Lactate-Modulated Immunosuppression of Myeloid-Derived Suppressor Cells Contributes to the Radioresistance of Pancreatic Cancer

Pancreatic cancer is a lethal malignancy and radiotherapy plays an important role in pancreatic ductal adenocarcinoma (PDAC) treatment (1). Although modern technology has made radiotherapy an effective tool, the benefits gained from radiation in PDAC remain controversial. For resectable, borderline resectable, and locally advanced unresectable pancreatic cancer, adjuvant or neoadjuvant chemoradiotherapy provides a substantial survival benefit compared with surgery without additional therapy (2). Some randomized trials compared the effect of adding radiotherapy to chemotherapy for treatment of unresectable pancreatic carcinoma (3). Although several earlier studies do not support the use of radiotherapy (4, 5), the dose and delivery of radiotherapy were suboptimal and chemotherapy regimens were not equivalent and standard in these studies. Two clinical trials demonstrated improved survival with chemoradiotherapy compared with chemotherapy alone (6, 7), suggesting the importance of selecting appropriate regimens to acquire a benefit from radiotherapy.

Radioresistance is a major impediment to successful cancer treatment. Radioresistance can be a result of tumor microenvironment components such as hypoxia or suppressive immune factors (5, 8). Immune responses triggered in tumor microenvironments are critical for determining the outcome of radiotherapy, but the effects of radiation on the immune microenvironment of tumors are still controversial (9, 10). Although some studies suggest that radiation improves the antitumor immunity of T cells (11, 12), growing evidence indicates that radiation may be more inclined to induce a suppressive immune microenvironment, in favor of radioresistance. Therefore, a better understanding of the complex interactions between radiation and the immune microenvironment will help achieve more effective disease control using radiotherapy.

As one of the major components of the suppressive immune microenvironment, myeloid-derived suppressor cells (MDSC) inhibit antitumor immunity through multiple mediators such as Arginase-1 (ARG1), S100A8, and S100A9, and influence tumor angiogenesis through matrix metalloproteinases (MMP) and VEGF (13). Circulating MDSCs may serve as potential biomarkers for predicting the efficacy of radiotherapy (14, 15). However, little is known about the influence of radiation on MDSCs in solid tumors. Radiation enhances cytokine-induced MDSC accumulation, which is responsible for radioresistance (16, 17). In addition to regulating chemotaxis, these radiotherapy-induced factors may influence the activation and function of immunosuppressive myeloid cells. Therefore, we explored the functional alterations of MDSCs after radiation and the mechanisms involved.

The Warburg effect is a characteristic metabolic feature of cancer cells and a key factor in PDAC development (18). Accumulated evidence suggests that the Warburg effect correlates with tumor response after radiation. Enhanced glycolysis and accelerated lactate production rates are seen in radioresistant cells and radiation confers radioresistance to tumor cells through alterations in glucose metabolism (19). The Warburg effect enhances DNA damage repair to promote radioresistance (20). Knockdown of pyruvate kinase M2 (PKM2), which encodes a key glycolysis enzyme, induces radiosensitivity in cancer cells (19). These findings suggest that radioresistance may be glycolysis dependent and lactate may be important for the intrinsic radioresistance of tumor cells. There is an interactive relationship between tumor cells, especially their glycolysis, and tumor immune microenvironments (21, 22). Glycolysis in triple-negative breast cancer cells affects G-CSF and GM-CSF expression, which induces immune suppressive MDSCs (22). However, the relationship between tumor cell metabolism and immunosuppressive microenvironment in pancreatic cancer has not been elucidated, especially in the context of radiotherapy. In addition, the molecular mechanisms underlying radioresistance linked to elevated glycolysis remain incompletely understood.

In this study, we found that radiation-upregulated lactate may be responsible for improving the immunosuppressive activity of MDSCs, which inhibited the antitumor T-cell response and accelerated the progression of PDAC. In addition, HIF-1α in myeloid cells was critical for lactate-regulated activation of MDSCs. These findings uncovered some of the immune mechanisms of radioresistance in PDAC and provided potential targets for new radiotherapy strategies.

