STING Contributes to Antiglioma Immunity via Triggering Type I IFN Signals in the Tumor Microenvironment

Gliomas are the most common primary malignant brain tumors and carry a dismal prognosis despite current treatments, and new therapies are needed. Immunotherapies are promising in this regard. However, the successful development of immunotherapy for gliomas requires detailed understanding of factors critical for antiglioma immunity.

In addition to the ability of type I IFNs to interfere with viral infection, they also enhance antitumor host immunity. Indeed, loss of type I IFN signaling promotes tumorigenesis in a variety of tumor types, such as sarcomas (1), melanomas (2, 3), and gliomas as we have reported previously (4). Although a growing body of evidence suggests that endogenously produced type I IFNs participate in antitumor immune responses at the level of host hematopoietic cells (5, 6), the molecular mechanisms responsible for inducing the type I IFN in the sterile tumor microenvironment remain elusive. Furthermore, the impact of type I IFN on immune cell populations participating in the antitumor response in vivo needs to be elucidated. In this regard, CD8α⁺ dendritic cells (DC) have been shown to require type I IFNs for effective antitumor immunity (2, 3). Type I IFNs directly enhance in vivo clonal expansion of CD4⁺ T cells following immunizations against lymphocytic choriomeningitis viruses (7), promote the survival of CD8⁺ T cells, and stimulate the development of cytolytic functions, including the production of IFNγ (8). Although we have previously demonstrated a critical role of type I IFNs on maturation of glioma-infiltrating CD11c⁺ DCs (4), it still remains to be elucidated how type I IFNs are induced in the glioma microenvironment and whether they directly affect T-cell functions.

Stimulator of IFN genes (STING) has recently been identified as one of the critical adaptors for cytosolic DNA sensing. It plays a critical role in host defense against viral and intracellular bacteria by regulating type I IFN signaling and innate immunity (9–12). STING is stimulated downstream of DNA sensors, such as helicase DDX41 (DExD/H-box helicases 41; ref. 13), and cyclic dinucleotides (CDN), such as c-di-GMP, c-di-AMP, cGMP-AMP (cGAMP), or 10-carboxymethyl-9-acridanone (CMA; refs. 14–18), thereby leading to production of type I IFNs. STING-deficient mice or cells show increased susceptibility to infection by several microbes and diminished levels of type I IFNs in response to several microbes and CDNs (19).

Considering that there are abundant dying tumor cells that release their genomic (g)DNA in the tumor microenvironment (20), we evaluated our hypothesis that STING-mediated DNA sensing is involved in type I IFN production in the glioma microenvironment, and stimulation of STING with its agonist enhances antiglioma immunity including T-cell responses.

Wild-type (WT) C57BL/6J (H-2K^b) and C57BL/6-background Sting^Gt/Gt mice (C57BL/6J-Tmem173^gt/) were purchased from The Jackson Laboratory. B6.129(Cg)-Gt(ROSA)26Sor^tm4(ACTB-tdTomato,-EGFP)Luo/J mice (“tdTomato” mice) were generated by breeding B6.Cg-Tg(Mx1-cre)1Cgn/J mice with B6.129(Cg)-Gt(ROSA)26Sor^tm4(ACTB-tdTomato,-EGFP)Luo/J mice (21). All mice were maintained and handled in accordance with the Animal Facility at the University of Pittsburgh (Pittsburgh, PA) per an Institutional Animal Care and Use Committee–approved protocol.

The following monoclonal antibodies (mAb) were obtained from BD Biosciences: anti-CD11c (HL3), anti-CD11b (M1/70), and anti-Gr-1 (RB6-8C5). The following mAbs were obtained from eBioscience: anti-CD4 (GK1.5), anti-CD8 (53-6.7), anti-CD3 (145-2C11), anti-CD19 (eBio1D3), anti-IFNγ (XMG1.2), anti-CD25 (7D4), and anti-FoxP3 (NRRF-30). The H-2K^b-binding OVA_257–264 (SIINFEKL) peptide was synthesized in the University of Pittsburgh Peptide Synthesis Facility. For Western blotting, an ISG54-specific polyclonal antibody (22) and actin-specific mAbs (Sigma-Aldrich) were used. For positive control, WT macrophages treated with 25 μg/mL of polyI:C (for 48 hours) were used.

