Tumor microenvironment following Gemcitabine treatment favors differentiation of immunosuppressive Ly6C^high myeloid cells

Abstract

Regulations of myeloid-derived suppressor cells (MDSC) by ongoing inflammation following repeated chemotherapy remain elusive. Here, we show that multi-dose clinical regimen of Gemcitabine (GEM) treatment enhances the immunosuppressive function of monocytic (M)-MDSC although tumor development is delayed in E0771 tumor-bearing mice. Accordingly, effector IFN-γ-producing CD4 and CD8 T cells significantly decrease in the tumor microenvironment (TME) of GEM-treated mice. The conditioned medium (CM) of GEM-treated tumor cells enhances differentiation of mouse BM cells and human PBMC into immunosuppressive M-MDSC. Cytokine profiling of GEM-treated tumor cells identifies GM-CSF as one of the most differentially expressed cytokines. Blockade or knockdown of GM-CSF can partially reduce immunosuppression of Ly6C^high cells induced by GEM-CM. Knockdown of GM-CSF in tumor cells also delays tumor progression with decreased accumulation of M-MDSC in TME. Mechanistically, enhanced production of reactive oxygen species (ROS) and activation of NF-κB are observed in GEM-treated tumor cells. Treatment with the mitochondrial targeted antioxidant and inhibitor of NF-κB signaling can abrogate GEM-induced hyperexpression of GM-CSF in E0771 cells. In addition, the phagocytic clearance of apoptotic tumor cells (efferocytosis) enhances the immunosuppressive function of BM Ly6C^high myeloid cells. Further, GEM treatment results in metabolic changes in residual tumor cells leading to the resistance to T-cell mediated killing. Together, our results define an undesired effect of repeated GEM treatment promoting immunosuppression in TME via upregulation of GM-CSF and efferocytosis as well as deregulation of lipid metabolism in residual tumor cells.

Introduction

Gemcitabine is a pyrimidine antimetabolite which has been wildly used as an anti-cancer chemotherapeutic agent for various solid tumors including pancreatic cancer and metastatic breast cancer (1, 2). However, chemoresistance still remains a major hurdle to successful therapy. Recent advances in tumor immunology have demonstrated the importance of extrinsic mechanisms in chemoresistance, which include the interplay between tumor cells and immune or stromal cells in the tumor microenvironment (TME) (3–5). An effective immune system makes a crucial contribution to the anti-tumor effects of chemotherapy whereas immunosuppressive environment promotes chemoresistance.

Chemotherapy also exerts immunostimulatory or immunosuppressive effects depending on different drugs, doses, schedule of administration, and cancer types (5–7). Previous studies have demonstrated that one-time treatment of Gemcitabine (GEM) or 5-Fluorouracil (5FU) can selectively deplete myeloid-derived suppressor cells (MDSC) in vivo (8–10). However, it is reported that GEM and 5-FU treatment decreases anti-cancer efficacy by activating the inflammasome pathway in MDSC (11). Therefore, a decrease in the number of MDSC does not necessarily reduce their immunosuppressive function. In addition to the direct effect of GEM on MDSC, certain chemotherapeutic agents, such as doxorubicin (DOX), cisplatin, and GEM, also result in the accumulation of immunosuppressive MDSC and M2 macrophages by modifying the TME. Several soluble factors, including IL-34, IL-6, prostaglandin E2, GM-CSF, IL-8, and extracellular vesicles have been shown to play important roles in promoting immunosuppression following chemotherapy (12–16). However, the underlying mechanisms of such responses have not been well defined.

Reactive oxygen species (ROS) are mainly generated inside mitochondria and able to oxidize biological molecules including DNA, proteins, and lipids. Mitochondria ROS (mtROS) have a dual role and contradictory effects in cancer (17, 18). Many studies have demonstrated that ROS may promote tumorigenesis and survival by triggering activation of transcription factors. On the other hand, anti-cancer effect of certain chemotherapy is due to the induction of oxidative stress and ROS-mediated cell injury (18, 19). In addition, ROS also acts as signal-transducing molecules that drive inflammation via production of proinflammatory cytokines (20–22). Emerging studies also reveal that residual tumor cells following chemotherapy promote chemoresistance and tumor recurrence. These cells exhibited elevated ROS and oxidative phosphorylation (OXPHOS) as well as altered lipid metabolism (23, 24). The mechanisms by which chemotherapy induce mtROS in tumor cells leading to ongoing inflammation and promotion of immunosuppression in TME remain to be explored.

In the present study, we used triple negative breast cell lines and demonstrated that repeated GEM treatment promotes the immunosuppressive activity of tumor Ly6C^high monocytic (M)-MDSC through the up-regulation of GM-CSF expression in residual tumor cells and phagocytosis of apoptotic tumor cells (efferocytosis). We found that enhanced production of mtROS and activation of NF-κB lead to hyperproduction of GM-CSF by tumor cells in response to GEM treatment. In addition, GEM treatment resulted in deregulation of lipid metabolism which was associated with decreased sensitivity to T cell-mediated tumor cell killing. These findings reveal an undesired effect of repeated GEM treatment promoting immunosuppression in TME, which might hinder the efficacy of GEM treatment and contribute to extrinsic chemoresistance.

Materials and Methods

Mice and tumor cells

C57BL/6J mice and Ovalbumin (OVA) T-cell receptor (TCR) Tg OT-II mice were purchased from the Jackson Laboratory. Rag2 deficient OVA TCR Tg OT-I mice were purchased from Taconic Biosciences. All animals were maintained under specific pathogen-free conditions and handled in accordance with the protocols approved by the Institutional Animal Care and Use Committee of the University of Louisville. Two TNBC cell lines E0771 (mouse) and MDA-MB-231 (human) were cultured in the complete DMEM medium containing 10% FBS. GM-CSF- and ICAM-1-knockdown E0771 cells were generated using mouse GM-CSF and ICAM-1 CRISPR plasmids (Santa Cruz Biotechnology). Mouse lymphoma cell lines EL4 and E.G7-OVA (derivative of EL4) were cultured in the complete RPMI 1640 medium. To generated GEM-resistant tumor cells, E0771 and E.G7-OVA cells were exposed to gradually increasing GEM concentrations starting at 1μM and 0.1μM, respectively. Prior to experiments, the GEM-resistant cells were maintained in drug-free medium for at least one week.

