Metformin-induced reduction of CD39 and CD73 blocks myeloid-derived suppressor cell activity in patients with ovarian cancer

Abstract

Metformin is a broadly prescribed drug for type 2 diabetes that exerts antitumor activity, yet the mechanisms underlying this activity remain unclear. We show here that metformin treatment blocks the suppressive function of myeloid-derived suppressor cells (MDSC) in patients with ovarian cancer (OC) by downregulating the expression and ectoenzymatic activity of CD39 and CD73 on monocytic and polymononuclear MDSC subsets. Metformin triggered activation of AMP-activated protein kinase α (AMPKα) and subsequently suppressed hypoxia-inducible factor-α (HIF-1α), which was critical for induction of CD39/CD73 expression in MDSC. Furthermore, metformin treatment correlated with longer overall survival in diabetic patients with OC, which was accompanied by a metformin-induced reduction in the frequency of circulating CD39⁺CD73⁺MDSC and a concomitant increase in the antitumor activities of circulating CD8⁺T cells. Our results highlight a direct effect of metformin on MDSC and suggest that metformin may yield clinical benefit through improvement of antitumor T cell immunity by dampening CD39/CD73-dependent MDSC immunosuppression in OC patients.

Introduction

Immune suppression is mediated by factors released from the tumor or by infiltration of different immunosuppressive cells in the tumor suppressive microenvironment such as myeloid-derived suppressor cells (MDSCs)(1-4)and regulatory T cells (Tregs)(5). MDSCs expand in tumor-bearing individuals and suppress T cell responses via various mechanisms. Increased number of MDSCs have been observed in the blood, lymph nodes, bone marrow (BM) and tumor sites of patients with cancer, acute and chronic inflammation, traumatic stress, and transplantation(1, 6). In a variety of human tumor types, MDSCs are mainly defined as two subpopulations of cells, including monocytic (M-MDSC) and polymononuclear MDSC (PMN-MDSC)(7-9).Accumulation of functional MDSCs has been demonstrated to be a main contributor to tumor-induced immunosuppression(10-12).Thus, a number of therapeutic approaches to overcome MDSC-induced immunosuppression are under clinical development(13-15).However, the molecular mechanisms that regulate MDSCs in human cancer immunopathogenesis is still remain obscure.

Ectonucleotidases are localized on the tumor cells and different populations of immune cells that hydrolyze ATP or ADP into AMP by the CD39 (ectonucleoside triphosphate diphosphohydrolase-1) and further processed into adenosine by the CD73 (ecto-5′-nucleotidase) in the tumor microenvironment(16-18). We and others have demonstrated the pivotal role of CD39/CD73-mediated adenosinergic pathways in tumor growth and metastasis through regulation of antitumor T cell immunity(19-21). More recently, we show that CD39/CD73-mediated ectoenzymatic activity contributes significantly to immune suppression by both mouse(22) and human MDSCs(23).

Metformin (dimethylbiguanide), a prescribed drug for type 2 diabetes, has been reported to have anti-cancer effects and reduce the frequency and mortality of specific cancers particularly in long-term users(24, 25). Metformin has been shown capable of facilitating cell cycle arrest and inducing apoptosis. It also appears to preferentially kill cancer-initiating/stem cells via AMP-activated protein kinase (AMPK) activation(26-28). Despite the chemo-preventive activity, little is known about the effect of metformin on antitumor immune responses.

Apart from the inhibitory effect of metformin on cancer cells, we here demonstrate the direct action of metformin on MDSCs, which is linked to improvement of antitumor CD8⁺ T cell functionality and increased overall survival of patients with ovarian cancer (OC). Importantly, MDSC-mediated immune suppression is diminished by metformin treatment, depending on inhibition of CD73/CD39 expression on MDSCs through activation of AMP-activated protein kinase(AMPKα) and inhibition of HIF-1α pathway. Our study thus reveals a novel antitumor mechanism of metformin administration that may restore antitumor T cell immunity by impairing the suppressive function of MDSCs to yield clinical benefit.

Material and Methods

Donor recruitment and sample preparation

From November 2013 to November 2015, 52 patients with OC from the First Affiliated Hospital of Zhengzhou University were enrolled. The patients have not been treated with anticancer drugs, radiotherapy or surgery in the beginning of the study and have no other systemic diseases. Clinical stages were classified according to the International Union against Cancer. Peripheral blood mononuclear cells (PBMCs) were obtained from 52 patients with OC and 25 healthy donors with similar gender and age distribution, respectively. Tumor tissues were obtained from 31 patients with OC for the isolation of tumor-infiltrating leukocytes (TILs). The overall survival analysis with adjustments for confounders, including age, histological subtype, grade, BMI, smoking, type of surgery, postoperative residual disease, and chemotherapy drug delivery approaches was obtained from those OC patients with diabetes who received metformin (n=50) and didn’t (n=34) in electronic medical records from from January 2013 to December 2016 in the First Affiliated Hospital of Zhengzhou University. The peripheral blood of 18 patients with primary diabetes was measured pre- and post-metformin treatment in the department of endocrinology. Fresh human PBMCs and TILs were harvested as previously described(4). All patients gave written informed consent. The whole consent procedure was in accordance with Declaration of Helsinki and the standard defined by Institutional Review Boards of the First Affiliated Hospital of Zhengzhou University.

