Targeting the Cytosolic Innate Immune Receptors RIG-I and MDA5 Effectively Counteracts Cancer Cell Heterogeneity in Glioblastoma§


  • Author contributions: M.G.: conception and design, financial support, provision of study material, collection/assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript; C.C. and D.T.: conception and design, collection/assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript; J.D.: conception and design, data analysis and interpretation, and final approval of manuscript; M.S. and H.N.: provision of study material, data analysis and interpretation, and final approval of manuscript; P.K., M.L., and A.P.: conception and design, provision of study material, data analysis and interpretation, and final approval of manuscript; J.M. and T.Q.: conception and design, provision of study material, data analysis and interpretation, manuscript writing, and final approval of manuscript; R.G. and A.W.: data analysis and interpretation and final approval of manuscript; R.R., R.F. and U.H.: data analysis and interpretation, manuscript writing, and final approval of manuscript; K.R.: provision of study material, data analysis and interpretation, manuscript writing, and final approval of manuscript; M.S. and R.B.: conception and design, data analysis and interpretation, manuscript writing, and final approval of manuscript; O.B.: administrative support, data analysis and interpretation, manuscript writing, and final approval of manuscript; G.H.: conception and design, financial support and final approval of manuscript; B.S.: conception and design, financial support, administrative support, provision of study material, collection/assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript. M.G., C.C., and D.T. contributed equally to this article. J.D. and M.S. contributed equally to this article.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    first published online in STEM CELLS EXPRESS February 8, 2013.


Cellular heterogeneity, for example, the intratumoral coexistence of cancer cells with and without stem cell characteristics, represents a potential root of therapeutic resistance and a significant challenge for modern drug development in glioblastoma (GBM). We propose here that activation of the innate immune system by stimulation of innate immune receptors involved in antiviral and antitumor responses can similarly target different malignant populations of glioma cells. We used short-term expanded patient-specific primary human GBM cells to study the stimulation of the cytosolic nucleic acid receptors melanoma differentiation-associated gene 5 (MDA5) and retinoic acid-inducible gene I (RIG-I). Specifically, we analyzed cells from the tumor core versus “residual GBM cells” derived from the tumor resection margin as well as stem cell-enriched primary cultures versus specimens without stem cell properties. A portfolio of human, nontumor neural cells was used as a control for these studies. The expression of RIG-I and MDA5 could be induced in all of these cells. Receptor stimulation with their respective ligands, p(I:C) and 3pRNA, led to in vitro evidence for an effective activation of the innate immune system. Most intriguingly, all investigated cancer cell populations additionally responded with a pronounced induction of apoptotic signaling cascades revealing a second, direct mechanism of antitumor activity. By contrast, p(I:C) and 3pRNA induced only little toxicity in human nonmalignant neural cells. Granted that the challenge of effective central nervous system (CNS) delivery can be overcome, targeting of RIG-I and MDA5 could thus become a quintessential strategy to encounter heterogeneous cancers in the sophisticated environments of the brain. STEM Cells 2013;31:1064–1074


Glioblastoma (GBM) is a brain tumor that accounts for one of the most devastating types of cancer. Optimized standards of care include surgical debulking combined with subsequent radiotherapy and chemotherapy. Yet, not every tumor cell seems susceptible—recurrence always occurs, and the median overall survival time is only 15 months [1, 2]. Key factors of GBM malignancy are cellular heterogeneity, the infiltrative nature of tumor cells, as well as a variety of resistance and immune escape mechanisms [3–5]. It is commonly accepted that successful therapy should address every one of these aspects. However, a “unifying” treatment approach has not yet been defined.

An appealing strategy that emerged during the last years is the mimicry of a viral infection. Viral nucleic acids are sensed by ubiquitously expressed cytosolic nucleic acid receptors that upon engagement trigger the release of antiviral and proinflammatory cytokines [6, 7]. In brain tumors, immune stimulation mimicking viral infections has so far mainly been explored by stimulation with synthetic long double-stranded RNA, specifically polyinosinic-polycytidylic acid (p(I:C)) with and without poly L-lysine (-LC) stabilization that mediate proinflammatory and anticancer signals via the endosomal toll-like receptor 3 (TLR3) [8–10] and, at least in animal models, the cytosolic innate immune receptor melanoma differentiation-associated gene 5 (MDA5) [11, 12].

In contrast to previous work, the major focus of this study was on the effects induced by specific activation of cytosolic innate immune receptors MDA5 and retinoic acid-inducible gene I (RIG-I) in human GBM. Application of respective ligands, that is, cytosolic p(I:C) and 5′-triphosphate RNA (3pRNA) has recently shown to induce promising anticancer effects in preclinical studies on extraneural types of solid cancer [13, 14]. An advantage of this approach is its potential to additionally stimulate apoptosis, which would represent a double-tracked antitumor mechanism encompassing immune response and cell death [13, 15, 16]. It is tempting to speculate that this strategy could also work efficiently in human brain cancer. Prerequisite to this approach, however, would be to know how specific and how sensitive heterogeneous tumor cell phenotypes, including those with stem-like properties, would respond and how susceptible nonmalignant neural cells are to the approach.

We recently introduced an experimental platform with which these questions can be addressed in a reductionist setting. Patient- and GBM-specific characteristics are mirrored in short-term expanded primary cell cultures (pGBMs) offering access to controlled in vitro analysis [17]. Ten of these pGBMs were applied in our study to specifically reflect and analyze the impact of cellular heterogeneity. pGBMs from routinely resected tumor tissue were studied next to pGBMs representing “residual GBM cells” derived from the surgical resection margin (pGBM Residual [primary “residual” GBM cultures];17). Additionally, primary cultures with (pGBM CSC) and without (pGBM w/o CSC) stem cell qualities were chosen for analysis based on their content of self-renewing, multipotent and tumorigenic cells (supporting information Fig. 1). For comparison, and as a control, we investigated a variety of human nonmalignant cellular specimens, that is, embryonic stem cell-derived neural stem cells (ESCd NSCs), induced pluripotent cell derived (iPSCd) neural stem cells (iPSCd NSC), microglia (iPSCd MC), and astrocytes (iPSCd Astro) as well as primary adult human neural progenitor cells (pAHNPs) (see Methods; supporting information Fig. 1).


