Metabolism is reprogrammed in cancer to fulfill the demands of malignant cells for cancer initiation and progression. Apart from its effects within cancer cells, little is known about whether and how reprogramed metabolism regulates the surrounding tumor microenvironment (TME). Myeloid-derived suppressor cells (MDSC) are key regulators of the TME and greatly affect tumor progression and therapeutic responses. In this study, our results revealed that retinol metabolism–related genes and enzymes were significantly downregulated in human colorectal cancer compared with adjacent colonic tissues, and tumors exhibited a defect in retinoic acid (RA) synthesis. Reduced ADH1-mediated retinol metabolism was associated with attenuated RA signaling and accumulated MDSCs in colorectal cancer tumors. Using an in vitro model, generating MDSCs from CD34+ myeloid precursors, we found that exogenous RA could abrogate the generation of polymorphonuclear MDSCs (PMN-MDSC) with negligible impact on myeloid differentiation. Mechanistically, RA could restrain the glycolytic capacity of myeloid cells, which in turn activated the AMP-activated protein kinase (AMPK) pathway, further impairing the suppressive capacity of myeloid cells. Supplementation with RA could significantly delay tumor growth, with reduced arginase-1–expressing myeloid cells and increased CD8+ and granzyme B+ T cells in both colitis-associated and implanted MC38 mouse colorectal cancer models. Our results indicated that the defect in ADH1-mediated RA synthesis could provide a possible mechanism that fosters the generation of PMN-MDSCs in colorectal cancer and that restoring RA signaling in the TME could serve as a promising therapeutic strategy to abrogate the generation of PMN-MDSCs.

Reprogrammed metabolism is now recognized as hallmark of cancer, and it contributes to fulfilling the bioenergetic, biosynthetic, and redox demands of malignant cells (1, 2). Many studies have investigated the role of reprogrammed metabolism in promoting tumor initiation and progression through cancer cell–intrinsic effects (3–5). Malignant cell–derived metabolites are not only intermediates or end products but also important signaling molecules in the tumor microenvironment (TME; refs. 6, 7). Although much is known about the roles and underlying mechanisms by which tumor-derived proteins educate the TME and facilitate disease progression (8, 9), little is known about the role of metabolites in the cross-talk between cancer cells and surrounding immune cells. Therefore, it is crucial to elucidate the impact of reprogrammed tumor metabolism on infiltrating immune cells in order to achieve a better understanding of tumor-induced immunosuppression.

Myeloid-derived suppressor cells (MDSC) are prominent components of the TME, which can suppress the antitumor T-cell response, enhance cancer stemness, and promote resistance to radiotherapy and immune-checkpoint blockade therapy (10–14). These cells originate from aberrant myelopoiesis, which is induced by tumor-secreted hematopoietic growth factors, such as granulocyte colony-stimulating factor (G-CSF; refs. 15–17). These growth factors are also responsible for expanding myeloid cells under physiologic conditions but without T-cell–suppressive activity (18, 19). These facts indicate that other factors in the TME might synergistically foster the suppressive function of MDSCs. An increasing body of evidence indicates that tumor cell–derived metabolites have an impact on the function of immune cells in the TME (20–22). For example, isocitrate dehydrogenase mutations in tumor cells induce a neomorphic enzymatic production of the oncometabolite (R)-2-hydroxyglutarate (R-2-HG), leading to the suppression of antitumor T-cell activity (20). To date, little is known about whether and how metabolic changes in tumor cells affect the generation and function of MDSCs in the TME.

Retinoic acid (RA) is synthesized from retinol through two-step oxidation (23). Previous studies have shown that RA regulates hematopoiesis, particularly, RA plays an important role in neutrophil maturation (24). For example, RA in the bone marrow contributes to the fine-tuned activation and expansion of hematopoietic stem and progenitor cells (25). RA has been previously applied to optimize the antileukemic efficacy in acute promyelocytic leukemia by inducing the differentiation of leukemic promyelocytes (26). However, despite its critical influence on myeloid differentiation, it remains unclear whether and how RA biosynthesis in tumors regulates the suppressive capacity of infiltrating myeloid cells in colorectal cancer.

Here, we identified that retinol metabolism–related enzymes were significantly downregulated in tumors compared with adjacent normal colonic tissues, indicating that RA synthesis was reduced in colorectal cancer. This decrease in alcohol dehydrogenase 1 (ADH1)–mediated RA synthesis led to attenuated RA signaling in tumor-infiltrating myeloid cells and accumulation of MDSCs. We used an in vitro model to induce polymorphonuclear MDSCs (PMN-MDSC) from CD34+ myeloid precursors, and our results revealed that exogenous RA could abrogate the generation and suppressive function of PMN-MDSCs. Mechanistically, RA could effectively restrain the glycolytic capacity of myeloid cells, leading to activation of the AMPK pathway and inhibition of the suppressive molecule arginase-1 (ARG1). Supplementation with RA attenuated the suppressive function of myeloid cells, improved antitumor T-cell responses, and delayed tumor growth both in colitis-associated colorectal cancers and in an implanted mouse colorectal cancer model.

Materials

Details of the materials used in this study are summarized in Supplementary Table S1.

Cell lines and cell culture

SW480 (ATCC, CCL-228), DLD1 (ATCC, CCL-221), and intestinal epithelial cell CCD 841 CoN (ATCC, CRL-1790) were obtained in 2019. MC38 cells were obtained from Professor Rui-Hua Xu (Sun Yat-sen University Cancer Center) in 2019. The cells were cultured with Dulbecco's modified Eagle medium or RPMI-1640 medium (Gibco) supplemented with 10% fetal bovine serum (FBS; Gibco), penicillin (100 U/mL, Gibco), and streptomycin (100 mg/mL, Gibco) at 37°C in a humidified atmosphere containing 5% CO2 and used for experiments within 10 passages. The cell lines were tested to ensure that it was negative for Mycoplasma. The cell lines were not reauthenticated within the past year.

Mice

Female C57BL/6 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 Institutional Animal Care and Use Committee of Sun Yat-sen University.

Tumor challenge and treatment experiments

Colitis-associated colorectal cancer was induced with azoxymethane (AOM; Sigma-Aldrich) and dextran sodium sulfate (DSS; MP Biomedicals) as described previously (27). Mice were sacrificed 3 weeks after the last DSS treatment. Multiple tumors occurred in the colon of one mouse. The volume of each single tumor was measured, and then the volume of all tumors was added up to calculate the total tumor volume. The average tumor size represented the mean volume of all tumors of one mouse. In the MC38 tumor model, 1 × 106 cells were subcutaneously injected into the flank of C57BL/6 mice, and tumor growth was monitored for up to 21 days. Tumor dimensions were measured with caliper every other day once when tumors were palpable. Tumor volumes were calculated using the equation (l2 × w)/2. Mouse cage and treatment were randomized 7 days after tumor implantation. For RA supplementation, mice were treated with 200 μg RA (Sigma-Aldrich) or vehicle (PBS) i.p. every other day for 2 or 3 weeks as indicated in therapy regimen (28–31). Tumor tissues and control samples were collected for flow cytometry and IHC as described below.

Human subjects

Peripheral blood (n = 6), cord blood (n = 18), and tumor tissues (n = 34) were obtained from the Cancer Center or the First Affiliated Hospital of Sun Yat-sen University, Guangzhou, China. All cancer patients were pathologically confirmed without previous anticancer therapy, and individuals with concurrent autoimmune disease, HIV, or syphilis were excluded from participation. All the samples were coded anonymously in accordance with local ethical guidelines (as stipulated by the Declaration of Helsinki), and written informed consents were obtained from participants prior to study onset. The protocol was approved by the Review Board of Sun Yat-sen University Cancer Center. Paired fresh tumor and nontumor tissues obtained from patients with colorectal cancer between 2016 and 2019 (n = 34) were used for microarray analysis, RT-PCR, IHC, immunofluorescence, and isolation of tissue-infiltrating leukocytes, as indicated below. Before processing, samples were shortly stored at 4°C.

Isolation of mononuclear cells from tissues and peripheral blood

Peripheral leukocytes were isolated by Ficoll density gradient centrifugation. Infiltrating leukocytes were obtained from fresh tumor and nontumor tissues as described in our previous studies (32). 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% FBS (Gibco) for one hour at 37°C. Dissociated cells were then filtered through 150 μm mesh, and the mononuclear cells (MNC) were obtained by Ficoll density gradient centrifugation. MNCs were washed and resuspended in media supplemented with 1% heat-inactivated fetal calf serum for flow cytometry.

Expansion of human CD34+ progenitor cells

CD34+ cells were expanded as described in our previous study (33, 34). Briefly, CD34+ cells (purity >90%) were purified from fresh human cord blood of healthy donors using a direct CD34 progenitor cell isolation kit, according to the manufacturer's instructions (Miltenyi Biotec; cat. #130-046-702). The CD34+ cells were plated in six-well plates at a concentration of 5 × 104 cells/mL with 3 mL/well of hematopoietic stem cell expansion media (StemSpan SFEM, Stem Cell Technologies) supplemented with stem cell factor (SCF, 100 ng/mL), Fms-like tyrosine kinase-3 ligand (FLT-3L, 100 ng/mL), thrombopoietin (TPO, 100 ng/mL), and of IL3 (20 ng/mL; R&D Systems). The cells were cultured at 37°C in 5% CO2 for 7 days. The medium was replaced with fresh expansion medium on days 3 and 5. The cell density was maintained at approximately 3–5 × 105 cells/mL. Information regarding the media and cytokines is summarized in Supplementary Table S1.