The procedure for intracerebroventricular DNA injection has been described previously (23). Briefly, the following DNA plasmids were mixed with in vivo compatible DNA transfection reagent, In vivo-JetPEI (Polyplus Transfection): pT2/C-Luc//PGK-SB100 (0.06 μg/mouse), Sleeping beauty (SB) transposon-flanked pT2/CAG-NRasV12 (0.12 μg/mouse), and pT2/shp53/mPDGF (0.12 μg/mouse), and injected into the right lateral ventricle of neonate. Intracranial injection of glioma cell lines has been described previously (24).

The procedure has been described previously (24).

The procedure has been described previously (24). Luciferin was obtained from Caliper Life Sciences.

The GL261 mouse glioma cell line was kindly provided by Dr. Robert Prins (University of California Los Angeles, Los Angeles, CA). The GL261-luc cell line was generated by transfection of GL261 cells (24) with a plasmid vector pcDNA3.1 encoding luciferase cDNA, followed by selection with G418 (Sigma), limiting dilution and selection of a clone based on the highest luciferase expression level using luminometer in the presence of luciferin in culture. Survival of syngeneic mice bearing GL261-luc cells was confirmed to be comparable with survival of those bearing parental GL261 cells (not shown). The Quad-GL261 cell line, kindly provided by Dr. John R. Ohlfest (University of Minnesota, Minneapolis, MN), expresses OVA_257–264, OVA_323–339, human gp100_25–33, and mouse I-Eα_52–68 (25). Stable expression of transgenes was maintained by G418 in the culture, and monitored every 3 months by evaluating their susceptibility against antigen-specific cytotoxic T lymphocytes, such as Pmel-1 cells, which were derived from B6.Cg-Thy1a/Cy Tg(TcraRcrb)8Rest/J mice (The Jackson Laboratory). The RMA-S mouse thymoma cell line was kindly provided by Dr. Walter J. Storkus (University of Pittsburgh). All cell lines were tested to be Mycoplasma free. No other authentication assay was performed.

Primers and probes for the following genes were obtained from Applied Biosystems: Ifna6 (Mm01703458_s1), Ifnb1 (Mm00439552_s1), Foxp3 (Mm00475162_m1), Tgfb1 (Mm01178820_m1), Tbx21 (Mm00450960_m1), Ifng (Mm01168134_m1), Ccl5 (Mm01302427_m1), Cxcl10 (Mm00445235_m1), and Gapdh (Mm99999915_g1). In some experiments, following primers were used: mouse pan Ifna forward: CCTGAGAAGAGAAGAACACAGCC, reverse: GGCTCTCCAGACTTTCTGCTCTG; mouse pan Ifnb forward: CCGAGCA GAGATCTTCAGGAA; reverse: CCTGCAACCACCACTCATTCT; mouse Gapdh forward: TCACCACCATGGAGAAGGC, reverse: GCTAAGCAGTTGGTGGTGCA. Gapdh was used as an internal control. Relative expression levels compared with control samples were calculated in each experiment using the ΔΔC_t method.

gDNA was isolated from GL261 and NIH 3T3 cell lines using the Wizard Genomic DNA Purification Kit from Promega. The final gDNA suspension was made in TE buffer (10 mmol/L Tris–Cl and 1 mmol/L EDTA). Aliquots of CD11b⁺ bone marrow–derived macrophage cells (5 × 10⁵ cells/mL) from WT or Sting^Gt/Gt mice were stimulated with gDNA (1 or 5 μg/mL). At 48 hours, cells were harvested and total RNA was extracted. Quantitative real-time PCR (qRT-PCR) analyses were performed with SsoFast EvaGreen Supermix (Bio-Rad), and data were analyzed with CFX manager 2.0 software from Bio-Rad.

These procedures have been described previously (24).