GBM Tissue Samples and Cell Culture

GBM patient characteristics, tissue and cellular derivation procedures were described recently (patient numbers: 015, 016, 035, 046, 078, 106; [17]). Primary cultures from these samples were classified as “pGBM w/o CSCs” (numbers 015, 016) or “pGBM CSCs” (numbers 035, 046, 078, 106). In contrast to pGBMs w/o CSCs, the latter contain cells with self-renewing, multipotent, and tumor-initiating potential (supporting information Fig. 1). Hippocampus tissue (patient numbers 155; 157) was obtained from epilepsy surgery at the Department of Neurosurgery, University of Bonn Medical Centre. The studies were approved by the local Ethics committee; all patients provided informed consent. Tissue diagnosis and tumor grading was done based on the current classification of the World Health Organization [3] and was confirmed by two independent neuropathologists at the Department of Neuropathology, University of Bonn Medical Centre (the National Reference Center of Neuropathology). Handling of fresh biopsy samples, derivation of pGBMs, hippocampus tissue-derived pAHNPs (pAHNP-1, number 155; pAHNP-2, number 157) as well as methods for cellular propagation of primary cultures under defined in vitro conditions were described recently [17, 18].

U87 cells were maintained and analyzed in Dulbecco's modified Eagle's medium (DMEM)/F12-based 10% fetal calf serum (FCS) (Hyclone, Logan, UT, adherent conditions in standard plastic ware. The following human stem cell-derived cultures and their derivatives were used as nonmalignant neural controls in this study.

Human NSCs

Long-term self-renewing NSCs (lt-NES) were originally derived from the human ESC line H9.2 (ESCd NSC-1) and from the human iPSC line PKa (iPSCd NSC-1). Conditions for the maintenance of lt-NES cells are detailed in [19, 20].

Gliogenic NSCs

Additionally, stably proliferating and clonogenic gliogenic neural stem cells were derived from H9.2 human embryonic cells (ESCd NSC-2) and from the iPS line iLB-C-35m-r4 (iPSCd NSC-2). These cells were generated using a retinoic acid-based differentiation paradigm followed by CD133-based fluorescence-activated cell sorting (FACS) (T.Q., unpublished data). Expansion was performed under adherent conditions in defined serum-free medium and in the presence of fibroblast growth factor-2 and epidermal growth factor. Astrocytic differentiation of iPSCd NSCs was achieved by growth factor withdrawal, in the presence of 10% FCS for 10–12 days (iPSCd Astrocytes).

Microglial Cells

Human microglial cells (iPSCd MC) were generated from the iPS line MP1 obtained by reprogramming from skin fibroblasts as previously described [21]. These cells proliferate without addition of growth factors and they were passaged 1:3 twice a week. The microglia phenotype was confirmed by flow cytometry (CD11b+, CD11c+, CD16/32+, CD36+, CD45+, CX3CR1+).

Generation of Natural Killer Cells and Coculture Experiments

Peripheral blood buffy coats from healthy human donors were received from the local blood bank. Natural killer (NK) cells within peripheral blood mononuclear cells (PBMC) were obtained by Ficoll-Hypaque density gradient centrifugation (Biochrom, Berlin, Germany, 4 × 105 PBMC were cultured overnight in 96-well flat-bottom wells in duplicates in 100 μl RPMI 1640 (Biochrom, Berlin, Germany), supplemented with 10% (v/v) heat-inactivated FCS (Invitrogen, Karlsruhe, Germany,, 1 mM L-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin (all from Sigma-Aldrich, Munich, Germany, with or w/o addition of 100 μl pGBM CSC supernatants (#046). For blocking of type I interferon (IFN) signaling, the mouse-anti human IFNα/β receptor chain 2 antibody (clone MMHAR-2; PBL, Piscataway, NJ) or the mouse-anti human IgG2A control antibody (R&D Systems, Wiesbaden, Germany, were used at final concentrations of 10 ng/μl each. For the NK degranulation assay and the lysis assay (see below), coculture experiments were performed by adding 4 × 104 pGBM CSCs to each well. The following conditions w/o addition of pGBM CSC were used as controls: Medium, untransfected pCA (short RNA with the sequence CACACACACACACACACACA), pCA used with Mirus transfection reagent, and pCA used with RNAiMAX.

NK Cell Degranulation Assay

Anti-CD107a (BD Pharmingen, Heidelberg, Germany, was added to each well (1 μl per well) of PBMC-pGBM CSC coculture. After 1 hour, monensin (Riedel-de Haen, Seelze, Germany) was additionally supplied and 3 hours later cells were washed and stained with Horizon-conjugated anti-CD3 and allophycocyanin (APC)-conjugated anti-CD56 antibody (BD Biosciences, Heidelberg, Germany, http://www. CD107a expression of CD56+/CD3 NK cells was determined by standard FACS analysis (see below).

Lysis Assay

DiO (0.03 μM; Invitrogen)-prelabeled pGBM CSCs were added to activated NK cells (see above) and the extent of pGBM CSC cell lysis was determined by FACS analysis after 24 hours of coculture as previously described [22]. Briefly, cells were labeled with Hoechst 33258 (0.05 μg/ml, Sigma) prior to flow cytometry. NK-mediated pGBM CSC lysis was determined as DiO+Hoechst33258+/(DiO+Hoechst33258++DiO+Hoechst33258) cells followed by subtracting spontaneous pGBM CSC lysis.

