Oncolytic Reovirus Inhibits Immunosuppressive Activity of Myeloid-Derived Suppressor Cells in a TLR3-Dependent Manner

Reovirus is a nonenveloped virus that contains 10 segments of dsRNA genome (1). Reovirus has been actively studied as an oncolytic virus that specifically replicates in tumor cells, but not normal cells, leading to efficient tumor cell killing (2). Many clinical studies, including phase III trials, have been ongoing internationally to clarify the effects of reovirus against various types of tumors (3, 4), based on the efficient tumor-killing efficiency, superior safety profile, and other promising properties of reovirus as an oncolytic virus. Moreover, reovirus either used clinically or encountered by common infection rarely induces severe symptoms in adults (2). In addition to these promising properties of reovirus, it has become clear that reovirus-mediated activation of antitumor immunity contributes to the antitumor activity of reovirus (e.g., facilitation of Ag presentation by dendritic cells [DCs]) (5, 6); however, the mechanism of the reovirus-mediated activation of antitumor immunity has not been fully understood.

Activation of antitumor immunity is known to be crucial for tumor regression (7, 8), and thus cancer immunotherapy has been garnering attention for more than a decade (9, 10). Several clinical studies of cancer immunotherapies, including T cell–based immunotherapies (11) and cancer vaccines (12), have been carried out in humans. However, the expected antitumor effects were obtained in only a limited number of clinical studies (13, 14), probably because antitumor immunity is significantly suppressed in cancer patients via various mechanisms (14), such as programmed death-1–mediated and CTLA-4–mediated signaling (15). The blockade of immune checkpoint molecules by using anti–programmed death-1 and anti–CTLA-4 Abs has provided a significant breakthrough in cancer immunotherapy (16). These findings indicate that overriding the immunosuppressive environment in cancer patients is important for effective tumor regression via the activation of antitumor immunity.

Among several known immunosuppressive mechanisms, myeloid-derived suppressor cells (MDSCs) have recently been demonstrated to play crucial roles in the immunosuppression in cancer patients (17). MDSCs are an immature and heterogeneous population that exert immunosuppressive functions via various mechanisms (18, 19). For example, MDSCs produce NO and reactive oxygen species, leading to suppression of T cell proliferation and induction of regulatory T cells (Tregs) (19, 20). Mouse MDSCs are Gr-1⁺/CD11b⁺, whereas human MDSCs lack HLA d-related and express either or both CD11b and CD33 (21). Previous studies demonstrated that a decrease in immunosuppressive MDSCs leads to the activation of host immunity, and that several chemotherapy drugs, including fluorouracil and gemcitabine, reduced the numbers of MDSCs in tumor-bearing hosts (22, 23). In addition, immunostimulatory molecules, including CpG oligodeoxynucleotide and polyinosinic-polycytidylic acid (polyI:C), which are ligands of TLRs and/or retinoic acid–inducible gene-I (RIG-I)–like receptors, inhibit the immunosuppressive functions of MDSCs and/or induce differentiation of MDSCs into immune-activating cells, including macrophages (24–26). These data suggest that MDSCs are a promising target for cancer immunotherapy, and that anticancer reagents capable of both directly killing tumor cells and also inhibiting the immunosuppressive functions of MDSCs exhibit superior antitumor effects via both mechanisms.

In this study, we hypothesized that reovirus inhibits the immunosuppressive functions of MDSCs after i.v. administration into tumor-bearing hosts, and this effect contributes to the reovirus-mediated tumor regression. We found that the suppressive activity of MDSCs on T cell proliferation was significantly reduced after i.v. injection of reovirus into tumor-bearing mice. The reduction in the immunosuppressive activity of MDSCs by reovirus was observed in IFN-β promoter stimulator-1 (IPS-1)-knockout (KO) mice, but not in TLR3 KO mice. These results indicate that reovirus inhibits the immunosuppressive functions of MDSCs via the TLR3 signaling pathway, followed by activation of antitumor immunity.

