Immunostimulatory RNA Oligonucleotides Induce an Effective Antitumoral NK Cell Response through the TLR7¹

Viruses are recognized by the innate immune system through endosomal and cytosolic receptors that selectively detect conserved pathogen structures (1). Molecular patterns specific to nucleic acids from viruses are recognized by the cytoplasmic helicases retinoic acid-inducible gene I (RIG-I)⁴ and melanoma differentiation-associated gene 5 (MDA-5) and by endosomal receptors of the TLR family, the TLRs 3, 7, 8, and 9 (1, 2, 3, 4, 5, 6, 7, 8). Binding of nucleic acid ligands to their respective receptor activates intracellular signaling cascades that rapidly induce innate and adaptive immune responses (1).

An important goal of therapeutic cancer vaccines is the induction of Ag-specific responses that mediate protective immunity against tumors. Because of their immunostimulatory properties, nucleic acid ligands for TLRs are powerful tools that can be used to boost tumor-specific immune responses. Indeed, inclusion in cancer vaccines of CpG oligodeoxynucleotides, that bind TLR9, leads to the production of IFN-γ and potentiates immune responses to tumor Ags in patients (9). Within the immune system, human TLR9 is expressed mainly on B cells and a subset of dendritic cells (DC), the plasmacytoid DC. The receptors TLR7 and TLR8 that are activated by ssRNA of viral origin are expressed in addition by human myeloid DC and monocytes that are essential for Ag presentation and for the initiation of immune responses against tumor Ags (5, 6, 10, 11). We have recently described RNA sequences that stimulate immune responses through the TLRs 7 and 8 (12). We have further shown that ssRNA oligonucleotides that activate TLR7 can trigger the generation of CTL by inducing Th1-type immunity (13).

In addition to the Ag-specific immune responses effected by CTL, a major component of antitumoral immunity is the innate NK cell response (14). NK cells are involved both in tumor immunosurveillance and in the rejection of established tumors and can prevent the dissemination of metastases (14, 15). The ideal NK cell targets are tumor cells that have lost expression of MHC class I or that overexpress ligands for the activating receptor NKG2D (16). Upon interaction with target cells, NK cells exert cytotoxic functions that are determined by a balance of multiple activating and inhibitory signals (16). Two important antitumoral effector mechanisms triggered by target cell interaction are the direct cytolytic activity of NK cells and the production of IFN-γ. Strategies to exploit NK cell-mediated immunity may thus hold strong potential for cancer therapy.

The therapeutic potential of TLR7 agonists is supported by encouraging results with a recently developed class of antitumor agents, the imidazoquinolines, that acts in part through the activation of TLR7 (17). The lead compound, imiquimod, is however only approved for the treatment of skin tumors by topical use (18). Although some imidazoquinoline compounds support antitumoral NK responses in i.v. disseminated tumor models (19), in solid tumors injection of imiquimod is only effective locally but not at distant sites (20). Due to the short half-life of these small molecules in vivo, frequent applications are necessary.

In this study, we show for the first time that immunostimulatory RNA oligonucleotides elicit an effective antitumoral NK response in vivo through TLR7. We demonstrate that RNA oligonucleotide treatment selectively inhibits growth of MHC class I-negative tumors but not growth of MHC class I-expressing tumors. Stimulation of mice with immunostimulatory RNA oligonucleotides leads to an activated NK phenotype with enhanced IFN-γ production and increased cytotoxicity. These effector functions are triggered indirectly through activation of DC to produce IL-12 and type I IFN. We further show that oligonucleotide-activated DC are sufficient to promote NK activation in vivo. Thus, the therapeutic application of RNA oligonucleotides represents a promising strategy to stimulate NK cell responses to induce effective antitumor immunity.