Tumor model and chemotherapy in vivo treatment

To establish subcutaneous (s.c.) tumor, 1×10⁶ E0771 tumor cells were suspended in PBS and inoculated s.c. in the flank of female B6 mice. Tumor sizes were measured twice a week with a caliper. When subcutaneous tumors reached a diameter between 5–6 mm, mice were intraperitoneally treated with GEM (60 mg/kg body weight) 4 times on days 10, 14, 18, and 22 or a single treatment on day 22. Tissues were collected two days later after last treatment.

In vitro cell differentiation and tumor M-MDSC purification

To examine the effects of tumor cell-derived factors on BM Ly6^high myeloid cell and human monocyte differentiation, we collected conditioned medium (CM) from GEM-treated E0771 and MDA-MB-231 cells as described previously (14). Briefly, E0771 and MDA-MB-231 cells were exposed to GEM (3–10 μmol/L) for 60 minutes followed by washing with PBS, and cultured in fresh medium for 3 days. Culture supernatants were collected as CM used in the consequent experiments. For in vitro cell differentiation assay, BM cells from naïve mice or human CD14⁺ monocytes were cultured in complete RPMI 1640 medium containing 25% CM from E0771 cells or MDA-MB-231cells for 6 days. Half of media was replaced every other day. On day 6, phenotype of differentiated cells was evaluated by flow cytometry and quantitative RT-PCR. Immunosuppression assay of BM Ly6C^high cells was performed by co-culturing with splenocytes from OT-I or OT-II mice in the presence of OVA. Immunosuppressive function of differentiated human monocytes was measured by co-culturing with autologous T cells in the presence of anti-human CD3/CD28 beads (Thermo Fisher Scientific). In some experiments, cytokines in the CM were neutralized using anti-mouse GM-CSF (clone MP1–22E9; BioLegend), anti-mouse CD54 (clone YN1/1.7.4; BioLegend), anti-human GM-CSF (clone BVD2–23B6; BioLegend). For tumor M-MDSC purification, exercised tumors were minced into small pieces and digested with medium containing 0.5 mg/ml collagenase A, 0.2 mg/ml hyaluronidase (type V), and 0.02 mg/ml DNase I for 20 min on a rotating platform. The M-MDSC cells (CD11b⁺Ly6C⁺Ly6G⁻) from tumors were sorted by using BD FACS Aria III. Fixable viability dye was used to exclude dead cells.

Antibodies and flow cytometry

Single-cell suspensions were blocked in the presence of anti-CD16/CD32 at 4°C for 15 min and stained on ice with the appropriate antibodies and isotype controls in PBS containing 1% FBS. The fluorochrome-labeled antibodies against mouse CD11b, Ly6G, Ly6C, PD-L1, IDO1, MerTK, CD4, CD8, IFN-γ, and their corresponding isotype controls were purchased from BioLegend. Fixable Viability Dye eFluor™ 780 was from Thermo Fisher Scientific. Annexin V Apoptosis Detection Kit was from BD Pharmingen. Cleaved Caspase-3 (Asp175)-PE was from Cell Signaling. For intracellular staining, the cells were fixed and permeabilized following surface staining. The samples were acquired using FACSCanto cytometer (BD Bioscience) and analyzed using FlowJo software.

Cytokine array and ELISA

The cytokine/chemokine profile of tumor CM was determined using proteome profiler mouse cytokine array kit (R&D Systems). The expression levels of cytokine/chemokine were measured by pixel density of each dot using the ImageJ software. Mouse and human GM-CSF kits were purchased from BioLegend.

Confocal analysis and Western blotting

For imaging assay of NF-κB p65 nuclear translocation, cells cultured on glass coverslips were fixed with 3.7% paraformaldehyde (15 min, room temperature) and washed with Tris-buffered saline (TBS). Coverslips were incubated with blocking buffer for 30 min at room temperature and anti-NF-κB p65 antibody (Cell Signaling) overnight at 4°C. The coverslips were washed and then incubated with PE-anti-Donkey Ig 1 h at room temperature. Nuclei were counterstained with nuclear stain DAPI (Sigma Aldrich) for 10 min. The fluorescence images were captured on confocal microscope. For Western blotting analysis, whole cell lysates were prepared using Triton X-100 lysis buffer containing protease and phosphatase inhibitors. The total protein lysates were resolved using 10% SDS-PAGE and then transferred to polyvinylidene difluoride membrane (GE Healthcare). Membranes were probed with primary antibodies against target molecules followed by reaction with secondary antibodies conjugated to horseradish peroxidase for appropriate incubation time. Antibodies against phospho-Stat5, stat5, phosphor-NF-κB p65, NF-κB p65, phospho-p70 S6 kinase, p70 S6 kinase, phospho-4E-BP1, 4E-BP1 were purchased from Cell Signaling Technology. The blots were developed with ECL Plus Western blotting detection reagents (GE Healthcare).

Quantitative real-time PCR

For qRT-PCR, RNA was extracted with TRIzol reagent (Invitrogen) and transcribed to cDNA with a reverse transcription kit (Bio-Rad). Quantitative real-time PCR reactions were performed using SYBR Green Supermix (Bio-Rad) with the relevant primers. We normalized gene expression level to β2-microglobulin (β-MG) housekeeping gene and represented data as fold differences by the 2^−∆∆Ct method (25). The primer sequences used for real-time PCR were as follows: Arg1: F: 5’-TTTTAGGGTTACGGCCGGTG-3’ and R: 5’-CCTCGAGGCTGTCCTTTTGA-3’; iNOS: F: 5’-GGGACTGAGCTGTTAGAGACAC-3’ and R: 5’-CCAAATCCAACGTTCTCCGT-3’; FASN: F: 5’-AAGCAGGCACACACAATGG and R: 5’-AGTGTTCGTTCCTCGGAGTG-3’; CPT1A: F:5’-GCATAAACGCAGAGCATTCC-3’and R: 5’-TCCATCCTCTGAGTAGCCCA-3’; VEGFA: F:5’-TTACTGCTGTACCTCCACC-3’ and R: 5’-ACAGGACGGCTTGAAGATG-3’.