C57BL/6J mice were obtained from Nippon SLC (Hamamatsu, Japan). IPS-1 KO mice of a C57BL/6 background were kindly provided by K.J. Ishii (National Institute of Biomedical Innovation, Health and Nutrition, Osaka, Japan) and S. Akira (Osaka University, Osaka, Japan). TLR3 KO mice of a C57BL/6 background were obtained from OrientalBioService (Kyoto, Japan). OT-I transgenic C57BL/6 mice were purchased from Jackson Laboratories (Bar Harbor, ME). GFP transgenic mice were kindly provided by M. Ikawa (Osaka University, Osaka, Japan) (27). All mice were used at 5–7 wk of age. All animal experimental procedures used in this study were performed in accordance with the institutional guidelines for animal experiments at Osaka University. L929 cells (a mouse fibroblast cell line) were maintained with MEM Joklik’s modified (56449C; SAFC Biosciences, Lenexa, KS) supplemented with 5% FBS (Thermo Fisher Scientific, Grand Island, NY). EL4 (a mouse lymphoma cell line) and B16 (a mouse melanoma cell line) cells were maintained with RPMI 1640 medium (R8758-500ML; Sigma-Aldrich, St. Louis, MO) containing 10% FBS. All mediums described earlier were supplemented with streptomycin (100 μg/ml), penicillin (100 U/ml) (Nacalai Tesque, Kyoto, Japan), and GlutaMAX (35050-061; Thermo Fisher Scientific). All cells were routinely monitored for changes in morphology and growth rate, and tested for mycoplasma. Mammalian reovirus type 3 Dearing was amplified in L929 cells and purified by CsCl density gradient centrifugation as previously described (28). Viral titers were determined by a plaque assay using L929 cells. UV-irradiated reovirus (UV-Reo) was prepared by exposing live virus to UV light for 15 min. Loss of infection ability of reovirus was checked by a plaque assay using L929 cells. Information about Abs used in this study are available upon request.

B16 and EL4 cells were seeded at a density of 1 × 10⁴ cells per well in a 96-well plate. On the following day, cells were infected with reovirus for 48 h. Cell viability was evaluated by an AlamarBlue assay.

EL4 tumor cells (2 × 10⁵ cells per mouse) were s.c. transplanted into mice. On day 7, reovirus (2 × 10⁷ PFUs per mouse) and UV-Reo (3 × 10⁸ PFUs per mouse) were i.v. administered to the tumor-bearing mice via the tail vein. The tumor volume was calculated using the following formula: tumor volume (mm³) = 1/2 × (width)² × (length).

Mice were s.c. transplanted with EL4 cells (2 × 10⁵ cells per mouse) or B16 cells (5 × 10⁵ cells per mouse). Two to three weeks after transplantation, reovirus was i.v. administered at a dose of 3 × 10⁸ PFUs per mouse into tumor-bearing mice. Two days after administration, MDSCs were isolated from the spleen using mouse/human CD11b MicroBeads (130-049-601; Miltenyi, Bergisch Gladbach, Germany) and used for the subsequent experiments, including T cell proliferation assay, as described later.

Splenocytes recovered from OT-I transgenic mice were stained with 50 μM CFSE (341-07401; Dojindo, Kumamoto, Japan). CFSE-stained splenocytes were seeded at a density of 1 × 10⁵ cells per well in a 96-well plate and were stimulated with 20 ng/ml H-2K^b OVA peptide (TS-5001-P; MBL, Nagoya, Japan) in the presence or absence of MDSCs. Three days after coculture, CFSE levels in splenic CD8⁺ T cells were determined by using flow cytometry (MACS Quant; Miltenyi) to evaluate the proliferation of CD8⁺ T cells.

Mouse MDSCs were differentiated from bone marrow cells as previously described (29). In brief, mouse bone marrow cells were flushed out from the femur and tibia of mice, after which 2 × 10⁶ cells were cultured in RPMI 1640 medium supplemented with 10% FBS, 100 U/ml penicillin G, 100 μg/ml streptomycin, GlutaMAX, and 40 ng/ml GM-CSF (315-03; PeproTech, Rocky Hill, NJ) for 4 d. Cells were gently collected and inoculated with reovirus or UV-Reo at a concentration equivalent to a multiplicity of infection (MOI) of 5. Twenty-four hours later, the cells were collected, washed, and transferred at a dose of 4 × 10⁶ cells per mouse via tail vein injection to EL4 tumor-bearing mice that received tumor cell transplantation 7–8 d before MDSC transfer. The tumor volume was calculated as described earlier. To evaluate the distribution of MDSCs in the tumor and spleen after transfer in tumor-bearing mice, we differentiated MDSCs from bone marrow cells of GFP-transgenic mice (GFP-MDSCs). GFP-MDSCs were treated with reovirus and transferred to EL4 tumor-bearing mice as described earlier. Spleen and tumor were recovered 24 h after transfer, followed by flow cytometric analysis.