Female C57BL/6 mice were purchased from Harlan-Winkelmann. TLR7-deficient mice (C57BL/6 background) were a gift from S. Akira (Osaka University, Osaka, Japan) and bred in the animal facilities of the Medical Department of the Ludwig-Maximilian University of Munich (Munich, Germany). Female β₂-microglobulin (β₂m)-deficient mice and IL-12p40-deficient mice (both C57BL/6 background) were from Charles River Laboratories. IFNAR-deficient mice (129Sv background) and wild-type (wt) 129Sv mice as controls were provided by Dr. H. J. Anders (Ludwig-Maximilian University of Munich, Munich, Germany). Mice were 5–12 wk of age at the onset of the experiments. All animal studies were approved by the local regulatory agency (Regierung von Oberbayern, Munich, Germany). The C57BL/6-derived T cell lymphoma cell line RMA and the TAP-deficient variant RMA-S were provided by Dr. J. Charo (Berlin, Germany).

The 20-mer oligoribonucleotides 9.2dr (5′-UGUCCUUCAAUGUCCUUCAA-3′) and poly(A) in both the unmodified phosphodiester (PD) and fully phosphorothioate (PTO) forms were purchased from CureVac. To protect against degradation, all oligoribonucleotides used were PTO unless indicated otherwise. The PTO-modified CpG oligodeoxyribonucleotide 1826 (5′-TCCATGACGTTCCTGACGTT-3′) was obtained from the Coley Pharmaceutical Group (Langenfeld, Germany). Oligonucleotides were complexed with N-[1-(2,3-dioleoyloxy)propyl]-N,N,N-trimethylammonium methylsulfate (DOTAP; Roche) before use by incubation for 20 min at a ratio of 1:2–1:5 (w/w).

For stimulation of human NK cells, the sequence 9.3as (5′-UGGUAAUUGAAGGACAGGU-3′; CureVac) was used. Two hundred nanograms of RNA was complexed with the polycationic polypeptide poly-l-arginine P7762 (Sigma-Aldrich). Stimulation with RNA was compared to stimulation with the small molecule TLR7 and TLR8 agonists 3M-001, 3M-002, and 3M-003 (3M Pharmaceuticals).

To prepare bone marrow-derived DC (BMDC), bone marrow cells from C57BL/6 mice were cultured in complete RPMI 1640 (10% FCS, 2 mM l-glutamine, 100 μg/ml streptomycin, and 1 IU/ml penicillin) supplemented with 20 ng/ml GM-CSF and 20 ng/ml IL-4 (Tebu Bio) (DC medium). On day 7, loosely adherent cells were harvested and washed. Splenic DC were isolated by MACS of splenocytes after labeling with anti-CD11c microbeads (Miltenyi Biotec). DC-depleted splenocytes contained <2% of CD11c⁺ cells. NK cells were isolated from total splenocytes by sorting with anti-DX5 (panNK) microbeads (purity >95%; Miltenyi Biotec).

Human PBMCs were prepared by Ficoll-Hypaque density gradient centrifugation of heparinized blood from healthy donors. NK cells were prepared with a NK Isolation Kit II (Miltenyi Biotec). Additional anti-CD4 microbeads were added to the kit to exclude contamination with monocytes and DC. Evaluation of the NK cell population showed purity of >97% (SEM, 2.5%) and absence of CD4⁺ cells (<0.1%). Monocytes were prepared with a Monocyte Isolation Kit II (Miltenyi Biotec). Anti-CD56 microbeads were added to the kit to exclude contamination with NK cells. The purity of isolated monocytes (CD11c⁺, HLA-DR⁺, and CD14⁺) was 96% (SEM, 2%); NK cells were absent (<0.1%). Omitting the additional CD4⁺ depletion (for NK cells) or CD56⁺ depletion (for monocytes) resulted in an up to 2% contamination with the respective cell population. Cells were cultured in complete RPMI 1640 supplemented with 10 mM 4-(2-hydroxyethyl)-1-piperazine ethanesulfonic acid (Sigma-Aldrich).