Antibodies and flow cytometry

The number of 1×10⁵ cells were harvested from fresh human PBMCs and TILs that were stained with anti-CD3 (PerCP-Cy 5.5 or APC-conjugated BD Bioscience), anti-CD8 (PerCP-conjugated from BD Bioscience), anti-7AAD (PerCP-conjugated BD Bioscience), anti-CD14 (APC-Cy7 or Percp-conjugated from BD Bioscience), and anti-CD11b (PE-cy7-conjugated from BD Bioscience). The following fluorescence-conjugated antibodies were also used:CD15 (Percp), CD33 (FITC and PE), HLA-DR (APC and PE), CD39 (APC), CD73 (FITC), CD124 (PE), CD56 (PE), CD19 (FITC), Ki-67 (PE), AnnexinV (APC), GranzymeB (PE), IFN-γ (APC), and Perforin (FITC) obtained from BD Biosciences (Supplementary TableS1). Among them, GranzymeB, IFN-γ and Perforin were stained intracellularly as follows: cells were first fixed with 2% parafolmaldehyde and permeabilized with 0.1% saponine–PBS buffer. Next, cells were incubated for 15min on ice with antibodies labeled with fluorochrome in the darkness. For surface assessment, cells were incubated with fluorochrome-labeled antibodies directly. In each case, isotypic control was performed. PBMCs were stained according to the manufacturer’s instructions. The appropriate isotype-matched control antibodies were purchased from BD Bioscience. Cells were analyzed using a FACS Canto II flow cytometer (BD, USA) and Diva analysis software (BD).

Cell isolation and sorting

Ovarian biopsy specimens (n=31) were cut into small pieces and digested in RPMI 1640 (Gibco, USA) supplemented with 0.25% trypsin (Gibco), 0.002% DNaseI (Gibco), and 20% fetal bovine serum (FBS, Gibco) at 37°C for 20 min. TILs within dissociated cells were filtered through a 100μm mesh and isolated by Ficoll-Hypaque density gradient centrifugation (Beijing Chemical Reagent Company, China). The mononuclear cells were washed and resuspended in medium supplemented with 10% heat-inactivated FBS for FACS analysis. PBMCs were isolated by Ficoll-Hypaque density gradient centrifugation, sequentially using the anti-CD14 and anti-CD11b MACS magnetic sorting system (Miltenyi Biotec, Germany) within 2h of sample collection. PMN-MDSCs (CD14⁻CD11b⁺) and M-MDSCs (CD14⁺CD11b⁺) were enriched according to the manufacturer’s instructions. The purity of the two subsets of cells was >95%. CD8⁺ T cells were further purified by MACS magnetic sorting system according to the manufacturer’s protocol (MiltenyiBiotec, Auburn, CA, USA) and the purity was also >95%. Then, CD8⁺ T lymphocytes were magnetically isolated, as confirmed by flow cytometry.

Quantitative RT-PCR

Total RNA from PMN-MDSCs and M-MDSCs was extracted using the TRIZOL Reagent (Invitrogen Life Technologies, CA, USA), RNA using the total of 5 μg cDNA was synthesized from and single-strand PrimeScript RT reagent kit with gDNA eraser (TaKaRa, Japan). The concentration and purity of all total RNA samples was verified using NanoDrop2000 (Thermo Scientific). The cDNA was obtained using a PrimeScript TM RT reagent kit (TaKaRa, Japan) according to instruction. Briefly, samples containing 1μg total RNA were incubated with 1μLgDNA Eraser, 2μL 5xcDNA Eraser Buffer and RNase free dH₂O at 42°C for 2 min. After adding the enzyme mix to the annealed total RNA sample, the reaction was incubated for 15min at 37°C. Quantitative real-time using SYBR Premix Ex Taq II (Takara) and assessed by outcarried PCRs were Agilent Mx3005P. Samples were amplified using the following conditions: 40 cycles of 95°C/30sec, 95°C/5sec, 60°C/30sec. Relative gene expression was determined by normalizing the expression of each target gene to GAPDH. The data were analyzed by 2-ΔΔCt. Primer sequences for all gene-specific amplifications were shown in supplementary Table S2.