Efferocytosis assay and differentiation of BM Ly6C^high cells with apoptotic tumor cells

To induce apoptosis, E0771-GFP tumor cells were treated with GEM (10 μM) for 48 h. The efferocytosis assay was performed by co-culture of apoptotic cells and BM-derived Ly6C⁺ cells (1:1 ratio) for 2 hours and analyzed using flow cytometry. The phagocytosis of apoptotic cells by Ly6C⁺ cells was further confirmed by Amnis ImageStream Flow Cytometer. For in vitro differentiation assay, BM cells were cultured in the presence of E0771 CM for 4 days prior to adding apoptotic cells. BM-derived Ly6C^high cells was purified after additional 48 h of incubation.

Measurement of ROS and lipid droplets

For quantification of intracellular ROS levels, cells were loaded with 1 μM cell permeant reagent 2’,7’-dichlorofluorescin diacetate (DCFDA, Abcam) or 5 μM MitoSOX Red (Thermo Fisher Scientific) for 15 min at 37 °C in PBS. For lipid droplet staining, cells were stained with BODIPY™ 493/503 (Thermo Fisher Scientific) for 15 min at 37 °C. All samples were then stained with viability dye for 20 min at 4 °C. After washing the fluorescence was recorded by flow cytometry.

In vitro T cell-mediated cytotoxicity assay

Spleen cells of OT-I mice were cultured with 25 μg/ml of OVA for 48 hours. CD8⁺ T cells were sorted as effector cells. For 6-hour LDH release assays, 1×10⁴ tumor cells (target cells) were incubated with effector T cells in 96-well round-bottom plates at the E/T ratio of 40 in 200 μl culture medium. The supernatants were harvested and T cell cytotoxicity was measured using LDH cytotoxicity assay kit (Thermo Scientific). The apoptosis of target cells was determined using Annexin V/7-AAD staining. The cleaved Caspase-3 in target cells was measured using antibody of Cleaved Caspase-3 (Asp175)-PE (Cell Signaling).

Statistical analysis

Data were analyzed using GraphPad Prism5.0 software (GraphPad Software). An unpaired Student t test was used to calculate significance. All graph bars are expressed as mean ± SEM. Significance was assumed to be reached at p < 0.05. The p values were presented as follows: *p<0.05, **p<0.01, and ***p<0.001.

Results

Multiple dose of GEM treatment enhances the immunosuppressive activity of M-MDSC although tumor development is delayed

Previous studies have demonstrated that one-time in vivo treatment of 5-FU and GEM results in a decrease of MDSC (8–10). These studies used Gr-1⁺CD11b⁺ to define MDSC. However, MDSC consist of two subpopulations: polymorphonuclear (PMN-MDSC, CD11b⁺Ly6G⁺Ly6C^low) and monocytic (M-MDSC, CD11b⁺Ly6G⁻Ly6C^high). M-MDSC are more suppressive than PMN-MDSC on a per cell basis (26). In addition, chemotherapy in clinic typically involves a number of cycles. In order to better understand the effects of chemotherapy on the differentiation and immunosuppressive function of MDSC in a clinically relevant scenario, murine breast cancer E0771-bearing mice were injected 4-times of GEM during two weeks or a single dose of GEM. As shown in Fig. 1A, multiple dose of GEM treatment significantly delayed tumor progression. Consistent with previous studies, a single dose of GEM resulted in a decrease of M-MDSC in spleens and tumors. However, multiple dose of GEM treatment led to the accumulation of spleen M-MDSC and no reduction of tumor M-MDSC (Fig. 1B). Expression of PD-L1 and indoleamine 2,3 dioxygenase 1 (IDO1) was significantly elevated on M-MDSC upon GEM treatment (Fig. 1C). Consequently, M-MDSC from mice treated with multiple dose of GEM displayed stronger immunosuppressive function for T cell proliferation (Fig. 1D).

Figure 1.

GEM treatment enhances the immunosuppressive activity of M-MDSC. (A) Female B6 WT E0771 tumor-bearing mice (n=6 per group) were intraperitoneally treated with GEM (60 mg/kg body weight) 4 times on days 10, 14, 18, and 22 or a single treatment on day 22. The tumor size was measured. (B) The mice were sacrificed 48 hours later after the last treatment. The percentages of spleen Ly6C^high M-MDSC within CD45⁺ population and tumor M-MDSC within CD11b⁺ population were summarized. Each dot represents a single mouse. (C) Expression of PD-L1 and IDO1 in tumor M-MDSC was evaluated by flow cytometry. (D) Tumor M-MDSC (CD11b⁺Ly6C^highLy6G⁻) were co-cultured with CFSE-labeled OT-I splenocytes and OVA (25 μg/ml) for 3 days. The CD8⁺ T cell proliferation was measured by flow cytometry. (E) The percentages of spleen CD4⁺ and CD8⁺ T cells within CD45⁺ population. The data are pooled from two experiments (n = 10–11 mice per group). (F) Intracellular IFN-γ production by spleen CD4⁺ and CD8⁺ T cells. The data are pooled from two experiments (n = 10–15 mice per group). (G) The percentages of tumor infiltrating CD4⁺ and CD8⁺ T cells within CD45⁺ population. (H) Intracellular IFN-γ production by tumor-infiltrating CD4⁺ and CD8⁺ T cells. Data in G and H represent at least two independent experiments. *p<0.05; **p<0.01; ***p<0.001; ns: not significant.