Immunostimulatory RNAs and Small Interfering RNAs

The synthetic dsRNA poly(I:C) was purchased from InvivoGen (Toulouse, France, For generation of DNA-template-dependent in vitro-transcribed 3pRNA, the T7-promoter region 5′-CAGTA ATAGGACTCACTATAG-3′ was hybridized with the promoter + template strand (5′-TTGTAATACGACTCACTATAGGGACG CTGACCCAGAAGATCTACTAGAA ATAGTAGATCTTCTGG GTCAGCGTCCC) and directly used as a template for in vitro transcription reaction. Nonstimulatory 20mer pCA-RNA (5′-CACACACACACACACACACA-3′) was purchased from Biomers (Ulm, Germany, Poly L-lysine stabilized p(I:C) (p(I:C)-LC) was a friendly gift from Hideho Okada, M.D. Ph.D., Brain Tumor Center, Pittsburgh. For quantification of transfection efficacy, we used a nonstimulatory control RNA (5′-UAAUUUAUGCGGCC CAAGAC-3′) hybridized with the matching antisense strand from Biomers and labeled with Cy3. Small interfering RNAs (siRNAs) and nonsilencing control siRNAs were designed as described recently [13] (specific sequences are listed in supporting information Table 1).

Transfection Procedures

3pRNA was transfected at a concentration of 500 ng/ml, unless otherwise noted (see Figs. 1, 3; supporting information Fig. 2) after 15 minutes incubation with RNAiMAX (0.15 μl per 100 ng RNA in 20 μl Opti-MEM) (Life Technologies, Darmstadt, Germany, Untransfected p(I:C) and p(I:C)-LC were supplied in Opti-MEM (Life Technologies) and applied to cells at the respectively indicated concentrations. p(I:C) was transfected using TransIT-LT1 (Mirus Bio LCC, Madison, 0.67 μl TransIT-LT1 was incubated with 100 ng RNA in 50 μl Opti-MEM for 20 minutes before added to cells in indicated concentrations. Opti-MEM was used for mock transfection. pCA served as nonstimulatory control (transfected and untransfected, as indicated above).

Figure 1.

Immunostimulation and ligand recognition. (A): Quantitative reverse transcription-polymerase chain reaction data showing levels of putative RNA molecule receptors before (−IFN β) and after (+IFN β) stimulation. At baseline, pGBM CSCs (n = 4, triplicates, mean ± SEM) and U87 cells express little TLR3, MDA5, and RIG-I mRNA transcripts and no TLR7. The expression of TLR3, MDA5, and RIG-I transcripts was upregulated by stimulation with IFN β. (B): Stimulation of pGBM CSCs with p(I:C) (concentration c1, 500 ng/ml; c2, 10 μg/ml), transfected p(I:C), and 3pRNA (both at 500 ng/ml) led to production of CXCL10, whereas of pIC-LC showed no immunostimulatory effect. Note: CXCL10 is a chemokine that attracts many immune cells (e.g., T and NK cells) (*, p < .05). (C): Secretion of type I IFN from pGBM CSCs could only be induced by stimulation with transfected p(I:C) and 3pRNA. Stimulation of the U87 glioma cell line did not lead to measurable levels of CXCL10 and type I IFN (data not shown). (D): pGBM CSCs were treated with siRNA specific for RIG-I, MDA5, and TLR3 (see insets for documentation of specificity) or control siRNA. CXCL10 levels were subsequently measured after stimulation with 3pRNA (left) or transfected p(I:C) (right). Data imply that 3pRNA mediates effects via RIG-I and transfected p(I:C) via MDA5 receptors (*, p < .05; ***, p < .001). Abbreviations: IFN, interferon; MDA-5, melanoma differentiation-associated gene 5; p(I:C), polyinosinic-polycytidylic acid; p(I:C)-LC, poly L-lysine stabilized p(I:C); pGBM CSC, stem cell-enriched primary GBM cultures; RIG-I, retinoic acid-inducible gene I; siRNA, small interfering RNA; TLR3, toll-like receptor 3; 3pRNA, 5′-triphosphate RNA.

siRNAs were transfected in 96-well plates at 20 nM in 150 μl medium with 0.125 μl Lipofectamine RNAiMAX (Life Technologies). 48 hours later, transfection of 3pRNA or p(I:C) was conducted in parallel with a second siRNA treatment (both as described above).

Analysis of Cellular Viability

Experimental stimulation was conducted 24 hours after seeding 104 pGBM or U87 cells per well on laminin/poly-L-ornithine (Life Technologies)-coated 96-well plates. The following seeding concentrations were used for human nonmalignant cells (cells per well): ESCd NSC-1, 3 × 104; ESCd NSC-2, 3 × 104; iPSCd NSC-1, 3 × 104; iPSCd NSC-2, 4 × 104; iPSCd Astro, 4 × 104; iPSCd MC, 4 × 103; pAHNP-1, 104; pAHNP-2, 104. Unless otherwise indicated, metabolic activity was determined as a measure of cellular viability using the alamarBlue assay according to the manufacturer's recommendations (Life Technologies) at 24 and 96 hours after stimulation. Fluorescence was measured using an Infinite 200 microplate reader (Tecan) at λex = 540 nm and λem = 590 nm. Experiments were performed in triplicates for each sample. Medium, untransfected pCA, pCA used with Mirus transfection reagent, and pCA used with RNAiMAX served as a control.

Annexin V-Based FACS Analysis

1 × 105 cells were collected at 24 hours following stimulation, suspended in 100 μl AnnexinV buffer and incubated with 5 μl Annexin V-fluoresceinisothiocyanat (FITC) for 1 hour at room temperature. To distinguish between living and dead cells, labeling with 1.2 μg/ml Hoechst 33258 (H33258) was used. Analysis was performed using a LSRII equipped with FACSDiva Software (BD Bioscience) on standard settings. 2 × 104 cells were counted per measurement. Vital cells were classified as Annexin V/H33258.

