NKG2D-Dependent Antitumor Effects of Chemotherapy and Radiotherapy against Glioblastoma

Glioblastoma is the most common malignant primary brain tumor in adults with a dismal prognosis (1). The first-line treatment in patients below 70 years of age includes surgical resection as feasible, radiotherapy, and concomitant and maintenance chemotherapy with temozolomide (TMZ), an alkylating agent that induces DNA damage (2, 3). In addition to these treatment modalities, several promising immunotherapeutic approaches against glioblastoma are currently being evaluated (4, 5). These efforts are supported by the observation that glioma cells express molecules that allow for an interaction with cells of the immune system such as major histocompatibility complex (MHC) class I and class II molecules (6) as well as MHC class I–like ligands that bind to the activating immune cell receptor natural-killer group 2 member D (NKG2D; ref. 7). In humans, NKG2D ligands (NKG2DL) comprise the MHC class I–related chains (MIC) A and B (MICA, MICB) and the UL16 binding proteins (ULBP) 1–6 (8). These ligands are expressed on human glioma cells in vitro (9) and in vivo (10) as well as on glioma-initiating cells (GIC), a subpopulation of cells with stem cell properties (11, 12). In mice, NKG2DL comprise the retinoic acid early inducible-1 (RAE-1) proteins, members of the H60 family (H60a, H60b, and H60c) and the murine UL16-binding protein like transcript-1 (MULT-1), which are also expressed by mouse glioma cells (9, 13). All NKG2DL bind to the NKG2D receptor, which is one of the major activating receptors on natural killer (NK) cells (8). In addition to NK cells, this receptor is constitutively expressed on NKT cells, αβ CD8 T cells, and γδ T cells (8, 14). Furthermore, its expression is induced on CD4 T cells by TNF-α and IL15 (15, 16). However, various glioma-derived humoral and cellular immunosuppressive mechanisms preclude an efficient antitumor immune response, including the expression of transforming growth factor (TGF)-β (17), prostaglandin E2 (PGE2; ref. 18), IL10 (19), growth and differentiation factor (GDF)-15 (20), lectin-like transcript 1 (LLT1; ref. 21), indoleamine 2,3-dioxygenase (IDO; ref. 22), programmed death ligand 1 (PD-L1; ref. 23), as well as the presence of immunosuppressive regulatory T cells (Tregs; ref. 24) and M2-polarized microglia (25). Enhancing the immunogenicity of glioma cells may be achieved either by inhibition of these immunosuppressive mechanisms (26) or by promoting immune activating signals such as the NKG2DL (27). Because various cellular stress stimuli including malignant transformation of cells or DNA damage can induce NKG2DL (8), we explored whether TMZ or irradiation (IR) as part of the standard treatment for glioblastoma increase NKG2DL levels on glioma cells and whether this promotes their immunogenicity. We also defined the molecular mechanisms underlying the TMZ- and IR-induced NKG2DL expression in glioma cells. Finally, we investigated the significance of the NKG2D system for the survival benefit gained with TMZ and IR in several immunocompetent mouse glioma models.