Generation of CD34+ cell–derived MDSCs

Suppressive myeloid cells were induced by 40 ng/mL each of GM-CSF and G-CSF as described in our previous work (33). To obtain CD34+ precursor–derived myeloid cells, the expanded CD34+ cells were plated in 24-well plates at a density of 2.5 × 105/well in complete Dulbecco's modified Eagle medium supplemented with 40 ng/mL each of GM-CSF and G-CSF, and cultured at 37°C in 5% CO2-humidified atmosphere for 3 to 4 days. Information regarding cytokines is summarized in Supplementary Table S1. For chemical treatment, MDSCs were induced by combined cytokines for 3 to 4 days in the presence of vehicle, RA (1 μmol/L), 2-deoxyglucose (2-DG, 1 mmol/L), dorsomorphin (2 μmol/L), or A-769662 (10 μmol/L).

Coculture of myeloid cells with pan T cells

Myeloid cells were resuspended in PBS and centrifuged twice to discard chemicals in the culture medium. Washed cells were cocultured with carboxyfluorescein diacetate succidimidyl ester (CFSE)-labeled T cells at the ratio of 1:1 or 1:2 in the presence of coated anti-CD3 and soluble anti-CD28. Briefly, pan T cells were purified from the peripheral blood of healthy donors using a Pan T-cell isolation kit (Miltenyi Biotec; cat. #130-096-535). Splenocytes were isolated from spleens of healthy C57BL/6 mice by Ficoll density gradient centrifugation. Pan T cells and splenocytes were separately 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 alone or cocultured with washed myeloid cells (at a 2:1 or 1:1 ratio) in the presence of coated anti-CD3 (2.5 μg/mL; human: eBioscience, cat. #16-0037-85; mouse: BioLegend, cat. #100314) and soluble anti-CD28 (5 μg/mL; human: eBioscience, cat. #16-0289-85; mouse: BioLegend, cat. #102112) at 37°C in a 5% CO2-humidified atmosphere for 6 days. After incubation, the cells were collected, stained with surface markers, and analyzed by flow cytometry.

RNA preparation and quantitative real-time PCR

Total RNA was isolated from tumor samples or cultured MDSCs using TRIzol reagent (Life Technologies, cat. #15596-018) and used for real-time RT-PCR. Briefly, aliquots containing 1 μg of total RNA was reverse-transcribed with 5 × All-In-One RT MasterMix (ABM; cat. #G492). Quantitative real-time (qRT)-PCR was performed in triplicate using SYBR Green real-time PCR MasterMix (Toyobo; cat. #QPS-201) in a Roche LightCycler 480 System. Gene expression was determined by threshold cycle (Ct) values normalized to 18S (ΔCt) and calculated with 2–ΔCt. All the results are presented in arbitrary units relative to the control group. The specific primers used in RT-PCR are summarized in Supplementary Table S1.

Microarray analysis

Total RNA was prepared from human colorectal cancer tissues and paired colonic tissues and amplified and labeled by the Low Input Quick Amp Labeling Kit, One-Color (Agilent Technologies; cat. #5190-2305). Labeled cRNA were purified by an RNeasy mini kit (QIAGEN; cat. #74106). Agilent SurePrint G3 Human Gene Expression Microarray 8 × 60K was used for the microarray analysis. Each slide was hybridized with 600 ng Cy3-labeled cRNA. Data were extracted with Feature Extraction software 10.7. Raw data were normalized by Quantile algorithm, limma packages in R. Heat maps were generated by MeV 4.9.0. All data are publicly available in the GEO database (GSE156355).

IHC and immunofluorescence staining

Cancer tissues from human patients and mice were embedded in paraffin, cut into 4-μm sections, and processed for IHC or immunofluorescence staining, as previously described (35). In brief, the sections were sequentially deparaffinized and rehydrated with xylene and a decreasing ethanol series. The slides were then soaked in 0.3% H2O2 for 10 minutes to quench endogenous peroxidase activity as appropriate and boiled in 1 × citrate buffer (pH 6.0, Sigma-Aldrich; cat. #C9999) or Tris-EDTA buffer (pH 8.5, Sigma-Aldrich; cat. #E1161) for 10 minutes for heat-induced epitope retrieval. Tissues were stained with rabbit anti-ADH1 (dilution 1:400, Abcam; cat. #ab108203), rabbit anti-RDH5 (dilution 1:200, Novus; cat. #NBP2-15097), rabbit anti-DHRS9 (dilution 1:100, Novus; cat. #NBP1-89378), rabbit anti-CD11b (dilution 1:1,000, Abcam; cat. #ab133357), rabbit anti–RIG-I (dilution 1:2,000, OriGene; cat. #TA506141), rabbit anti–granzyme B (dilution 1:1,000, Abcam; cat. #ab4059), and rabbit anti–arginase-1 (dilution 1:200, CST; cat. #93668S). For IHC staining, the signal was visualized with diaminobenzidine (DAB) by horseradish peroxidase–conjugated anti-rabbit/mouse Dako REAL EnVision detection systems (Dako; cat. #K5007) and counterstained with Mayer's hematoxylin. For immunofluorescence staining, the signal was visualized with AlexaFluor-488 and AlexaFluor-555 TSA Kits (Invitrogen), and nuclei were visualized with Prolong Gold Antifade Reagent with 4′,6-diamidino-2-phenylindole (DAPI; Invitrogen).

The tissue sections were viewed under an optical microscope (Nikon) or a scanning confocal microscope (Zeiss) at low-power field (100×), then three representative high-power fields (400× or 200×) were selected and analyzed manually (36). For evaluating the expression of proteins, H-score was calculated by ΣI × PI, where, “I” was the intensity of staining on specific tissues and cells of interest scored from 0 to 3 (0, negative; 1, weak; 2, moderate; 3, strong), and “PI” was the percentage of cells at each staining intensity ranging from 0% to 100% (37). All the stained slides were evaluated in a blinded fashion by two observers. The average score by two investigators was applied in the following analysis to minimize interobserver variability. For evaluating the density of immune cells, the number of nucleated positive cells was counted manually and expressed as cells per field. A summary of the antibodies used is presented in Supplementary Table S1.

Immunoblotting

Proteins were extracted from induced MDSCs with RIPA buffer (Thermo Fisher Scientific; cat. #89900). Protein concentrations of the cell lysates were determined using a BCA protein assay kit (Thermo Scientific, cat. #23225). Equal amounts of cellular proteins (15 μg) were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and immunoblotted with antibodies against arginase-1, C/EBPβ, p-AMPK, AMPK, and β-actin and visualized by ChemiDoc Imaging Systems (Bio-Rad) with an enhanced chemiluminescence (ECL) kit (Millipore; cat. #WBKL S0500). The antibodies used for the immunoblotting are summarized in Supplementary Table S1.

Flow cytometry

Flow cytometry was performed as previously described (38). For surface marker staining, MNCs were prepared from peripheral blood and fresh tissue biopsy specimens as described above, or in vitro–cultured MDSC cells and suspended in phosphate-buffered saline (PBS) supplemented with 1% heat-inactivated FBS. In some experiments, the cells were stained with surface markers, fixed, and permeabilized with 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 antibodies. Data were acquired on a Cytoflex S flow cytometer (Beckman Coulter). Cell population analysis and t-SNE analysis were performed with FlowJo 10.3 software or CytoExpert 2.0 (Beckman Coulter). The fluorochrome-conjugated antibodies used were summarized in Supplementary Table S1.

XF extracellular flux analyzer experiments

Single-cell suspensions of myeloid cells were plated at a density of 1 × 105 cells/well in XF24 Cell Culture Microplates (Seahorse Bioscience) precoated with 1 μg/well Cell-Tak solution (BD Biosciences, Corning; cat. #354240). To quantify glycolytic metabolism, the cells were incubated in XF Base Medium supplemented with 2 mmol/L glutamine before injections using the glycolytic test kit (Seahorse Bioscience). Experiments were performed in an XFe24 analyzer (Seahorse Bioscience). The data were analyzed by Seahorse Wave 2.3. The information of the assay kit is summarized in Supplementary Table S1.

Data set analysis

The Kyoto Encyclopedia of Genes and Genomes (KEGG) gene set (c2.cp.kegg.v5.2.symbols) from the Molecular Signatures Database-MsigDB was applied for gene set enrichment analysis between colorectal cancer tissues and adjacent colonic tissues expression data from our microarray analysis (GSE156355). Cytoscape and Enrichment map were used for visualization of the GSEA results.

The GEO dataset (GSE 17536) from colorectal cancer patients (n = 177) with survival data was used for survival analysis. The overall survival curve was plotted by Kaplan–Meier methods and analyzed by log-rank test with GraphPad Prism 7.0.

Statistical analysis

All experiments were performed using at least three different samples. The Student t test and one-way analysis of variance (ANOVA) with Bonferroni adjustment were performed to analyze data, and a P value of <0.05 was considered statistically significant. All analyses were performed using SPSS version 19.0 or GraphPad Prism 7.0.

