Cancer Stem Cell-Secreted Macrophage Migration Inhibitory Factor Stimulates Myeloid Derived Suppressor Cell Function and Facilitates Glioblastoma Immune Evasion

Abstract

Shifting the balance away from tumor-mediated immune suppression toward tumor immune rejection is the conceptual foundation for a variety of immunotherapy efforts currently being tested. These efforts largely focus on activating antitumor immune responses but are confounded by multiple immune cell populations, including myeloid-derived suppressor cells (MDSCs), which serve to suppress immune system function. We have identified immune-suppressive MDSCs in the brains of GBM patients and found that they were in close proximity to self-renewing cancer stem cells (CSCs). MDSCs were selectively depleted using 5-flurouracil (5-FU) in a low-dose administration paradigm, which resulted in prolonged survival in a syngeneic mouse model of glioma. In coculture studies, patient-derived CSCs but not nonstem tumor cells selectively drove MDSC-mediated immune suppression. A cytokine screen revealed that CSCs secreted multiple factors that promoted this activity, including macrophage migration inhibitory factor (MIF), which was produced at high levels by CSCs. Addition of MIF increased production of the immune-suppressive enzyme arginase-1 in MDSCs in a CXCR2-dependent manner, whereas blocking MIF reduced arginase-1 production. Similarly to 5-FU, targeting tumor-derived MIF conferred a survival advantage to tumor-bearing animals and increased the cytotoxic T cell response within the tumor. Importantly, tumor cell proliferation, survival, and self-renewal were not impacted by MIF reduction, demonstrating that MIF is primarily an indirect promoter of GBM progression, working to suppress immune rejection by activating and protecting immune suppressive MDSCs within the GBM tumor microenvironment. Stem Cells 2016;34:2026-2039.

Keywords: Cancer stem cells; Glioblastoma; Immunotherapy; MIF; Macrophage migration inhibitory factor; Myeloid-derived suppressor cells; Tumor immune suppression; Tumor microenvironment.

Figure 1

MDSCs are present in glioblastoma tumors adjacent to cancer stem cells and predict poor survival. Virtual triple immunohistochemistry of GBM tumors (A) demonstrates the presence of MDSCs (** = HLA-DR⁻, CD33⁺) preferentially associated with CSCs (* = CD133⁺ or SOX2⁺). HLA-DR was stained in brown, CD33 was stained in red (upper right panel) or brown (lower right panel), CD133 was stained in brown, and SOX2 was stained in red. Scale bar = 50 μm. These results are quantified in (B) *** = p <.001 by Students’ unpaired t-test. Bioinformatics analysis of The Cancer Genome Atlas (C) indicated that an MDSC signature (CD11b⁺, CD33⁺, CD45⁺, CD244⁺, CXCR2⁺) negatively correlates with GBM patient prognosis. Statistical significance of survival was based on log-rank analysis. Abbreviations: CSCs, cancer stem cells; GBM, Glioblastoma; MDSC, Myeloid-derived suppressor cell.

Figure 2

MDSCs promote tumor development. Schematic depicting treatment paradigm (A), in which MDSCs were targeted using a low-dose 5-fluorouracil (5-FU) treatment strategy during tumor engraftment and growth in the GL261 syngeneic glioma model. Kaplan–Meier survival curves (B) demonstrate that tumor-bearing mouse survival was increased with low-dose 5-FU treatment (10 mg/kg, red) compared with higher-dose (50 mg/kg, blue) and vehicle control (black) groups. Examples of tumor burden as measured by bioluminescence (C) show differences in low-dose 5-FU compared with the control treatment group. Analysis of peripheral blood MDSC levels (based on CD244.2 and GR-1 positivity) in tumor-bearing mice using flow cytometry (D) demonstrates that 5-FU treatment reduces MDSCs. Analysis of brain tissue in tumor-bearing mice using flow cytometry (E) reveals that 10 mg/kg 5-FU increased intratumoral CD8-positive T cells (CD3⁺/CD8⁺) and reduced T regulatory (T_Reg) cells (CD3⁺/CD4⁺/CD25⁺/Foxp3⁺). Micrographs of tumor-bearing mice (F) confirm a reduction in MDSCs (GR-1⁺/ARG-1⁺, yellow arrows in top panels), neutrophils (GR-1⁺/ARG-1⁻, yellow arrows in top panels), T_Regs (FOXP3⁺/CD4⁺, yellow arrows in bottom panels), and CD4⁺ activated T cells (FOXP3⁻/CD4⁺, brown arrow bottom panel) with a concomitant increase in CD8⁺ T cells with 10 mg/kg 5-FU treatment compared with vehicle control. These immune cell populations were not detected after treatment with 50 mg/kg 5-FU. Values shown are means ±standard deviation, scale bar = 20 μm, and nuclei were counterstained with DAPI. *, p <.05, **, p <.01, and ***, p <.001 by one-way ANOVA, and statistical significance for in vivo studies was calculated using log-rank analysis. Abbreviations: IC Tx, intracranial transplantation; IP, intraperitoneal; MDSC, Myeloid-derived suppressor cell.