Splenocytes, splenic DC, or BMDC were stimulated with 10 μg/ml 9.2dr or poly(A) or 1 μg/ml CpG complexed with DOTAP for 6 h at 1 × 10⁶/ml in 200 μl of complete medium. For NK-DC cocultures, stimulated DC were washed twice and 2 × 10⁵ purified NK cells or 2 × 10⁵ total splenocytes in 200 μl of fresh medium/well were added. For cultures of NK cells with DC supernatant, 200 μl of BMDC supernatant was added to 2 × 10⁵ purified NK cells or total splenocytes. Cell activation was measured 18 h later. Neutralizing Abs against IL-2 (20 μg/ml, clone JES6-1A12; R&D Systems), IL-6 (5 μg/ml, clone MP520F3; R&D Systems), IL-12/IL-23p40 (40 μg/ml, clone C17.8; BioLegend), and IL-15 (2 μg/ml, clone AIO.3; MBL) were added to the supernatant, as indicated, 2 h before coincubation with NK cells.

Human NK cells (1 × 10⁵ in 200 μl of medium) were cocultured with accessory cells (monocytes) at the indicated ratios. Cells were stimulated for 12 or 36 h with 1 μg/ml 9.3as or 10 μM 3M-001, 3M-002, or 3M-003 before cell activation or IFN-γ production were measured.

For in vivo immunostimulation, 20–40 μg of oligonucleotides complexed with DOTAP were injected i.v. or 100 μg of uncomplexed CpG were injected s.c. into TLR7-deficient or wt C57BL/6 mice. Single-cell suspensions were prepared from spleen 22 h after injection unless indicated otherwise. For DC transfer, TLR7-deficient mice were injected twice i.v. with 2.5 × 10⁷ wt BMDC in 200 μl of Hepes-buffered saline at a 2-h interval. Wild-type BMDC were either unstimulated or stimulated in vitro with 1 μg/ml 9.2dr for 6 h before transfer. Two hours after the cell transfer, mice were injected i.v. with RNA oligonucleotides. For tumor therapy, groups of five mice were injected on day 0 with 10⁶ TAP-deficient RMA-S or wt RMA tumor cells s.c. Mice were treated i.v. with RNA oligonucleotides on days 0, 3, and 6 (RMA-S) and additionally on days 9 and 12 (RMA). For depletion of NK cells, 0.5 μg of the IL-2R chain-specific mAb TMβ1 was given i.p. 2 days before and 2 days after tumor challenge as described previously (21). For depletion of CD8 T cells, 0.1 μg of the RmCD8 mAb was given i.p. 2 and 1 days before and 7 days after tumor challenge. Efficacy of depletion was confirmed by flow cytometry. Tumor size was expressed as the product of the perpendicular diameters of individual tumors. For assessing NK activation after repeated injections of RNA oligonucleotides in tumor-bearing mice, 4 × 10⁵ B16 cells were injected i.v. and animals were treated with RNA oligonucleotides eight times at 3-day intervals.

Single-cell suspensions from splenocytes or blood were treated with ammonium chloride buffer to lyse erythrocytes. RMA-S tumors were mechanically disrupted and incubated with collagenase (Sigma Aldrich) at 37°C for 20 min, then passed through a 40-μm cell strainer (BD Biosciences) to obtain single-cell suspensions. For analysis of activation markers, cells were stained with fluorochrome-coupled mAbs and analyzed by flow cytometry. Anti-mouse CD3-allophycocyanin, NK1.1-PerCP, and IFN-γ-FITC, anti-mouse and anti-human CD69-PE, and anti-human leukocyte Ag DR-PerCP were from BD Biosciences.

To assess IFN-γ production by NK cells, splenocytes were isolated 4 h after RNA injection and incubated for 4 h with brefeldin A at a concentration of 1 μg/ml. Cells were stained with PerCP-conjugated anti-NK1.1 (BD Biosciences), fixed with 2% paraformaldehyde, and permeabilized (0.5% BSA, 0.5% saponin, and 0.02% sodium azide in PBS). Fixed cells were stained with FITC-conjugated anti-IFN-γ Ab (BD Biosciences) for 25 min. The percentage of IFN-γ-positive NK cells was determined by flow cytometry. Levels of human or murine IFN-γ in the supernatant were quantified by ELISA (BD Biosciences) according to the manufacturer’s protocol.