The human glioma cell lines LN-18 and LN-229 were kindly provided by Dr. N. de Tribolet (Centre Hospitalier Universitaire Vaudois, Lausanne, Switzerland). LN-229-R cells were generated by repetitive exposure to TMZ resulting in a shift of the EC₅₀ (28). Generation of LNT-229_MGMT and LNT-229_neo (29) and LN-18_shMGMT and LN-18_puro cells (30) has been described. SMA-560 glioma cells were obtained from Dr. D. Bigner (Duke University Medical Center, Durham, NC) and GL-261 were obtained from the National Cancer Institute (Frederick, MD). SMA-560_Turbo650 and GL-261_NirFP were created by lentiviral transduction of GL-261 and SMA-560 cells with plasmids encoding near-infrared fluorescent proteins Turbo650 and NirFP (Evrogen) and selection by fluorescence-activated cell sorting (FACS). Adherent cell lines were maintained in Dulbecco's Modified Eagle Medium (DMEM; Invitrogen), containing 2 mmol/L l-glutamine (Gibco Life Technologies), and 10% fetal calf serum (FCS; Biochrom KG). The GIC cell lines S-24 and ZH-305 were generated from human glioblastoma patient specimens (31). After tumor removal, tissue was dissociated using a papain system and a gentleMACS Dissociator (Miltenyi Biotec). These cells were then maintained as suspension cultures in Neurobasal Medium with B-27 supplement (20 μL/mL) and glutamax (10 μl/mL) from Invitrogen and fibroblast growth factor (FGF)-2, epidermal growth factor (EGF; 20 ng/mL each; Peprotech) and heparin (32 IE/mL; Ratiopharm). All cell lines were routinely tested for Mycoplasma using PCR (last test in December 2016). For all experiments described herein, the adherent cells were allowed to attach over a 24-hour period. Subsequently, the experiments were carried out in serum-free medium. KU-60019 (Selleckchem) is a potent and specific ataxia-telangiectasia mutated (ATM) inhibitor, and concentrations < 1.5 μmol/L ensure specificity for ATM. TMZ, kindly provided by Schering-Plough, was prepared in stock solutions (100 mmol/L) in dimethylsulfoxide (DMSO). N-(2-chloroethyl)-N′-cyclohexyl-N-nitrosourea (CCNU) was kindly provided by Medac. Cells were irradiated using a cobalt-60 source (Sulzer), and for different fractionations the approximative biological effective dose and dose per fraction according to the linear quadratic model (32) were determined using the R package DVHmetrics (https://cran.r-project.org/web/packages/DVHmetrics/index.html) under the assumption of an α/β ratio of 10 for human glioma cell lines. Thiazolyl blue tetrazolium bromide (MTT) was obtained from AxonLAb.

The following monoclonal antibodies (mAb) were used for the assessment of cell surface expression of MICA, MICB, ULBP2, ULBP3, RAE-1, MULT-1, H60 or blocking of NKG2D: MICA (AMO1, mouse IgG1), MICB (BMO1, mouse IgG1), ULBP2 (BUMO1, mouse IgG1), and ULBP3 (CUMO3, mouse IgG1). Their generation has been described previously (33). RAE-1_FITC and MULT-1_PE and blocking anti-human NKG2D (clone 149810) were obtained from R&D Systems Europe. H60_PerCP was obtained from Novus Biologicals. Blocking but not depleting anti-mouse NKG2D (clone C7) was obtained from eBioscience. As controls, we used isotype-matched antibodies from Sigma-Aldrich. The PE-conjugated goat anti-mouse IgG from Dako was used as secondary antibody where appropriate. Cells were detached with Accutase (Life Technologies), preincubated in phosphate-buffered saline (PBS) with 2% FCS, and stained with specific mAbs (10 μg/mL) or matched mouse Ig isotype for 30 minutes on ice, followed by incubation with PE-conjugated secondary antibody for 30 minutes where appropriate. After washing, flow cytometric analyses were performed using a BD FACSVerse Analyzer (BD). In case of intracellular staining for ATM^ser1981, Fix/Perm Buffer Set from BioLegend was used. For flow cytometric assessment of tumor-infiltrating lymphocytes, live/dead staining with FVS 510, anti-CD3_ PerCP-Cy5.5, anti-CD4_FITC, anti-CD8_APC-H7, anti-NKp46_PE, anti-IFNγ_APC, and anti-TCRγ/δ_BV421 from BioLegend were used. Specific fluorescence indexes (SFI) were calculated by dividing median fluorescence obtained with the specific antibody by median fluorescence obtained with isotype control antibody. For in vivo experiments, fluorescence intensity was expressed as median fluorescence intensity. Data were analyzed with FlowJo software (TreeStar).

We used a flow cytometry-based cytotoxicity assay to determine immune-mediated glioma cell lysis (34). Specific lysis was expressed as percentage of cell death of the PKH-26 labeled targets. Percentage of target cell lysis was corrected for spontaneous background lysis by subtracting the percentage of dead cells in control samples (targets alone) from the percentage of dead cells within the test samples. As effector cells, we used either splenocytes isolated from mice, human NK cells isolated from PBMC by negative selection using an NK cell isolation kit (Miltenyi Biotec) or NKL cells obtained from M.J. Robertson (Indiana University School of Medicine, Indianapolis, IA). For blocking experiments, NKL cells were preincubated for 2 hours at 4°C with anti-NKG2D or IgG1 isotype control, and the antibody was also present during the coincubation of target and effector cells. All experiments were done in triplicates.

