can further promote the activation of these cells to generate functional IMCs. Given the predominant role of suppressive myeloid cells in regulating immune responses in cancer, the present study provides important new insights into the mechanisms by which TME regulates the generation of IMCs. It is, therefore, reasonable to predict that selectively targeting TME and IMC differentiation might provide a novel strategy for immune-based anticancer therapies.

No potential conflicts of interest were disclosed.

Conception and design: H.-W. Sun, L. Zheng

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W.-C. Wu, H.-W. Sun, J. Chen, H.-Y. OuYang, H.-T. Chen, Z.-Y. Shuang, M. Shi, Z. Wang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W.-C. Wu, H.-W. Sun, J. Chen, L. Zheng

Writing, review, and/or revision of the manuscript: W.-C. Wu, H.-W. Sun, L. Zheng

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X.-J. Yu, H.-T. Chen, Z.-Y. Shuang, M. Shi, Z. Wang

Study supervision: L. Zheng

This work was supported by project grants from the National Key R&D Program of China (2017YFA0505803 and 2018ZX10302205), the National Natural Science Foundation of China (81502455, 81730044, 81802663, and 91842308), the Health Medical Collaborative Innovation Program of Guangzhou (201400000001–3), and the Fundamental Research Funds for the Central Universities (16lgjc46 and 171gjc32).

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.