CD8⁺T cell suppression assay

PMN-MDSCs (Lin⁻CD14⁻CD11b⁺CD15⁺HLA-DR^low/neg) and M-MDSCs (Lin⁻CD14⁺CD11b⁺CD15⁻HLA-DR^low/neg) purified from peripheral blood of ovarian cancer patients or healthy donors as targets were incubated in a 24-well U bottom plate, and at effector: targeted (E: T) ratios of 0:1, 1:1 and 4: 1 with autologous CD8⁺ T cells (10⁵ per well) as effector cells stimulated by 30ng/ml anti-CD3/anti-CD28 (Miltenyi Biotech). 2-Methoxyestradiol (2-MeOE2), a HIF-1α inhibitor and 120μmol/L cobalt (II) chloride (CoCl_2,Sigma-Aldrich) were added to the 24-well U bottom plate at the 0:1, 1:1, and 4:1 E: T ratios, and incubated for 24h. To examine the importance of CD39/CD73 activity, MDSCs were pretreated with pretreated with ARL67165, a CD39 inhibitor (250μmol/L, Sigma-Aldrich), and/or APCP, a CD73 inhibitor (100 μmol/L, Sigma-Aldrich) for 30 min before adding the CD8⁺T cells. To measure the function of CD8⁺T suppressed by MDSCs, CD8⁺T cells were stimulated with 1mg/ml phorbolmyfismteacetate (PMA, Sigma-Aldrich), 50ng/ml ionomycin (Sigma-Aldrich), and 5mg/ml brefeldin A solution (BFA, Biolegend) for 4h at 37°C in 5% CO₂ atmosphere prior to harvesting, then intracellular IFN-γ, GranzymeB and Perforin production (Biolegend, USA) were determined by flow cytometry.

Animal experiment

Animal protocols were approved by the Review Board of the First Affiliated Hospital of Zhengzhou University. Female BALB/c-nu mice (15-18g, 4-6 weeks old) were obtained from the animal facility (Beijing Vital River Laboratory Animal Technology Co., Ltd.) and randomly segregated into 2 groups with 3 mice in each group. The First Affiliated Hospital of Zhengzhou University provides human primary ovarian solid tumor tissues. Tumor tissue was mechanically disrupted using a scissors to generate clusters or aggregates of cells that passed through a size 50 stainless steel wire mesh. The resulting tumor-derived cell aggregates were washed once in RPMI 1640 and resuspended in PBS. Suspensions of tumor-derived cell aggregates were injected subcutaneously into the left flank of recipient BALB/c-nu mice. Each mouse received between 100 to 300 mg of the cell aggregates in a volume of 1.0ml until about two weeks. The treatment group of BALB/c-nu mice received 10mM metformin (Sigma-Aldrich) in PBS by oral gavage every day till day 35. The control mice received vehicle alone. To analyze the effects of metformin treatment, TILs were collected from the tumor tissue of experimental and control mice. The tissue from each mouse was grinded and diluted 10-fold with the RBC lysis buffer. The cells were washed with PBS followed by a second wash in FACS buffer [2% heat inactivated fetal bovine serum and 2mM ethylenediaminetetraacetic acid (EDTA)] in PBS. The cells collected from each mouse were re-suspended in equal volume of FACS buffer. Fluorophore-conjugated human specific antibodies CD3-Percp, CD14-APC-Cy7, CD11b-PE-Cy7, CD39-APC, CD73-FITC were purchased from Invitrogen (Biolegend). The samples to be analyzed were first incubated with human FcR blocking reagent at 4℃ in dark for 15min (#130-059-90, Miltenyi Biotech). The cells were then washed and appropriate antibodies were added to each set of cells and the samples were incubated on ice for 30 min in dark. The samples were then washed and re-suspended in 300μL of FACS buffer. The cell phenotype was determined using cytofluorimetric analysis by flow cytometer (BD Biosciences).

Western blot analysis

PMN-/M-MDSCs were extracted into cold lysis buffer containing 50mM Tris-HCl (pH 7.5), 150 mM NaCl, 1mM EDTA, 1mM MgCl₂, 0.5% Triton X-100, phosphatase inhibitor mix (1mM NaF, 1 mMNa₃VO₄ and 1mM β-glycerol phosphate), and protease inhibitor mix (1mM PMSF, 2μg/mL Roche (Cocktail: Inhibitor Protease aprotinin, 1μg/mL leupeptin, and 1μg/mL pepstatin A). Complete Cell lysates were clarified by centrifugation at 12,000×gcontent for 15min and subjected to SDS-PAGE (8-10% polyacrylamide gels), transferred onto PVDF membranes (Millipore, Bedford, MA). The membrane was blocked in TBS-T buffer (20mM Tris-Hcl, pH 7.5, 150mM NaCl and 0.05% Tween-20) containing 5% (w/v) non-fat milk at room temperature (22°C) for 1 h and then probed with antibodies for total and phosphorylated AMPKα (Cell Signaling Technology), total HIF-1α (Cell Signaling Technology, D2U3T) and GAPDH (Santa Cruz Biotech) at 4°C overnight, followed by incubation with horseradish peroxidase-conjugated anti-IgG for 1h at room temperature. Detection was performed with the SuperSignal West Femto Maximum Sensitivity Substrate Trial Kit (Pierce, Rockford, IL, USA). The band images were digitally captured and quantified with a Fluor Chem FC2 imaging system (Alpha Innotech, San Leandro, CA, USA).