Total RNA was isolated using the NucleoSpin RNA II system from Macherey-Nagel and cDNA was prepared using the iScript cDNA Synthesis Kit from Bio-Rad Laboratories AG. For real-time PCR, gene expression was measured in an ABI Prism 7000 Sequence Detection System (Applied Biosystems) with SYBR Green Master Mix (Thermo Fisher Scientific) and primers (Microsynth AG) at optimized concentrations. Primers for MICA, MICB, ULBP2, and ULBP3 have been published (35). Primers used to detect murine NKG2DL were RAE-1 forward 5′-TTTGGGAGCACAACCACAGAT-3′, reverse 5′-TAAAGTTGGCGGGCTGAAAGA-3′, MULT-1 forward 5′-CTGCCAGTAACAAGGTCCTTTC-3′, reverse 5′-GCTGTTCCTATGAGCACCAATG-3′, H60a forward 5′-CTGAGCTATCTGGGGACCATAC-3′, and reverse 5′-AGTCTTTCCATTCACTGAGCAC-3′. As reference gene, we used human HPRT1: forward 5′-TGAGGATTTGGAAAGGGTGT-3′, reverse 5′-GAGCACACAGAGGGCTACAA-3′, and mouse HPRT1: forward 5′- TTGCTGACCTGCTGGATTAC-3′, reverse 5′- TTTATGTCCCCCGTTGACTG-3′ respectively. The conditions were 40 cycles at 95°C/15 seconds and 60°C/1 minute. Standard curves were generated for each gene. Relative quantification of gene expression was determined by comparison of threshold values. All results were normalized to HPRT1 and calculated with the ΔCTT method for relative quantification.

For determination of cytotoxicity, 5 × 10³ cells were seeded per well in 96-well plates, allowed to attach for 24 hours (adherent cells) and irradiated or exposed to TMZ, CCNU, or staurosporine as indicated for 72 hours in serum-free medium. Percentage of living cells was determined by flow cytometry after live/dead staining with Zombie Aqua Fixable Viability Kit (BioLegend). For acute growth inhibition assays, we used the same experimental setting but either crystal violet staining (for adherent cells) or MTT (for suspension cells) as read-out. Clonogenic survival assays were performed by seeding 10² cells per well in 96-well plates. After 24 hours, the cells were irradiated or exposed to TMZ, CCNU, or staurosporine as indicated for 24 hours in serum-free medium, followed by observation for 20 days. As readout methods, we again used either crystal violet staining or MTT.

For the detection of proteins in cell lysates, cells were lysed and processed as previously described (28). Thirty micrograms of protein was used per lane and visualization of protein bands was accomplished using horseradish peroxidase (HRP)-coupled secondary antibodies (Sigma-Aldrich) and enhanced chemiluminescence (Pierce/Thermo Fisher).

Cells were cultured in chamber slides with polystyrene-treated glass (BD Biosciences), fixed with 4% paraformaldehyde and permeabilized with 0.5% Triton X-100 (Sigma-Aldrich). Blocking with 3% FCS was followed by incubation with anti-ATM protein kinase pS1981 monoclonal antibody (Rockland; diluted 1:100) overnight at 4°C. Donkey anti-mouse IgG Alexa Fluor 488–labeled secondary antibody (Life Technologies) was used at 1:200. Slides were mounted in Vectashield Mounting Media with DAPI, and images were acquired by using a Leica TCS SP5 confocal microscope.

All experiments were done in accordance with the guidelines of the Swiss federal law on animal protection and were approved by the cantonal veterinary office. C57BL/6 mice were purchased from Charles River Laboratories. VM/Dk mice were bred in pathogen-free facilities at the University of Zurich. NKG2D^−/− mice have been previously described (36) and were kindly provided by D. H. Busch (Munich, Germany). Mice of 6 to 12 weeks of age were used in all experiments in groups of 7 to 10 mice. For intracranial tumor implantation, SMA-560 cells (5 × 10³) or GL-261 cells (2 × 10⁴) were stereotactically implanted into the right striatum at day 0. Mice were observed daily and sacrificed as indicated or in the survival experiments when developing neurologic symptoms. If indicated, local cranial radiotherapy with a single dose of 12 Gy was performed at day 10 after tumor implantation using a Gulmay 200 kV X-ray unit at 1 Gy/minute at room temperature. If indicated, mice received TMZ (10 mg/kg/day) per oral gavage from days 7 to 11 after tumor implantation. MRI was performed with a 4.7 T small animal magnetic resonance imager (Pharmascan; Bruker Biospin) at day 13 after tumor implantation. Coronal T2-weighted images were acquired using Paravision 6.0 (Bruker BioSpin). Mean ± SD of the tumor volume in mm³ from 5 mice/group were determined by the formula (length × width × depth)/2.