Retinol metabolism–related enzymes are reduced in colorectal cancer tissues

To study metabolic reprogramming in colorectal cancer, we performed microarray and KEGG pathway analysis with paired tumor and adjacent normal colonic tissues obtained from six patients with pathologically confirmed colorectal cancer. Metabolic pathways that are well known for supporting cancer cell proliferation, such as mismatch-repair and DNA replication pathways, were significantly enriched in colorectal cancer tissues, whereas genes related to retinol metabolism were enriched in normal colonic tissues (Supplementary Fig. S1A). Gene set enrichment analysis confirmed a significant decrease in the expression of retinol metabolism–related genes in colorectal cancer compared with adjacent nontumor tissues (Fig. 1A). Retinol metabolism is critical for the biosynthesis of RA, an important immune-regulating metabolite (39). Alcohol dehydrogenase 1 (ADH1) facilitates the first-step oxidation of retinol to retinaldehyde and then to RA. Results of our microarray analysis found that ADH1 expression was the most decreased among retinol metabolism–related genes (Fig. 1B). This reduction in ADH1 expression was confirmed by qRT-PCR and IHC analyses (Fig. 1CE). Similar to the case in human samples, ADH1 expression was reduced in an AOM-DSS-induced colorectal cancer model in comparison with colonic epithelial cells (Fig. 1F). In addition to ADH1, the expression of other retinol metabolism–related enzymes, including retinol dehydrogenase 5 (RDH5) and dehydrogenase/reductase 9 (DHRS9), was also significantly downregulated in human colorectal cancer cell lines and tumor tissues (Supplementary Fig. S1B–S1H). Collectively, these data indicated that RA biosynthesis process might be impaired in both human and mouse colorectal cancer. ADH1A expression was found to have a positive association with favorable overall survival (hazard ratio: 0.347; 95% confidence interval: 0.204–0.590) in colorectal cancer patients in the Gene Expression Omnibus (GEO) dataset (GEO: GSE 17536; Fig. 1G).

Figure 1.

Retinol metabolism–related enzymes are significantly downregulated in colorectal cancer. A, Gene set enrichment analysis for the retinol metabolism–related gene set in colorectal cancer. Gene expression was determined by microarray with colorectal cancer tissues (T) or adjacent colonic tissues (N) from six patients. NES, normalized enrichment score. B, Heat map of differentially expressed genes in retinol metabolism–related gene set in colorectal cancer tissues (T) and adjacent colonic tissues (N). Normalized and log-transformed data are shown. C, Relative ADH1 expression in colorectal cancer (T) and adjacent colonic tissues (N) determined by qRT-PCR. Data represent paired samples from 12 patients of two independent experiments and are shown as mean ± SEM. The statistical significance of differences between groups was determined using a paired Student t test; **, P < 0.01. D, Representative IHC staining of ADH1 expression in human colorectal cancer tissue. E, ADH1 expression on tumor (T) and adjacent nontumor tissues (N) from eight patients was scored and summarized as mean ± SEM. The statistical significance of differences between groups was determined using paired Student t test; ****, P < 0.0001. F, Representative IHC analyses of ADH1 expression on colorectal cancer tissue obtained from an AOM-DSS–treated mouse. Zoomed images, 200×. G, Overall survival of human colorectal cancer patients from the GEO dataset (GSE17536). Patients were assigned into ADH1Alow (n = 88) and ADH1Ahigh (n = 89) cohorts based on the median of ADH1A expression. The statistical significance in survival was determined by the log-rank test.

Figure 1.

Retinol metabolism–related enzymes are significantly downregulated in colorectal cancer. A, Gene set enrichment analysis for the retinol metabolism–related gene set in colorectal cancer. Gene expression was determined by microarray with colorectal cancer tissues (T) or adjacent colonic tissues (N) from six patients. NES, normalized enrichment score. B, Heat map of differentially expressed genes in retinol metabolism–related gene set in colorectal cancer tissues (T) and adjacent colonic tissues (N). Normalized and log-transformed data are shown. C, Relative ADH1 expression in colorectal cancer (T) and adjacent colonic tissues (N) determined by qRT-PCR. Data represent paired samples from 12 patients of two independent experiments and are shown as mean ± SEM. The statistical significance of differences between groups was determined using a paired Student t test; **, P < 0.01. D, Representative IHC staining of ADH1 expression in human colorectal cancer tissue. E, ADH1 expression on tumor (T) and adjacent nontumor tissues (N) from eight patients was scored and summarized as mean ± SEM. The statistical significance of differences between groups was determined using paired Student t test; ****, P < 0.0001. F, Representative IHC analyses of ADH1 expression on colorectal cancer tissue obtained from an AOM-DSS–treated mouse. Zoomed images, 200×. G, Overall survival of human colorectal cancer patients from the GEO dataset (GSE17536). Patients were assigned into ADH1Alow (n = 88) and ADH1Ahigh (n = 89) cohorts based on the median of ADH1A expression. The statistical significance in survival was determined by the log-rank test.

Close modal

ADH1 expression is associated with RA signaling and MDSCs in colorectal cancer

To determine whether loss of ADH1 expression would affect RA signaling in colorectal cancer, we analyzed the expression of RIG-I (DDX58), a well-known retinoic acid-inducible gene (40). The results showed that reduced ADH1 expression was correlated with lower RIG-I expression in colorectal cancer (Fig. 2A). Multiplex immunofluorescence revealed that RIG-I expression was significantly lower in myeloid cells in tumor tissues compared with adjacent normal colon tissues (Fig. 2B and C). Together, these data indicated that RA signaling was diminished in tumor tissue, especially in tumor-infiltrating myeloid cells.

Figure 2.

Retinol metabolism affects RA signaling and fosters the accumulation of MDSCs in colorectal cancer. A, Correlation between ADH1A and the RA-inducible gene RIGI (DDX58). ADH1A and RIGI expression was analyzed using our microarray dataset of colorectal cancer samples. Pearson correlation was calculated. B, Multiplex staining of RIG-I (green), CD11b (red), and DAPI (blue) in each region (colorectal cancer tissue and adjacent normal colon tissue) of paraffin-embedded colorectal cancer sections analyzed by confocal microscopy. Zoomed area, 400×. C, RIG-I expression on CD11b+ cells in tumor (T) and adjacent nontumor tissues (N) from four patients were scored and summarized as means ± SEM. Paired t tests were performed to analyze the data; **, P < 0.01. D, Correlation between ADH1A and the signature genes of MDSCs. Expression of CD15 (FUT4), S100A8, S100A9, and VEGFA was analyzed based on our microarray dataset of colorectal cancer samples. Pearson correlation was calculated. E–G, Tumor-infiltrating immune cells were isolated from paired colon tumor (T) and nontumor (N) tissues and then analyzed by flow cytometry. E and F, CD45+ cells from tumor and nontumor tissues were visualized by t-SNE analysis (E) with five main subsets expressing different intensities of CD3/19/56, CD133, CD11b, CD33, CD14, and CD15 (F). G, The immune cell subsets from colon tumor (T) and nontumor (N) tissues displayed separately. H, Tumor-infiltrating myeloid subsets in CD45+CD3CD19CD567AADPHLA-DRlow/− MDSCs were identified on the basis of the expression of CD33, CD11b, CD14, and CD15. MPs, myeloid precursors, CD133+CD11b−/low; IMCs, immature myeloid cells, CD33+CD14CD15; M-MDSCs, monocytic MDSCs, CD33+CD14+CD15; PMN-MDSCs, CD33+CD14CD15+. The proportions of different myeloid subsets from four patients of four independent experiments were summarized as the means ± SEM. Paired t tests were performed to analyze the data; *, P < 0.05; **, P < 0.01.

Figure 2.

Retinol metabolism affects RA signaling and fosters the accumulation of MDSCs in colorectal cancer. A, Correlation between ADH1A and the RA-inducible gene RIGI (DDX58). ADH1A and RIGI expression was analyzed using our microarray dataset of colorectal cancer samples. Pearson correlation was calculated. B, Multiplex staining of RIG-I (green), CD11b (red), and DAPI (blue) in each region (colorectal cancer tissue and adjacent normal colon tissue) of paraffin-embedded colorectal cancer sections analyzed by confocal microscopy. Zoomed area, 400×. C, RIG-I expression on CD11b+ cells in tumor (T) and adjacent nontumor tissues (N) from four patients were scored and summarized as means ± SEM. Paired t tests were performed to analyze the data; **, P < 0.01. D, Correlation between ADH1A and the signature genes of MDSCs. Expression of CD15 (FUT4), S100A8, S100A9, and VEGFA was analyzed based on our microarray dataset of colorectal cancer samples. Pearson correlation was calculated. E–G, Tumor-infiltrating immune cells were isolated from paired colon tumor (T) and nontumor (N) tissues and then analyzed by flow cytometry. E and F, CD45+ cells from tumor and nontumor tissues were visualized by t-SNE analysis (E) with five main subsets expressing different intensities of CD3/19/56, CD133, CD11b, CD33, CD14, and CD15 (F). G, The immune cell subsets from colon tumor (T) and nontumor (N) tissues displayed separately. H, Tumor-infiltrating myeloid subsets in CD45+CD3CD19CD567AADPHLA-DRlow/− MDSCs were identified on the basis of the expression of CD33, CD11b, CD14, and CD15. MPs, myeloid precursors, CD133+CD11b−/low; IMCs, immature myeloid cells, CD33+CD14CD15; M-MDSCs, monocytic MDSCs, CD33+CD14+CD15; PMN-MDSCs, CD33+CD14CD15+. The proportions of different myeloid subsets from four patients of four independent experiments were summarized as the means ± SEM. Paired t tests were performed to analyze the data; *, P < 0.05; **, P < 0.01.