Next, we examined T cells in spleens and tumors from GEM-treated mice. Compared to a single dose treatment, multiple dose of GEM treatment significantly decreased frequency of CD4 and CD8 T cells in spleens (Fig. 1E) and tumors (Fig. 1G). Further, the remarkable decrease of IFN-γ-producing CD4 and CD8 T cells was observed in the mice received multiple dose of GEM, especially in tumor tissues (Fig. 1F and H). These data suggest that multi-dose GEM treatment may enhance the immunosuppressive activity of Ly6C^high M-MDSC resulting in decreased effector T cell function.

Conditioned medium (CM) of GEM-treated tumor cells promotes the differentiation of immunosuppressive mouse BM Ly6C^high cells and human M-MDSC

In tumor-bearing hosts, MDSC are generated in the bone marrow and migrate to peripheral lymphoid organs and contribute to formation of TME (27). To investigate how GEM treatment alters the phenotype of tumor M-MDSC, we determined the role of tumor-derived soluble factors in the in vitro differentiation of mouse BM cells. Addition of CM from non-treated E0771 cells (Control CM), but not GEM alone, was sufficient to induce Ly6C^high cell expansion. Major population displayed the phenotype of M-MDSC (CD11b⁺Ly6C^highLy6G⁻) (Fig. 2A). Notably, gene expression analysis revealed that MDSC signature gene arginase 1 (Arg1) and inducible nitric oxide synthase (iNOS) was enhanced in Ly6C^high cells differentiated from GEM-CM (Fig. 2B). The differentiated BM Ly6C^high cells did not express MHC class II, but express PD-L1 and IDO1. Flow cytometry analysis revealed the expression levels of PD-L1 and IDO1 were higher in Ly6C^high cells differentiated from GEM-CM (Fig. 2C). Consequently, Ly6C^high cells from GEM-CM exhibited more immunosuppressive activity (Fig. 2D). Next, we examined whether the CM from GEM-treated human breast cancer cells promotes human PBMC differentiation into immunosuppressive human M-MDSC. To this end, human CD14⁺ monocytes were cultured for 6 days in the presence of CM. Both CM from control and GEM-treated MDA-MB-231 cells induced human monocytes differentiation into human M-MDSC with similar immunosuppressive function for human T cell activation (Fig.2E). However, we observed more HLA-DR^low cells within CD11b⁺CD14⁺CD33⁺ cell population in PBMC cultured in the presence of GEM-CM (CD14⁺CD33⁺HLA-DR^low) (Fig. 2F). Together, these data indicate that tumor cell-derived soluble factors may contribute to the GEM-induced enhanced differentiation of M-MDSC.

Figure 2.

Conditioned medium (CM) of GEM-treated tumor cells promotes the differentiation of M-MDSC. Mouse E0771 and human MDA-MB-231 breast cancer cells were treated with GEM (3–10 μM) for 1 h, followed by wash and 3-days culture for CM collection. (A) Naive mouse BM cells were cultured in complete RPMI 1640 media with GEM alone (3 μM) or 25% E0771 CM for 6 days. Half media were replaced every 2 days. The phenotype of differentiated cells was analyzed by gating on CD11b⁺ populations. (B) BM Ly6C^high cells were sorted after 6 days culture and gene expression of Arg1 and iNOS was determined by qRT-PCR. (C) Expression of PD-L1 and IDO1 in BM Ly6C^high cells was tested by flow cytometry. (D) Immunosuppressive function of BM Ly6C^high cells was measured by co-culturing with CFSE-labeled OT-I splenocytes and OVA (25 μg/ml) for 3 days. T cell proliferation was measured by CFSE dye dilution. (E) Human CD14⁺ monocytes were cultured with MDA-MB-231 CM for 6 days, and half media were replaced every 2 days. The immunosuppression assay was performed by co-culturing with autologous CD3⁺ T cells stimulated with anti-CD3/28 beads for 72 hours. T cell proliferation was measured by flow cytometry. (F) Human PBMC were cultured in the presence of CM for 6 days, and half media were replaced every 2 days. The differentiated cells were analyzed by flow cytometry based on the phenotypic markers of human M-MDSC (CD14⁺CD33⁺HLA-DR^low). The percentage of HLA-DR^low cells within CD14⁺CD33⁺ population was summarized. Data represent two to five independent experiments. *p<0.05; **p<0.01; ***p<0.001; ns: not significant.

GEM treatment enhances GM-CSF production by tumor cells

Inflammatory signals play a critical role in the regulation of differentiation and immunosuppressive function of MDSC (27–29). We therefore next evaluated the possibility that GEM therapy modifies inflammatory profiles in tumor cells. To examine the soluble factors in the CM responsible for the differentiation of BM Ly6C^high cells and human monocytes into immunosuppressive monocytes, we measured the cytokine/chemokine profile. Cytokine array analysis showed enhanced expression of GM-CSF and soluble intercellular adhesion molecule-1 (sICAM-1) in the GEM-CM from E0771 cells (Fig. 3A). ELISA confirmed this result (Fig. 3B). Further, in vivo treatment of E0771 tumor-bearing mice with GEM resulted an increase in GM-CSF expression in tumor cells (Viable CD45⁻ population) (Fig. 3C).

Figure 3.