Female C57BL/6 mice were purchased from the Chinese Academy of Sciences (Shanghai, China). Conditional LSL-Kras^G12D/+ and Pdx-1-Cre strains (008179 and 014647, respectively, The Jackson Laboratory) were interbred to obtain LSL-Kras^G12D/+_; Pdx-1-Cre (KC) mice, characterized by spontaneous pancreatic neoplasia (23). Hif-1α^fl/lysM mice were generated by crossing Hif-1α^flox mice (a gift of Professor Ming Zhang, Shanghai Jiaotong University, Shanghai, China) with LysM^Cre mice (004781, The Jackson Laboratory). For the orthotopic PDAC model, 1 × 10⁶ of Panc-02-luciferase cells in 50 μL PBS were injected into the body of the pancreas by laparotomy. For the subcutaneous challenge, 5 × 10⁵ of LTPA or Panc-02 cells in 100 μL PBS were inoculated on the right flank of mice. The animals were kept in a specific pathogen-free environment. All animal experiments were approved by the Institutional Animal Care and Use Committee of Fudan University (Shanghai, China). The time interval between radiotherapy and analysis of tissue samples was 6 weeks in the KC mouse model and 15 days in the orthotopic model.

Irradiation was carried out with X-ray using an ONCOR linear accelerator (Siemens), of which the dose rate is about 3.6 Gy/minute. In the endogenous KC model and orthotopic PDAC model, mice were treated with a minilaparotomy and fiducials were placed in the peripancreatic adipose tissue or surrounding the tumor to mark the target area. A cone-beam CT scan using the Inveon PET/CT (Siemens) was employed for treatment planning. In subcutaneous PDAC model, mice were placed in individual lead boxes with a cut-out that allowed the position of tumor to be irradiated, with complete shielding of the rest of the body (24). Radiation was performed with 3 doses of 6 Gy at 48-hour intervals (25). The subcutaneous tumor volume was calculated using the formula: volume = length × width²/2. The treatment of neutralization antibody and inhibitor started 1 day before irradiation and continued until animals were sacrificed. Anti–Gr-1 neutralizing antibody was given at a dose of 100 μg/mouse every 4 days. GSK2837808A, a selective lactate dehydrogenase A (LDHA) inhibitor, was administered at a dose of 100 mg/kg daily. For bioluminescence imaging, mice received an intraperitoneal injection of luciferin (150 mg/kg, Promega) and were imaged under anesthesia with isoflurane. Images of luciferase activity were acquired using a Xenogen IVIS imaging system (Xenogen).

Human PDAC cell lines (Aspc-1, Panc-1, and SW1990) were purchased from the ATCC in 2017 and murine PDAC cell lines (LTPA and Panc-02) were generous gifts from Professor Jing Xue (Shanghai Jiaotong University, Shanghai, China) in 2017. The cells were tested to confirm the absence of Mycoplasma contamination. Panc-02 cells were transfected with a vector encoding the firefly luciferase gene (luc-pcDNA3.0) by using Lipofectamine 2000 (#11668019, Thermo Fisher Scientific) according to the manufacturer's instructions. The transfected cells screened with G418 (#10131027, Thermo Fisher Scientific) to generate stably transfected Panc-02 cell line (Panc-02-luc). PDAC cell lines were irradiated with 2, 6, or 10 Gy of X-rays (26). After 72 hours, fresh medium was added to replace the culture medium and conditioned medium (CM) was collected 24 hours later to stimulate MDSCs. Fractionation of the CM was passed through ultracentrifugal filters. The fraction that >3 kDa remained above the filter and <3 kDa passed through to the bottom chamber. The CM fraction that <3 kDa was treated with proteinase K for 1 hour at 37°C. To inhibit lactate production, tumor cells were treated with 90 mmol/L oxamic acid (Sigma) for 24 hours. In some experiments, MDSCs were treated with 5 mmol/L lactate (Sigma) for 6 hours. To inhibit various pathways, cells were pretreated with siRNA for 12 hours or inhibitors (500 nmol/L AZD8055; 50 nmol/L rapamycin; 20 μmol/L LY294002) for 1 hour.

Murine MDSCs isolated from bone marrow or tumor tissues of tumor-bearing mice by using the Myeloid-Derived Suppressor Cell Isolation Kit (#130–094–538, Miltenyi Biotec) according to the manufacturer's instructions. Tumor tissues were minced and digested with 1 mg/mL collagenase IV at 37°C for 1 hour to prepare single-cell suspensions. For isolating MDSCs, 1 × 10⁸ bone marrow cells or single-cell suspensions were centrifuged at 300 × g for 10 minutes at 4°C. The cells resuspended in 400 μL of PBS with 0.5% BSA and incubated with FcR blocking reagent for 15 minutes at 4°C. One-hundred microliters of biotin-conjugated Gr-1 antibody was added to the cell suspension and incubated for a further 15 minutes at 4°C. The labeled cells were washed by adding 10 mL of buffer and resuspended, then incubated with 100 μL of anti-biotin microbeads for 10 minutes at 4°C. Cells were washed and resuspended in 500 μL of buffer. Magnetic separation was then performed as the instructions supplied by the kit. After the magnetic separation, the isolated cells were maintained in RPMI1640, supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin for further studies. The purity of MDSCs was analyzed by flow cytometry after labeling with PE-conjugated CD11b and FITC-conjugated Gr-1 antibodies.