Close modal

To evaluate the potential downstream impact of the absence of RA biosynthesis on intestinal immunity, we analyzed the correlation between ADH1 expression and myeloid markers with microarray data. Reduced ADH1 expression was correlated with increased expression of MDSC markers (CD15 [FUT4], S100A8, S100A9, and VEGFA; Fig. 2D). We also performed flow cytometry and t-SNE analysis to determine the types of tumor-infiltrating immune cells. The tumor tissue contained significantly enriched subsets of MDSCs, mainly PMN-MDSCs and CD133+ myeloid precursor cells (Fig. 2EH; Supplementary Fig. S2), indicating that reduced ADH1 expression was accompanied by decreased RA signaling and accumulation of PMN-MDSCs in colorectal cancer.

RA abolishes the generation of functional PMN-MDSCs

To elucidate the impact of the absence of RA synthesis on the generation of PMN-MDSCs, we used human hematopoietic progenitors and the established short-term culture system to reliably induce MDSCs (33, 34). In this system, GM-CSF and G-CSF specifically induce the generation of arginase-1–expressing PMN-MDSCs from CD34+ myeloid precursors, rather than inducible nitric oxide synthase (iNOS)–expressing monocytic MDSCs (M-MDSC; Supplementary Fig. S3). RA did not affect the differentiation of PMN myeloid cells (CD15+ cells; Fig. 3A and B), and it decreased the expansion of immature myeloid cells (Supplementary Fig. S4). RA significantly decreased the protein, as well as RNA expression, ARG1 (arginase-1), a key suppressive molecule of PMN-MDSCs (Fig. 3CE). We next evaluated the suppressive function of myeloid cells by coculture with activated T cells. Induced myeloid cells were cocultured with CFSE-labeled T cells in the presence of coated anti-CD3 and soluble anti-CD28 for 5 days and analyzed by flow cytometry. Treatment with RA almost completely abolished the capacity of PMN-MDSC to inhibit the proliferation of both helper and cytotoxic T cells (Fig. 3F and G; Supplementary Fig. S5A–S5E). Accordingly, IFNγ production in T cells was also recovered when cells were cocultured with myeloid cells induced in the presence of exogenous RA (Fig. 3H and I; Supplementary Fig. S5F). Taken together, our data indicated that treatment with RA could abolish the suppressive function of MDSCs without interfering with myeloid differentiation. Cytokine-induced MDSCs could also promote the expansion of regulatory T cells (Treg); however, RA treatment did not affect the frequency of Tregs after coculturing with MDSCs (Supplementary Fig. S5G and S5H).

Figure 3.

RA abrogates the suppressive capacity of PMN-MDSCs. A and B, CD34+ precursors were cultured in medium alone (Med) or with combined cytokines (GM-CSF and G-CSF, GM + G) in the presence of vehicle (Veh) or RA for 3 days. The proportion of CD15+ polymorphonuclear cells was monitored by flow cytometry (A) and summarized as mean ± SEM (B); paired t tests were performed to analyze the data; n.s., not significant. FSC, forward scatter. C, Myeloid cells induced as described in A were collected for RNA isolation. ARG1 expression was measured by qRT-PCR. Data from four independent experiments are shown as the means ± SEM. **, P < 0.01. D, Immunoblotting analysis for ARG1 expression in myeloid cells induced with indicated conditions. E, Relative expression intensity of ARG1 from three samples is shown as the means ± SEM. **, P < 0.01. F and G, Myeloid cells induced with the indicated conditions were cocultured with CFSE-labeled T cells at the ratio 1:2 in the presence of coated anti-CD3 and soluble anti-CD28 for 5 days. F, T-cell proliferation was then examined by flow cytometry to assess the suppressive capacity of myeloid cells. G, Data represent 11 independent samples and are shown as the means ± SEM. Paired t tests were performed to analyze the data; **, P < 0.01. H and I, Myeloid cells induced with the indicated conditions were cocultured with anti-CD3/CD28–activated T cells for 5 days. IFNγ expression in CD4+ T cells was examined by flow cytometry. Data from three independent experiments are summarized as mean ± SEM. The statistical significance was determined by a paired Student t test; *, P < 0.05.

Figure 3.

RA abrogates the suppressive capacity of PMN-MDSCs. A and B, CD34+ precursors were cultured in medium alone (Med) or with combined cytokines (GM-CSF and G-CSF, GM + G) in the presence of vehicle (Veh) or RA for 3 days. The proportion of CD15+ polymorphonuclear cells was monitored by flow cytometry (A) and summarized as mean ± SEM (B); paired t tests were performed to analyze the data; n.s., not significant. FSC, forward scatter. C, Myeloid cells induced as described in A were collected for RNA isolation. ARG1 expression was measured by qRT-PCR. Data from four independent experiments are shown as the means ± SEM. **, P < 0.01. D, Immunoblotting analysis for ARG1 expression in myeloid cells induced with indicated conditions. E, Relative expression intensity of ARG1 from three samples is shown as the means ± SEM. **, P < 0.01. F and G, Myeloid cells induced with the indicated conditions were cocultured with CFSE-labeled T cells at the ratio 1:2 in the presence of coated anti-CD3 and soluble anti-CD28 for 5 days. F, T-cell proliferation was then examined by flow cytometry to assess the suppressive capacity of myeloid cells. G, Data represent 11 independent samples and are shown as the means ± SEM. Paired t tests were performed to analyze the data; **, P < 0.01. H and I, Myeloid cells induced with the indicated conditions were cocultured with anti-CD3/CD28–activated T cells for 5 days. IFNγ expression in CD4+ T cells was examined by flow cytometry. Data from three independent experiments are summarized as mean ± SEM. The statistical significance was determined by a paired Student t test; *, P < 0.05.

Close modal

RA impairs glycolysis and activates the AMPK pathway in myeloid cells

In our previous study, we demonstrated that glycolysis is essential for the generation of functional MDSCs from hematopoietic precursors (34). To elucidate the mechanisms by which RA abrogates the suppressive capacity of myeloid cells, we investigated the glycolysis capacity of myeloid cells induced in medium with or without exogenous RA. First, qRT-PCR revealed that RA decreased the expression of genes encoding multiple transporters and enzymes involved in glycolysis, including glucose transporter 3 (GLUT3, SLC2A3), hexokinase 3 (HK3), and 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 (PFKFB3; Fig. 4A). Next, we examined the effect of RA on the bioenergetic profiling of myeloid cells using the Seahorse XF-analyzer platform. As evidenced by the expression of glycolysis-related genes, the extracellular acidification rate (ECAR) and glycolysis capacity were significantly attenuated in RA-treated myeloid cells (Fig. 4B and C).

Figure 4.

RA impairs glycolysis in MDSCs and activates the AMPK pathway. A, Relative gene expression of glycolysis genes in myeloid cells induced with medium alone (Med) or with cytokines in the presence of vehicle (GM + G Veh) or RA (GM + G RA). Data from three independent experiments are summarized as means ± SEM. *, P < 0.05; **, P < 0.01. B, ECAR of myeloid cells induced with indicated conditions was assessed by Seahorse XFe24 analyzer with the addition of glucose (Glc), oligomycin (Oligo), and 2-deoxyglucose (2-DG) at the indicated times. C, Glycolysis capacity of myeloid cells from five independent samples are shown as means ± SEM. The statistical significance was determined by a paired Student t test; **, P < 0.01. D, Immunoblotting analysis for phosphorylated (p)-AMPK, AMPK, C/EBPβ, and ARG1 in myeloid cells induced with medium alone (Med) or with cytokines in the presence of vehicle (GM + G Veh), RA (GM + G RA), or 2-DG (GM + G 2-DG). E and F, Relative gene expression of ARG1 (E) and CEBPB (F) in myeloid cells induced with the indicated conditions. Data from four independent samples are summarized as means ± SEM. Paired t tests were performed to analyze the data; *, P < 0.05. G, Immunoblotting analysis for p-AMPK, AMPK, C/EBPβ, and ARG1 in myeloid cells induced with indicated conditions. H, Immunoblot analysis for C/EBPβ and ARG1 in myeloid cells induced with medium alone (Med) or cytokines with or without RA or A-769662 at the indicated concentrations. I, Myeloid cells were induced for 3 days with medium alone (Med) or cytokines in the presence of vehicle (Veh) or A-769662. The myeloid cells were then cocultured with CFSE-labeled T cells in the presence of anti-CD3 and anti-CD28 for 5 days. T-cell proliferation was detected by flow cytometry to assess the suppressive capacity of myeloid cells. The proportion of nondividing T cells is shown as means ± SEM. n = 3; statistical significance was determined by a paired Student t test; *, P < 0.05; **, P < 0.01.

Figure 4.

RA impairs glycolysis in MDSCs and activates the AMPK pathway. A, Relative gene expression of glycolysis genes in myeloid cells induced with medium alone (Med) or with cytokines in the presence of vehicle (GM + G Veh) or RA (GM + G RA). Data from three independent experiments are summarized as means ± SEM. *, P < 0.05; **, P < 0.01. B, ECAR of myeloid cells induced with indicated conditions was assessed by Seahorse XFe24 analyzer with the addition of glucose (Glc), oligomycin (Oligo), and 2-deoxyglucose (2-DG) at the indicated times. C, Glycolysis capacity of myeloid cells from five independent samples are shown as means ± SEM. The statistical significance was determined by a paired Student t test; **, P < 0.01. D, Immunoblotting analysis for phosphorylated (p)-AMPK, AMPK, C/EBPβ, and ARG1 in myeloid cells induced with medium alone (Med) or with cytokines in the presence of vehicle (GM + G Veh), RA (GM + G RA), or 2-DG (GM + G 2-DG). E and F, Relative gene expression of ARG1 (E) and CEBPB (F) in myeloid cells induced with the indicated conditions. Data from four independent samples are summarized as means ± SEM. Paired t tests were performed to analyze the data; *, P < 0.05. G, Immunoblotting analysis for p-AMPK, AMPK, C/EBPβ, and ARG1 in myeloid cells induced with indicated conditions. H, Immunoblot analysis for C/EBPβ and ARG1 in myeloid cells induced with medium alone (Med) or cytokines with or without RA or A-769662 at the indicated concentrations. I, Myeloid cells were induced for 3 days with medium alone (Med) or cytokines in the presence of vehicle (Veh) or A-769662. The myeloid cells were then cocultured with CFSE-labeled T cells in the presence of anti-CD3 and anti-CD28 for 5 days. T-cell proliferation was detected by flow cytometry to assess the suppressive capacity of myeloid cells. The proportion of nondividing T cells is shown as means ± SEM. n = 3; statistical significance was determined by a paired Student t test; *, P < 0.05; **, P < 0.01.