ROS detection

Intracellular ROS were determined using a fluorescent probe DCFH-DA. Cells were pretreated with 4μM carboxy-2′, 7′-dichlorodihydrofluorescein diacetate (H2DCFDA, Sigma, USA), then stained with surface markers to monitor the ROS level in cells.

Measurement of CD39/CD73 activity by HPLC-MS analysis

To determine the effect of metformin on enzyme activity of CD39 and CD73 in MDSCs, we measured the conversion of extracellular AMP to adenosine (ADO; CD73 activity) and extracellular ATP and ADP to AMP (CD39 activity) by HPLC-MS analysis as described previously(29). In brief, purified PMN-MDSCs and M-MDSCs in the absence or presence of metformin (final concentration 2mM) were incubated at 37°C for 12h with exogenous 100μM ATP (Sigma-Aldrich) and ARL67165 (CD39 inhibitor, 250μmol/L) or 100μM AMP (Sigma-Aldrich) and APCP (CD73 inhibitor, 100μmol/L). Supernatants from the cell cultures were harvested by centrifugation at 200×g for 6 min at 4°C to remove the cells. Supernatants were filtered (0.22μm), and frozen (–20°C) until analysis via LC-MS. The chromatography was operated on an ACQUITY™ UPLC system (Waters Corp., Milford, MA, USA) with cooling autosampler and column oven. The separation was conducted on an ACQUITY™ UPLC BEH C₁₈ column (2.1mm×50mm, 1.7μm) at 40℃. The mobile phase was consisted of acetonitrile (A) and 0.1% formic acid water (B). The elution program was as follows: 0–5min, 10% A. The flow rate was 0.2mL/min. Mass spectrometric detection was carried out on a Xevo TQD triple-quadrupole tandem mass spectrometer (Waters Corp., Milford, MA, USA) equipped with an electrospray ionization (ESI) interface. Quantitative analysis was performed in selected ion monitoring (SIR) mode and the quasi-molecular ion were m/z ADO (267.70), m/z AMP (347.50), m/z ADP (427.90), and m/z ATP(507.70). CD39/CD73 activity was expressed as percent AMP conversion in this time frame.

Statistical analysis

Based on the distribution level, data were reported as mean ± SD and correlation analysis were evaluated with parametric (Independent-Sample or paired T test and Spearman’s test) or non-parametric (Wilcoxon and Spearman’s r test) tests. Significant differences (P<0.05) were determined using 1-way ANOVA, 2-tailed t-test or 1-tailed t-test, as indicated in figure legends. Outliers were detected by a Grubb’s test. Significant correlations were determined using 2-tailed (or 1-tailed where designated) Pearson correlation calculations. All statistics and graph preparations were performed using Prism 6 (Graph Pad Software Inc.).

Results

Elevated expression of functional CD39/CD73 on MDSCs in patients with ovarian cancer

The percentages of CD14⁻CD11b⁺ and CD14⁺CD11b⁺ cells were determined in Lin⁻ (CD3⁻CD56⁻CD19⁻) mononuclear cells. The PMN-MDSCs were indicated as Lin⁻CD14⁻CD11b⁺CD15⁺HLA-DR^low/negative and M-MDSCs as Lin⁻CD14⁺CD11b⁺CD15⁻HLA-DR^low/negative (Fig.S1A) as described previously(30). In line with the previous studies (4, 31), the frequencies of PMN-MDSCs (Fig.S1B) and M-MDSCs (Fig.S1C) were both higher in peripheral blood and tumor tissues of patients with OC than those in normal donors. Interestingly, the frequency of PMN-MDSCs was higher than that of M-MDSCs in OC patients, but lower in healthy donors (Fig.S1D). As expected, MDSCs from OC patients expressed higher level of typical MDSC-associated suppressive markers including TNF-α, COX-2, IL-10, arginase1 (ARG1) and TGFβ than those from healthy donors (Fig.S2A). We also confirmed that both subsets of MDSCs mediated their inhibitory effect on CD8⁺ T cell proliferation and functions(Fig.S2B).

We and others have reported that MDSCs suppress T cell immunity through the enzymatic activity of CD39/CD73 in mouse models(22, 32). To determine the role of CD39/CD73 on MDSC in human tumor immunopathology, we showed that the proportion of CD39⁺ and CD73⁺ MDSCs in peripheral blood from OC patients was significantly higher than that from healthy donors (Fig.1A and 1B). There were the markedly elevated levels of CD39/CD73 expression on both MDSCs subsets in tumor tissues compared with peripheral blood from patients or healthy donors. Furthermore, expression levels of CD39 on both MDSC subsets were positively correlated with the levels of CD73 (Fig.1C and 1D). We purified both MDSC subsets from OC patients and cocultured with autologous CD8⁺T cells in the presence or absence of CD39/CD73 enzymatic inhibitors (ARL67156 and/or APCP). As expected, the production of IFN-γ or perforin from CD8⁺ T cells was inhibited by both MDSC subsets. However, this effect was diminished by blockade of CD39 and CD73 activities using their inhibitors (Fig.1E and 1F). These findings indicate that CD39 and CD73 ectonucleotidase activities may contribute to MDSC-mediated immune suppression in vitro.