Murine MDSCs isolated from bone marrow of tumor-bearing mice prestimulated with 5 mmol/L lactate for 6 hours or conditioned medium (CM) for 12 hours. For the suppression assay, T cells isolated from the spleens of tumor-bearing mice by using The Pan T Cell Isolation Kit (#130–095–130, Miltenyi Biotec) according manufacturer's instructions. Purified T cells in RPMI1640/10% FBS coated by anti-CD3 and anti-CD28 antibodies were cocultured with MDSCs at a ratio of 1:4. BrdU (#000103, Thermo Fisher Scientific) was added 12 hours before detection and cells were stained with CD3-FITC antibody and BrdU-PE antibody to analyze BrdU⁺ T cells by flow cytometry as described below. The supernatants were collected to examine the IFNγ secretion. The IFNγ protein was measured by using IFNγ ELISA Kit (#88–7314–86, Invitrogen) according to the manufacturer's instructions.

For cell subgroup analysis, single cells suspended in 100 μL PBS with 0.5% BSA and incubated with FcR blocking reagent for 15 minutes at 4°C, followed by addition of proper concentration of fluorescent-conjugated antibody in 100 μL PBS/0.5% BSA for 15 minutes at 4°C. The antibodies are listed in Supplementary Table S1. The labeled cells washed and resuspended with 500 μL buffer and analyzed by flow cytometry (Celesta, BD Biosciences). The data were analyzed by FlowJo software.

Total RNAs were extracted with TRIzol (Invitrogen). Five-hundred nanograms of RNA was reverse transcribed by using PrimeScript RT Master Mix (Takara) and qPCR was performed using the Power SYBR Green PCR Kit (Takara) with the LightCycler 480 instrument (Roche). Relative expression of target gene was calculated using the comparative method by normalization to the internal control β-actin. Data were presented as fold change by normalization to control group. The relative expression of genes was calculated using the 2^−ΔΔC_t method. The primer sequences of mouse genes are shown in Supplementary Table S2.

For Western blot analysis, tissues or cells were lysed in lysis buffer containing 50 mmol/L Tris-HCl, 2% SDS, protease inhibitor cocktail (Roche), and 1 mmol/L phenylmethylsulfonylfluoride. The protein concentrations were measured by BCA Protein Assay Kit (Thermo Fisher Scientific). Equal amounts of proteins were subjected to SDS-PAGE and the membranes were blocked with PBST containing 5% nonfat skim milk for 1 hour. The membranes were incubated with primary antibodies at 4°C overnight. After washing with PBST, the membrane was incubated with a secondary antibody for 2 hours at room temperature and developed with Enhanced Chemiluminescence (ECL) Detection System (Pierce/Thermo Fisher Scientific). The antibodies used for Western blotting are listed in Supplementary Table S1. After incubation with secondary antibodies, bands were visualized using an ImageQuant LAS 4000 (GE).

Human and mouse PDAC cell lines were starved in serum-free DMEM for 6 hours prior to the indicated stimulation. The culture medium was replaced by fresh medium at different time points and collected 1 hour later to assess lactate production. Tumor tissue was homogenized and the supernatant was collected after centrifugation to detect the concentration of lactate. The L-Lactate Assay Kit (#K607), Pyruvate Colorimetric Assay Kit (#K609), and Glucose Uptake Colorimetric Assay kits (#K676, BioVision) were used according to the manufacturer's protocol.

siRNA and shRNA were synthesized by GenePharma (GenePharma). The detailed sequences are listed in Supplementary Table S3. For in vivo experiments, shRNAs were constructed in a pSIH1-H1-copGFP vector according to the previous study (27). shRNA lentiviruses were generated and transduced into LTPA or Panc-02-luc cell lines according to the manufacturer's instructions. The efficiency of knockdown was confirmed by Western blotting.