For in vivo blockade of NKG2D signaling, mice were injected i.p. with 100 μg of the blocking but not depleting anti-NKG2D antibody (clone C7; ref. 37) or with isotype control in PBS. Antibodies were given either 1 day before and 1 day after tumor implantation or at days 6 and 7 after tumor implantation and were reinjected every 7 days until the mice were sacrificed. Time of antibody administration is indicated in the figure legends.

Isolation of orthotopic tumor cells was performed on day 12 after tumor implantation. Brains were harvested after transcardial perfusion with ice-cold PBS to remove all circulating leukocytes from the CNS. Tumor cells were separated from myelin and red blood cells using a Percoll gradient suspension (Sigma-Aldrich). Cells were washed with PBS and stained with Zombie Aqua Fixable Viability Kit and fluoroconjugated antibodies specific to indicated cell surface markers for flow cytometry.

Studies were approved by the Institutional Review Board (KEK-StV-Nr.19/08), and informed consent was received prior to inclusion to the study. Twenty-one pairs of primary (before chemoradiation) and recurrent glioblastoma (variable time points after chemoradiation) specimens from patients who underwent brain tumor resection between 2000 and 2014 at the Department of Neurosurgery, University Hospital Zurich (Zurich, Switzerland) were collected. Immunohistochemistry was performed as described (31) using anti-MICA, anti-MICB, anti-ULBP2, or anti-ULBP3 antibodies from Sino Biological (Lucerna-Chem AG) or antiprogrammed death-ligand 1 (PD-L1) from Cell Signaling Technology. Images were analyzed in an unsupervised and blinded fashion using TMARKER, a software toolkit for histopathologic staining estimation (38).

Data were presented as means ± SD. Experiments were repeated at least three times, if not indicated differently. Viability and acute and clonogenic cell growth studies were performed at least in triplicates. Statistical analyses were performed in GraphPad Prism using multiple two-tailed Student t tests and correction for multiple comparisons using the Holm–Sidak method. For analysis of tissue microarray data, we used Wilcoxon matched-pairs signed rank test. For analysis of heterogeneity of immunohistochemically stained NKG2DL, we calculated the intraclass correlation coefficient (ICC; ref. 39) as a statistical measure to assess staining variation for 2 tissue cores from each tumor sample by using the R-package ICC (https://cran.r-project.org/web/packages/ICC/index.html). Kaplan–Meier survival analysis was performed to assess survival differences among the treatment groups and P values were calculated with the Gehan–Breslow–Wilcoxon test. Throughout all figures, significance was concluded at *, P < 0.05 and **, P < 0.01.

Exposure of glioma cells to TMZ has growth-inhibitory as well as cytotoxic effects. To define the sensitivity of LN-18 and LN-229 cells to TMZ, we treated these cells with a broad range of TMZ concentrations and determined cell death, acute growth inhibition, and clonogenic cell survival (Fig. 1A). Because NKG2DL are upregulated in response to various stress stimuli, we explored in a next step whether TMZ induces the expression of NKG2DL in these cells. We observed an induction of several NKG2DL on mRNA and protein cell surface level over a wide concentration range (Fig. 1A; Supplementary Fig. S1A–S1B) including low concentrations with minor cytotoxic and growth-inhibitory effects as well as clinically relevant concentrations around plasma levels of 30 to 80 μmol/L of TMZ that are achieved in human patients (40). To evaluate the effect on other activating immune cell receptor ligand systems, we assessed CD112 and CD155 as ligands of the human DNAX accessory molecule-1 (DNAM-1, CD226) activating immune cell receptor. In contrast to NKG2DL, the cell surface expression of CD112 and CD155 was unaffected by TMZ (Supplementary Fig. S1C). Next, we examined the effect of TMZ on NKG2DL expression in GIC, a subpopulation of glioma cells with stem-like properties that are associated with resistance to chemotherapy and irradiation (41). S-24 cells were relatively resistant to TMZ with an EC₅₀ value of 267 μmol/L in clonogenic survival assays, whereas ZH-305 cells were more sensitive with an EC₅₀ of 7.3 μmol/L (Fig. 1B). TMZ induced several NKG2DL on mRNA and cell surface protein levels in both GIC lines. Again, there was no induction of DNAM-1 ligands (Supplementary Fig. S1D). Furthermore, we determined the expression of NKG2DL on mouse glioma cells and their induction by TMZ. GL-261 and SMA-560 cells differed in their sensitivity to TMZ. The EC₅₀ for clonogenic cell survival was ≈50 μmol/L for GL-261 and >500 μmol/L for SMA-560 (Fig. 1C). Similar to human cells, exposure to TMZ resulted in an upregulation of NKG2DL in both murine glioma cell models. H60a is not expressed in C57BL/6 mice and the syngeneic GL-261 cells (42) and was therefore not detected in this cell line.