For in vivo determination of cytotoxicity, target splenocytes from wt and β₂m-deficient mice were labeled for 15 min at room temperature with 15 and 0.15 pM CFSE, respectively. Cells were washed twice with PBS and resuspended at 1 × 10⁸/ml. The two populations were mixed at a 1:1 ratio and 1 × 10⁷ cells were injected i.v. into wt C57BL/6 mice injected 4 h previously with 9.2dr. CFSE staining in splenocytes was analyzed by flow cytometry 4–10 h after target cell injection. Specific lysis was calculated as follows: specific lysis (percent) = 100 − [100 × (CFSE^low cells in stimulated mice/CFSE^high cells in stimulated mice)/(CFSE^low cells in unstimulated mice/CFSE^high cells in unstimulated mice)] (22).

For determination of cytotoxicity in vitro, splenic NK cells were either isolated from treated mice 4 h after oligonucleotide injection or were activated with 100 μl of supernatant from RNA-stimulated DC for 4 h. NK cells were plated at 5 × 10⁵ cells/100 μl in complete RPMI 1640 without phenol red (PAA). YAC-1 target cells (1 × 10⁶ cells/ml) were labeled with 10 μM AM-calcein (Invitrogen) for 30 min at 37°C, washed three times with PBS, and resuspended at 5 × 10⁴ cells/ml. NK cells were incubated with labeled target cells for 4 h at 37°C, 150 μl of supernatant was transferred to a new plate, and fluorescence was measured at 485/535 nm as previously described (23). The background cytotoxic activity of TLR7-deficient NK cells from untreated mice was similar to that of TLR7-deficient NK cells from RNA-treated mice.

Comparisons in tumor size among groups were made using the Mann-Whitney U test for various time points. For cytokine ELISAs and cytotoxicity assays, significance was assessed by Student’s t test. Statistical analysis was performed using SPSS software. Error bars indicate SEM.

We have recently described immunostimulatory sequences within RNA oligonucleotides that activate a Th1-type immune response, an important effector of antitumoral immunity (12, 13). To assess whether immunostimulatory RNA oligonucleotides can suppress tumor growth through activation of NK cells, the second major cellular component of antitumoral immunity, we examined the efficacy of RNA oligonucleotides in the treatment of RMA-S tumors. The RMA-S cell line is a MHC class I-negative variant of the RMA lymphoma that is selectively targeted by NK cells, unlike the MHC-sufficient parent cell line (24). Mice were treated with the highly immunostimulatory oligonucleotide 9.2dr (12, 13) after inoculation of either NK-sensitive RMA-S cells or wt RMA cells. Three injections of the 9.2dr oligonucleotide complexed with DOTAP at 3-day intervals suppressed progression of RMA-S tumors compared with untreated mice (Fig. 1,A). In contrast, after injection of a control poly(A) sequence of the same length, no effect on tumor growth was observed. Treatment with RNA oligonucleotides did not prevent tumor progression in the MHC-sufficient RMA tumors (Fig. 1,B), suggesting that NK cells are the critical effectors of the antitumor effect of immunostimulatory RNA. Within the tumor, NK cells represented 11.6% of small lymphocytes (±2.4%; n = 5). Results for one tumor are shown in Fig. 1,C. Indeed, during treatment with 9.2dr, RMA-S tumors progressed more rapidly in NK-depleted mice than in CD8-depleted mice or undepleted mice (Fig. 1 D).

To characterize the effect of RNA oligonucleotides on NK cells, mice were injected i.v. with RNA oligonucleotides complexed with DOTAP. In addition to oligonucleotides comprising a PTO backbone, oligonucleotides of the same sequence with an unmodified PD backbone were used. As shown in Fig. 2,A, a single application of 9.2dr PTO resulted in up-regulation of the early activation molecule CD69 on >80% of splenic NK cells. CD69 expression was also up-regulated on a small proportion of splenocytes in mice injected with the unmodified 9.2dr PD oligonucleotide. Furthermore, 9.2dr PTO enhanced production of IFN-γ by NK cells as early as 4 h after injection (Fig. 2 B). Notably, this increase in IFN-γ production was observed directly ex vivo without the need for restimulation in culture by target cells. Activation of NK cells was clearly sequence dependent, since NK cells were not activated by treatment with poly(A) oligonucleotides of the same length in either the PTO or PD form. Because of their stronger immune-activating capacity, PTO oligonucleotides were used for all additional experiments.