Tissues were fixed in 4% (w/v) paraformaldehyde and embedded in paraffin to prepare tissue sections. For histologic analysis, sections were cut 8-μm thick and then stained with hematoxylin and eosin (H&E). For IHC, the sections were analyzed using the avidin–biotin–peroxidase complex method. After antigen retrieval, the section was incubated with the antibodies listed in the Supplementary Table S1 at 4°C overnight followed by secondary antibody incubation (GK500705, GeneTech) according to the manufacturer's protocol. Images were acquired by using Leica Qwin Plus v3 software. The area of positive staining was measured and normalized to the total area of image by using Image-Pro Plus v6.2 Software (Media Cybernetics Inc.).

Statistical analyses were performed using GraphPad Prism 7 software. All data were presented as means ± SEM. Comparisons between two groups were performed with a two-tailed Student t test. Multiple-group comparisons were performed using one-way ANOVA followed by Tukey multiple comparisons test or two-way ANOVA followed by a Bonferroni correction to compare each group. Comparison of the survival curves was calculated by the Cox log-rank test. Correlation analysis was performed using Pearson correlation. P < 0.05 was considered statistically significant (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant).

Given the influences of MDSCs on the immunosuppressive microenvironment during tumor progression, we studied the interactional relationship between radiotherapy and MDSCs in PDAC mouse models. To validate the facilitation of radiotherapy for MDSC activation, 16-week-old KC mice with scattered early pancreatic intraepithelial neoplasia (PanIN) lesions or orthotopic PDAC mice were treated with radiotherapy (Fig. 1A). Consistent with a previous study (25), radiation-treated KC mice exhibited accelerated PDAC progression, as evidenced by higher frequencies of advanced PanIN lesions (Fig. 1B). Radiation significantly increased the expression of functional genes such as S100a8, S100a9, Arg1, and Mmps in CD11b⁺Gr-1⁺ MDSCs isolated from tumor tissues of irradiated KC mice and orthotopic PDAC mice (Fig. 1C; Supplementary Fig. S1A). Expression of tumor-promoting cytokines, such as Il10, Arg-1, and Vegf, were significantly elevated in tumor-associated macrophages (TAM) after irradiation, indicating activation of TAMs (Supplementary Fig. S1B). The Ly6G⁺ polymorphonuclear MDSC subset was the major population (Supplementary Fig. S1C). Radiation diminished the antitumor T-cell response in KC mice, as evidenced by decreased IFNγ⁺ T cells and lower IFNγ protein (Fig. 1D and E). To better characterize the impact of local irradiation and exclude the direct effect of radiotherapy on myeloid cells, we examined the infiltration of MDSCs in tumor tissues at different time points after irradiation for insights on the dynamics of myeloid cells in the orthotopic model. MDSCs continued todecline during treatment and for 2 days after radiotherapy. After that, the number of cells began to increase and reached a peak 1 week later (Supplementary Fig. S1D). In addition, depleting MDSCs with Gr-1-neutralizing antibody significantly inhibited tumor growth (Fig. 1F) and restored the antitumor T-cell response (Fig. 1G and H) in orthotopic PDAC mice, implying that MDSCs were required for radioresistance in PDAC. The efficiency of Gr-1⁺ cells' depletion was determined by flow cytometric analysis of tumor tissues (Supplementary Fig. S1E). Taken together, these results demonstrated that radiotherapy enhanced the tumor-promoting activity of MDSCs, which was essential for radioresistance in PDAC.

Tumor cell–derived soluble signals initiate signal transduction pathways that commit myeloid cells to become immunosuppressive (28). Hence, we postulated that radiation caused multiple changes within tumor cell–derived factors and further orchestrated the immunosuppressive microenvironment. To test our hypothesis, MDSCs were isolated from tumor-bearing mice and stimulated with CM collected from tumor cells that had received different doses of radiation. The purity of MDSCs was over 99% (Supplementary Fig. S2A). The immunosuppressive potency of the isolated myeloid cells was evaluated by coculturing with T cells at different ratios (Supplementary Fig. S2B). We optimized radiation dose by treating tumor cells with different doses of irradiation and selected the most effective dose (6 Gy) for in vitro experiments (Supplementary Fig. S2C). Consistent with the augmented activation of MDSCs in vivo, functional genes of MDSCs were significantly upregulated after treatment with irradiated CM (Fig. 2A). We cocultured T cells with CM-treated MDSCs and observed a significant decrease in Brdu⁺ T cells and IFNγ production, suggesting that supernatants from irradiated tumor cells enhanced the immunosuppressive activity of MDSCs (Fig. 2B and C; Supplementary Fig. S2D). These results revealed that soluble signals secreted by irradiated tumor cells might be required for improving MDSC immunosuppression.