Total RNA was extracted from cells using ISOGEN (319-90211; Nippon Gene, Tokyo, Japan). After the treatment with RNase-free DNase I (M0303L; New England Biolabs, Ipswich, MA), cDNA was synthesized from 1 μg of total RNA using a Superscript VILO cDNA synthesis kit (11754250; Invitrogen). Quantitative PCR analysis was carried out using THUNDERBIRD qPCR Mix (QPS-101; TOYOBO, Osaka, Japan) and StepOnePlus Real-Time PCR systems (Life Technologies). The mRNA levels of the indicated genes were normalized by GAPDH mRNA levels. Sequences of the primers are available upon request.

Cells were washed with cold PBS and lysed with radioimmunoprecipitation assay buffer (89901; Thermo Fisher Scientific, Waltham, MA) containing phosphatase and protease inhibitor mixture (P8340; Sigma-Aldrich) and DL-DTT solution (646563-10X; Sigma-Aldrich). The cell lysates were analyzed by SDS-PAGE. The proteins were electrotransferred onto a polyvinylidene difluoride membrane. Western blot analysis were carried out by using anti–α-tubulin mAb (clone B-7, sc-5286; Santa Cruz Biotechnology, Santa Cruz, CA), anti-reovirus σ3 mAb (clone 4F2; Developmental Studies Hybridoma Bank, Iowa City, IA), and anti–β-actin mAb (clone AC-15, A5441; Sigma-Aldrich) as a primary Ab and Chemi-Lumi One Super (02230; Nacalai Tesque). Images were captured using an LAS-3000 system (Fujifilm, Tokyo, Japan).

Statistical significance was determined using Student t test. Data are presented as the mean ± SD.

To evaluate whether reovirus induced the activation of immune cells, including NK cells and T cells, after i.v. administration into tumor-bearing mice, we measured CD69 expression levels on NK cells and T cells 2 d after i.v. administration of reovirus. CD69 is well known as an activation marker of immune cells (30). We used EL4 and B16 cells for the formation of s.c. tumors in this study because Gr-1⁺/CD11b⁺ cells, which included MDSCs, were efficiently increased in the spleen of EL4 and B16 tumor-bearing mice (see later), although EL4 and B16 cells were less susceptible to reovirus infection in vitro than the reovirus-susceptible tumor cells we previously reported (31). The finding that reovirus had only small cytotoxic effects on these cells was partly due to the relatively low expression levels of the primary receptor for reovirus, junctional adhesion molecule-A (JAM-A) (32), on EL4 and B16 cells (Supplemental Fig. 1). The expression levels of CD69 on the cell surface were significantly upregulated not only on NK cells and T cells in the spleen (Fig. 1A, 1B), but also on CD8⁺ T cells in the tumors of reovirus-treated mice (Fig. 1C, 1D). The average numbers of CD4⁺/CD69⁺ T cells in the tumors were also increased by reovirus administration, although not to a statistically significant degree. The numbers of CD69⁺ NK cells in the tumors were comparable between reovirus-treated and PBS-treated tumor-bearing mice at this time point (2 d after administration). In addition, the numbers of Tregs (CD25⁺/Foxp3⁺ cells) in the spleen were reduced by >40% compared with those in PBS-treated mice on day 8 after reovirus administration (Fig. 1E, 1F). Splenic CD8⁺ T cells expressing CD107a, which is a degranulation maker and is used as a surface marker of activated T cells (33), increased by ∼2-fold on day 8 after reovirus administration (Fig. 1E). UV-Reo also significantly activated NK and T cells in the spleen after i.v. administration (Fig. 1G). Probably because of reovirus-mediated activation of immune cells, not only live reovirus but also UV-Reo significantly suppressed the s.c. EL4 tumor growth after systemic administration, although 15-fold higher amounts of UV-Reo were required to suppress the tumor growth at levels similar to live reovirus (Fig. 1H). These results indicate that antitumor immune cells are significantly activated after reovirus administration in tumor-bearing mice, leading to the suppression of tumor growth.

To examine whether reovirus administration resulted in a reduction in the numbers of MDSCs in tumor-bearing mice, we determined the frequencies of Gr-1⁺/CD11b⁺ cells in the spleen. Compared with naive mice, the percentages of Gr-1⁺/CD11b⁺ cells in the spleen increased by >4-fold in the EL4 tumor-bearing mice (Fig. 2A, 2B). i.v. administration of reovirus at this dose (3 × 10⁸ PFUs per mouse) did not alter the percentages of Gr-1⁺/CD11b⁺ cells in the spleen, compared with those in PBS-treated mice, although the flow cytometric patterns of Gr-1 and CD11b expression on the splenocytes were slightly different between PBS- and reovirus-injected mice. We further examined the frequencies of granulocytic MDSCs (Ly6C^mid/Ly6G⁺) and monocytic MDSCs (Ly6C^high/Ly6G⁻) in the spleen after reovirus administration. Similar levels of granulocytic MDSCs were found in the spleen of reovirus-treated and PBS-treated mice, whereas the percentages of monocytic MDSCs were significantly lower in the spleen of reovirus-treated mice than in those of PBS-treated mice (Fig. 2C, 2D).