Neurosphere Assay

To estimate the frequency of self-renewing clonogenic cells, the neurosphere assay was used according to established protocols [17, 23]. Neurospheres were quantified at 21 days in culture, triturated to a single cell suspension, and replated for analysis of the secondary and tertiary neurospheres. Multipotency was determined by plating a representative fraction of secondary and tertiary neurospheres onto laminin/poly-L-ornithine-coated glass coverslips allowing differentiation for 2–3 weeks before fixation in 4% paraformaldehyde. Immunofluorescence analysis was performed according to standard protocols [23, 24] using antibodies against ßIII tubulin (Promega, Madison, WI,; monoclonal mouse, 1:1,000) and glial fibrillary acidic protein (GFAP) (DAKO, , Glostrup, Denmark,, polyclonal rabbit, 1:600). Cell nuclei were visualized with 4',6-diamidino-2-phenylindole (DAPI) (Sigma).

Quantitative Reverse Transcription-Polymerase Chain Reaction

Total RNA was isolated using the RNeasy Mini Kit (Qiagen, Hilden, Germany, according to the manufacturer's instructions. Quantification of RNA concentration was performed with a Nanodrop (Peqlab, and cDNA was synthesized using the Expand Reverse Transcriptase (Roche, Basel, Switzerland, Quantitative reverse transcription-polymerase chain reaction (qRT-PCR) was performed on a realplex 4 Mastercycler Epp Gradient S (Eppendorf) using the following PCR conditions: 95°C for 2 minutes, followed by 40 cycles of 95°C for 15 seconds, 60°C for 20 seconds, and 72°C for 30 seconds. Primers for quantification of Musashi, Nestin, and Sox-2 were described recently [17]. For RIG-I, MDA5, TLR3 and TLR7 cDNA was amplified in a total volume of 20 μl using LightCycler 480 System (Roche). Primers and Probe designs were performed using Universal Probe Library (Roche). Probe #88 for TLR3, #18 for TLR7, #82 for MDA5, and #18 for RIG-I (all Roche) were used. PCR conditions: 95°C for 10 minutes, followed by 50 cycles of 95°C for 10 seconds, 60°C for 30 seconds, and 72°C for 1 second. The following primer pairs obtained from Metabion (Martinsried, Germany, were used: RIG-I: left 5′-gaa aga ctt ctt cag caa tgt cc-3′, right 5′-ctt tct agt tcc tgc agc ttt tct-3′; MDA5: left 5′-ccg aga gaa gat gat gta taa agc ta-3′, right 5′-ttt gca tct gta att cca aaa tct-3′; TLR3: left 5′-aag gct agc agt cat cca aca-3′, right 5′-agc aac ttc atg gct aac agt g-3′; TLR7: left 5′-gaa cga caa tga cat ctc ttc ct-3′, right 5′-aaa aca tct aag tga ttt cct ctg aat-3′. To induce receptor expression cells were preincubated with 300 IU/ml recombinant IFNβ (Milteny Biotech, Bergisch Gladbach, Germany,

Data analysis was performed with the Mastercycler Epp Realplex Software (Eppendorf). Mean values were calculated from triplicate qRT-PCR reactions. Expression was normalized to GAPDH.

Single Nucleotide Polymorphism Analysis

The analysis of genomic aberrations was conducted using the Illumina Human610-Quad Bead Chip and methods described recently [17].

Protein Preparation and Immunoblot Analysis

Adherent and supernatant cells were lysed in buffer containing 50 mM Tris, pH 7.4, 0.25 M NaCl, 1 mM EDTA, 0.1% Triton X-100, 0.1 mM EGTA, 5 mM Na3VO4, 50 mM NaF, and protease inhibitors (Complete, Mini, EDTA-free; Roche). Gel electrophoresis and blotting were carried out with 5–10 μg denatured protein lysate by using the Xcell SureLock Mini-Cell apparatus with 4%–12% gels in MES SDS buffer and PVDF membranes according to the manufacturer's protocol (Invitrogen). After blocking (Western Blocking Reagent in phosphate buffered saline (PBS); Roche), blots were incubated with primary antibodies for 1 hour at room temperature or overnight at 4°C. Following washes with 0.1% Tween 20 (Sigma-Aldrich) in PBS and 1 hour of incubation with horseradish peroxidase (HRP)-conjugated secondary antibodies, blots were washed and visualized (ECL Plus Western Blotting Detection System; Amersham, GE Healthcare, Protein levels of β-actin were analyzed as a control for constant loading and transfer. The following antibodies were used: anti-caspase-3, anti-Bcl-xL, Bcl-2, Bcl-w, and HRP-conjugated secondary antibodies (New England Biolabs, Ipswich, MA,; Mcl-1 (BD Pharmingen); anti-Noxa (N-15; Santa Cruz Biotechnology Inc., Heidelberg, Germany,; anti-β-actin (AC-15; Sigma-Aldrich).

Measurement of Cytokines and Chemokines

After the indicated time points, cell culture supernatants were collected and stored at −20°C. Concentrations of IP-10 were detected using the BD OptEIA IP-10 ELISA set (BD Biosciences) according to the manufacturer's recommendations and analyzed by Apollo 8 LB 912 Micro Plate Reader (Berthold Technologies, Bad Wildbad, Germany, Type I IFN was calculated using HEK-BlueTM IFN-α/β cells from InvivoGen (Toulouse, France). 5 × 104 cells were cultured in 96 flat bottom wells in DMEM plus FCS (both Life Technologies) and Ciprofloxacin (Bayer, Leverkusen, Germany, with addition of 20 μl of cell culture supernatant or different concentrations of recombinant IFN β (Miltenyi Biotech, Bergisch Gladbach, Germany, serving as standard. After 24 hours, a final concentration of 1 mg/ml pNPP substrate (Sigma-Aldrich) was added to supernatant and analyzed after 30 minutes shaking at RT at 405 nm in Apollo 8 LB 912 Micro Plate Reader (Berthold Technologies).