Although inflammatory factors are important for inducing oncogenic function of MDSCs, we found that radiotherapy did not induce higher expression of Il1β, Il6, Ccl2, Ccl18, and Cxcl1/2 in PDAC mice (Fig. 2D and E). The activation of NFκB signaling was also unaffected in tumor tissues of PDAC mice (Fig. 2F). Consistent with the lack of proinflammatory signaling, radiotherapy neither induced an inflammatory phenotype and architectural alteration in the pancreases nor induced mucositis in the small intestine and colon (Supplementary Fig. S3A and S3B). Consistent with this result, radiotherapy fails to induce tumor-promoting inflammation in KC mice (25). In addition, no discrepancy in proinflammatory factor expression (Supplementary Fig. S3C and S3D) and NFκB activation (Fig. 2G) was observed in human PDAC cell lines (Aspc-1, Panc-1, and SW1990) and murine PDAC cell lines Panc-02 at 72 hours after radiation. These results indicated that radiotherapy did not induce sustained inflammation in PDAC and inflammatory factors might not have major functions in the long-term activation of MDSCs.

To ascertain the main suppressive factors that act on MDSCs in the irradiation system, we subfractionated CM from irradiated tumor cells using a 3-kDa molecular weight cut-off filter. We found that the filtrate induced immunosuppressive function in MDSCs, whereas the retentate did not (Fig. 2H and I). The filtrate retained the capability to promote the activity of MDSCs after treatment with proteinase K (Fig. 2H and I), suggesting that functional soluble factors were small molecules rather than peptides. Taken together, our results indicated that nonproteinaceous small factors derived from tumor cells might be the major effective molecules in promoting an immunosuppressive phenotype in MDSCs after radiotherapy.

The Warburg effect is a fundamental feature of tumors, which is characterized by enhanced glycolysis and increased lactate production. Therefore, we examined the alteration of lactate, the product of the Warburg effect, in the radiotherapy system of PDAC. Lactate levels were consistently elevated in tumor tissues of KC mice and orthotopic PDAC mice after radiotherapy (Fig. 3A). Radiation upregulated the expression of LDHA, which catalyzes the conversion of pyruvate to lactate, in tumor tissues of PDAC mice (Fig. 3B and C). In line with lactate and LDHA results, HIF-1α, an essential enzyme responsible for the Warburg effect, also significantly increased after radiotherapy, suggesting an enhanced Warburg effect (Fig. 3C). Consistent with the in vivo results, lactate was elevated in human and murine PDAC cell lines by radiotherapy for at least 120 hours (Fig. 3D and E). In addition, Western blots showed that expression of PKM2 and LDHA was significantly upregulated in LTPA at 72 hours after different doses of radiation (Fig. 3F). The expression of the genes for 13 critical glycolytic enzymes in the Warburg effect, including PKM2 and LDHA, all increased after radiation (Supplementary Fig. S4A).

HIF-1α is a transcriptional regulator that modulates a series of glycolysis enzymes involved in the Warburg effect (29). We found that HIF-1α was elevated in PDAC mice and LTPA after irradiation (Fig. 3C and F). Knocking down Hif-1α (Hif-1α^kd) in LTPA reduced the elevated lactate induced by irradiation (Fig. 3G), accompanied by decreased glucose consumption and pyruvate (Supplementary Fig. S4B). In addition, Hif-1α knockdown did not significantly affect tumor cell growth (Supplementary Fig. S4C). LDHA and PKM2 in irradiated LTPA also decreased after knocking down Hif-1α (Fig. 3H). The expression of Pkm2, Ldha, and 11 other glycolytic enzymes was downregulated after Hif-1α knockdown (Fig. 3I). In accordance with observations in cell lines, lactate was decreased in orthotopic PDAC mice inoculated with Hif-1α^kd Panc-02 (Fig. 3J). Genetic silencing of Hif-1α significantly reduced the tumor burden and weight in orthotopic PDAC mice (Fig. 3K). These results demonstrated that irradiation consistently upregulated the Warburg effect and lactate production both in vitro and in vivo, and HIF-1α played a role in the regulation of radiotherapy on lactate.