To corroborate our findings that the upregulation of NKG2DL is not a general response pattern of glioma cells to cell death induction but rather a specific response to alkylating chemotherapy, we exposed LN-18 and LN-229 cells to different concentrations of staurosporine. Despite its strong effect on glioma cell viability, none of the NKG2DL was upregulated by staurosporine (Supplementary Fig. S1E). However, CCNU, another alkylating agent commonly used in patients with recurrent glioblastoma (43) also induced NKG2DL already at low concentrations, close to those typically achieved in the plasma of patients (3.4–3.8 μmol/L; ref. 44; Supplementary Fig. S1F).

Because radiotherapy belongs to the standard of care for glioma patients, we also assessed the effect of IR on NKG2DL expression in different glioma models. LN-18 cells were more sensitive to irradiation than LN-229 cells with an EC₅₀ value of 4 vs. 11 Gy in clonogenic survival assays. In both cell lines, IR induced the expression of several NKG2DL on mRNA and cell surface protein level (Fig. 2A). The induction of NKG2DL cell surface expression following IR was also confirmed when different fractionation schemes were applied (Supplementary Fig. S2A). Consistent with the TMZ data, there was no induction of DNAM-1 ligands upon irradiation (Supplementary Fig. S2B). In S-24 and ZH-305 GIC, irradiation had minor cytotoxic effects with an EC₅₀ value of > 20 Gy but clear effects on clonogenic survival with EC₅₀ values ≈ 5 Gy. A clinically relevant single fraction in the range of 2 to 4 Gy increased NKG2DL mRNA and cell surface protein levels (Fig. 2B).

We also confirmed the irradiation-mediated induction of NKG2DL in GL-261 and SMA-560 mouse glioma cells. In both cell lines, irradiation upregulated NKG2DL on the mRNA as well as on cell surface protein level (Fig. 2C).

MGMT promoter methylation predicts benefit from alkylating chemotherapy with TMZ in glioblastoma. To explore whether the TMZ-mediated induction of NKG2DL is influenced by MGMT, we used subcell lines of LN-18 with a stably silenced MGMT gene (30) or LNT-229 cells that stably overexpress MGMT (29). The modulation of MGMT expression affected the sensitivity to TMZ (Fig. 3A), but not to IR (Supplementary Fig. S3A). Furthermore, MGMT expression significantly decreased TMZ-mediated NKG2DL induction. This was demonstrated by an increased NKG2DL induction upon shRNA-mediated MGMT silencing in LN-18 glioma cells that naturally express MGMT and a diminished NKG2DL induction in MGMT-overexpressing LNT-229 cells compared with MGMT-deficient wild-type LN-229 cells (Fig. 3B). The IR-mediated upregulation of NKG2DL was unaffected by the MGMT status (Supplementary Fig. S3B). Glioma cells can also acquire resistance to TMZ independent from MGMT expression. Mechanistically, this is linked, amongst others, to the downregulation of DNA mismatch–repair proteins (28). Because this acquired resistance is a challenge in clinical practice that needs alternative treatment options, we assessed the induction of potentially immune-activating NKG2DL in a glioma cell line with acquired TMZ resistance (28). Also in these cells, TMZ or IR induced the cell surface protein level of NKG2DL. The same effect was observed after IR (Supplementary Fig. S3C–S3D).