Figure 1. CD39/CD73 expressed on MDSCs in OC patients mediates MDSCs suppressive activity.

(A)The representative flow cytometry analysis of CD39 and CD73 expression on PMN-MDSCs (CD3⁻CD19⁻CD56⁻CD11b⁺CD14⁻HLA-DR^low/neg) and M-MDSCs (CD3⁻CD19⁻CD56⁻CD11b⁺CD14⁺HLA-DR^low/neg) from patients with OC. (B)The percentages of CD39 and CD73 expression on PMN-MDSCs and M-MDSCs from peripheral blood (n=42) and matched tumor tissues (n=26) of patients with OC. The peripheral blood of healthy donors (n=16) were measured as control. The correlation between the expression levels of CD39 and CD73 on PMN-MDSCs(C) or M-MDSCs(D) were established by Spearman’s rho analysis.*P<0.05. Sorted PMN-MDSCs(E) or M-MDSCs(F) from OC patients were cultured with autologous CD8⁺ T cells at a ratio of 1:4 in the absence or presence of CD39 inhibitors (ARL67156) and/or CD73 inhibitors (APCP) for 3 days (n=4). Production of IFN-γ or perforin in CD8⁺T cells was measured by flow cytometry. Data are given as means ± SD.*P<0.05.

Metformin reduces the expression and activity of CD39 and CD73 on MDSCs

Both MDSC subsets were purified from OC patients and subject to metformin effect on CD39/CD73 expression by MDSCs. We found reduced expression levels of CD39/CD73 on both MDSC subsets 24h after metformin treatment in vitro by flow cytometry (Fig.2A and 2B). Likewise, metformin treatment decreased mRNA expression of CD39/CD73 on both MDSC subsets by qRT-PCR (Fig.2C). However, there were no significant differences in expression levels of a number of inhibitory molecules including COX-2, TNFα, IL-10, TGFβ, iNOS and ARG1 and ROS production in MDSCs between the metformin treatment and control groups (Fig.S3A and 3B). We further asked whether metformin can affect the ectoenzymatic activity of CD39/CD73 on MDSCs. CD39, an ecto-apyrase that converts ATP/ADP to AMP, and extracellular AMP is metabolized to adenosine (ADO) by MDSCs-expressed CD73(29). The typical production massspectra and selected reaction monitoring chromatograms of ATP, ADP, AMP, ADO were detected by HPLC-MS using quantitative analysis in selected ion monitoring (SIR) mode (Fig.2D and 2E). As expected, metformin decreased the production of AMP from extracellular ATP and ADO from extracellular AMP in supernatants of MDSC cultures, indicating that metformin inhibited the enzymatic activity of CD39 in the conversion of ATP to AMP and CD73 activity in the conversion of AMP to ADO (Fig.2F and 2G). To examine the in vivo effect of metformin on MDSCs, we grew the human OC tissues on nude mice after subcutaneous inoculation. Consistent with the in vitro results above, we found decreased expression of CD39 and CD73 on tumor-infiltrating MDSCs 35 days after metformin treatment (Fig.2H).

Figure 2. Metformin reduces the expression and enzymatic activity of CD39 and CD73 on MDSCs.

(A)The representative flow cytometry analysis of CD39/CD73 expression on purified OC PMN-MDSCs and M-MDSCs treated with 2mM metformin or control solvents for 48h. (B)The expression levels of CD39 and CD73 on PMN-MDSCs or M-MDSCs were determined at indicated times by flow cytometry. (C)The gene expression of CD39 and CD73 in purified PMN-MDSCs and M-MDSCs 24h after treatment of metformin or control solvents was measured by qRT-PCR. Relative gene expression was determined by normalizing the expression of each target gene to GAPDH. Purified PMN-MDSCs and M-MDSCs were treated with 2mM metformin in the presence ofexogenous ATP and ARL67165 (CD39 inhibitor) or AMP and APCP (CD73 inhibitor) or control solvents for 12h. Representative HPLC-MS graphs show the peak resolution(D) and retention time(E) of metabolites of adenylate, including ATP, ADP, AMP and adenosine (ADO). The CD39 and CD73 enzymatic activity was determined by HPLC-MS analysis of extracellular ATP conversion to ADP(F) and extracellular AMP conversion to ADO(G), respectively. (H)Flow cytometry analysis of CD39/CD73 expression on human ovarian tumor-infiltrating PMN-MDSC and M-MDSC in nude mice (n=3) treated by 10mM metformin or control vehicles. Data are representative of three independent experiments. Data are given as means ± SD.*P<0.05.