On the basis of the findings that irradiation boosted the Warburg effect and the function of MDSCs, we investigated whether increased lactate was responsible for activating MDSCs in radiotherapy. Lactate-stimulated MDSCs exhibited upregulated expression of tumor-promoting functional genes (Fig. 4A) and more potent immunosuppressive activity toward T cells (Fig. 4B). In addition, acidifying medium with 1% hydrochloric acid did not suppress the expression of functional genes in MDSCs, eliminating the effects of decreased pH due to accumulation of lactate and confirming the lactate-mediated activation effect (Supplementary Fig. S5). Pharmacologic blockade of lactate production by oxamic acid reversed irradiated CM-induced MDSC activation (Fig. 4C). We found that improved immunosuppressive activity of MDSCs induced by irradiated CM was impaired after Ldha knockdown in LTPA and lactate supplementation reversed the impaired activation (Fig. 4D).

We further verified that tumor cell–derived lactate promoted MDSC activation in the irradiated orthotopic PDAC model. Radiotherapy slightly promoted tumor progression in orthotopic mouse models (Fig. 4E and F). The expression of MDSC functional genes significantly increased and T-cell immunity was significantly inhibited after radiotherapy (Fig. 4G and H). Ldha knockdown resulted in significantly delayed tumor growth in PDAC mice after radiotherapy (Fig. 4E and F). Ldha deficiency in tumor cells inhibited expression of tumor-promoting genes in MDSCs and ultimately promoted the antitumor function of T cells after radiation (Fig. 4G and H). Consistent with this result, Ldha knockdown delayed tumor growth in the subcutaneous PDAC model (Supplementary Fig. S6A). LDHA blockade by treating tumor-bearing mice with an LDHA-specific inhibitor (GSK 2837808A) also significantly inhibited tumor growth and MDSC activation (Fig. 4I; Supplementary Fig. S6B) and increased the antitumor immunity of T cells in PDAC mice after radiotherapy (Fig. 4J). These results suggested that lactate was necessary for the radiation-induced immunosuppressive phenotype of MDSCs.

We elucidated the signaling pathway involved in the progress of lactate-induced activation of MDSCs. Lactate can evoke signals via its receptor, GPR81 (30). We confirmed the expression of GPR81 on MDSCs (Fig. 5A). Lactate upregulated GPR81 (Fig. 5B) and Gpr81 knockdown completely reversed the increase in tumor-promoting gene expression induced by lactate (Fig. 5C). This result suggested that GPR81 was required for the lactate-induced activation of MDSCs. HIF-1α is essential for macrophage activation and polarization induced by lactate (31, 32). Hence, we examined whether HIF-1α was also engaged in the regulation of MDSCs by lactate. HIF-1α was significantly upregulated in MDSCs after lactate treatment (Fig. 5B) and Hif-1α knockdown abrogated the lactate-induced upregulation of functional gene expression in MDSCs (Fig. 5C). Consistent with this result, the increase in tumor-promoting gene expression induced by lactate was reversed in MDSCs isolated from Hif-1α^fl/lysM mice in which Hif-1α was specifically deficient in the myeloid cell lineage (Fig. 5D). STAT3 is one of the most important transcription factors for MDSC function (33). Phosphorylated-STAT3 (p-STAT3) was significantly upregulated in MDSCs after lactate treatment (Fig. 5B), suggesting activation of MDSCs. Hif-1α siRNA inhibited elevation of p-STAT3 induced by lactate (Fig. 5E), indicating a connection between them. These results confirmed that HIF-1α mediated the lactate-regulated activation of MDSCs.

To determine the signaling pathway involved in the effects of lactate on HIF-1α, we detected the activity of the mTOR signaling pathway. mTOR regulates HIF-1α in the cell metabolism of monocyte and MDSCs (34, 35). We found elevated mTOR signaling in lactate-treated MDSCs, as shown by increased phosphorylated Akt, mTOR, and S6 (Fig. 5F), indicating activation of the mTOR pathway. PI3K inhibitor (LY294002) and mTOR inhibitor (AZD8055 and rapamycin) significantly reduced HIF-1α and p-STAT3 induced by lactate in MDSCs (Fig. 5F). Consistent with this finding, Akt siRNA or Raptor siRNA also reduced HIF-1α and p-STAT3 in lactate-treated MDSCs (Fig. 5E). Both LY294002 and rapamycin inhibited the upregulation of tumor-promoting genes induced by lactate (Fig. 5G). In addition, Gpr81 siRNA reduced p-Akt, p-S6, HIF-1α, and p-STAT3 (Fig. 5E). Consistent with in vitro findings, radiotherapy also upregulated GPR81, p-Akt, p-mTOR, p-S6, HIF-1α, and p-STAT3 in pancreatic tumor tissues of KC mice (Fig. 5H), further supporting the effects of the GPR81/mTOR/HIF-1α/STAT3 pathway on MDSC regulation in vivo. These results indicated that lactate activated MDSCs through the GPR81/mTOR/HIF-1α/STAT3 pathway after radiotherapy in PDAC.