To assess the cytotoxicity of NK cells following RNA oligonucleotide application, an in vivo assay was performed (22). In this assay CFSE-labeled splenocytes from β₂m-deficient mice were injected i.v. as target cells. Splenocytes from wt mice labeled with a higher CFSE concentration were injected simultaneously as reference. Blood was collected from recipients 10 h after target cell injection and PBMC were analyzed for CFSE staining by flow cytometry. Lysis of CFSE^low β₂m-deficient cells was selectively increased in 9.2dr PTO-treated mice with a specific lysis of >60% (Fig. 2, C and D). Indeed, an increased specific lysis was detected as early as 4 h after target cell injection (data not shown).

We then examined CD69 expression on NK cells from tumor-bearing mice. As with healthy animals, CD69 was up-regulated in mice bearing RMA-S tumors upon treatment with 9.2dr PTO (Fig. 2,E). Within the tumor, NK cells also showed increased CD69 expression following treatment with 9.2dr PTO (Fig. 2,E). CD69 up-regulation was also seen after the last of three RNA injections, suggesting that activation of NK cells by RNA oligonucleotides is not subject to a decreased response after repeated administration. This was confirmed in another tumor model after i.v. injection of B16 melanoma cells: NK activation was not impaired even after eight applications of 9.2dr PTO over 21 days (Fig. 2 F). Thus, RNA oligonucleotides activate NK cells in vivo in a sequence-dependent manner, without the need for extensive restimulation by target cells.

RNA from viral origin is recognized by a variety of different receptors of the innate immune system, including TLR7 (1). To determine whether NK cell activation by RNA oligonucleotides is mediated through TLR7, TLR7-deficient and wt mice were injected with 9.2dr complexed with DOTAP. As shown in Fig. 3,A, TLR7-deficient mice injected with 9.2dr showed no up-regulation of the early activation molecule CD69 on NK cells. In contrast, injection of the CpG oligodeoxynucleotide 1826, a ligand for TLR9, resulted in expression of CD69 on 80% of NK cells in TLR7-deficient mice, demonstrating that TLR7-deficient NK cells can respond to activation through another TLR. Also, the percentage of IFN-γ-producing NK cells was not significantly increased by 9.2dr in TLR7-deficient mice (Fig. 3,B). Furthermore, NK cells isolated 4 h after RNA oligonucleotide injection showed TLR7-dependent cytotoxicity. NK cells from 9.2dr-treated wt mice effected a specific lysis of >60% against the NK-sensitive cell line YAC-1 (Fig. 3 C) in a standard in vitro cytotoxicity assay (25). In contrast, NK cells from TLR7-deficient mice showed no cytotoxic activity upon 9.2dr stimulation. Thus, NK cell activation by RNA oligonucleotides is TLR7-dependent.

We have previously shown that stimulation with RNA oligonucleotides leads to DC activation through TLR7 (12, 13). Because NK cells can be efficiently activated by DC (26, 27), we investigated whether DC play a role in NK-cell activation by RNA oligonucleotides. When total splenocytes from naive mice were stimulated with 9.2dr, NK cells showed strong CD69 up-regulation (Fig. 4,A). In contrast, when DC were depleted from splenocytes by MACS before stimulation, CD69 up-regulation on NK cells was abolished. We then activated BMDC with 9.2dr, washed them extensively to remove oligonucleotides, and cocultured them with naive splenocytes. The NK cell fraction within these splenocytes showed an activated phenotype that was associated with dose-dependent production of IFN-γ (Fig. 4,B). Similar results were observed when 9.2dr-stimulated splenic DC were used to activate splenocytes (data not shown). Isolated NK cells were also activated by RNA-stimulated DC and induced to produce high amounts of IFN-γ (Fig. 4 B). Thus, DC are necessary and sufficient to induce NK cell activation and IFN-γ production through RNA oligonucleotides.