Next, we evaluated the immunosuppressive activity of MDSCs recovered from reovirus-treated mice by T cell proliferation assay. T cell proliferation assay is the most commonly used assay for evaluation of the immunosuppressive activity of MDSCs (34, 35). Almost all of the CD11b⁺ cells (>90%) that we recovered from the spleen by a magnetic sorting expressed Gr-1 (data not shown); therefore, the CD11b⁺ cells recovered from the spleen are designated as Gr-1⁺/CD11b⁺ cells hereafter. As shown in Fig. 2E, MDSCs from PBS-treated mice significantly suppressed the proliferation of CD8⁺ T cells in a manner dependent on the ratios of MDSCs. Gr-1⁺/CD11b⁺ cells from tumor-free mice did not suppress T cell proliferation (data not shown). In contrast, T cell proliferation was not significantly suppressed by MDSCs from reovirus-treated tumor-bearing mice. A reduction in the suppressive activities against T cell proliferation was also found in MDSCs of UV-Reo–treated mice (Supplemental Fig. 2A). Similar results were obtained in MDSCs isolated from B16 tumor-bearing mice (Supplemental Fig. 2B). These results indicate that i.v. administration of reovirus results in inhibition in the immunosuppressive activity of MDSCs in tumor-bearing mice.

Next, to examine whether reovirus was taken up by MDSCs after i.v. administration, we examined whether reovirus genomic RNA was present in the MDSCs recovered from the spleen of reovirus-treated tumor-bearing mice by RT-PCR analysis. More than 90% of MDSCs in the spleen expressed JAM-A, which is an infection receptor for reovirus (36) (Fig. 3A). Reovirus genome was detected in the splenic MDSCs from reovirus-treated mice, but not PBS-treated mice (Fig. 3B). These data suggested that reovirus was internalized into MDSCs in the spleen after i.v. administration. To examine whether reovirus was replicated in MDSCs, we examined outer capsid protein σ3 levels in the splenic MDSCs by Western blotting 48 h after i.v. administration of reovirus and UV-Reo. Detectable levels of the σ3 protein expression were not found in MDSCs of any of the mice (Fig. 3C). In addition, the numbers of viral plaques were comparable between 24 and 48 h after inoculation with reovirus in MDSCs that were induced from mouse bone marrow cells in vitro by cultivation in the presence of GM-CSF (in vitro–generated MDSCs) (29) (Fig. 3D). These data indicate that neither detectable levels of virus protein expression nor progeny virus production occur in MDSCs after inoculation with reovirus.

Next, in vitro–generated MDSCs were treated with reovirus to examine whether reovirus directly inhibited the immunosuppressive activity of MDSCs after internalization in MDSCs. After a 4-d cultivation in the presence of GM-CSF, >85% of bone marrow cells exhibited high-level expressions of Gr-1 and CD11b (Fig. 4A). Nontreated in vitro–generated MDSCs significantly suppressed the CD8⁺ T cell proliferation (Fig. 4B), indicating that MDSCs were successfully differentiated from bone marrow cells. Treatment of MDSCs with reovirus led to a significant restoration of CD8⁺ T cell proliferation. The immunosuppressive activity of MDSCs on CD8⁺ T cell proliferation was also completely diminished by UV-Reo. Treatment with reovirus did not appear to alter the expression profiles of Gr-1 or CD11b on in vitro–generated MDSCs (Fig. 4C). Reovirus did not greatly reduce the viabilities of MDSCs after inoculation (Fig. 4D). Taken together, these data indicate that reovirus is directly taken up by MDSCs after i.v. administration, leading to a reduction in the suppressive activity of MDSCs independently of virus replication.