We further corroborated the contribution of lactate-induced HIF-1α signaling to MDSC activation in radiotherapy in vivo. We generated Hif-1α^fl/lysM mice by crossing Hif-1α flox mice with Lysm-Cre to specifically delete Hif-1α in myeloid cells. Hif-1α^fl/lysM mice displayed significantly slower tumor growth after irradiation than wild-type (WT) mice (Fig. 6A and B; Supplementary Fig. S7A). Depletion of Hif-1α in myeloid cells downregulated radiation-induced tumor-promoting gene expression in tumor tissues (Fig. 6C), indicating abolished immunosuppressive activity of MDSCs. In addition, significantly increased frequencies of IFNγ⁺ and Ki67⁺ T cells, as well as higher IFNγ, were found in Hif-1α^fl/lysM groups (Fig. 6D–F). These results implied that lactate-induced HIF-1α in MDSCs inhibited the antitumor T-cell response after radiation. We also found decreased angiogenesis-related gene expression and microvessel density in Hif-1α^fl/lysM mice (Supplementary Fig. S7B and S7C). Targeting HIF-1α in myeloid cells resulted in comparable antitumor and proimmune effects as Ldha knockdown in tumor cells. Hif-1α^fl/lysM mice inoculated with Ldha^kd tumor cells resulted in little radioresistance (Fig. 6A and B), less expression of MDSC functional genes (Fig. 6C), more Ki67⁺ T cells and IFNγ⁺ T cells and higher IFNγ levels after radiotherapy (Fig. 6D–F). This finding implied that blocking the Warburg effect in tumor cells and HIF-1α signaling in myeloid cells may have additive effects in improving radiotherapy efficiency.

Finally, we analyzed the correlation between LDHA and functional genes HIF1A, S100A8, and S100A9 in patients with PDAC using The Cancer Genome Atlas (TCGA) and International Cancer Genome Consortium (ICGC) databases (36). LDHA expression positively correlated with expression of HIF1A, S100A8, and S100A9 (Supplementary Fig. S8A). Log-rank (Mantel–Cox) analysis showed that high LDHA or HIF1A expression significantly correlated with poor survival (Supplementary Fig. S8B and S8C). Data from the TCGA databases showed that high LDHA, PKM2, or HIF1A expression correlated with poorer clinical response in irradiated PDAC patients (Supplementary Fig. S8D), confirming that the Warburg effect may decide the clinical outcome of radiotherapy (Supplementary Fig. S8D).

Taken together, these data suggested that radiotherapy-induced lactate can promote tumor progression by reprogramming the antitumor T-cell response to an immunosuppressive state through activating HIF-1α signaling in MDSCs (Fig. 6G). The status of tumor metabolism, the levels of HIF-1α, and the activity of MDSCs have substantial prognostic value for the radiotherapy of human PDAC.

Despite progress in treatment delivery modes, postradiotherapy relapse caused by radioresistance still frequently occurs in pancreatic cancer (3). It is increasingly clear that multiple changes in the tumor immune microenvironment are pivotal to radioresistance (9, 10). In this study, we uncovered several new insights into the postradiotherapy immunosuppressive microenvironment. We found that radiation promoted the activation of MDSCs, which was dependent on enhanced lactate secretion. HIF-1α mediated the activity of lactate in regulating the MDSC function, which reprogrammed the tumor microenvironment into a more immunosuppressive phenotype.

We observed that lactate was highly produced and persisted for a long time after radiotherapy. Inflammatory cytokines, such as IL6, have key roles in regulating immune cell function and induce radiation tolerance and tumor relapse (16). Radiation-induced secretion of proinflammatory cytokines did not last long (37, 38). Radiotherapy fails to induce higher inflammation in KC mice (25). The expression of acute-phase proteins returned to baseline within a week in patients with oral cancer undergoing radiation (39). Consistently, the inflammatory response after radiotherapy was not associated with survival with pancreatic cancer (40). This study confirmed that radiotherapy did not induce long-term inflammation in patients with PDAC, various PDAC mouse models, or PDAC cell lines. Thus, inflammation was not a sustained phenomenon and surviving tumor cells did not have more inflammation after radiation. Other factors, such as lactate, may dominate the immunosuppressive microenvironment after radiotherapy.