Tumor Xenograft Experiments

pCA (control)-, transfected p(I:C)-, or 3pRNA-stimulated pGBMs (#046; #106) were harvested, counted, and resuspended in 0.1% DNase/PBS. Cellular viability was confirmed via trypan blue exclusion. Cells were injected stereotactically into the striatum of 12-week-old Scid Beige−/− mice (0.5 mm anterior, 1.5 mm lateral, 3 mm deep). Mice were monitored daily and euthanized upon presentation with signs of distress/neurological symptoms or evident weight loss. For histological analysis, anesthetized mice underwent vascular perfusion, brains were removed, cryoprotected, and cut on a cryostat (Leica) at 20 μm thickness. Standard methods were applied for histological phenotyping of xenograft tissue [17]. The local regulatory agency (Landesamt für Natur, Umwelt und Verbraucherschutz, Recklinghausen) approved all studies involving animals. All animals were cared according to their guidelines. SCID Beige mice were purchased from the Jackson Laboratory (Charles River, Sulzfeld, Germany,

Statistical Analysis

SPSS v.19.0 software was used for statistical analysis. Data presented with error bars represent mean ± SEM. Two-sided Student's t test was used to determine the statistical significance of quantitative variables between two groups. For multiple comparisons, ANOVA and Tukey's post hoc test were used. The Bonferroni correction was applied for multiple-comparison correction.


Stimulation of Cytosolic RNA Receptors RIG-I and MDA5 in Human Primary GBM Cells Triggers an Innate Immune Response

To demonstrate that primary GBM cells are accessible for immune stimulatory mechanisms, we first investigated the expression and function of putative receptors. pGBM CSCs (n = 4) revealed low baseline levels of TLR3, MDA5, and RIG-I, but expression levels of these three innate immune receptors strongly increased upon activation of the type I IFN-mediated feedback loop (18–60-fold; Fig. 1A; left panel). An upregulation of TLR7, another endosomal immune receptor for single and double stranded RNA [7], was not detected. Similar findings were apparent in U87 glioma cells (Fig. 1A; right panel). Furthermore, a substantial release of the immune cell attracting and antiangiogenic chemokine CXCL10 was observed following application of stimulatory RNA molecules. The secretion of CXCL10 could be triggered by transfected p(I:C), untransfected p(I:C), and 3pRNA (Fig. 1B), respectively. By contrast, only transfected p(I:C) and 3pRNA stimulated the release of type I IFN (Fig. 1C). Strongest effects were elicited by 3pRNA in these experiments; very little specific responses were recorded following application of p(I:C)-LC (supporting information Fig. 2). For verification of receptor-ligand specificity, siRNA was designed to effectively knockdown the expression of RIG-I, MDA5, or TLR3. In 3pRNA-treated pGBM CSCs, only the inhibition of RIG-I led to a decreasing CXCL10 response (Fig. 1D, left). Treatment with transfected p(I:C), by contrast, revealed a reduction of CXCL10 levels upon inhibiting MDA5 but not TLR3 (Fig. 1D, right). This suggested 3pRNA as ligand for RIG-I and MDA5 as a target of transfected p(I:C) in pGBM CSCs, and, that their specific interactions led to an induction of CXCL10 secretion. Together, the data indicated that primary GBM cells could be stimulated to increase the expression of the cytosolic RNA molecule receptors MDA5 and RIG-I, and that these receptors can be targeted by transfected p(I:C) and 3pRNA, respectively, to mediate characteristic immune responses.

Many innate and adaptive immune cells could be impacted by this mechanism. To gain first insights, we focused on NK cells because this population is known to contribute significantly to the clearance of experimentally induced malignant glioma [25]. Naïve CD56+/CD3− human donor NK cells were exposed to supernatants from stimulated pGBM CSCs and subsequently to untreated pGBM CSCs (supporting information Fig. 3A, left). CD107a-FACS analysis was used to measure the extent of toxic degranulation as an indicator for increased NK cell activity [26]. Results implied that neither naïve pGBM CSCs alone (data not shown) nor the supernatants of stimulated pGBM CSCs alone were sufficient to activate NK cells. To activate NK cells, naïve pGBM CSCs had to be provided together with the supernatant of 3pRNA-stimulated pGBM CSCs. Stimulation with p(I:C)-LC or transfected p(I:C), by contrast, yielded only modest effects in this setting (supporting information Fig. 3A, right). Type I IFN could be identified as a contributing factor in this process. In the presence of naïve pGBM CSCs addition of type I IFN mediated NK cell activation, while neutralizing antibodies attenuated the immunostimulatory activity on NK cells (supporting information Fig. 3B). Moreover, activated NK cells accomplished pGBM cell lysis as a functional consequence (supporting information Fig. 3C, left). This lysis was facilitated by addition of type I IFN and it was inhibited with respective neutralizing antibodies (supporting information Fig. 3C, right).

Transfected p(I:C) and 3pRNA Additionally Induce Apoptosis in GBM CSCs

In a second series of experiments, we determined the capability of RNA molecules to induce cell death. Four pGBM CSCs (see Methods, supporting information Fig. 1) and, for reference, the glioma cell line U87 were supplied with transfected 3pRNA and p(I:C) as well as with untransfected p(I:C) and p(I:C)-LC. Metabolic activity was determined as a measure of cellular viability using the alamarBlue assay at 24 and 96 hours postexposition (Fig. 2A). In these experiments, only transfected 3pRNA and transfected p(I:C) were shown to impact on cell survival and metabolism. At 24 hours, a decline of cellular viability was most evident after application of transfected p(I:C). At 96 hours, p(I:C) and 3pRNA exposition led to near abolishment of tumor cell viability. By contrast, untransfected p(I:C) and p(I:C)-LC did not affect the cell's viability, even at 20-fold increased concentrations (Fig. 2A, inset, c2) that are known to effectively activate TLR3 [27]. FACS-based analysis confirmed this demonstrating a strongly increased frequency of apoptotic and dead cells in samples treated with transfected 3pRNA and p(I:C) but not after application of untransfected p(I:C) (Fig. 2B). This suggested that apoptosis was mediated by activation of MDA5 and RIG-I but not TLR3. Cell viability of pGBM CSCs was monitored for several days thereafter (Fig. 2C) and the onset of apoptotic cell death could be confirmed by demonstrating activated caspase 3 (Fig. 2D, left). In melanoma cell lines, these effects are known to be mediated by activated Noxa, a key protein among the BH3-only proteins of the Bcl-2 family that are important initiators of mitochondrial apoptosis [13]. Similarly, we here found an induction of Noxa in 3pRNA- and p(I:C)-transfected pGBM CSCs (Fig. 2D, right).