GEM treatment enhances GM-CSF production by tumor cells. (A) The cytokine/chemokine profile in the CM of GEM-treated or non-treated E0771 cells were examined by mouse cytokine array. Pixel density in each spot of the array was determined using ImageJ. (B) GM-CSF expression in the CM of GEM-treated E0771 cells was measured by ELISA. (C) E0771 tumor-bearing mice (n=3 per group) were treated with GEM (60 mg/kg, IP). Tumor tissues were harvested after 24 hours later, and viable tumor cells (viability dye⁻ CD45⁻) were sorted. The gene expression of GM-CSF in tumor cells was determined by qRT-PCR. (D) BM cells from naïve mice were cultured in the presence of control CM or GEM-CM w/wo GM-CSF, ICAM (CD54) blocking Abs. Immunosuppressive function of Ly6C^high cells was measured by culturing with OT-I T cells and OVA (25 µg/ml) for 3 days. T cell proliferation was determined by measuring CFSE dye dilution. (E) BM cells from naïve mice were cultured in the presence of CM of control or GM-CSF KO E0771 cells for 6 days. Immunosuppressive function of Ly6C^high cells was measured. (F) BM-derived macrophages were stimulated with CM for indicated times, the activation of Stat5 and mTORC1 pathways were examined by Western blot. (G) Tumor progression of GM-CSF KO and control E0771 cells in female B6 WT mice (n=5 per group). (H) Tumor M-MDSC within CD11b⁺ cells from tumor tissues of GM-CSF KO and control E0771 tumor-bearing mice. (I) Tumor progression of ICAM-1 KO and control E0771 cells in female B6 WT mice (n=4–5 per group). (J) GEM treatment enhanced GM-CSF expression by human breast cancer cell line MDA-MB-231. (L) Human CD14⁺ monocytes were cultured in the presence of GEM-CM w/wo GM-CSF blocking Abs for 6 days. Immunosuppressive function of in vitro differentiated monocytes was measured by co-culturing with autologous CD3⁺ T cells stimulated with anti-CD3/28 beads for 72 hours. Data represent two to four independent experiments. *p<0.05; **p<0.01; ***p<0.001.

Next, we tested the function and intracellular signaling events of GM-CSF and sICAM-1 in promoting Ly6C^high cell differentiation. In vitro blocking of GM-CSF or ICAM-1 (CD54) with neutralizing Abs could partially reduce immunosuppression of Ly6C^high cells induced by GEM-CM (Fig. 3D). The Ly6C^high cells differentiated from CM of GM-CSF KO E0771 cells exhibited decreased immunosuppressive function compared to Ly6C^high cells from control CM (Fig. 3E). Stat5 is a major signaling element for GM-CSF receptor (30). GM-CSF requires activation of the AKT/mTOR/mTORC1 signaling cascade in licensing monocytes for suppressor function (29). Therefore, we examined the activation of Stat5 and mTORC1 in BM-derived cells stimulated with tumor CM. As shown in Fig. 3F, the GEM-CM significantly increased the levels of phospho-Stat5 and 4E-BP1, one of best-characterized downstream molecules of mTORC1. To ascertain the importance of GM-CSF and sICAM-1 in in vivo tumor development, GM-CSF knockdown and ICAM-1 knockdown E0771 cells were subcutaneously injected into female B6 WT mice. As expect, depletion of GM-CSF in E0771 cells resulted in a delay in tumor progression (Fig. 3G) and decrease of Ly6C^high M-MDSC (Fig. 3H). In contrast, ICAM-1 deficiency in tumor cells promoted tumor progression (Fig. 3I), ruling out the possibility that sICAM-1 promotes immunosuppressive function of M-MDSC in in vivo GEM treatment setting.

In human breast tumor cells, GM-CSF was detected at higher levels in cell line MDA-MB-231, but not in other tested cell lines including MDA-MB-483, ZR-75–1, BT474, and MCF-7 (Data not shown). Importantly, GM-CSF production by MDA-MB-231 was further increased upon GEM treatment (Fig. 3J). GM-CSF neutralizing antibody significantly blocked the immunosuppressive function of human M-MDSC differentiated from GEM-CM (Fig. 3K). Together, these data suggest that GM-CSF plays a critical role in mediating GEM treatment-induced differentiation of immunosuppressive myeloid cells.

GEM treatment increases mtROS and NF-κB activation in tumor cells

NF-κB is a critical transcriptional factor contributing to the regulation of inflammatory cytokines including GM-CSF expression (13, 14). In order to investigate the roles of NF-κB activation in GEM-induced GM-CSF hyperexpression, we first studied the activation of NF-κB in E0771 cells following GEM treatment. As shown in Fig. 4A, GEM in vitro treatment significantly induced nuclear translocation of p65, a marker for NF-κB activation. Western blot analysis revealed that GEM treatment led to enhanced and persistent phosphorylation of NF-κB p65 (Fig. 4B). Pre-treatment with NF-κB inhibitor Bay 11–7082 could completely abrogate GEM-induced GM-CSF hyperexpression in E0771 cells (Fig. 4C). These data suggest that GEM-induced NF-κB activation contributes to the enhanced production of GM-CSF in tumor cells following GEM treatment.

Figure 4.

GEM treatment increases mtROS production and NF-κB activation in tumor cells. (A) E0771 cells were stimulated by GEM (10 µM) for 4 hours. NF-κB p65 nuclear translocation was determined by confocal microscopy. Cells with p65 translocation to the nucleus were indicated with arrows. (B) E0771 cells were stimulated by GEM (10 µM) for 24 and 48 hours. The phosphorylation of NF-κB p65 was determined by Western blot. (C) E0771 cells were pre-treated with NF-κB inhibitor Bay 11–7082 for 1 hour and then treated with GEM for CM collection as described above. GM-CSF levels in CM were measured with ELISA. (D and E) E0771 cells were treated with GEM (10 µM) for indicated times. Production of mtROS was determined with mitochondrial probe MitoSOX Red. (F) E0771 cells were treated with DOX (1 µM) for 6 hours. Production of mtROS was determined with MitoSOX Red. (G) E0771 tumor-bearing mice were treated with GEM (60 mg/kg, IP). ROS levels in tumor cells were determined using H2DCFDA fluorescent dye and flow cytometry 24 hours later. (H) E0771 cells were pre-treated with NF-κB inhibitor Bay 11–7082, mitochondria-targeted antioxidants MitoTEMPO or MitoQ for 2 hour and then GEM for 6 hours. Gene expression of GM-CSF in E0771 cells was determined using qRT-PCR. (I) E0771 cells were pre-treated with MitoQ and then GEM for 4 hours. NF-κB p65 nuclear localization was determined by confocal microscopy. Cells with p65 translocation to the nucleus were indicated with arrows. Percentages of cells with NF-κB p65 nuclear location were summarized. Data represent at least two independent experiments. **p<0.01; ***p<0.001.