Radiation leads to pronounced immune suppression by CSF-1–induced expansion of TAMs (25). We observed increased infiltration of MDSCs and TAMs in KC mice that received radiotherapy. Expression of tumor-promoting cytokines was significantly elevated in MDSCs and TAMs after irradiation. Studies showed that GM-CSF and CSF-1 were more important for promoting proliferation than maintaining the function of myeloid cells (16, 31, 32). Multiple signals such as IL6, IL1β, or IL17 are required for myeloid cells to gain immunosuppression function (41). Hence, lactate may be more responsible for reeducating myeloid cells to develop an immunosuppressive phenotype than promoting proliferation after radiation.

In addition to the regulatory role of HIF-1α in tumor cells, we found that lactate-increased HIF-1α in myeloid cells was essential for the immunosuppressive phenotype of the tumor microenvironment after radiotherapy. Previous studies showed that mTOR regulates HIF-1α (42). We demonstrated that PI3K inhibitor and mTOR inhibitor significantly reduced HIF-1α induced by lactate in MDSCs. Interfering with HIF-1α slightly reduced the activation of Akt and S6 in the context of lactate treatment, suggesting a positive feedback loop between mTOR signal and HIF-1α.

HIF-1α regulates the function and differentiation of myeloid cells under hypoxia in the tumor microenvironment (43). However, accumulating evidence indicates that HIF-1α also participates in the regulation of immune cells normally residing in normoxic environments. HIF-1α participates in regulating T-cell activation and the inflammatory response of dendritic cells (44, 45). HIF-1α–dependent glycolytic activity regulates the differentiation and functions of MDSCs in different inflammatory diseases (46, 47). Myeloid cells located in the tumor margins, which are considered to be less hypoxic than the center of tumor mass, are important for the prognosis of patients with cancer (48, 49). Hence, the upregulation of HIF-1α by radiotherapy under normoxia may be important for the immunoregulation of MDSCs and need to further attention. In this study, we found that lactate was a key mediator for HIF-1α accumulation in MDSCs after irradiation.

MDSCs are a current research focus for targeted PDAC therapy. Studies show that targeted depletion of granulocytic MDSCs in PDAC increases adaptive immunity (50, 51). In addition, blocking CXCR2, an important receptor for MDSC recruitment, enhances the effect of chemotherapy and immunotherapy in PDAC (48, 49). Because lactate produced by tumor cells contributes to MDSC-induced immune suppression, targeting lactate may be a new way to abrogate the immunosuppressive activity of MDSCs. LDHA is generally considered a safe therapeutic target and several LDHA inhibitors have been tested for their anticancer activity 52, 53). Therefore, reversing the Warburg effect by LDHA inhibitors, which further reprogram MDSCs, may provide another strategy to overcome radioresistance.

Collectively, our results demonstrated that radiation-increased lactate induced HIF-1α signaling in myeloid cells, which created a resistant microenvironment during radiotherapy. Because the radiotherapy strategy of patients with PDAC is more complex, it is difficult to find suitable patient samples to verify this hypothesis. However, our study suggested that reeducating the tumor microenvironment by modulating tumor cell metabolism and targeting inhibitory myeloid cells may result in a more effective and durable response than irradiation alone in patients with PDAC.

No potential conflicts of interest were disclosed.

X. Yang: Conceptualization, resources, data curation, formal analysis, funding acquisition, writing–original draft, writing–review and editing. Y. Lu: Data curation, methodology, writing–original draft. J. Hang: Resources, data curation, funding acquisition. J. Zhang: Resources, formal analysis. T. Zhang: Resources, formal analysis. Y. Huo: Resources, formal analysis. J. Liu: Resources, formal analysis. S. Lai: Resources, formal analysis. D. Luo: Resources, funding acquisition. L. Wang: Conceptualization, resources. R. Hua: Resources, writing–original draft. Y. Lin: Conceptualization, data curation, funding acquisition, writing–original draft, writing–review and editing.

This work was supported by Major Special Projects of the Ministry of Science and Technology (2018ZX10302207), grants from National Natural Science Foundation of China (81972627, 91942313, 81900591, 81902955, 81702844, and 81970810), Shanghai Sailing Program (19YF1449500), and Jiangsu Province Science Foundation for Youths (BK20190161). The authors acknowledge Professor Rui He (Fudan University, Shanghai, China) for her selfless help throughout the project.

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.