To elucidate the molecular mechanisms mediating the treatment-induced NKG2DL induction in glioma cells, we assessed the ATM pathway as part of the DNA damage response to genotoxic stress induced by TMZ (45). In LN-229 and S-24 cells, we detected an increase of active phospho-ATM^Ser1981 upon exposure to TMZ (Fig. 3C). Inhibition of ATM using RNA interference (Supplementary Fig. S3E) or KU-60019, a specific ATM inhibitor that inhibited ATM at 1.25 μmol/L with little toxicity (Supplementary Fig. S3F), abrogated the TMZ-induced upregulation of MICA and MICB in LN-229, S-24 cells (Fig. 3D). We confirmed this also for ZH-305 cells (Supplementary Fig. S3G). Furthermore, we observed this ATM dependency also for irradiation-mediated NKG2DL induction (Fig. 3D; Supplementary Fig. S3G).

To investigate functional effects of the TMZ- or RT-induced NKG2DL induction, we performed cytotoxicity assays using polyclonal human NK cells or NKL cells (46) as immune effectors. Preexposure of LN-229 or S-24 cells to TMZ resulted in an enhanced immune cell–mediated cytolysis (Fig. 4A; Supplementary Fig. S4A). In contrast, exposure of MGMT-overexpressing LNT-229 cells to TMZ at the same concentrations did not enhance immune cell–mediated cytolysis (Supplementary Fig. S4B). Preincubation of effector cells with blocking but not depleting anti-NKG2D antibodies abrogated the TMZ-induced glioma cell susceptibility to immune cell killing (Fig. 4A). Similarly, LN-229 or S-24 cells that were preirradiated with 2 Gy were more susceptible to immune cell–mediated cytolysis in an NKG2D-dependent manner (Fig. 4B).

To study the effect of TMZ and irradiation on glioma-associated NKG2DL in vivo, we generated GL-261_niRP and SMA-560_TurboFP650 mouse glioma cells, which stably express near-infrared fluorescent proteins and which are syngeneic to C57BL/6 or VM/Dk mice. This allowed for the detection of these cells by flow cytometry (Fig. 5A) and the specific assessment of NKG2DL protein levels on the cell surface ex vivo. After orthotopic tumor cell injection, we treated mice either with a single dose of local IR at day 10 or with TMZ per oral gavage for 5 consecutive days starting at day 7 after tumor cell inoculation. At day 12, mice were euthanized, and the tumors were explanted. TMZ and irradiation led to an upregulation of NKG2DL in both orthotopic murine glioma cell models with a more pronounced effect in the SMA-560 model (Fig. 5A).

To study the effect of chemo- and radiotherapy on glioma-associated NKG2DL in human glioblastoma patients, we created a tissue microarray (TMA) encompassing 21 paired formalin-fixed samples of human glioblastoma specimens obtained before and after treatment with TMZ and/or radiotherapy. From 9 of these paired samples, we could also isolate RNA. Compared with basal expression, we detected increased levels of several NKG2DL on mRNA as well as on cell surface protein level after treatment with TMZ or IR or both (Fig. 5B). Based on 2 cores from each tumor, we found a heterogeneous expression of NKG2DL within tumors, particularly for ULBP2 and ULBP3 (Supplementary Fig. S5A). We did not observe correlations between NKG2DL and survival or NKG2DL and the immunosuppressive ligand PD-L1 in this small patient population. There were also no significant differences in PD-L1 expression between primary and recurrent human glioblastoma samples (Supplementary Fig. S5B–S5D).