Figure 3

CSC conditioned media increases arginase-1 production and survival in MDSCs. Schematic (A) depicting the coculture strategy employed to determine the effect of CSC and NSTC conditioned media on MDSCs. Flow cytometry plots (B) of overall MDSC levels (red region) with control (unconditioned) media (null), NSTC conditioned media, and CSC conditioned media. Arginase-1 expression in MDSCs derived from the marrow of tumor-bearing mice was elevated as assessed by flow cytometry with CSC conditioned media compared with NSTC conditioned and control media. These results were repeated using mice implanted with two separate patient-derived human GBM specimens, hGBM 4121 (C) and hGBM 10 (D). The ratio of live to dead cells as assessed by Annexin V in MDSCs derived from the marrow of tumor-bearing mice was elevated with CSC conditioned media and compared with NSTC conditioned and control media. Experiments were repeated with conditioned media generated from the CSCs and NSTCs isolated from hGBM 4121 (E) and hGBM 10 (F) xenografts. Values shown are means ±standard deviation, *, p <.05, **, p <.01, and ***, p <.001 by one-way ANOVA. Abbreviations: CM, conditioned media; CSCs, cancer stem cells; MDSC, Myeloid-derived suppressor cell; NSTC, Nonstem tumor cell.

Figure 4

MDSCs exposed to CSC conditioned media promote immune suppression. Schematic (A) depicting the coculture strategy employed to determine the effect of MDSCs exposed to CSCs and NSTC conditioned media on T cell function. The ratio of CD4⁺ to CD8⁺ cells (B) was altered by MDSCs exposed to CSC conditioned media compared with NSTC conditioned media. The activation of T cells as reported by interferon-γ (IFNγ) expression (C) was suppressed when MDSCs were exposed to CSC conditioned media compared with NSTC conditioned media. Values shown are means ±standard deviation, *, p <.05, **, p <.01, and ***, p <.001 by one-way ANOVA. Activated T cells and neutrophils were used as controls. Abbreviations: CSC, cancer stem cell; MDSCs, Myeloid-derived suppressor cells; NSTC, Nonstem tumor cell.

Figure 5

CSC conditioned media is enriched in macrophage MIF, which increases arginase-1 production in MDSCs in a CXCR2-dependent manner. Dot blots (A) and quantitation (B) of cytokine arrays. Red box: MIF; blue box: G-CSF. Immunoblots of MIF in GBM CSCs (CD133+) and NSTCs (CD133−) in human GBM cells from two patient specimens (C). MDSC ARG-1 production as assessed by flow cytometry after the addition of exogenous MIF (D) or MIF receptor neutralizing antibodies (E) to conditioned media. (F) MDSC ARG-1 production in the presence of neutralizing antibodies to the CXCR2 and CXCR4 receptors. Values shown are means ±standard deviation, *, p <.05, **, p <.01, and ***, p <.001 by one-way ANOVA. Abbreviations: CM, conditioned media; CSC, cancer stem cell; MDSCs, Myeloid-derived suppressor cells; MIF, migration inhibitory factor; NSTC, Nonstem tumor cell.

Figure 6

MIF depletion attenuates tumor immune suppression and confers a survival advantage in vivo. Immunoblot for MIF after MIF shRNA treatment of GL261 cells (A). Survival of mice intracranially injected with control (black) or MIF KD (red) GL261 cells, p <.01 by log-rank test (B). Median survival is indicated. MDSCs in the peripheral blood of mice bearing control and MIF KD GL261 intracranial xenografts (C). Intratumoral CD8⁺ T cells (CD3⁺/CD8⁺), **, p <.01 by pairwise t-test (D) and T_Reg cells (CD3⁺/CD4⁺/CD25⁺/FOXP3⁺), ***, p <.001 by pairwise t-test (E) in MIF KD GL261-injected mice vs. controls. Immunofluorescence of intracranial mouse xenografts (F). (Left) Yellow arrows: MDSCs (GR-1⁺/ARG-1⁺); brown arrows: neutrophils (GR-1⁺/ARG-1⁻). (Right) Yellow arrows: T_Regs (FOXP3⁺/CD4⁺); brown arrows: CD4⁺ activated T cells (FOXP3⁻/CD4⁺). Bioinformatics analysis of MIF mRNA expression levels in human GBM pathological specimens (G) compared with nonmalignant control tissues (Left Panel, p = .03 by pairwise t-test). MIF mRNA expression levels in primary versus recurrent human GBM (Central Panel, p = .043 by pairwise t-test). Kaplan–Meier analysis of MIF expression levels informs human GBM patient survival (Right Panel, p = .0026 by log-rank test). Cartoon summary of the CSC/MDSC/T Cell interaction that takes place in the GBM TME (H). Scale bar = 20 μm. Abbreviations: GBM, glioblastoma; MDSCs, myeloid-derived suppressor cells; MIF, migration inhibitory factor.