Figure 2.

Antineoplastic activity. (A): Graphs represent alamarBlue-based cellular viability data following application of immunostimulatory RNA molecules to pGBM CSCs (left, n = 4, triplicates, mean ± SEM) and to U87 glioma cells (right, triplicates, mean ± SEM). The inset depicts data generated with two concentrations of p(I:C) and p(I:C)-LC, that is, standard (c1, 500 ng/ml) and high (c2, 10 μg/ml) (*, p < .05; ***, p < .001). (B): Graphs represent Hoechst/Annexin V-based fluorescence activated cell sorting analysis data to assess cell death and apoptosis following stimulation of pGBM CSCs (left, n = 4, triplicates, mean ± SEM) and U87 glioma cells (right, triplicates, mean ± SEM). Insets show characteristic flow cytometric measurements (exemplified by pGBM CSC case #046). (C): Time- and concentration-dependent monitoring of the course of metabolic activity of pGBM CSCs following application of transfected p(I:C) or 3pRNA (n = 4, triplicates, mean ± SEM). (D): Caspase-3 and its active cleaved subunits (17, 19 kDa) as well as Noxa protein were visualized by immunoblotting. β-Actin served as loading control. (E): The graphs demonstrate quantitative data from the NSA (left; inset: NS-size calculation) and quantitative reverse transcription-polymerase chain reaction analysis on stem cell-typic mRNA expression levels of pGBM CSCs that were collected 5 days postapplication of immunostimulatory RNA molecules. Note: Results are calculated as percent of the nonstimulatory pCA (RNA)-control. Abbreviations: FITC, Fluoresceinisothiocyanat; NSA, neurosphere assay; pGBM CSC, stem cell-enriched primary GBM cultures; p(I:C), polyinosinic-polycytidylic acid; 3pRNA, 5′-triphosphate RNA.

GBM malignancy has frequently been attributed to the activity of stem-like, that is, self-renewing, multipotent, and tumorigenic cancer cells [28]. We thus sought to decipher and discriminate the influence of stimulatory RNA molecule application on the stem-like tumor cell population in our paradigm. Experiments based on the rationale that decreasing or increasing frequencies of stem-like cells in pGBM CSCs following application of stimulatory RNA molecules would indicate a selective targeting of tumor cells either with or without stem cell characteristics. Noteworthy, pGBM CSCs are commonly referred to as “cancer stem cell cultures.” These cultures are enriched for stem-like identities, but they are nevertheless heterogeneous. pGBM CSCs also contain and maintain tumor cell identities without self-renewing, multipotent, and tumor-initiating potential. For analysis, cells from four pGBM CSCs were collected at two time points after 3pRNA and p(I:C) application: (a) at the maximum of their inhibition (i.e., 5 ± 1 days) and (b) after re-establishing confluent cell cultures (i.e., at 28 ± 14 days). As the population of stem-like tumor cells may encompass more than just one phenotype, several investigational strategies were applied. The neurosphere assay [17, 23] was used to estimate the frequency of self-renewing and multipotent cells; qRT-PCR to evaluate the gene expression of typical neural stem cell markers; xenograft transplantation to assess for tumorigenicity. Not one of these assays, however, revealed a difference between treated and untreated pGBM CSCs (Fig. 2E; supporting information Fig. 4). Thus, while 3pRNA and p(I:C) strongly reduced the overall cell number via induced apoptosis, the relative number of stem-like cells in the surviving GBM population remained unchanged. We concluded from these data that tumor cells with and without stem cell properties can be targeted similarly by application of 3pRNA and p(I:C).

The Targeting Efficacy of Transfected p(I:C) and 3pRNA Can Be Validated in a Variety of pGBM Cell Platforms

Even though pGBM CSCs, as used above, can mirror patient-specific traits in vitro [17, 29, 30] and even though human cells are broadly considered for evaluating novel treatment approaches, their overall predictive value still needs to be established [31]. Considering this, and additionally the known heterogeneity in GBM cell and tissue samples, we additionally investigated two alternative pGBM cell platforms: (a) pGBMs w/o CSCs, representing patient-specific primary cultures that do not contain cells with self-renewing activity (see Patients and Methods section). Recent studies have indicated that cell populations like these might carry a variety of unexpected malignancy traits [32]. (b) pGBM Residual cells. These primary cultures exemplify a distinct population of tumor cells that remain in the patient after completion of standard tumor resection [17]. Experimentation with both of these alternative GBM cell platforms revealed a high targeting efficacy of transfected p(I:C) and to a lesser degree of 3pRNA. The viability of pGBM Residual cells and of pGBMs w/o CSCs was found decreased at 24 hours and even more decreased at 96 hours after exposition (Fig. 3A, 3B, left panel). Also, the application of transfected p(I:C) and 3pRNA led to a strong increase of CXCL10 secretion (Fig. 3A, 3B, right panel). Untransfected p(I:C), by contrast, induced an immune response in treated cells but exhibited only moderate effects on cellular viability. p(I:C)-LC application showed very little impact on both endpoints.

Figure 3.