Close modal

Glycolysis is one of the predominant ATP-generation pathways in activated myeloid cells. Blocking glycolysis could increase the AMP:ATP ratio, which in turn would activate the AMPK pathway, the central pathway for nutrient sensing (41). Indeed, RA could effectively activate the AMPK pathway in myeloid cells to a level similar to glycolysis inhibitor 2-deoxyglucose (2-DG; Fig. 4D; Supplementary Fig. S6A). A previous study revealed that AMPK activation can repress CEBP/β expression in triple-negative breast cancer cells (42). CEBP/β is an important regulator of the suppressive capacity of MDSCs (43). Accordingly, in our study, RA or 2-DG decreased the expression of C/EBPβ and ARG1 in PMN-MDSCs (Fig. 4DF; Supplementary Fig. S6B and S6C). In summary, RA activated the AMPK pathway and attenuated the suppressive capacity of myeloid cells by inhibiting glycolytic capacity.

To further explore the role of the AMPK pathway in the suppressive function of myeloid cells, we incubated RA-treated myeloid cells with the specific AMPK phosphorylation inhibitor dorsomorphin (44). Treatment inhibited AMPK phosphorylation, while recovering the expression of C/EBPβ and ARG1 in RA-treated cells (Fig. 4G; Supplementary Fig. S6D–S6F). Activating the AMPK pathway with the specific agonist A-769662 (45) significantly downregulated C/EBPβ and ARG1 expression in PMN-MDSCs (Fig. 4H; Supplementary Fig. S6G and S6H). Furthermore, the AMPK agonist A-769662 could impair the suppressive capacity of cytokine-induced myeloid cells (Fig. 4I; Supplementary Fig. S6I–S6K). Our data suggested that RA could inhibit the glycolytic capacity of myeloid cells, which in turn activated the AMPK pathway, further inhibiting the suppressive capacity of myeloid cells.

RA reduces the MDSCs and tumor burden in AOM-DSS–induced mouse colorectal cancer

Given the profound impact of RA on attenuating the suppressive capacity of myeloid cells in vitro, we aimed to assess the effect of RA on PMN-MDSCs and tumor burden in a mouse model. A colorectal cancer mouse model was established using the colonotropic carcinogen AOM and the inflammatory agent DSS to induce colorectal cancer in mice (Fig. 5A; ref. 27). In this AOM-DSS model, ADH1 expression was significantly decreased in cancer cells compared with adjacent colonic epithelial cells (Fig. 1F), with accumulated myeloid precursor cells and MDSCs in tumor tissues (Supplementary Fig. S7A and S7B). These results revealed similar characteristics between the AMO-DSS–induced colorectal cancer in mice and human colorectal cancer. To further investigate the role of RA in colorectal cancer development, AOM-DSS mice were given intraperitoneal RA injections every other day. RA supplementation was found to decrease the incidence of tumors, and the tumor number was lower and tumor burden decreased by more than half in the RA-treated mice versus control vehicle-treated mice (Fig. 5BE).

Figure 5.

RA reduces MDSCs and tumor burden in AOM-DSS–induced colorectal cancer. A, Therapy regimen. Mice were i.p. injected with vehicle or 200 μg RA every other day for 3 weeks following AOM-DSS treatment. B, Representative image of colons from an AOM-DSS–treated mouse with the indicated therapy. C–E, Total tumor volume (C), tumor number (D), and average tumor size (E) in AOM-DSS mice treated with RA compared with vehicle. Results from six mice/group are represented as mean ± SEM. Unpaired t tests were performed to analyze the data; *, P < 0.05. F, Representative flow-cytometric analysis and quantification of CD11b+ myeloid cells, PMN-MDSCs, and CD8+ T cells in tumor tissues of an AOM-DSS mouse treated with vehicle or RA. G and H, Frequency of CD11b+ myeloid cells (G) and PMN-MDSCs (H) in the total CD45+ cell population from tumor tissues after therapy (n = 5 or 6/group). Data from two independent experiment are summarized as mean ± SEM. *, P < 0.05. I, IHC staining for arginase-1 and granzyme B expression in tumor sections from vehicle- or RA-treated AOM-DSS mice. J, Number of arginase-1+ from I was quantified and summarized as the means ± SEM. n = 6/group; the statistical significance was determined by Student t test; *, P < 0.05. K, Frequency of CD8+ T cells in the total CD45+ cell population from tumor tissues after therapy (n = 5 or 6/group). Data are from two independent experiment summarized as mean ± SEM. The statistical significance was determined by Student t test; *, P < 0.05. L, Number of granzyme B+ cells from I were quantified and summarized as the mean ± SEM. n = 6/group; unpaired t tests were performed to analyze the data; *, P < 0.05.

Figure 5.

RA reduces MDSCs and tumor burden in AOM-DSS–induced colorectal cancer. A, Therapy regimen. Mice were i.p. injected with vehicle or 200 μg RA every other day for 3 weeks following AOM-DSS treatment. B, Representative image of colons from an AOM-DSS–treated mouse with the indicated therapy. C–E, Total tumor volume (C), tumor number (D), and average tumor size (E) in AOM-DSS mice treated with RA compared with vehicle. Results from six mice/group are represented as mean ± SEM. Unpaired t tests were performed to analyze the data; *, P < 0.05. F, Representative flow-cytometric analysis and quantification of CD11b+ myeloid cells, PMN-MDSCs, and CD8+ T cells in tumor tissues of an AOM-DSS mouse treated with vehicle or RA. G and H, Frequency of CD11b+ myeloid cells (G) and PMN-MDSCs (H) in the total CD45+ cell population from tumor tissues after therapy (n = 5 or 6/group). Data from two independent experiment are summarized as mean ± SEM. *, P < 0.05. I, IHC staining for arginase-1 and granzyme B expression in tumor sections from vehicle- or RA-treated AOM-DSS mice. J, Number of arginase-1+ from I was quantified and summarized as the means ± SEM. n = 6/group; the statistical significance was determined by Student t test; *, P < 0.05. K, Frequency of CD8+ T cells in the total CD45+ cell population from tumor tissues after therapy (n = 5 or 6/group). Data are from two independent experiment summarized as mean ± SEM. The statistical significance was determined by Student t test; *, P < 0.05. L, Number of granzyme B+ cells from I were quantified and summarized as the mean ± SEM. n = 6/group; unpaired t tests were performed to analyze the data; *, P < 0.05.

Close modal

We then examined the tumor-infiltrating immune cells in AOM-DSS mice treated with RA. Consistent with the findings in the clinical sample analysis and in vitro study, PMN-MDSCs constituted a major proportion of MDSCs. Treating mice with RA significantly decreased the frequency of infiltrating myeloid cells as well as their expression of ARG1 (Fig. 5FJ). RA treatment also simultaneously increased infiltrating CD8+ T cells and granzyme B-expressing cells in colorectal cancer tumors (Fig. 5F and I–L). Thus, treating mice with AOM-DSS–induced colorectal cancer with RA effectively attenuated the suppressive capacity of PMN-MDSCs and elicited a robust CD8+ T-cell response. No significant difference in the infiltration of Tregs in the TME of RA-treated mouse was seen (Supplementary Fig. S7C and S7D).

RA impairs the suppressive capacity of MDSCs and delays tumor growth in implanted colorectal cancer

To further confirm the role of RA deficiency in cancer progression, we investigated the therapeutic benefit of RA in an MC38 subcutaneous tumor model. Mice with MC38 tumors were intraperitoneally injected with RA on alternate days for 2 weeks (Fig. 6A). RA was found to significantly delay tumor growth and caused a 50% reduction in tumor burden (Fig. 6B and C). Consistent with the results in the AOM-DSS model, RA treatment also significantly decreased the percentage of PMN-MDSCs and attenuated ARG1 expression in tumors, without significant influence on Treg infiltration (Fig. 6DF; Supplementary Fig. S7C–S7F). Ex vivo analysis confirmed that RA supplementation abrogated the suppressive capacity of tumor-infiltrating myeloid cells (Fig. 6G and H; Supplementary Fig. S7G and S7H). At the same time, RA treatment significantly increased the proportion of CD8+ T cells and granzyme B+ cells in tumors (Fig. 6D and I; Supplementary Fig. S7E and S7I). Collectively, our data indicated that supplementation of RA is a promising therapeutic strategy for cancers with RA synthesis deficiency by modulating the function of tumor-infiltrating PMN-MDSCs.

Figure 6.