Emerging studies suggest that mtROS also acts as signal-transducing molecules that drive inflammation via production of pro-inflammatory cytokines (20, 21). To understand the importance of mtROS in GEM-induced GM-CSF hyperexpression, we next evaluated ROS production in E0771 cells by using MitoSOX-based assay as mitochondria are a main source of ROS (30). GEM treatment resulted in low level but a significant increase of mtROS production in E0771 cells (Fig. 4D and E), whereas Doxorubicin (DOX) treatment induced a higher level of ROS (Fig. 4F). In addition, GEM in vivo treatment enhanced ROS generation in tumor cells as evidenced by staining with a standard probe of ROS H2DCFDA (Fig. 4G). Pre-treatment with ROS inhibitors, MitoTEMPO or MitoQ, could partially abrogate GM-CSF hyperexpression in GEM-treated E0771 cells (Fig. 4H). To establish a causative relationship between ROS and NF-κB activation, we used MitoQ pre-treatment to reduce ROS and assessed NF-κB activation. Indeed, pre-treatment with MitoQ could decrease nuclear translocation of NF-κB p65 (Fig. 4I). Collectively, these results suggest that ROS can activate the pathway of NF-κB leading to GM-CSF hyperproduction in tumor cells following GEM treatment.

Efferocytosis enhances immunosuppressive function of BM Ly6C^high cells

Cell death is a common event in solid tumors during tumor progression, and anti-tumor effects of chemotherapy are mainly due to the induction of apoptotic tumor cell death. Immunogenic cell death (ICD) has been acknowledged to induce anti-tumor immunity because of immunogenicity of dying cancer cells. However, cancer cell death can be nonimmunogenic. Recent studies also demonstrate that phagocytic clearance of apoptotic cells (efferocytosis) promotes immunosuppression of tumor macrophages (31–33). We therefore assessed the effect of efferocytosis on the differentiation of M-MDSC. Treatment with GEM significantly induced E0771 apoptosis (Fig. 5A). Mer tyrosine kinase (MerTk) is an important receptor for efferocytosis. Flow cytometry revealed that expression of MerTK was upregulated in M-MDSC following GEM in vivo treatment (Fig. 5B). Next, we addressed the ability of Ly6C^high monocytes to uptake apoptotic tumor cells. Apoptotic E0771-GFP cells were mixed with in vitro differentiated BM Ly6C^high cells and efferocytosis was assessed using Flow cytometry. As shown in Fig. 5C, BM Ly6C^high cells were able to acquire apoptotic cells. ImageStream flow cytometric analysis revealed that apoptotic tumor cells are indeed internalized by Ly6C⁺ cells (Fig. 5D). To determine the effect of efferocytosis on the immunosuppressive function of Ly6C^high monocytes, apoptotic E0771 cells were co-cultured with BM cells in the presence of CM of E0771. Gene expression analysis revealed that the expression levels of Agr1, iNOS, and VEGF were increased in BM Ly6C^high monocytes co-cultured with apoptotic tumor cells (Fig. 5E). Consequently, these monocytes exhibited an enhanced immunosuppressive activity on CD8 OT-I T cell proliferation (Fig. 5F). These results indicate that, in addition to cell-derived soluble factors, GEM treatment may promote the immunosuppression of M-MDSC via the phagocytosis of apoptotic cells in TME.

Figure 5.

Efferocytosis enhances immunosuppressive function of BM Ly6C^high cells. (A) E0771 cells were treated with GEM (10 µM) for 48 hours. Apoptotic cells (Annexin V⁺7-AAD⁻) were determined using PE Annexin V/7-AAD detection kit. (B) MerTK expression in tumor M-MDSC from E0771 tumor-bearing mice treated with 4 times of GEM (n=4–5 per group). (C) BM cells were cultured with E0771 CM for 4 days and then co-cultured with apoptotic E0771-GFP cells at the ratio of 1:1 for 2 hours. The efferocytosis of BM Ly6C^high cells was determined using flow cytometry. (D) Representative images of phagocytosis of Ly6C^high cells captured by the Amnis Imagestream flow cytometry. (E) BM cells were cultured with E0771 CM for 4 days and then co-cultured with apoptotic E0771 cells at the ratio of 1:1 for additional 2 days. Gene expression levels of Arg1, iNOS, and VEGF were determined by qRT-PCR. (F) The immunosuppressive function of BM Ly6C^high cells was evaluated by co-culturing with CFSE labeled OT-I splenocytes in the presence of OVA (25 µg/ml) for 3 days. Data represent at least two independent experiments. *p<0.05; ***p<0.001.

GEM treatment results in deregulation of lipid metabolism in tumor cells leading to a decreased sensitivity to T cell-mediated killing

Mitochondria are the main site of lipid oxidation, and oxidation of fatty acids is the source of increased mtROS. Next, we investigated the significance of lipid metabolism in tumor cells in response to GEM treatment. Our analysis revealed that GEM in vitro treatment resulted in a temporal decrease in gene expression of fatty acid synthase (FASN) and carnitine palmitoyltransferase 1A (CPT1A), two key genes related to fatty acid metabolism (Fig. 6A). To further study the lipid metabolism in tumor cells following chemotherapy, we established GEM-resistant E0771 cells (E0771-GEM-R) mimicking residual tumor cells of in vivo chemotherapy treatment. ROS production was increased in E0771 GEM resistant cells (Fig. 6B). Importantly, both FASN and CPT1A was upregulated in E0771 GEM resistant cells (Fig. 6C) as well as in residual tumor cells following 4 times of GEM treatment (Fig. 6D).

Figure 6.