To investigate whether activation of NK cells by DC was mediated by soluble factors, BMDC were activated with 9.2dr, washed to remove oligonucleotides, and cultured for a further time period of 18 h. Supernatant from the culture was then added to unstimulated splenocytes or to purified NK cells. In both cases, supernatant from activated DC but not from unstimulated DC led to up-regulation of CD69 on NK cells in a dose-dependent manner (Fig. 4,C). Supernatant from DC activated with the higher dose of RNA oligonucleotides led to IFN-γ production by both splenocytes and purified NK cells (Fig. 4,C). Furthermore, supernatant from activated DC but not from DC stimulated with poly(A) induced NK cell cytotoxicity (Fig. 4 D). Thus, NK cells are stimulated through soluble factors produced by activated DC.

Using a panel of neutralizing Abs, we showed that IFN-γ production by NK cells was entirely IL-12 dependent (Fig. 5,A). In contrast, neutralizing Abs for IL-2, IL-6, and IL-15 had no effect on IFN-γ production. To confirm the role of IL-12 in NK activation, splenocytes from IL-12p40-deficient mice were stimulated with RNA oligonucleotides. Unexpectedly, activation of NK cells by 9.2dr stimulation based on CD69 expression was not suppressed in IL-12-deficient splenocytes (Fig. 5,B, left panel). IFN-γ production was clearly dependent on IL-12, as IFN-γ secretion was strongly reduced in NK cells stimulated with DC from IL-12-deficient mice (Fig. 5,B, middle panel). In contrast, cytotoxicity of NK cells activated with the culture supernatant of RNA-stimulated DC from IL-12-deficient mice was not impaired (Fig. 5 B, right panel).

Since type I IFN has also been implicated in the activation of NK cells (28, 29), we measured the activation of NK cells from type I IFN receptor-deficient mice upon RNA stimulation. Both CD69 up-regulation and cytotoxicity of NK cells were significantly reduced compared with wt NK cells (Fig. 5 C). In contrast, IFN-γ production in type I IFN receptor-deficient NK cells was preserved. We conclude that the effector functions of NK cells following RNA oligonucleotide application are differently regulated. Although IFN-γ production is dependent on IL-12, cytotoxicity depends on type I IFN. Thus, DC-derived IL-12 and type I IFN are both critical cytokines for the activation of NK cells by RNA oligonucleotides.

In contrast to mice in which TLR7 is the sole receptor for ssRNA in the endosome, in humans TLR7 and TLR8 share the detection of ssRNA (5, 6). Small molecule TLR7 and TLR8 agonists, the imidazoquinolines, have been shown to activate NK cells (30). We compared the effect of immunostimulatory RNA with three small molecule agonists on human NK cells. Peripheral blood monocytes were used as accessory cells, because these cells express TLR8 and can differentiate to monocyte-derived DC. Immunostimulatory RNA up-regulated CD69 expression on up to 60% of NK cells in a monocyte-dependent manner (Fig. 6,A). Stimulation with small molecule TLR7/8 agonists also increased CD69 expression, albeit at lower levels. An increase in CD69 expression was also seen when NK cells were cocultured with plasmacytoid DC for both immunostimulatory RNA and imidazoquinolines (data not shown). Furthermore, stimulation of the monocyte-NK coculture with immunostimulatory RNA induced IFN-γ production by NK cells as early as 12 h after stimulation (Fig. 6,B). IFN-γ induction by imidazoquinolines occurred later and did not reach the levels obtained by immunostimulatory RNA (Fig. 6 C). For both imidazoquinolines and immunostimulatory RNA, stimulation was clearly dependent on the presence of monocytes.

To determine whether DC also mediate NK cell activation by RNA oligonucleotides in vivo, TLR7-deficient mice were reconstituted with wt BMDC before treatment with 9.2dr. CD69 expression on NK cells was up-regulated in these mice upon 9.2dr treatment (Fig. 7, A and B). In contrast, we have shown that TLR7-deficient mice without transferred DC show no NK cell activation following 9.2dr injection (Fig. 3,A). NK cell activation was also seen when DC were stimulated with 9.2dr before transfer, in addition to the in vivo 9.2dr treatment (Fig. 7). Transfer of DC alone without RNA stimulation did not induce NK activation. Thus, DC are sufficient to mediate activation of NK cells by RNA oligonucleotides in vivo.