Metformin modulates the expression of CD39 and CD73 on MDSCs via AMPKα and HIF-1αpathway

To search the molecular mechanism of metformin-mediated regulation of CD39 and CD73 expression on MDSCs, we first measured the expression levels of AMPKα and HIF-1α in purified PMN-MDSCs and M-MDSCs treated with metformin (AMPK activator) or Compound C (AMPK inhibitor) by qRT-PCR (Fig.3A-C) and Western blot (Fig.3D), As expected, metformin augmented AMPKα phosphorylation while reduced HIF-1α expression in both MDSC subsets in a dose-dependent manner. In contrast, treatment of Compound C reduced AMPKα phosphorylation while augmented HIF-1αexpression in both MDSC subsets in a dose-dependent manner. The data suggest that HIF-1α is likely a downstream target of AMPKα triggered by metformin.

Figure 3. Metformin inhibits HIF-1α via activation of AMPKα in MDSCs.

(A, B, C)The gene expression of AMPKα and HIF-1α in purified PMN-MDSCs and M-MDSCs from peripheral blood of OC patients was measured 24h after treatment of metformin or Compound C (AMPK inhibitors) at indicated doses using qRT-PCR. Relative gene expression was determined by normalizing the expression of each target gene to GAPDH. (D)The protein expression of total and phosphorylated AMPKα and total HIF-1α in purified PMN-MDSCs and M-MDSCs from peripheral blood of OC patients 24h after treatment of metformin or Compound C at indicated doses was measured by Western blot. Anti-GAPDH was used as a loading control. Data are representative of three independent experiments. Data are given as means ± SD.*P<0.05.

Given the pivotal role of HIF-1α for induction of CD73/CD39 expression, we thus tested if metformin modulated CD73/CD39 expression on MDSCs via AMPKα and HIF-1α pathways. Metformin treatment decreased the expression of CD39 and CD73 on both MDSC subsets, but the effect was abrogated by addition of Compound C (Fig.4A). To examine the role of HIF-1α in this process, we confirmed the increased HIF-1α expression 24h after CoCl₂ induced-hypoxia but attenuated HIF-1α expression by addition of HIF-1α inhibitor 2-MeOE2 in purified MDSCs using Western blot (Fig.4B) and qRT-PCR (Fig.4C). In line with these results, we showed that treating MDSCs with CoCl₂ increased the expression of CD39 and CD73 by qRT-PCR (Fig.4D) and flow cytometry (Fig.4E), while the effect was abrogated by addition of either 2-MeoE2 or metformin (Fig.4D and 4E). Moreover, compared to metformin treatment alone, CoCl₂ rescued the inhibitory effect of metformin on CD39/CD73 in MDSCs(Fig.4D and 4E). Taken together, these data suggest that metformin decreases the expression of CD39 and CD73 on MDSCs through activation of AMPKα and inhibition of HIF-1α pathway.

Figure 4. Metformin reduces the expression of CD39 and CD73 to limit the suppressive activity of MDSCs via AMPKα/HIF-1α pathway.

(A)Flow cytometry analysis of CD39 and CD73 expression on purified PMN-MDSCs and M-MDSCs from peripheral blood of OC patients with OC in the absence or presence of 2mM metformin and/or 100μMCompound C for 24h. Expression levels of HIF-1α in purified PMN-MDSCs and M-MDSCs were measured by Western blot(B) and qRT-PCR(C) 24h after treatment of CoCl₂ and/or 2-MeoE2 (HIF-1α inhibitor) in vitro. Expression of CD39 and CD73 on purified PMN-MDSCs and M-MDSCs treated with reagents as indicated for 24h was analyzed by qRT-PCR(D) and flow cytometry(E). Data (n=5) are given as means ± SD.*P<0.05.

Metformin inhibits MDSC suppressive activity by decreasing CD39 and CD73 expression onMDSCs

We have demonstrated that ectoenzymatic activity of CD39/CD73 on MDSCs contributes to MDSC suppressive function (Fig.1E and 1F)(22, 23). Given the ability of metformin to inhibit CD39/CD73 expression and enzymatic activity of MDSCs, we tested if metformin could impair MDSC suppressive activity by decreasing CD39/CD73 expression on MDSCs. Indeed, we found that metformin treatment diminished the suppressive effect of both MDSC subsets on production of IFN-γ(Fig.5A,B), Granzyme B(Fig.5A,C)or perforin (Fig.5A,D)from CD8⁺ T cells in vitro. As metformin modulated CD73/CD39 expression on MDSCs via AMPKα and HIF-1α pathways (Fig. 5), we next examined the involvement of these pathways in the mechanism of MDSC-mediated suppressive activity towards T cells. Treatment of MDSCs with either CoCl₂ or Compound C alone potentiated the inhibitory effect of MDSCs on CD8⁺ T cell responses, while the effect was diminished by addition of metformin(Fig.6A). Notably, compared to metformin treatment alone, addition of CoCl₂ or Compound C compromised the inhibitory action of metforminon on MDSC-mediated suppressive activity, consistent with the results on their regulation on CD39/CD73 expression on MDSCs HIF-1α expression. We further confirmed the role of HIF-1α in MDSC-mediated suppressive activity. As shown in Fig. 6B, C, addition of HIF-1α inhibitor 2-MeoE2 also significantly attenuated the suppressive effect of both MDSC subsets in the presence of CoCl₂ on production of Granzyme B and perforin from CD8⁺ T cells in vitro in a dose-dependent manner. In addition, we detected higher expression levels of HIF-1α mRNA in ovarian tumor tissues than matched nontumor ovarian tissues by qRT-PCR (Fig.6D). There was also elevated HIF-1αexpression in the purified two subsets of MDSCs from OC patients compared to that in healthy donors (Fig. 6E)