Targeting GBM cell heterogeneity. (A): Graphs represent alamarBlue-based cell viability (left) and CXCL10 secretion (right) data following application of immunostimulatory RNA molecules to #015 and #016 cells (triplicates, mean ± SEM). Both cases represent patient-specific primary GBM cells that did not contain stem-like tumor cells (*, p < .05; **, p < .01; ***, p < .001). (B): Similar data presentation for Residual GBM cells (n = 4 cases, mean ± SEM, triplicate analysis). Note, that untransfected p(I:C) and p(I:C)-LC were analyzed at two concentrations, that is, at standard (c1, 500 ng/ml) and at high concentration (c2, 10 μg/ml). Controls: Medium, untransfected pCA, pCA used with Mirus transfection reagent, and pCA used with RNAiMAX (*, p < .05; **, p < .01). Abbreviations: pGBM w/o CSC, primary GBM cultures without stem cell properties; pGBM Residual, primary residual GBM cultures; p(I:C), polyinosinic-polycytidylic acid; p(I:C)-LC, poly L-lysine stabilized p(I:C); 3pRNA, 5′-triphosphate RNA.

Transfected P(I:C) and 3pRNA Induce Only Marginal Toxicity in Human Nonmalignant Cells

It has recently been shown that melanoma cell lines respond more efficiently to RIG-I- and to MDA5-induced apoptosis than control cultures of nonmalignant melanocytes, fibroblast, and keratinocytes [13]. Consequently, we investigated the responses of human nonmalignant neural cells to RIG-I- and MDA5-induced apoptosis signaling. Compared to the high sensitivity of the various pGBM cell populations to transfected p(I:C) and 3pRNA (Figs. 2 and 3), only a moderate decline of cellular viability was observed in the various nonmalignant human neural stem/progenitor cell sources investigated here, that is, in ESCd-NSCs, iPSCd-neural progeny, and in pAHNPs (Fig. 4A–4C, left panel). Furthermore, with exception of iPSCd MC (microglia), human nonmalignant neural cells were capable to respond to immune stimulation as evident from increasing CXCL10 levels in the supernatant (Fig. 4A–4C, right panel). Thus both, malignant and nonmalignant brain cells were accessible to RNA-mediated immune stimulation, but only the pGBM cells showed a substantial decrease of their viability.

Figure 4.

Toxicity/tumor cell specificity. Graphs represent alamarBlue-based metabolic (left) and CXCL10 secretion-based (right) response data following application of immunostimulatory RNA molecules to human nontumor cells (triplicates, mean ± SEM). (A): Two independent ESCd NSCs were investigated (*, p < .05; **, p < .01). (B): Among the tested iPSCd cell populations were NSCs (n = 2), as well as astrocytes, and microglia (*, p < .05; **, p < .01). (C): Two independent, patient-specific epilepsy surgery (hippocampectomy)-derived pAHNP populations were investigated (case #155, pAHNP-1; #157, pAHNP-2). Untransfected p(I:C) and p(I:C)-LC were analyzed at two concentrations, that is, at standard (c1, 500 ng/ml) and at high concentration (c2, 10 μg/ml). Controls: Medium, untransfected pCA, pCA used with Mirus transfection reagent, and pCA used with RNAiMAX served as a control (*, p < .05; **, p < .01). (D): Western blot analysis of Bcl-2 expression levels in naïve pGBM CSCs (left three lanes; numbers 035, 046, 078) versus naïve nonmalignant human control cells (right three lanes; iPSCd NSC cell populations). Note, a substantial decrease of Bcl-2-expression was observed in eight of nine investigated pGBM cultures (data not shown). Abbreviations: ESCd NSC, embryonic stem cell-derived neural stem cells; iPSCd NSC, induced pluripotent stem cell derived neural stem cells; p(I:C), polyinosinic-polycytidylic acid; p(I:C)-LC, poly L-lysine stabilized p(I:C); pGBM CSC, stem cell-enriched primary GBM cultures; 3pRNA, 5′-triphosphate RNA; pAHNP, primary adult human neural progenitor cells.

Similar to pGBMs, nonmalignant human neural cells responded to application of transfected p(I:C) and 3pRNA with an increased expression of the BH3-only protein Noxa, indicating activation of the mitochondrial apoptosis pathway (supporting information Fig. 5A). Antiapoptotic Bcl-xL has been identified as a critical factor in the rescue of nonmalignant skin cells from Noxa-mediated apoptosis [13]. However, protein expression analysis in nonmalignant neural cells and pGBMs suggested that Bcl-xL levels did not change after application of p(I:C) or 3pRNA (supporting information Fig. 5A, 5B). Also, Bcl-xL was not found increased in naïve nonmalignant human neural cells, and it was not consistently decreased in naïve pGBMs (supporting information Fig. 5C). By contrast, investigation of other antiapoptotic proteins (i.e., Mcl-1, Bcl-w, Bcl-2) intriguingly revealed that Bcl-2 was expressed at higher levels in nonmalignant neural populations whereas pGBMs displayed low Bcl-2 levels (Fig. 4D; supporting information Fig. 5).

To finally assess the uptake efficacy of RNA molecules in malignant versus nonmalignant cell populations, cells were provided with Cy3-labeled nonstimulatory control RNA (Patients and Methods section). FACS analysis demonstrated a striking enrichment of fluorescent RNA in both cell populations (>90%, respectively), irrespective of the transfection reagents used (RNAiMAX or Mirus) (supporting information Fig. 6). Based on the similar transfection efficacy data, the divergent responses of malignant and nonmalignant brain cells to RNA-mediated immune stimulation likely result from an employment of different signaling pathways.


Conventional concepts of brain tumor therapy aim at directing growth inhibition, death, or differentiation of cancer cells. Indirect approaches include, for example, the inhibition of tumor-related angiogenesis, vaccination, cytokine-mediated immune modulation, as well as unspecific immune stimulation via TLRs (for review see9, 33, 34). Not many approaches, if any, are able to integrate direct and indirect antitumor activity. In this regard, targeting the cytosolic immune receptor system could represent a valuable alternative, combining the stimulation of an innate immune response with a direct induction of mitochondrial apoptosis in cancer cells. Inspired from recent preclinical studies in melanoma and ovarian cancer [13, 14, 22], we here applied this strategy successfully to human brain cancer.