We have recently described immune-activating sequences within ssRNA oligonucleotides that can act through TLR7 (12). These oligonucleotides induce Th1-type cytokines in vivo and trigger the generation of CTL and Ab production when coinjected with Ag (13). In this study, we show for the first time that RNA oligonucleotides can inhibit tumor growth in a sequence-dependent manner. Treatment of mice with TLR7-activating oligonucleotides selectively suppressed growth of NK-sensitive tumors, demonstrating that RNA oligonucleotides activate NK immunity (31, 32). Two main mechanisms are responsible for NK-mediated antitumor immunity, IFN-γ production and direct cytotoxicity (16). IFN-γ is a crucial mediator of antitumor immunity in experimental models and elevated levels of IFN-γ are associated with disease outcome in several clinical studies (14, 33). The increase in IFN-γ production by NK cells upon RNA oligonucleotide treatment suggests that IFN-γ may play an important role in antitumor efficacy. Indeed, in a murine lung metastasis model, IFN-γ was essential for the NK-dependent antitumor effect of 3M-011, a TLR7 agonist (19). We further show that RNA oligonucleotides potentiate NK cytotoxic activity in vivo. The close contact of activated NK cells with target cells is a prerequisite for cytotoxic activity (34) and indeed we clearly show the presence of activated NK cells within the s.c. tumors. Thus, both IFN-γ production and direct cytotoxicity may synergize to provide NK-mediated antitumor immunity upon RNA oligonucleotide treatment.

In the absence of infection, NK cells present a naive phenotype and do not efficiently up-regulate effector functions upon contact with target cells expressing NK-activating receptors (35, 36). To achieve efficient protection against infections and tumors, NK cells therefore require an initial priming that can be effected by bacterial, viral, or even parasitic infectious agents (37, 38, 39). Primed NK cells are not directly activated to produce IFN-γ, but they respond to triggering by target cells (37). In contrast to this two-step process, we show here that a single injection of RNA oligonucleotides rapidly stimulates IFN-γ production by NK cells in vivo without the need for restimulation with target cells. Furthermore, an increase in cytotoxicity was seen in vivo as early as 4 h after injection of the target cells. Thus, RNA oligonucleotides rapidly activate NK cells in vivo while bypassing the need for additional restimulation. Activation was clearly sequence-specific, as a poly(A) sequence of the same length induced neither an activated phenotype nor IFN-γ production or cytotoxicity. We thus show for the first time that RNA oligonucleotides can activate NK cells in vivo in a sequence-dependent manner.

Because the oligonucleotide backbone itself, independently of the nucleotide sequence, may play a role in TLR7 and TLR9 recognition and activation (40, 41), we compared in vivo NK activation by oligonucleotides with the same sequence containing either an unmodified PD backbone or a PTO-modified backbone. We observed low activation with the PD backbone, while activation was strongly enhanced by the use of the PTO backbone, although stimulation remained sequence-dependent. Similarly, we previously observed stronger T and B cell activation with PTO RNA oligonucleotides (13). This probably relates to the increased stability provided by the PTO backbone, since RNA is highly susceptible to nuclease degradation (42).

RNA can be sensed by a variety of receptors of the innate immune system, including the TLRs 3 and 7, the cytoplasmic helicases RIG-I and MDA-5, the serine-threonine kinase PKR, and the NALP-3 inflammasome (3, 4, 5, 6, 7, 8, 43, 44, 45, 46). Importantly, we showed that NK activation by the 9.2dr RNA oligonucleotide is entirely TLR7-dependent, since the increase in NK effector function following RNA treatment is absent in TLR7-deficient mice. This demonstrates that cytosolic receptors do not participate in the in vivo recognition of these RNA oligonucleotides. Thus, we showed for the first time that RNA oligonucleotides activate NK cells in vivo through TLR7. In vitro studies show that also human NK cells within PBMC can be stimulated by a TLR7-activating ssRNA sequence derived from HIV-1 (47). This suggests that a therapeutic strategy using RNA oligonucleotides to target NK cells may also be effective in humans.