Figure 5. Metformin diminishes MDSC-mediated immune suppression in vitro.

PMN-MDSCs or M-MDSCs sorted from OC patients were cultured with autologous CD8⁺T cells at a ratio of 1:4 in the presence of 2mM metformin or control solvents for 24h (n=5). (A)The representative flow cytometrydotplotsshowingCD8⁺T cells secretory function and the summary graph showing production of IFN-γ(B), Granzyme B(C) or perforin(D) in CD8⁺ T cells was measured by flow cytometry. Data (means ± SD) are representative of three independent experiments. *P<0.05.

Figure 6. MDSC-mediated immune suppression depends on AMPKα/HIF-1α pathway.

(A)PMN-MDSCs or M-MDSCs sorted from OC patients were cultured with autologous CD8⁺T cells at a ratio of 1:4 in the absence or presence of 2mM metformin with or without 100μM Compound C and 120μM CoCl₂ for 24h, the summary graph showing production of IFN-γ, Granzyme B and perforin in CD8⁺ T cells was measured by flow cytometry. The flow cytometry analysis of CD8⁺T cells producing Granzyme B, or perforin cultured with/without MDSCs at the indicated ratios for 24h. The percentage of Granzyme B⁺CD8⁺T cells(B), or perforin⁺CD8⁺ T cells(C)was summarized (n=5). (D)The relative expression of HIF-1α from ovarian tumor tissues (n=26) and ovarian cyst tissues (n=18) was measured by qRT-PCR. (E)The relative expression of HIF-1α in purified PMN-MDSCs and M-MDSCs from peripheral blood of OC patients (n=16) versus healthy donors (n=13) was analyzed by qRT-PCR. Data are given as means ± SD. *P<0.05.

To analyze the in vivo effect of metformin on MDSCs, expression levels of CD39/CD73 on MDSCs were compared by flow cytometryin OC patients with diabetes between 2 groups before and after treatment of metformin. Consistent with the results described above, the percentages of CD39⁺or CD73⁺MDSCs were decreased in patients receiving metformin in this window of opportunity study, randomized trial (Fig.7A and 7B). In support further, we found increased production of Granzyme B, or perforin from CD8⁺ T cells in these patients after metformin treatment compared to before treatment (Fig.7C and 7D). Notably, overall survival of diabetic ovarian cancer patients receiving metformin treatment (n=50) was longer than those not receiving metformin (n=34, p=0.0068) (Fig.7E and Supplementary TableS3). These results suggest a possibility that metformin may yield clinical benefit from improvement of antitumor T cell immunity at least in part by dampening CD39/CD73-dependent MDSC suppressive activity in OC patients.

Figure 7. Metformin treatment decreases expression of CD39 and CD73 on MDSCs and increases CD8⁺T cell effector function in patients with diabetes.

(A, B)Expression of CD39 and CD73 on PMN-MDSCs or M-MDSCs from peripheral blood of patients with primary diabetes (n=18) was measured pre- and post-metformin treatment by flow cytometry. (C, D)Production of perforin, or Granzyme B in CD8⁺ T cells from peripheral blood of patients with primary diabetes (n=18) was measured pre- and post-metformin treatment by flow cytometry. Data are given as means ± SD. *P<0.05; ** P<0.01; *** P<0.001. (E)Metformin intake was associated with the improved survival in OC patients with diabetes. Overall survival was compared between OC patients with diabetes received metformin (n=50) and those unreceived (n=34). P=0.0068.

Discussion

Although a recent study has reported a potential role of metformin for the differentiation and immunosuppressive function of mouse MDSCs in vitro(33), to the best of our knowledge, our study is the first to demonstrate that metformin may be involved in the regulation of human MDSCs-mediated immunesuppression in vitro and in vivo. In this report, using human OC as a model system, we explored the direct effect of metformin on MDSCs suppressive activity. Our results indicate that in addition to tumor-intrinsic effects, metformin also enhances CD8⁺ T cell activities by counteracting the immunosuppressive activity of MDSCs, which may facilitate tumor development and immune evasion.