RA supplementation abolishes suppressive capacity of MDSCs and delays tumor growth in an MC38 mouse model. A, Therapy regimen. Tumor analyses of implanted MC38 mice i.p. injected with vehicle or 200 μg of RA every other day for 2 weeks. B, Representative image of MC38 tumors with the indicated therapy. C, Mean tumor volume of subcutaneous MC38 tumors in vehicle- versus RA-treated mice at the indicated time points. Results are shown as mean ± SEM. n = 6/group; **, P < 0.01. D, Flow-cytometric analysis and quantification of CD11b+ myeloid cells, PMN-MDSCs, and CD8+ T cells in MC38 tumors treated with vehicle or RA at day 21 after implantation. E and F, Frequency of CD11b+ cells (E) and PMN-MDSCs (F) in the total CD45+ MNCs in MC38 tumor tissues (n = 6/group). Data from two independent experiment are summarized as mean ± SEM. The statistical significance was determined by Student t test; *, P < 0.05. G and H, Tumor-infiltrating MDSCs from MC38 mice treated with vehicle or RA were purified and cocultured with CFSE-labeled splenocytes for 3 days in the presence of anti-CD3 and anti-CD28 antibodies. G, Flow cytometry was used to monitor the proliferation of CD8+ T cells. H, Data from three independent samples were summarized as the mean ± SEM. Student t tests were performed to analyze the data; *, P < 0.05. I, Frequency of CD8+ T cells in the total CD45+ MNCs in MC38 tumor tissues (n = 6/group). Data from two independent experiment are summarized as mean ± SEM. Statistical significance was determined by Student t test; *, P < 0.05.

Figure 6.

RA supplementation abolishes suppressive capacity of MDSCs and delays tumor growth in an MC38 mouse model. A, Therapy regimen. Tumor analyses of implanted MC38 mice i.p. injected with vehicle or 200 μg of RA every other day for 2 weeks. B, Representative image of MC38 tumors with the indicated therapy. C, Mean tumor volume of subcutaneous MC38 tumors in vehicle- versus RA-treated mice at the indicated time points. Results are shown as mean ± SEM. n = 6/group; **, P < 0.01. D, Flow-cytometric analysis and quantification of CD11b+ myeloid cells, PMN-MDSCs, and CD8+ T cells in MC38 tumors treated with vehicle or RA at day 21 after implantation. E and F, Frequency of CD11b+ cells (E) and PMN-MDSCs (F) in the total CD45+ MNCs in MC38 tumor tissues (n = 6/group). Data from two independent experiment are summarized as mean ± SEM. The statistical significance was determined by Student t test; *, P < 0.05. G and H, Tumor-infiltrating MDSCs from MC38 mice treated with vehicle or RA were purified and cocultured with CFSE-labeled splenocytes for 3 days in the presence of anti-CD3 and anti-CD28 antibodies. G, Flow cytometry was used to monitor the proliferation of CD8+ T cells. H, Data from three independent samples were summarized as the mean ± SEM. Student t tests were performed to analyze the data; *, P < 0.05. I, Frequency of CD8+ T cells in the total CD45+ MNCs in MC38 tumor tissues (n = 6/group). Data from two independent experiment are summarized as mean ± SEM. Statistical significance was determined by Student t test; *, P < 0.05.

Close modal

Reprogrammed metabolism in malignant cells occurs in order to meet the bioenergetic and biosynthetic demands of malignant cells. Studies have shown that tumor metabolites and nutrient competition could impair the antitumor immune response (6, 7). In the present study, we showed that RA synthesis–related enzymes were deficient in colorectal cancer, leading to attenuated RA signaling in tumor-infiltrating myeloid cells and an accumulation of PMN-MDSCs. Addition of RA could abrogate the generation of functional PMN-MDSCs by restraining glycolysis capacity, activating the AMPK pathway, and subsequently inhibiting the expression of ARG1 on these cells. RA supplementation also significantly reduced the accumulation of ARG1-expressing myeloid cells, improving the antitumor T-cell response and leading to delayed tumor growth in a mouse colorectal cancer model. Our results indicated that the defect in RA synthesis could provide a possible mechanism that fosters the generation of PMN-MDSCs in colorectal cancer and that normalizing the RA signaling in the TME could serve as a promising therapeutic strategy to abrogate the generation of PMN-MDSCs and improve antitumor responses (Fig. 7).

Figure 7.

Retinoic acid synthesis deficiency fosters PMN-MDSC generation in colorectal cancer (CRC). Retinol metabolism–related enzymes are significantly downregulated in human colorectal cancer compared with adjacent colonic tissues, indicating a defect in ADH1-mediated RA synthesis. RA biosynthesis deficiency could provide a possible mechanism that fosters the generation of PMN-MDSCs in colorectal cancer tumors. Supplementation with exogenous RA could abolish the generation of functional PMN-MDSCs with negligible impact on myeloid differentiation by restraining the glycolytic capacity and activating the AMPK pathway. Restoring RA signaling in the TME could abolish the suppressive function of tumor-infiltrating MDSCs and improve the antitumor response. Blue cross, suppression on antitumor response; arrows, production or promotion; blunted lines, inhibition; black symbol, existing effect; gray symbol, abrogated effect.

Figure 7.

Retinoic acid synthesis deficiency fosters PMN-MDSC generation in colorectal cancer (CRC). Retinol metabolism–related enzymes are significantly downregulated in human colorectal cancer compared with adjacent colonic tissues, indicating a defect in ADH1-mediated RA synthesis. RA biosynthesis deficiency could provide a possible mechanism that fosters the generation of PMN-MDSCs in colorectal cancer tumors. Supplementation with exogenous RA could abolish the generation of functional PMN-MDSCs with negligible impact on myeloid differentiation by restraining the glycolytic capacity and activating the AMPK pathway. Restoring RA signaling in the TME could abolish the suppressive function of tumor-infiltrating MDSCs and improve the antitumor response. Blue cross, suppression on antitumor response; arrows, production or promotion; blunted lines, inhibition; black symbol, existing effect; gray symbol, abrogated effect.

Close modal

Intensive studies have reported that the reprogrammed metabolism seen in cancer is characterized by enhanced aerobic glycolysis and glutaminolysis, which can fulfill the bioenergetic and biosynthetic demands of cancer cells (2, 46, 47). In the present study, we identified a reduction in the expression of retinol metabolism–related enzymes as a reprogrammed metabolic process that regulates the formation of an immunosuppressive TME. Microarray analysis of paired colorectal cancer and adjacent normal tissues from colorectal cancer patients suggested that retinol metabolism may be altered in colorectal cancer tissues. Due to limited human samples and cell lines examined, the generalizability of this reduced retinol metabolism in colorectal cancer still needs to be validated in more cohorts. Retinol is first oxidized to retinaldehyde by ADH1, and then to RA, a key natural regulatory retinoid. ADH1 expression was the most decreased among retinol metabolism–related genes, and reduced ADH1 expression correlated with increased MDSC markers in tumor cells. Consistent with these findings, RIG-I expression was significantly downregulated in tumor-infiltrating myeloid cells. The results of both the in vitro and in vivo studies showed that supplementation with RA effectively attenuated the generation of PMN-MDSCs. MDSCs are consisted of PMN-MDSCs and M-MDSCs, but PMN-MDSCs were significantly enriched and much higher than M-MDSCs in colorectal cancer tumor tissues, indicating that PMN-MDSCs were the dominant MDSC population in colorectal cancer tissues. Thus, this study focused on the impact of decreased retinoic acid synthesis on the generation of PMN-MDSCs and found that RA could abrogate the generation and function of PMN-MDSCs by modulating their glycolysis. It should be noted that the function of MDSCs could also be affected by other metabolic pathways, cytokines, or chemokines (10, 34, 48). However, it remains to be explored whether and how these pathways contribute to the regulation of MDSC function in human tumors with RA synthesis deficiency.

MDSCs originate from aberrant myelopoiesis induced by cytokines, such as G-CSF. However, these hematopoietic growth factors also contribute to the expansion of normal (non-suppressive) myeloid cells under physiologic conditions (18, 19). Although the precise regulatory mechanisms remain unclear, our previous study and other studies have shown that local environmental conditions determine the differentiation and generation of functional MDSCs (34, 49, 50). RA is an active vitamin A metabolite and an important regulator for embryonic development, cell differentiation, and immune response (51, 52). We found that impaired RA signaling was associated with the accumulation of MDSCs. Exogenous RA did not affect the differentiation of CD15+ cells; however, it abolished ARG1 expression and the suppressive function of PMN-MDSCs. Thus, the absence of RA synthesis might be an important mechanism by which myeloid cells in colorectal cancer tissues acquire suppressive ability.

Studies have shown that MDSCs exhibit high glycolytic activity, which is essential for the generation of functional MDSCs (34, 53). In our study, we demonstrated that RA could decrease the expression of multiple transporters and enzymes involved in glycolysis, leading to an impaired ECAR and glycolysis capacity of myeloid cells. RA attenuated glycolysis capacity and activated AMPK in PMN-MDSCs. AMPK activation in triple-negative breast cancer cells has been demonstrated to repress the expression of CEBP/β (42), which is a critical transcription factor that regulates the function of MDSCs, in tumor cells. Similarly, such a mechanism may also be involved in regulating CEBP/β in MDSCs. We observed that AMPK activation by RA was vital to abolish the suppressive capacity of myeloid cells by repressing CEBP/β and ARG1 expression. Our study identified that RA could effectively restrain glycolytic activity and abolish the suppressive capacity of myeloid cells.