GEM treatment results in deregulation of lipid metabolism in tumor cells leading to a decreased sensitivity to T cell-mediated killing. (A) E0771 cells were treated with GEM (10 µM) for 48 and 72 hours. Viable E0771 cells, as determined by Fixable Viability Dye eFluor™ 780, were sorted for qRT-PCR analysis gene expression of FASN and CPT1A. (B) E0771 GEM resistant cells were generated and ROS production in parent or GEM resistant E0711 cells was determined using H2DCFDA fluorescent dye. (C) Gene expression of FASN and CPT1A in parent or GEM resistant E0711 cells was determined using qRT-PCR. (D) Gene expression of FASN and CPT1A in E0771 tumor cells from tumor-bearing mice treated with 4 times of GEM (n=3 per group). (E) Lipid droplet in parent or GEM resistant E0711 cells was determined using Bodipy 493/503. (F) In vitro T cell cytotoxicity assay was performed by co-culturing of effector CD8⁺ T cells and OVA expressing E.G7 or E.G7-GEM-R tumor cells at the ratio of 40:1 for 6 hours. LDH in the supernatants was determined using LDH cytotoxicity assay kit. In some experiments, E.G7-GEM-R cells were pre-treated with Etomoxir (150 µM) for 48 hours or C75 (30 μM) for 24 hours. (G) E.G7 or E.G7-GEM-R tumor cells were co-cultured with CFSE-labeled effector CD8⁺ T cells for 4 hours. The apoptosis of tumor cells was determined by using Anexin V/7-AAD staining. The cleaved Caspase-3 in target tumor cells were measured by flow cytometry. Data represent at least two independent experiments. *p<0.05; **p<0.01; ***p<0.001.

Increase of fatty acid can lead to an increase in lipid droplets (34). To investigate whether GEM-treated tumor cells had increased stores of lipid, we used lipophilic fluorescent probe BODIPY 493/503 to label cellular neutral lipid contents. We observed an appreciable increase in the level of lipid droplets in E0771-GEM-R cells compared to parent E0771 cells (Fig. 6E). Previous studies have shown that chemotherapy sensitizes tumor cells to the cytotoxic effect of cytotoxic T lymphocytes (CTLs) by using tumor cells pre-treated with paclitaxel (TAX) or DOX for 18 hours (35, 36). In addition, CTL can eliminate tumor cells even with anti-apoptotic mutations conferring drug resistance (37). However, the chemotherapeutic agents may have a delayed toxic effect on target cells. Further, recent studies have demonstrated that chemoresistance is associated with altered lipid metabolism in tumor cells (38, 39). Released exosomes following chemotherapy exhibited inhibition of CD8⁺ T cell response (16). To address the impact of chemotherapy on T cell-mediated cytotoxicity, we generated GEM-resistant E.G7-OVA cells (E.G7-GEM-R). Antigen-specific T cells were generated by in vitro activated OT-I splenocytes with antigen OVA and used as effectors in CTL assays. For targets, we used OVA-negative EL4, OVA-positive E.G7, as well as E.G7-GEM-R cells. As shown in Fig. 6F, OVA-specific OT-I T cells displayed specific cytotoxicity to OVA-positive E.G7 compared to EL4 cells, but the E.G7-GEM-R cells exhibited a decrease in sensitivity to T cell-mediated cytotoxicity. Pre-treatment of E.G7-GEM-R cells with FAO inhibitor Etomoxir (ETO) and C75 partially restored the sensitivity of chemoresistant tumor cells to T cell-mediated cytotoxicity. This effect was further confirmed using an apoptosis assay with Annexin V/7-AAD staining. Compared to E.G7 cells, a little apoptosis was evident in E.G7-GEM-R cells co-cultured with antigen-specific T cells (Fig. 6G). In addition, significant increase of cleaved caspase-3 was observed in the E.G7 cells, but not in the E.G7-GEM-R cells (Fig. 6G). Taken together, these results further indicate that GEM treatment could impair the effectiveness of T cells via the induction of deregulation of lipid metabolism in residual tumor cells that survive from chemotherapy.

Discussion

Traditionally, conventional chemotherapy has been thought to act through the direct killing of tumor cells via drug toxicity. However, accumulating evidence indicates that anti-tumor immunity significantly contributes to the anti-tumor responses of chemotherapy (1–3). Therefore, it is expected that decrease of this immunity during chemotherapy has a negative effect on its efficacy. Thus, the impacts of chemotherapy on anti-tumor immunity need further thorough investigation in order for rationale-based combinatorial regimens to be developed (4, 5).

MDSC are a chief component of immunosuppressive networks. Tumor cells secrete a variety of factors to promote MDSC development and immunosuppressive function (27, 40). Previous studies using tumor-bearing mice have demonstrated that one-time GEM or 5-FU treatment can selectively deplete MDSC in vivo (8–10). However, chemotherapy in clinic typically involves a number of cycles. Previous evidence has revealed that pancreatic cancer patients receiving multi-dose of GEM (day 1, 8 and 15 in a 28-day cycle) display no reduction of M-MDSC in peripheral blood although PMN-MDSC showed temporal decrease after the first treatment (41). In this study, we show that repeated GEM treatment enhances accumulation and immunosuppressive function of M-MDSC. We believe that inflammation following multiple dose of GEM treatment leads to the observed effects. This inflammation may be derived from tumor cells in tumor microenvironment. However, host response following multiple dose of GEM could also contribute to the migration and differentiation of M-MDSC. Under chemotherapy condition, the demand for myeloid cells increases because of cell depletion induced by cytotoxic drugs. Therefore, myelopoiesis rates are increased to meet the additional need for myeloid cells (42, 43). Our future study will address how host response following chemotherapy modulates bone marrow progenitor cell myeloid differentiation potential, which may lead to accumulation of M-MDSC in spleen and tumor.