Numerous preclinical and clinical studies have demonstrated direct anti-tumor activities for metformin. However, the precise causes and extent of metformin’s antitumor effect remain elusive. Although metformin likely has pleiotropic effects on a wide array of cellular functions, its effect on immune system serves as an important proof-of-concept study. In particular, the role of metformin in tumor immunity is only beginning to be appreciated(34). There are several recent studies showing that metformin can promote CD8⁺memory T cell generation(35, 36) and protect them from apoptosis and exhaustionin the tumor using experimental mouse tumor models (37). It is now becoming clear that metformin can directly act on T cells. Indeed, administration of metformin inhibited the Th1- and Th17 cell-mediated immune responses in mice(38, 39). Metformin was found further to attenuate Th1- and Th17-derived IL-22 production and suppress tumor growth in an orthotopic mouse model of hepatocellular carcinoma(40).It will therefore be interesting to test whether metformin could exert dual effects on T cells and MDSCs to promote antitumor T cell immune responses. Moreover, the possibility that metformin might affect other immune cells (e.g. NK cells and dendritic cells) warrants further investigation.

Human MDSCs comprise a heterogeneous population of myeloid cells with the ability to suppress immune responses that have been poorly characterized (2, 41-43). We have previously reported that a distinct fraction among both PMN- (CD14⁻CD11b⁺) and M-MDSCs (CD14⁺CD11b⁺) that coexpresses functional CD39/CD73, linking their immunosuppressive and chemo-protective effects to disease progression in patients with non-small cell lung cancer(23). In line with results, we found that expression and enzymatic activity of CD39/CD73 on MDSCs inhibit CD8⁺T cells immunity similarly in OC patients, indicating the importance of CD39/CD73 in MDSC-mediated immune suppression is not limited to a specific type of cancer.

Although the mechanisms of regulation of CD39/CD73 expression on human MDSCs are largely unknown, the present study discovers a direct connection between metformin and the CD39/CD73-mediated adenosinergic effect to reverse MDSCs suppressive function. Metformin as the most widely used antidiabetic drug emerges recently as an anticancer agent via AMPK activation that is a major energy sensor of the energetic status of the cell, which may affect the metabolism of adenylates including ATP, ADP, AMP. This led us to identify possible cross-talk between metformin-induced AMPK activation and adenosingeric effect in MDSCs. In this study, we provide the new evidence that metformin is able to limit adenosine generation from ATP, ADP and AMP through reducing the MDSC expression and ectoenzyme activity of CD39/CD73. In an effort to reveal the molecular basis for metformin-mediated inhibitory effect on MDSCs activity, our data support a model in which metformin inhibits expression and activity of CD39/CD73 through activation of AMPKα and inhibition of HIF-1α pathway to impair MDSCs immunosuppressive activity, thereby improving antitumor T cell immune responses. Consistent with this notion, AMPK downstream-activated signaling pathways were found to be implicated in the maturation of bone-marrow cells to MDSCs (44). Furthermore, metformin-mediated AMPK activation also inhibited hypoxia-induced HIF-1αexpression in cancer cells(45-47). However, the role of AMPK activation in reducing HIF-1α expression seems to depend on the cell type. For example, hypoxia induces HIF-1 expression in mouse embryonic fibroblasts independent of AMPK(48). Tumor-infiltrating T cells from metformin-treated animals had similar levels of HIF-1α, even though they experienced less hypoxia. Previous studies reported the role of HIF-1α in regulating function and differentiation of MDSCs in the tumor microenvironment(49-51). Our present data thus build on this platform, suggesting that metformin may limit the function and fate of MDSCs in tumor-bearing hosts by orchestrating the CD39/CD73-dependent adenosinergic effect via AMPK/HIF-1α pathway.

Depletion of MDSCs or inhibiting MDSCs function mitigates cancer progression (52, 53).Our results indicate that MDSCs are likely one primary cellular target of metformin, consistent with previous results showing a selective, inhibitory effect of phenformin (a close analog of metformin) on MDSCs-driven immune suppression(54). Our data add to an increasing number of reports(22, 55-57) suggesting the immunomodulatory effects of chemopreventive reagents on MDSCs as adjuvants for cancer therapy. Along these lines, metformin may serve as one of potentiators of antitumor immunity. Thus, we hypothesize that many forms of immunotherapy that aim to act by reinvigorating T cells at the tumor site will be improved through modulation of this aspect of the microenvironment. This is supported by a report demonstrating improved intratumoral T-cell function and tumor clearance by combination of metformin with PD-1 checkpoint blockade in mouse tumor models(58). However, future studies will be required to understand how checkpoint blockade, stimulatory agonistic antibody therapy, or adoptive cell therapy might be improved by metformin treatment.

In this study, we highlight a new application for metformin as a potent immunomodulatory agent, which overcomes a major mechanism of tumor immune evasion by targeting CD39/CD73 on MDSCs. These data broaden our understanding about the action of metformin in cancer treatment, and provide a rational basis for future clinical testing of administration of metformin in combination with cancer immunotherapeutics.

Significance.

The antitumor activity of an anti-diabetes drug is attributable to reduced immunosuppressive activity of myeloid-derived tumor suppressor cells.