Finally, we asked whether the NKG2D system plays any role for the survival benefit gained from TMZ or irradiation in murine glioma models. We inhibited the NKG2D system in fully immune-competent, orthotopic SMA-560 glioma-bearing mice by repetitive intraperitoneal injections of a blocking but not depleting anti-NKG2D antibody (37). Its biological activity reflecting target inhibition was verified by decreased ex vivo cytolysis of SMA-560 cells upon TMZ exposure or irradiation by immune effector cells isolated from anti–NKG2D-treated mice (Supplementary Fig. S6A). At the treatment schedules used, either IR or TMZ prolonged survival, but this effect was more prominent for IR. Administration of the anti-NKG2D antibody abrogated the survival benefit conferred by TMZ and attenuated the IR-mediated survival benefit in SMA-560 glioma-bearing mice (Fig. 6A). This NKG2D-dependent effect of TMZ or IR in SMA-560 glioma-bearing mice was also present when the anti-NKG2D antibody was administered at days 6 and 7 post tumor implantation when tumors had already been established (Supplementary Fig. S6B). To confirm the importance of an intact NKG2D system for the efficacy of TMZ and IR in glioma in a second syngeneic setting, we used NKG2D knockout (NKG2D^−/−) mice as an even more robust model. These mice were treated with the same regimen of TMZ or IR. In addition, we also included the combination of both treatments, reflecting the current standard of care for human glioblastoma patients. There was no difference in median survival of glioma-bearing NKG2D^−/− or NKG2D-intact mice when no treatment was administered. TMZ or irradiation prolonged the median survival of GL-261 tumor–bearing mice and the combination of both therapies further increased the survival (Fig. 6B). However, the survival gain conferred by TMZ, irradiation, or the combination of both was significantly reduced in NKG2D^−/− mice (Fig. 6B). The survival data were corroborated by MRI. At day 6 post tumor implantation, we could not clearly delineate the tumor due to superimposing post-surgery alterations, but at day 13 post tumor implantation, we observed reduced activity of the antitumor treatments with regard to tumor growth in NKG2D-deficient mice (Fig. 6C; Supplementary Fig. S6C). Finally, we analyzed tumor-infiltrating immune cells. TMZ alone significantly reduced NK and CD4 T cells, and IR as well as the combination of TMZ and IR reduced NK cells within the tumor microenvironment (Fig. 6D). There was no difference in the composition of tumor-infiltrating immune cells in NKG2D^−/− vs. NKG2D-intact mice. However, the activation status of infiltrating immune cells, which did not differ in untreated NKG2D^−/− or NKG2D-intact mice, was impaired in NKG2D^−/− mice upon treatment. NK cells as well as CD4 and CD8 T cells produced more IFNγ in NKG2D-intact mice following treatment with TMZ, IR, or the combination of TMZ and IR, and this induction was attenuated in NKG2D^−/− mice. In NKG2D-intact mice, γδ T cells produced more IFNγ upon treatment with TMZ, IR, or the combination of TMZ and IR compared with NKG2D^−/− mice. In NKG2D^−/− mice, we observed more IFNγ production in γδ T cells only upon IR (Fig. 6D; Supplementary Fig. S6D–S6F).

TMZ chemotherapy and radiotherapy constitute the standard treatment modalities in patients with newly diagnosed glioblastoma (43). The antiglioma effects of TMZ and IR comprise different molecular mechanisms such as induction of cell-cycle arrest, senescence, and apoptosis (47, 48). Furthermore, there is increasing evidence that cell death upon exposure to TMZ or IR can promote antitumor immune responses by releasing tumor-associated antigens or damage-associated molecular pattern molecules such as calreticulin, adenosine triphosphate, or high-mobility group box 1 protein (49–51). In addition to these soluble and potentially immune-stimulating molecules, glioma cells express membrane-bound ligands to the activating immune cell receptor NKG2D which basically enables target cell killing without prior sensitization and irrespective of MHC restriction.

We observed an upregulation of several NKG2DL on mRNA and protein level upon exposure to TMZ, CCNU, or IR in several mouse and human glioma cells, including stem-like cells (Figs. 1 and 2; Supplementary Figs. S1 and S2). NKG2DL induction by TMZ has previously been reported in four other human glioma cell lines (52). We found that the NKG2DL induction was independent from cytotoxic or growth inhibitory effects and was achieved at clinically relevant concentrations of chemotherapeutic agents and low doses of IR. Furthermore, we confirmed the upregulation of NKG2DL on glioma cells upon treatment with TMZ or IR in vivo in two orthotopic mouse glioma models (Fig. 5A). The use of fluorescently labeled glioma cells excluded contaminating signals from immune cells which could also express NKG2DL (53). These findings were further corroborated by an analysis of paired samples of human glioblastoma tissue specimens obtained from patients during initial surgery and at tumor recurrence following radio- and/or chemotherapy. The increased NKG2DL expression levels after alkylating chemotherapy or radiotherapy support our in vitro data as well as findings from the mouse studies (Fig. 5B). Changes in glioma cell NKG2DL levels may be confounded by other factors than treatment such as passenger mutations that occur during the course of the disease (54, 55). Despite these limitations, our data strongly suggest that NKG2DL expression levels are increased following radiochemotherapy. Together with the observation that NKG2DL can also be induced in TMZ-resistant cells (Supplementary Fig. S3C and S3D), this provides a rationale to investigate NKG2D-targeting therapies (27) also at tumor recurrence.