GFP-positive or GFP-negative CD4⁺ T cells derived from draining lymph nodes of glioma-bearing tdTomato mice were sorted by MoFlo AstrosTM (Beckman Coulter). CD8⁺ T cells were isolated from non–glioma-bearing WT mice and labeled with carboxyfluorescein succinimidyl ester (CFSE; Life Technologies) for 10 minutes in the incubator. After washing with medium, CD8⁺ T cells were cocultured with GFP-negative or GFP-positive CD4⁺ T cells in the presence of Dynabeads (Gibco by Life Technologies). After 60-hour incubation, samples were evaluated by BD Accuri C6.

OVA-specific cytotoxicity of brain-infiltrating leukocytes (BIL) was measured by 4-hour ⁵¹Cr-release assay as described previously (4, 24). Briefly, freshly isolated BILs were incubated with ⁵¹Cr-labeled GL261 cells loaded with or without OVA_257–264 peptide (10 μg/mL) for 4 hours. For reverse antibody-dependent cell-mediated cytotoxicity (ADCC) of lymphocytes, GFP-positive or GFP-negative CD8⁺ T cells derived from draining lymph nodes of glioma-bearing mice were sorted by MoFlo Astrios (Beckman Coulter), then incubated with ⁵¹Cr-labeled Fc receptor–positive RMA-S cells pretreated with or without anti-CD3 mAb (10 μg/mL, 145-2C11; BD PharMingen) for 4 hours. The percentage of cytotoxicity was calculated as described previously (26).

C-di-GMP (InvivoGen) was dissolved in physiologic water per the manufacturer's instruction. Mice bearing gliomas received intracranial injections of either c-di-GMP (4 μg/2 μL/dose) or mock treatment with solvent alone. Some mice received subcutaneous vaccinations with OVA_257–264 peptide (100 μg/dose) emulsified in incomplete Freund Adjuvant (Difco Laboratories) on the same day as the c-di-GMP treatment.

The statistical significance of differences between two groups was determined by the Student t test; one-way ANOVA with the Holm post hoc test was conducted for multiple group comparisons. The log-rank test was used to determine statistically significant differences in survival curves among groups. All mouse data were analyzed by R Environment version 2.10.1.

We first evaluated type I IFN mRNA levels in the mouse glioma microenvironment by qRT-PCR. Murine brain hemispheres bearing de novo glioma expressed significantly higher levels of Ifna6, Ifna8, and Ifnb1 compared with non–tumor-bearing contralateral hemispheres (Fig. 1).

To determine the effects of type I IFN expression in the glioma microenvironment, we used a novel reporter mouse model, in which type I IFN signaling induces the Mx1 (IFN-induced GTP-binding protein) promoter-driven Cre recombinase, which turns the expression of loxp-flanked tdTomato off, and turns GFP expression on, thereby enabling us to monitor the induction of IFN signaling in the glioma microenvironment. Under two-photon microscopy, glioma tissues showed higher levels of GFP signals compared with the normal (non–glioma-bearing) brain (Fig. 2A), further substantiating IFN induction in the glioma microenvironment. Using flow cytometry, we evaluated the percentage of GFP⁺ cells, in which IFN signaling has turned GFP signal on. In each of CD11b⁺ Gr-1⁺, CD11b⁺ CD11c⁺, CD19⁺, and CD3⁺ BIL subpopulations, glioma-bearing brains revealed a higher percentage of GFP⁺ cells compared with the spleen or inguinal lymph nodes (iLN; Fig. 2B and C). GFP⁺ RFP⁺ double-positive cells are thought to be the ones that have been exposed to IFN but still retain residual RFP. Because non–glioma-bearing brains do not contain sufficient numbers of BILs, we were unable to evaluate them. Nonetheless, the spleen and iLNs derived from non–glioma-bearing mice showed percentages of GFP⁺ cells that were similar to those derived from glioma-bearing mice (data not shown), suggesting that type I IFNs produced locally in the glioma tissue transmit their signals in BILs, but do not have significant impacts on cells in the spleen or iLN.