Therapy of brain tumors, particularly of malignant gliomas, is considered demanding for many reasons. One of the most critical challenges, that is, cancer heterogeneity might become resolvable based on the findings of our study. Heterogeneity of cancer cells involves phenotype, genotype, and function, for example, the coexistence of cancer cells with and without stem cell characteristics, which in concert are made responsible for therapy resistance, relapse, and the overall fatal prognosis of GBM patients [5, 35, 36]. To address these issues, we investigated an array of pGBMs encompassing a broad spectrum of possible cancer cell phenotypes, including stem-like and Residual pGBM cell populations. Notably, the data of our study implied that every pGBM cell population from this portfolio could effectively be engaged to promote cell-autonomous and immune cell-mediated antitumor responses. Cytosolic delivery of p(I:C) and 3pRNA could be established as most potent tool for the coinduction of an innate immune response and apoptosis. Data from siRNA knockdown experiments and from the comparison of transfected versus untransfected p(I:C) furthermore corroborated that these effects were specifically mediated by MDA5 and RIG-I, respectively. In this regard, our observations confirmed mechanistic insights gained from studies on other solid cancer entities [13, 14, 22]. Remarkably, even high concentrations of untransfected p(I:C) did not result in relevant proapoptotic signaling, despite relevant CXCL10 responses were detected in pGBMs. Thus, to induce a most effective anticancer effect, our data may highlight the need for an effective targeting of the cytosol.

The accessibility of pGBM cells for immune stimulatory mechanisms nevertheless surprised. Traditionally these cancer cells are known to evade detection by the immune system [34, 37]. The possibility to induce IFNs directly within GBM tissue may thus provide a unique option to bypass the immunological barrier posed by CNS tumors, and our study exposed MDA5 and RIG-I as promising therapeutic targets. Analysis of ligand-mediated effects in pGBMs suggested that p(I:C)-initiated declines of cellular viability occurred at lower concentrations and at earlier time points compared to the effects induced by 3pRNA. By contrast, 3pRNA exhibited a more pronounced immune response. Intriguingly, both approaches effectively targeted cancer cells without similarly affecting a variety of human nonmalignant neural cell populations. The underlying mechanism of this tumor cell specificity could not be explained based on published data from non-neural solid tissue cancer. In contrast to studies on melanoma and noncancerous skin cells [13], the investigation of Bcl-xL expression levels alone did not argue for an involvement in brain tumor cell specificity. Rather, we observed Bcl-2 to be strongly reduced in GBM cells compared to nonmalignant neural cells. This might serve as a first indication for an employment of different pathways in GBM and melanoma that result in similarly increased sensitivities to RIG-I- and MDA5-induced apoptosis. These data open new paths for future research and translational studies. Granted that the challenge of effective CNS delivery can be overcome, stimulating RIG-I and MDA5 could represent a valuable alternative in the spectrum of current GBM treatment strategies. The resulting double-tracked tumor defense mechanism of immune response and cell death separates this approach from the standard chemotherapy and irradiation regimens and from other therapeutic attempts involving mechanisms of pathogen recognition and activation of innate immunity (e.g., TLR-mediated). The marginal cytotoxic effects observed in nontumor human neural cell populations in vitro provide an additional benefit that encourages further exploration for future brain tumor therapy. As our study also reports that MDA5- and RIG-I mediated response mechanisms could impact on NK cells, further development of p(I:C)- and 3pRNA-treatment strategies for CNS disease certainly needs to consider long-term effects of toxicity. Future studies could also address the question to what degree the different cellular subpopulations of GBM are targeted. A single application of p(I:C), for instance, resulted in a more pronounced effect on pGBM Residual cells in our cellular system. On the other hand, it is noteworthy that 3pRNA would allow molecular refinement to optimize antitumor efficacy. In perspective, molecular integration of siRNA components could be explored to synergize RIG-I-mediated antitumor activity with additional gene silencing effects in analogy to Poeck et al. [38]. This idea would be supported by recent progress in establishing the chemical synthesis of 3pRNA [39].


In conclusion, our study provides first in vitro evidence for an effective targeting of cancer cell heterogeneity in GBM via stimulation of the cytosolic innate immune receptors RIG-I and MDA5. If the in vitro observed tumor cell specificity and the low levels of neurotoxicity could be confirmed in patients, new avenues would open for cancer research and GBM therapy.


This work was funded by the Lichtenberg program of the VW foundation to B.S. Additional support was received from a BONFOR program grant of the Medical Faculty of the University of Bonn to M.G. and by grants of the Deutsche Forschungsgemeinschaft SFB670, SFB704, SFB832, KFO177 as well as a BMBF Go-Bio grant to G.H.; K.R and H.N. were supported by the Deutsche Forschungsgemeinschaft (KFO177, SFB704, and FOR1336) and the Hertie-Foundation. R.B was supported by the German Cancer Aid (Grant 107805 and the Melanoma Research Network) and by the German Research Foundation (DFG) Grant GK 1202. We thank the Departments of Neurosurgery and Neuropathology at the University of Bonn Medical Center for their assistance in tumor procurement, and processing of tissue and paraffin-embedded samples. We thank M. Keller, R. Eisenreich, H. Höfer, A. Leinhaas, C. Kammerbauer, and B. Lüdenbach for their technical assistance. lt-NES® is a registered trademark of LIFE & BRAIN GmbH. p(I:C)-LC (Hiltonol) was a kind gift from ONCOVIR, Inc., WA. O.B. is director of Life & Brain, GMBH; A.P. and D.T are employed by LIFE & BRAIN, GMBH; and B.S. and M.G. conduct contract-based research with LIFE & BRAIN, GMBH. J.M. is currently affiliated with Laboratory of Genetics, The Salk Institute for Biological Studies, La Jolla, CA.


O.B.: leadership position, intellectual property rights/inventor/patent holder, and ownership interest (all LIFE & BRAIN GmbH); G.H.: intellectual property rights/inventor/patent holder of RIG-I ligand.