MDSCs are key regulators of antitumor immunity and affect virtually all types of cancer therapy (13, 14, 54). Several therapeutic approaches targeting the generation and function of MDSCs have been implemented in preclinical and clinical research (55). However, agents that target the myeloid cell lineage might hamper antitumor activities by some of these heterogeneous cell populations. Thus, selectively modulating the differentiation and functions of MDSCs could serve as an alternative strategy to improve the antitumor immune response. In the present study, it was demonstrated that RA could selectively attenuate the suppressive capacity of PMN-MDSCs with negligible impact on myeloid differentiation. Supplementation with RA was found to enhance cytotoxic T-cell responses and delay tumor growth. Given growing evidence regarding the deficiency of RA synthesis on several types of human cancers (56, 57), studying the mechanisms that selectively regulate retinol metabolism could provide a novel strategy to increase the response and efficacy of anticancer therapy. Due to the complexity and redundancy of immunoregulatory networks in the TME, monotherapies that block single immunosuppressive element/pathway could only result in limited or no clinical benefit, particularly in cancer patients (58, 59). In this context, RA as a single agent has failed to show a benefit in patients with solid tumors (60). Thus, understanding how the combination strategy, including RA, to effectively overcome the impact of tolerized conditions, would be critical for immunotherapy-induced robust antitumor responses.

No disclosures were reported.

H.-W. Sun: Conceptualization, formal analysis, funding acquisition, investigation, visualization, writing–original draft. J. Chen: Formal analysis, validation, investigation, writing–original draft. W.-C. Wu: Investigation. Y.-Y. Yang: Investigation. Y.-T. Xu: Investigation. X.-J. Yu: Resources. H.-T. Chen: Resources. Z. Wang: Resources. X.-J. Wu: Resources, supervision, funding acquisition, writing–review and editing. L. Zheng: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

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 (81730044, 81502459, and 91842308), the China Postdoctoral Science Foundation (2019M653190), the Science and Information Technology of Guangzhou (201904020040), and the Sun Yat-sen University Clinical Research 5010 Program Fund (2015024).

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.