We and others have previously demonstrated a critical role of type I IFN pathway in the function of tumor-infiltrating CD11c⁺ DCs as antigen-presenting cells (2–4). However, whether local production of type I IFNs directly affects the T cells in the glioma-bearing mice remains to be elucidated. The tdTomato mouse model allowed us to address this question in vivo during glioma progression. We sorted CD4⁺ and CD8⁺ T-cell populations from draining lymph nodes based on their GFP expression. CD4⁺ T cells that received the type I IFN signal (GFP⁺ cells) expressed significantly lower levels of Foxp3 and Tgfb1 compared with CD4⁺ T cells that did not receive the type I IFN signal (GFP⁻ cells; Fig. 3A), suggesting that the GFP⁻ population contains more regulatory T cells (Treg). Indeed, GFP⁻ cells inhibited CD8⁺ T-cell proliferation more profoundly than GFP⁺ cells in the coculture assay (Fig. 3B). Among the CD8⁺ T cells, GFP⁺ cells expressed significantly higher levels of Tbx21 and Ifnγ (Fig. 3C), suggesting that the type I IFN signal skews CD8⁺ cells toward type I effector cells. Accordingly, GFP⁺ CD8⁺ cells showed higher cytotoxic activity than GFP⁻ cells (Fig. 3D). Taken together, these results indicate that type I IFN signaling directly enhances antitumor activity of T cells in glioma-bearing mice.

Next, we focused on identifying the specific cells that are primarily responsible for producing IFN in glioma as a means to define the signaling mechanism of IFN induction in the “sterile” tumor microenvironment. As it was previously reported that CD11b⁺ and CD11c⁺ cells are responsible for IFN production (4), we isolated CD11b⁺ and CD11c⁺ cells from BILs derived from SB glioma-bearing mice. As shown in Fig. 4A, CD11b⁺ cells showed higher levels of Ifna6 and Ifnb1 expression than CD11c⁺ cells by qRT-PCR.

We next focused on the stimulus and the signaling pathway responsible for the observed Ifn induction. We excluded RNA sensors from our evaluation because, based on the literature (27), we thought it unlikely that high levels of immunostimulatory RNA, which can stimulate IFN production, would be induced in the glioma microenvironment. Although other receptors, such as high-mobility group protein B1 (28) and inflammasomes (29), have been implicated in antitumor immunity, these receptors do not directly cause strong IFN induction. We therefore hypothesized that gDNA derived from dying or dead cells can induce type I IFNs through STING-mediated signaling in glioma-infiltrating macrophages because both human and mouse glioma tissues contain necrotic areas that are heavily infiltrated by macrophages (23, 30). Indeed, we found upregulation of Sting and Ifi16, which is involved in DNA virus sensing (Supplementary Fig. S1; ref. 31). Another DNA sensor, Aim2, which is responsible for inflammasome activation in response to DNA (32), was not upregulated. We first tested this hypothesis in vitro by stimulating CD11b⁺ macrophages with gDNA derived from either GL261 glioma or NIH3T3 cells in vitro. We detected enhanced expression of pan Ifna and pan Ifnb at similar levels, which was abrogated by DNases (Fig. 4B). The induction of pan Ifna was partially abrogated in STING-deficient cells (Fig. 4C). These data indicate cell-derived gDNA, from either nonmalignant or malignant cells, induces type I IFNs at least partially in a STING-dependent manner, and led us to further investigate the role of STING in antiglioma immunity.

To determine the in vivo role of STING in glioma development, we induced de novo SB gliomas in WT or Sting^Gt/Gt mice. Total RNA extracted from glioma-bearing brains of Sting^Gt/Gt mice showed significantly lower levels of Ifna6 and Ifnb1 compared with total RNA derived from WT mice (Fig. 5A). Also, ISG54 protein, which is induced by type I IFNs (22), was detected at a lower amount in the right (i.e., glioma-bearing) hemisphere of Sting^Gt/Gt mice than in the counterpart in WT mice (Fig. 5B), indicating partial loss of IFN signaling. SB glioma-bearing Sting^Gt/Gt mice exhibited significantly shorter survival compared with WT mice (Fig. 5C). In BILs, Sting^Gt/Gt mice exhibited more CD11b⁺ Gr-1⁺ immature myeloid cells, which are likely myeloid-derived suppressor cells (MDSC; ref. 33), and CD25⁺ Foxp3⁺ CD4⁺ Tregs than WT mice. Furthermore, Sting^Gt/Gt mice had less IFNγ-producing CD8⁺ T cells compared with WT mice (Fig. 5D). These results suggest that STING is at least partially responsible for spontaneous type I IFN production, and affects the phenotype of a variety of BIL populations, including T cells, in the glioma microenvironment. These data also led us to evaluate whether augmentation of the STING-mediated signal via administration of a STING agonist would enhance the antiglioma immunity.