GM-CSF is a cytokine that plays a pivotal role in promoting myeloid cell differentiation. In tumor, GM-CSF exhibits controversial roles regarding anti-tumor immunity (44, 45). On one hand, GM-CSF may promote anti-tumor immunity through the education of immune stimulatory dendritic cells. On the other hand, GM-CSF promotes tumor development by facilitating differentiation of immunosuppressive myeloid cells (14, 29, 46, 47). We demonstrate that knockdown of GM-CSF expression in mouse tumor cells significantly delay tumor development, this is correlated with a decrease in Ly6C^high M-MDSC in TME. Both in vitro and in vivo GEM treatment enhance GM-CSF expression in tumor cells. For human breast cancer cells, GM-CSF is highly expressed in TNBC cell line MDA-MB-231, and GEM treatment can further increase the expression of GM-CSF. These data, along with other studies, provide a potential causal link between tumor-derived inflammation and development of immunosuppressive myeloid cells in chemotherapy setting (12–14). mTORC1 has been reported to provide novel signaling for licensing monocyte immunosuppressive function induced by GM-CSF (29). Our recent study also reveals critical role of mTORC1 in mediating suppressor function of M2 macrophages and tumor-associated macrophages (48). Here, we show that mTORC1 is also involved in promoting M-MDSC immunosuppressive function in chemotherapy setting. In support of this result, our data show enhanced mTORC1 pathway activation induced by CM of GEM treated tumor cells. Although sICAM-1 is highly expressed in the CM of GEM-treated cells and breast cancer patients (49), we exclude the roles of sICAM-1 in mediating GEM treatment-induced M-MDSC differentiation as knockdown of ICAM-1 in tumor cells promotes tumor progression. In addition, previous studies found that ICAM-1 can suppress tumor metastasis by inhibiting M2 macrophage polarization (50).

It is widely appreciated that NF-κB signaling plays important roles in mediating inflammation and cytokine expression. Many chemotherapeutic agents can promote inflammation via activation of NF-κB signaling, which may lead to failure of therapy and metastasis (51). Here, we show that GEM treatment enhances GM-CSF expression via the NF-κB signaling. ROS has a dual role in cancer (17, 18). First, ROS may promote tumorigenesis and survival by triggering activation of transcription factors. Second, high ROS levels are generally considered detrimental to cells. Numerous standard chemotherapies generate ROS in cancer cells, which is one of the proposed mechanisms by which chemotherapeutic agents induce tumor regression. It is becoming increasingly evident that ROS may also act as signal-transducing molecule that drives inflammation via production of proinflammatory cytokines (20, 52). GEM treatment results in a lower level of ROS compared to Doxorubicin. However, we present evidence that pre-treatment with mitoTEMPO or MitoQ significantly abrogates GEM-induced GM-CSF hyperexpression in E0771 cells. This effect may be through the modulation of NF-κB signaling as we show that MitoQ pre-treatment can partially reduce the NF-κB activation. Therefore, our data indicate that GEM-induced mtROS may act as a signaling molecule to enhance NF-κB activation leading to an upregulated GM-CSF production in tumor cells.

Clearance of apoptotic cells (efferocytosis) plays a major role in the resolution of inflammation. However, recent studies also demonstrate that efferocytosis increases immunosuppression in the TME (31–33, 53). Macrophage is a major cell population responsible for efferocytosis. LC3-associated phagocytosis in TAM promotes immunosuppression (31). In addition, monoclonal anti-Neu antibody (Lapatinib) treatment induces tumor cell apoptosis and widespread efferocytosis, which drive immune tolerance in TME through MerTK and IDO1 (33). In steady condition, Ly6C⁺ monocytes can efferocytose and cross-present cell-associated antigen to CD8⁺ T cells. Ligation of TLR7 can enhance cross-presentation by Ly6C⁺ monocytes (54). We show that GEM treatment enhances efferocytosis receptor MerTK expression in M-MDSC in tumor. Apoptotic tumor cells following GEM treatment promotes development of immunosuppression of BM Ly6^high monocytes. This raises the possibility that chemotherapy-induced tumor cell apoptosis might trigger immunosuppression within TME, which results in tumor cells evading treatment.

Combination of chemotherapy with immunotherapy aiming to activate T cells could be an attractive approach in the treatment of cancer. It has been shown that chemotherapy makes tumor cells permeable to granzyme B that are more susceptible to the cytotoxic effect of T cells (35). On the other hand, chemotherapy results in metabolic changes in tumor cells leading to the chemoresistance and tumor recurrence (23, 24, 38). The tumor metabolic reprogramming and MDSC-mediated immunosuppression has been established (47). However, whether tumor cell reprogramming following chemotherapy affects T cell mediated cytotoxicity is not well defined. In this work, we also focus on metabolic changes in tumor cells following repeated GEM treatment. We have shown that expression of fatty acid metabolism gene FASN and CPT1A is elevated in in vitro established GEM-resistant cell lines as well as residual tumor cells following GEM in vivo treatment, which leads to an increase in lipid droplets. Interestingly, we have made a link between tumor cell reprogramming and resistance to T cell-mediated tumor cell killing. Inhibition of FAO partially restores the sensitivity of the tumor cells to cytotoxic T cell killing. This result is consistent with a previous finding that surviving tumor cells following chemotherapy can escape immune killing by upregulation of PDL1, CD47, and CD73 (55).

Taken together, our results define an undesired effect of GEM treatment on ROS production and metabolic changes in tumor cells. These effects might promote the development of immunosuppressive myeloid cells via the up-regulation of GM-CSF and efferocytosis in the TME. The results would provide important information for development combination therapies using chemotherapy and ROS or FAO inhibitors.

Key Points.

Hyperexpression of GM-CSF in tumor cells promotes M-MDSC immunosuppression.

Efferocytosis enhances immunosuppression of Ly6C^high myeloid cells.

Chemoresistance decreases the sensitivity to CTL induced cytotoxicity.

Abbreviations used in this article:

Arg1

Arginase 1

CM

conditioned medium

CPT1A

carnitine palmitoyltransferase 1A

FASN

fatty acid synthase

GEM

Gemcitabine

GM-CSF

Granulocyte-macrophage colony-stimulating factor

iNOS

inducible nitric oxide synthase

MDSC

myeloid-derived suppressor cells

mTORC1

mechanistic target of rapamycin complex 1

PBMC

peripheral blood mononuclear cell

ROS

reactive oxygen species

VEGFA

vascular endothelial growth factor A