Small molecule agonists of the imidazoquinoline family have been shown to activate NK cells through TLR7 and TLR8 (30). Previous in vitro studies on the mechanisms of NK cell activation by imidazoquinolines have however been inconsistent. Direct activation of human NK cells by imidazoquinolines without the need for accessory cells has been shown in vitro but is dependent on the presence of the cytokines IL-12, IL-15, or IFN-α that are generally produced by accessory cells (48, 49). Contrasting reports claim that the induction of IFN-γ production in vitro by imidazoquinolines is dependent on activation of accessory cells such as monocytes (30, 50, 51). Furthermore, imidazoquinolines also activate the NALP3 inflammasome, which leads to the production of bioactive IL-1β and IL-18 (52). Indeed, IL-18 was shown to contribute to NK cell activation by imidazoquinoline compounds (51). In this study, we clearly show that reconstitution of TLR7-deficient mice with wt DC restores activation of NK cells by immunostimulatory RNA. Thus, we provide in vivo data demonstrating that the main mechanism for in vivo activation of NK cells by immunostimulatory RNA is entirely TLR7-dependent and is mediated by accessory cells. Indeed, the induction of NK responses to pathogens in vivo requires DC (37, 38, 39).

Although DC can activate NK cells by direct contact (29, 53), cell-cell contact is not necessary for activation of NK cells by RNA oligonucleotides. Instead, activation is mediated by DC-secreted factors, as the stimulatory effect on NK cells can be achieved by transferring supernatant of activated DC. Several cytokines, the most important being IL-2, IL-10, IL-12, IL-15, IL-18, and IFN-α, have been implicated in the regulation of NK activation (26, 54, 55, 56, 57). Type I IFN in particular plays an important role for the in vivo priming of NK cells in different infectious models (37, 38, 39). Depending on the pathogen, type I IFN affects NK cell priming by acting either on myeloid DC that are induced to produce IL-15 (37) or directly on NK cells (39). The cellular source of type I IFN necessary to activate NK cells upon TLR stimulation remains however unclear (37). Interestingly, depletion of plasmacytoid DC, the main producers of the type I IFN-α, does not impair activation of NK cells in vivo upon infection (38). We have previously shown that RNA oligonucleotides induce both IL-12 and IFN-α in vivo (13). We show here that the effector functions following RNA oligonucleotide treatment are differently regulated: whereas IFN-γ production is dependent on IL-12, cytotoxicity depends entirely on type I IFN.

The therapeutic potential of cancer treatment strategies targeting NK cells has gained momentum in recent years. NK cells are one of the cellular mediators of Ab-dependent cell-mediated cytotoxicity, an important effector mechanism for the therapeutic use of mAbs targeting tumor Ags. Other treatments targeting NK cells include NK cell adjuvants such as cytokines, soluble TRAIL, c-kit tyrosine kinase inhibitors, or even the transfer of preactivated NK cells (14). These therapeutic strategies could be profitably combined with RNA oligonucleotides to enhance NK cell activation in vivo.

In summary, RNA oligonucleotides possess an important therapeutic potential for the treatment of NK-sensitive tumors by triggering the activation of NK cells via the induction of cytokine production by DC. In addition to their direct antitumoral efficacy, NK cells favor the generation of CTL through IFN-γ production (21, 58). This mechanism may contribute to the ability of RNA oligonucleotides to induce an Ag-specific CTL response (13). Because both MHC-negative and MHC-positive tumor cells can thus be eliminated, the combined activation of CTL and NK cells by RNA-based therapies may help prevent tumor immune escape (59). Furthermore, RNA oligonucleotides can be designed to include other antitumoral properties in the same molecule: RNA oligonucleotides containing a 5′-triphosphate lead to antitumoral efficacy through activation of the RIG-I receptor (60). In addition, introduction of a specific inhibitory sequence (small interfering RNA) permits the knockdown of specific genes such as tumor-promoting genes (60). Thus, it will be possible to combine TLR7 activation with other potent antitumoral mechanisms in one RNA molecule, providing a promising mutifunctional therapeutic approach.

The authors have no financial conflict of interest.