Among various ligands that have been reported to activate STING, structure–function studies have indicated that the CDNs have been the most authentic and robust activators of STING (34). When we administered c-di-GMP intratumorally in tdTomato mice bearing gliomas, BILs from c-di-GMP–treated mice showed increased numbers of GFP⁺CD8⁺, GFP⁺CD4⁺, and GFP⁺CD11c⁺ cells compared with control mice treated with solvent alone (Supplementary Fig. S2). Treatment of glioma-bearing WT mice with c-di-GMP significantly prolonged survival (Fig. 6A), and upregulated Ccl5 and Cxcl10 levels compared with control treatment (Fig. 6B) in a STING-dependent manner. In BILs, c-di-GMP treatment also enhanced tumor-homing of CD4⁺ and CD8⁺ T cells as well as IFNγ-producing CD8⁺ T cells in a STING-dependent manner (Fig. 6C and D). In the de novo glioma model, administration of c-di-GMP also significantly inhibited glioma growth (Fig. 6E). These data indicate that direct intratumoral administration of c-di-GMP enhances antiglioma immunity by enhancing the recruitment of T cells into the brain tumor site.

Finally, to investigate whether c-di-GMP treatment would enhance the efficacy of vaccinations targeting a tumor-specific antigen, using the mouse Quad-GL261 glioma cell line expressing OVA_257–264 (25), we evaluated a combination of c-di-GMP and the OVA_257–264 peptide vaccine. Although monotherapy with c-di-GMP alone significantly prolonged the survival of mice compared with vaccine alone or negative control with mock treatment (P < 0.01), the combination treatment further enhanced the survival benefit, with 7 of 10 mice surviving longer than 70 days (P < 0.05 compared with c-di-GMP alone; Fig. 7A). All 10 mice treated with OVA_257–264 peptide vaccine alone died by day 47. In BIL analyses (Fig. 7B), while c-di-GMP monotherapy significantly enhanced the tumor-homing of CD8⁺ T cells compared with mice receiving mock treatment or vaccine alone, the combination regimen further enhanced the percentage of CD8⁺ cells compared with c-di-GMP alone. BILs obtained from mice receiving c-di-GMP monotherapy showed a modest but significant cytotoxic activity against both OVA_257–264 peptide–pulsed and nonpulsed GL261 cells when compared with the control treatment, suggesting that c-di-GMP therapy induces cytotoxic responses against endogenous antigens in GL261 cells (Fig. 7C). Furthermore, BILs obtained from mice receiving the combination therapy demonstrated a significantly higher OVA_257–264 peptide–specific cytotoxic activity compared with ones from mice receiving c-di-GMP alone, vaccine alone, or mock treatment (Fig. 7C). These data strongly support the development of a combination strategy with vaccine and a STING agonist.

This is, to our knowledge, the first study to describe the induction and roles of type I IFNs in the glioma microenvironment. These mechanistic evaluations also led us to demonstrate the efficacy of c-di-GMP as an adjuvant in glioma immunotherapy.

Mouse glioma tissues spontaneously expressed type I IFN mRNAs. Fuertes and colleagues (3) have demonstrated that tumor-resident CD11c⁺ CD8α⁺ DCs are the source of type I IFNs and these DCs are critical for induction of tumor-reactive T-cell responses. In human melanoma, Wenzel and colleagues (35) have demonstrated a presence of strong type I IFN signals in regressive melanocytic skin lesions. Our studies using tdTomato mice indicate that the type I IFN signal in the glioma microenvironment indeed promotes type I T-cell responses while inhibiting Tregs. On the other hand, the development of gliomas in the brain did not affect immune cells in the spleen and iLNs, suggesting that spontaneous immune response in the glioma site does not induce systemic immune responses at least through the type I IFN signals.