We demonstrate that the upregulation of NKG2DL upon TMZ or IR requires ATM (Fig. 3D; Supplementary Fig. S3G), which supports the concept that the DNA-damage response is one stimulus for the induction of NKG2DL (56). Consequently, ATM inhibitors may potentially counteract NKG2D-dependent antitumor immune effects. This needs to be considered in future trials evaluating the activity of such ATM inhibitors as radiosensitizers (57).

Although the net effect of NKG2D ligand induction is of rather small magnitude, it has important functional consequences. The induction of NKG2DL by TMZ or IR enhanced the immunogenicity of glioma cells, including GIC, and rendered the cells more susceptible to immune-mediated cytolysis (Fig. 4; Supplementary Fig. S4A). Chemotherapy and radiotherapy have various effects on tumor cells and the microenvironment comprising both immune-stimulatory and immune-suppressive mechanisms. Our study indicates that treatment-associated NKG2DL induction constitutes a relevant immune-stimulatory mechanism because inhibition of NKG2D signaling abrogated the enhanced cytolysis. Furthermore, tumor-infiltrating NK, CD4, CD8 T cells and to some extent also γδ T cells produced more IFNγ in an NKG2D-dependent manner upon treatment with TMZ and/or irradiation (Fig. 6D). This emphasizes the relevance of the NKG2D-mediated immune-stimulatory mechanism of TMZ and IR and adds another relevant mechanism to the concept of immunogenic cell death.

We did not only demonstrate that TMZ- and radiotherapy-mediated NKG2DL induction can be used as a strategy to render glioma cells more immunogenic but also that the full efficacy of TMZ and IR against glioblastoma depends on an intact NKG2D system. The survival benefit gained with these treatment modalities was diminished upon blockade of NKG2D signaling with an inhibitory but nondepleting antibody or in NKG2D knockout mice (Fig. 6A and B; Supplementary Fig. S6A–S6C). Inhibition or deficiency of NKG2D in mice did not result in a significant survival difference without additional treatment, suggesting that the basal expression levels of NKG2DL are too low to promote a relevant immune response (7, 10, 26, 34, 58).

The NKG2DL induction upon TMZ treatment or IR could provide a rationale for future studies investigating the synergistic application of these conventional treatment modalities with other NKG2D-based immunotherapeutic strategies (27). So far, one phase I study has used pure NK cells for adoptive immunotherapy in patients with recurrent malignant gliomas (59). However, no concomitant treatment with TMZ or IR was administered and additive or synergistic effects to this adoptive cell therapy need to be explored in future clinical trials.

In summary, the present dataset demonstrates the relevance of a so far unrecognized mechanism mediating antitumor effects of TMZ and IR that is likely to be clinically relevant. Based on our findings, further studies evaluating the combination of radiochemotherapy with additional NKG2D-based immunotherapeutic strategies should be considered for the treatment of glioblastoma.

P. Roth reports receiving speakers bureau honoraria from BMS, Novartis, and Novocure, and is a consultant/advisory board member for MSD, Roche, Virometix, Covagen, and Molecular Partners. No potential conflicts of interest were disclosed by the other authors.

Conception and design: T. Weiss, M. Weller, P. Roth

Development of methodology: T. Weiss, M. Pruschy, P. Roth

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Weiss, H. Schneider, M. Silginer, M. Pruschy, B. Polić

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Weiss, M. Pruschy, M. Weller

Writing, review, and/or revision of the manuscript: T. Weiss, H. Schneider, A. Steinle, M. Pruschy, B. Polić, M. Weller, P. Roth

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Weiss, A. Steinle, M. Weller, P. Roth

Study supervision: T. Weiss, M. Weller, P. Roth

This study was supported by grants from the Gertrud-Hagmann Foundation and the Swiss Cancer League (KFS-3478-08-2014) to P. Roth and “Hochspezialisierte Medizin Zurich” (HSM-2) to M. Weller and P. Roth.

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.