1.
Hanahan
D
,
Weinberg
RA
. 
Hallmarks of cancer: the next generation
.
Cell
2011
;
144
:
646
74
.
2.
DeBerardinis
RJ
,
Chandel
NS
. 
Fundamentals of cancer metabolism
.
Sci Adv
2016
;
2
:
e1600200
.
3.
Vander Heiden
MG
,
DeBerardinis
RJ
. 
Understanding the intersections between metabolism and cancer biology
.
Cell
2017
;
168
:
657
69
.
4.
McDonald
OG
,
Li
X
,
Saunders
T
,
Tryggvadottir
R
,
Mentch
SJ
,
Warmoes
MO
, et al
Epigenomic reprogramming during pancreatic cancer progression links anabolic glucose metabolism to distant metastasis
.
Nat Genet
2017
;
49
:
367
76
.
5.
Wang
T
,
Fahrmann
JF
,
Lee
H
,
Li
YJ
,
Tripathi
SC
,
Yue
C
, et al
JAK/STAT3-regulated fatty acid beta-oxidation is critical for breast cancer stem cell self-renewal and chemoresistance
.
Cell Metab
2018
;
27
:
136
50
.
6.
Pavlova
NN
,
Thompson
CB
. 
The emerging hallmarks of cancer metabolism
.
Cell Metab
2016
;
23
:
27
47
.
7.
Colegio
OR
,
Chu
NQ
,
Szabo
AL
,
Chu
T
,
Rhebergen
AM
,
Jairam
V
, et al
Functional polarization of tumour-associated macrophages by tumour-derived lactic acid
.
Nature
2014
;
513
:
559
63
.
8.
Kuan
EL
,
Ziegler
SF
. 
A tumor-myeloid cell axis, mediated via the cytokines IL-1alpha and TSLP, promotes the progression of breast cancer
.
Nat Immunol
2018
;
19
:
366
74
.
9.
Liao
W
,
Overman
MJ
,
Boutin
AT
,
Shang
X
,
Zhao
D
,
Dey
P
, et al
KRAS-IRF2 axis drives immune suppression and immune therapy resistance in colorectal cancer
.
Cancer Cell
2019
;
35
:
559
72
.
10.
Gabrilovich
DI
,
Nagaraj
S
. 
Myeloid-derived suppressor cells as regulators of the immune system
.
Nat Rev Immunol
2009
;
9
:
162
74
.
11.
Si
Y
,
Merz
SF
,
Jansen
P
,
Wang
B
,
Bruderek
K
,
Altenhoff
P
, et al
Multidimensional imaging provides evidence for down-regulation of T cell effector function by MDSC in human cancer tissue
.
Sci Immunol
2019
;
4
:
eaaw9159
.
12.
Wang
Y
,
Yin
K
,
Tian
J
,
Xia
X
,
Ma
J
,
Tang
X
, et al
Granulocytic myeloid-derived suppressor cells promote the stemness of colorectal cancer cells through exosomal S100A9
.
Adv Sci
2019
;
6
:
1901278
.
13.
Liang
H
,
Deng
L
,
Hou
Y
,
Meng
X
,
Huang
X
,
Rao
E
, et al
Host STING-dependent MDSC mobilization drives extrinsic radiation resistance
.
Nat Commun
2017
;
8
:
1736
.
14.
De Henau
O
,
Rausch
M
,
Winkler
D
,
Campesato
LF
,
Liu
C
,
Cymerman
DH
, et al
Overcoming resistance to checkpoint blockade therapy by targeting PI3Kgamma in myeloid cells
.
Nature
2016
;
539
:
443
7
.
15.
Waight
JD
,
Netherby
C
,
Hensen
ML
,
Miller
A
,
Hu
Q
,
Liu
S
, et al
Myeloid-derived suppressor cell development is regulated by a STAT/IRF-8 axis
.
J Clin Invest
2013
;
123
:
4464
78
.
16.
Dolcetti
L
,
Peranzoni
E
,
Ugel
S
,
Marigo
I
,
Fernandez Gomez
A
,
Mesa
C
, et al
Hierarchy of immunosuppressive strength among myeloid-derived suppressor cell subsets is determined by GM-CSF
.
Eur J Immunol
2010
;
40
:
22
35
.
17.
Welte
T
,
Kim
IS
,
Tian
L
,
Gao
X
,
Wang
H
,
Li
J
, et al
Oncogenic mTOR signalling recruits myeloid-derived suppressor cells to promote tumour initiation
.
Nat Cell Biol
2016
;
18
:
632
44
.
18.
Panopoulos
AD
,
Watowich
SS
. 
Granulocyte colony-stimulating factor: molecular mechanisms of action during steady state and ‘emergency’ hematopoiesis
.
Cytokine
2008
;
42
:
277
88
.
19.
Alshetaiwi
H
,
Pervolarakis
N
,
McIntyre
LL
,
Ma
D
,
Nguyen
Q
,
Rath
JA
, et al
Defining the emergence of myeloid-derived suppressor cells in breast cancer using single-cell transcriptomics
.
Sci Immunol
2020
;
5
:
eaay6017
.
20.
Bunse
L
,
Pusch
S
,
Bunse
T
,
Sahm
F
,
Sanghvi
K
,
Friedrich
M
, et al
Suppression of antitumor T cell immunity by the oncometabolite (R)-2-hydroxyglutarate
.
Nat Med
2018
;
24
:
1192
203
.
21.
Tyrakis
PA
,
Palazon
A
,
Macias
D
,
Lee
KL
,
Phan
AT
,
Velica
P
, et al
S-2-hydroxyglutarate regulates CD8(+) T-lymphocyte fate
.
Nature
2016
;
540
:
236
41
.
22.
Liu
Y
,
Liang
X
,
Dong
W
,
Fang
Y
,
Lv
J
,
Zhang
T
, et al
Tumor-repopulating cells induce PD-1 expression in CD8(+) T cells by transferring kynurenine and AhR activation
.
Cancer Cell
2018
;
33
:
480
94
.
23.
Molotkov
A
,
Duester
G
. 
Genetic evidence that retinaldehyde dehydrogenase Raldh1 (Aldh1a1) functions downstream of alcohol dehydrogenase Adh1 in metabolism of retinol to retinoic acid
.
J Biol Chem
2003
;
278
:
36085
90
.
24.
Canete
A
,
Cano
E
,
Munoz-Chapuli
R
,
Carmona
R
. 
Role of vitamin A/retinoic acid in regulation of embryonic and adult hematopoiesis
.
Nutrients
2017
;
9
:
159
.
25.
Cabezas-Wallscheid
N
,
Buettner
F
,
Sommerkamp
P
,
Klimmeck
D
,
Ladel
L
,
Thalheimer
FB
, et al
Vitamin A-retinoic acid signaling regulates hematopoietic stem cell dormancy
.
Cell
2017
;
169
:
807
23
.
26.
Wang
ZY
,
Chen
Z
. 
Acute promyelocytic leukemia: from highly fatal to highly curable
.
Blood
2008
;
111
:
2505
15
.
27.
Wirtz
S
,
Popp
V
,
Kindermann
M
,
Gerlach
K
,
Weigmann
B
,
Fichtner-Feigl
S
, et al
Chemically induced mouse models of acute and chronic intestinal inflammation
.
Nat Protoc
2017
;
12
:
1295
309
.
28.
Ma
HS
,
Greenblatt
SM
,
Shirley
CM
,
Duffield
AS
,
Bruner
JK
,
Li
L
, et al
All-trans retinoic acid synergizes with FLT3 inhibition to eliminate FLT3/ITD+ leukemia stem cells in vitro and in vivo
.
Blood
2016
;
127
:
2867
78
.
29.
Elstner
E
,
Muller
C
,
Koshizuka
K
,
Williamson
EA
,
Park
D
,
Asou
H
, et al
Ligands for peroxisome proliferator-activated receptorgamma and retinoic acid receptor inhibit growth and induce apoptosis of human breast cancer cells in vitro and in BNX mice
.
Proc Natl Acad Sci U S A
1998
;
95
:
8806
11
.
30.
Koshiuka
K
,
Elstner
E
,
Williamson
E
,
Said
JW
,
Tada
Y
,
Koeffler
HP
. 
Novel therapeutic approach: organic arsenical melarsoprol) alone or with all-trans-retinoic acid markedly inhibit growth of human breast and prostate cancer cells in vitro and in vivo
.
Br J Cancer
2000
;
82
:
452
8
.
31.
Bhattacharya
N
,
Yuan
R
,
Prestwood
TR
,
Penny
HL
,
DiMaio
MA
,
Reticker-Flynn
NE
, et al
Normalizing microbiota-induced retinoic acid deficiency stimulates protective CD8(+) T cell-mediated immunity in colorectal cancer
.
Immunity
2016
;
45
:
641
55
.
32.
Kuang
DM
,
Xiao
X
,
Zhao
Q
,
Chen
MM
,
Li
XF
,
Liu
RX
, et al
B7-H1-expressing antigen-presenting cells mediate polarization of protumorigenic Th22 subsets
.
J Clin Invest
2014
;
124
:
4657
67
.
33.
Wu
WC
,
Sun
HW
,
Chen
HT
,
Liang
J
,
Yu
XJ
,
Wu
C
, et al
Circulating hematopoietic stem and progenitor cells are myeloid-biased in cancer patients
.
Proc Natl Acad Sci U S A
2014
;
111
:
4221
6
.
34.
Wu
WC
,
Sun
HW
,
Chen
J
,
OuYang
HY
,
Yu
XJ
,
Chen
HT
, et al
Immunosuppressive immature myeloid cell generation is controlled by glutamine metabolism in human cancer
.
Cancer Immunol Res
2019
;
7
:
1605
18
.
35.
Fang
JH
,
Zhou
HC
,
Zhang
C
,
Shang
LR
,
Zhang
L
,
Xu
J
, et al
A novel vascular pattern promotes metastasis of hepatocellular carcinoma in an epithelial-mesenchymal transition-independent manner
.
Hepatology
2015
;
62
:
452
65
.
36.
Ding
T
,
Xu
J
,
Wang
F
,
Shi
M
,
Zhang
Y
,
Li
SP
, et al
High tumor-infiltrating macrophage density predicts poor prognosis in patients with primary hepatocellular carcinoma after resection
.
Hum Pathol
2009
;
40
:
381
9
.
37.
Ishibashi
H
,
Suzuki
T
,
Suzuki
S
,
Moriya
T
,
Kaneko
C
,
Takizawa
T
, et al
Sex steroid hormone receptors in human thymoma
.
J Clin Endocrinol Metab
2003
;
88
:
2309
17
.
38.
Li
XF
,
Chen
DP
,
Ouyang
FZ
,
Chen
MM
,
Wu
Y
,
Kuang
DM
, et al
Increased autophagy sustains the survival and pro-tumourigenic effects of neutrophils in human hepatocellular carcinoma
.
J Hepatol
2015
;
62
:
131
9
.
39.
Larange
A
,
Cheroutre
H
. 
Retinoic acid and retinoic acid receptors as pleiotropic modulators of the immune system
.
Annu Rev Immunol
2016
;
34
:
369
94
.
40.
Liu
TX
,
Zhang
JW
,
Tao
J
,
Zhang
RB
,
Zhang
QH
,
Zhao
CJ
, et al
Gene expression networks underlying retinoic acid-induced differentiation of acute promyelocytic leukemia cells
.
Blood
2000
;
96
:
1496
504
.
41.
Hardie
DG
,
Ross
FA
,
Hawley
SA
. 
AMPK: a nutrient and energy sensor that maintains energy homeostasis
.
Nat Rev Mol Cell Biol
2012
;
13
:
251
62
.
42.
Li
W
,
Tanikawa
T
,
Kryczek
I
,
Xia
H
,
Li
G
,
Wu
K
, et al
Aerobic glycolysis controls myeloid-derived suppressor cells and tumor immunity via a specific CEBPB isoform in triple-negative breast cancer
.
Cell Metab
2018
;
28
:
87
103
.
43.
Bronte
V
,
Brandau
S
,
Chen
SH
,
Colombo
MP
,
Frey
AB
,
Greten
TF
, et al
Recommendations for myeloid-derived suppressor cell nomenclature and characterization standards
.
Nat Commun
2016
;
7
:
12150
.
44.
Zhou
G
,
Myers
R
,
Li
Y
,
Chen
Y
,
Shen
X
,
Fenyk-Melody
J
, et al
Role of AMP-activated protein kinase in mechanism of metformin action
.
J Clin Invest
2001
;
108
:
1167
74
.
45.
Goransson
O
,
McBride
A
,
Hawley
SA
,
Ross
FA
,
Shpiro
N
,
Foretz
M
, et al
Mechanism of action of A-769662, a valuable tool for activation of AMP-activated protein kinase
.
J Biol Chem
2007
;
282
:
32549
60
.
46.
Gatenby
RA
,
Gillies
RJ
. 
Why do cancers have high aerobic glycolysis?
Nat Rev Cancer
2004
;
4
:
891
9
.
47.
Altman
BJ
,
Stine
ZE
,
Dang
CV
. 
From Krebs to clinic: glutamine metabolism to cancer therapy
.
Nat Rev Cancer
2016
;
16
:
619
34
.
48.
Hossain
F
,
Al-Khami
AA
,
Wyczechowska
D
,
Hernandez
C
,
Zheng
L
,
Reiss
K
, et al
Inhibition of fatty acid oxidation modulates immunosuppressive functions of myeloid-derived suppressor cells and enhances cancer therapies
.
Cancer Immunol Res
2015
;
3
:
1236
47
.
49.
Corzo
CA
,
Condamine
T
,
Lu
L
,
Cotter
MJ
,
Youn
JI
,
Cheng
P
, et al
HIF-1alpha regulates function and differentiation of myeloid-derived suppressor cells in the tumor microenvironment
.
J Exp Med
2010
;
207
:
2439
53
.
50.
Wu
C
,
Ning
H
,
Liu
M
,
Lin
J
,
Luo
S
,
Zhu
W
, et al
Spleen mediates a distinct hematopoietic progenitor response supporting tumor-promoting myelopoiesis
.
J Clin Invest
2018
;
128
:
3425
38
.
51.
Cunningham
TJ
,
Duester
G
. 
Mechanisms of retinoic acid signalling and its roles in organ and limb development
.
Nat Rev Mol Cell Biol
2015
;
16
:
110
23
.
52.
Grizotte-Lake
M
,
Zhong
G
,
Duncan
K
,
Kirkwood
J
,
Iyer
N
,
Smolenski
I
, et al
Commensals suppress intestinal epithelial cell retinoic acid synthesis to regulate interleukin-22 activity and prevent microbial dysbiosis
.
Immunity
2018
;
49
:
1103
15
.
53.
Deng
Y
,
Yang
J
,
Luo
F
,
Qian
J
,
Liu
R
,
Zhang
D
, et al
mTOR-mediated glycolysis contributes to the enhanced suppressive function of murine tumor-infiltrating monocytic myeloid-derived suppressor cells
.
Cancer Immunol Immunother
2018
;
67
:
1355
64
.
54.
Di Mitri
D
,
Toso
A
,
Chen
JJ
,
Sarti
M
,
Pinton
S
,
Jost
TR
, et al
Tumour-infiltrating Gr-1+ myeloid cells antagonize senescence in cancer
.
Nature
2014
;
515
:
134
7
.
55.
Fleming
V
,
Hu
X
,
Weber
R
,
Nagibin
V
,
Groth
C
,
Altevogt
P
, et al
Targeting myeloid-derived suppressor cells to bypass tumor-induced immunosuppression
.
Front Immunol
2018
;
9
:
398
.
56.
Guo
E
,
Wei
H
,
Liao
X
,
Xu
Y
,
Li
S
,
Zeng
X
. 
Prognostic value of alcohol dehydrogenase mRNA expression in gastric cancer
.
Oncol Lett
2018
;
15
:
5505
16
.
57.
Triano
EA
,
Slusher
LB
,
Atkins
TA
,
Beneski
JT
,
Gestl
SA
,
Zolfaghari
R
, et al
Class I alcohol dehydrogenase is highly expressed in normal human mammary epithelium but not in invasive breast cancer: implications for breast carcinogenesis
.
Cancer Res
2003
;
63
:
3092
100
.
58.
Smyth
MJ
,
Ngiow
SF
,
Ribas
A
,
Teng
MW
. 
Combination cancer immunotherapies tailored to the tumour microenvironment
.
Nat Rev Clin Oncol
2016
;
13
:
143
58
.
59.
Zou
W
,
Wolchok
JD
,
Chen
L
. 
PD-L1 (B7-H1) and PD-1 pathway blockade for cancer therapy: mechanisms, response biomarkers, and combinations
.
Sci Transl Med
2016
;
8
:
328rv4
.
60.
Costantini
L
,
Molinari
R
,
Farinon
B
,
Merendino
N
. 
Retinoic acids in the treatment of most lethal solid cancers
.
J Clin Med
2020
;
9
:
360
.