On the basis of our data with STING-deficient mice and cells, we surmised that STING is at least partially responsible for the production of type I IFNs responding to its ligands in the glioma microenvironment. On the basis of our data showing that both glioma (i.e., GL261)- and nontransformed fibroblast (i.e., NIH3T3)-derived gDNA equally induced type I IFN mRNAs in myeloid cells in a STING-dependent manner, we postulate that gDNA released from either glioma or nontumor stroma cells could be ligands for STING signaling. Indeed, necrosis is often observed in mouse de novo as well as human glioma lesions (23, 30). It is likely that STING is activated downstream of specific DNA sensors that are activated by gDNA, such as cyclic guanosine monophosphate-adenosine monophosphate (cGAMP) synthase (cGAS; ref. 36), Ifi16, and DDX41 (13, 15–18, 37), in the glioma tissue. cGAS may play a major role as cGAS produces cGAMP, which binds to and activates the adaptor protein STING, thereby inducing type I IFNs and other cytokines (36, 38). Further investigations are warranted to gain better understanding of STING activation in gliomas.

Cyclic-di-GMP had been demonstrated to be an effective vaccine adjuvant (39, 40) before it was found to be a ligand for STING in 2011 (16). Before our data in the current report, other groups have also demonstrated the role of c-di-GMP as an effective adjuvant. Ebensen and colleagues (41) have demonstrated that intranasal administrations of c-di-GMP in combination with vaccines induce significantly stronger humoral and cellular immune responses than the administration of the antigen alone. Moreover, Hu and colleagues (42) have demonstrated that subcutaneous administrations of c-di-GMP plus Staphylococcus aureus–associated antigens induce enhanced humoral immune responses in mice, leading to prolonged survival after a challenge with cognate bacteria. These studies administered c-di-GMP three times to observe protective effects in their disease settings. On the other hand, in our glioma model, one c-di-GMP injection was sufficient to induce significant antitumor effects. This may be because our treatment was directed to the tumor site as local therapy, whereas systemic protection against infections requires systemic enhancement of the immune system. Interestingly, during the preparation of this article, Miyabe and colleagues (43) reported that subcutaneous administration of c-di-GMP delivered in liposome, but not c-di-GMP alone, can induce high-level IFNβ and antitumor immunity. It has also been demonstrated that intravenous administrations of c-di-GMP suppress vaccine-induced responses (44). Following the submission of our original article, Chandra and colleagues (45) have reported efficacy of intraperitoneally administered c-di-GMP in a metastatic 4T1 mammary adenocarcinoma model. As strategies to induce type I IFNs, we (46, 47) and others (48–50) have conducted cancer immunotherapy clinical trials using Toll-like receptor ligands. On the basis of our data in this study, early-phase clinical studies are warranted to evaluate the safety and efficacy of intratumoral administration of a STING agonist in patients with glioma.

No potential conflicts of interest were disclosed.

Conception and design: T. Ohkuri, A. Ghosh, A. Kosaka, M. David, S.N. Sarkar, H. Okada

Development of methodology: T. Ohkuri, A. Ghosh, A. Kosaka

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Ohkuri, A. Ghosh, A. Kosaka, J. Zhu, M. Ikeura, M. David, S.C. Watkins, S.N. Sarkar

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Ohkuri, A. Ghosh, A. Kosaka, S.N. Sarkar, H. Okada

Writing, review, and/or revision of the manuscript: T. Ohkuri, A. Kosaka, H. Okada

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.N. Sarkar, H. Okada

Study supervision: S.N. Sarkar, H. Okada

The authors thank Drs. John R. Ohlfest, Adam J. Litterman (both at University of Minnesota) and Gary Kohanbash (University of Pittsburgh) for their technical and administrative assistance.

This study was supported by the NIH (2R01 NS055140, 2P01 NS40923, 1P01 CA132714, and 5U24AI082673) and Musella Foundation for Brain Tumor Research and Information. This project used UPCI shared resources (Animal Facility, Small Animal Imaging facility, Cell and Tissue Imaging Facility, and Cytometry Facility) that are supported, in part, by NIH P30CA047904.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.