Previous studies demonstrated that immunostimulatory molecules, including CpG oligodeoxynucleotide and polyI:C, which are ligands of TLRs and/or retinoic acid–inducible gene-I (RIG-I)–like receptors, induce differentiation of MDSCs into immune-activating cells (25, 26). In order to examine whether reovirus treatment induced differentiation of MDSCs into immune-activating cells, the expression levels of immune activation markers, including CD86 and MHC class II (MHC II), on MDSCs were examined after reovirus treatment. The expression levels of CD86 and MHC II were significantly elevated on the surface of splenic MDSCs after treatment with reovirus (see below and Supplemental Fig. 3A). In addition, reovirus treatment largely induced type I IFN expression in the splenic MDSCs and in vitro–generated MDSCs (Fig. 6, Supplemental Fig. 3B). Next, to examine whether reovirus-treated MDSCs exhibited antitumor effects, in vitro–generated MDSCs were incubated with reovirus, followed by transfer into tumor-bearing mice. The tumor growth showed a tendency to be promoted in the mice receiving nontreated MDSCs (Fig. 4E). In contrast, transfer of reovirus-treated MDSCs significantly suppressed the tumor growth compared with transfer of mock-treated MDSCs. The growth of s.c. tumors was also significantly reduced by transfer of UV-Reo–treated MDSCs. We did not find significant differences in the tumor growth between the reovirus- and UV-Reo-treated groups. These results suggest that reovirus directly inhibits the immunosuppressive activity of MDSCs and induces differentiation of MDSCs into immune-activating cells after internalization, irrespectively of virus replication. Reovirus-mediated differentiation of MDSCs into immune-activating cells would contribute to a decline in tumor growth.

Next, to examine the biodistribution of in vitro–generated MDSCs after transfer to tumor-bearing mice via tail vein injection, we generated GFP-MDSCs. GFP-MDSCs were treated with reovirus and then transferred into tumor-bearing mice. Flow cytometric analysis demonstrated that similar levels of GFP-MDSCs were detected in the spleen of mice receiving reovirus-treated and mock-treated GFP-MDSCs (Supplemental Fig. 4A). In contrast, the percentages of reovirus-treated GFP-MDSCs in the tumors were almost 3-fold higher than those of mock-treated GFP-MDSCs (Supplemental Fig. 4B). Expression levels of CCR5 and CX3CR1, which have been shown to be involved in the monocytic migration (37, 38), on in vitro–generated MDSCs were upregulated by reovirus treatment (Supplemental Fig. 4C). These results might partly explain the efficient accumulation of reovirus-treated GFP-MDSCs in the tumor. These data indicate that reovirus treatment enhances the tumor accumulation of in vitro–generated MDSCs after i.v. transfer. In addition to the differences in the immune properties between reovirus-treated and mock-treated MDSCs, the differences in the tumor growth after MDSC transfer might be partly due to the differences in the biodistribution of reovirus-treated and mock-treated MDSCs after transfer to tumor-bearing mice.

To examine the mechanisms of reovirus-mediated inhibition of the immunosuppressive activity of MDSCs, we focused on the reovirus dsRNA genome. dsRNA is known to efficiently activate innate immunity via recognition by pattern recognition receptors and to work as an immunogenic adjuvant (39). Reovirus genome has been demonstrated to be recognized by RIG-I, melanoma differentiation–associated protein 5 (MDA5), and TLR3, leading to innate immune responses (39–41). IPS-1 KO mice and TLR3 KO mice were administered reovirus to examine the involvement of innate immune responses in reovirus-induced inhibition of the immunosuppressive activity of MDSCs. IPS-1 is an adaptor protein that mediates RIG-I– and MDA5-mediated type I IFN induction (42). Both splenic MDSCs and in vitro–generated MDSCs expressed TLR3 on the endosomal membrane (Fig. 5A). Expressions of RIG-I, MDA5, and IPS-1 were detected in the MDSCs by real-time RT-PCR analysis (Fig. 5B). The mRNA levels of RIG-I, MDA5, and TLR3 were upregulated in MDSCs after reovirus administration, although the expression level of IPS-1 was not significantly altered.

Next, we examined the immunosuppressive activity of MDSCs recovered from IPS-1 KO and TLR3 KO mice after reovirus administration. Splenic MDSCs recovered from PBS-injected IPS-1 KO and TLR3 KO mice significantly suppressed T cell proliferation at levels similar to those from wild-type (WT) mice (Fig. 5C, 5D). As observed for the MDSCs of WT mice, the suppressive activity of MDSCs of IPS-1 KO mice on CD8⁺ T cell proliferation was ablated by reovirus administration (Fig. 5C). In contrast, administration of reovirus did not alter the suppressive activity of MDSCs of TLR3 KO mice on CD8⁺ T cell proliferation (Fig. 5D), indicating that TLR3 was involved in reovirus-induced inhibition of immunosuppressive activity of MDSCs. We further performed adoptive MDSC transfer experiments to examine the role of TLR3 on reovirus-induced inhibition of the immunosuppressive activity of MDSCs. UV-Reo–treated MDSCs differentiated from TLR3 KO bone marrow cells did not suppress the tumor growth after transfer into tumor-bearing WT mice, whereas tumor growth was significantly slowed by transfer of UV-Reo–treated MDSCs differentiated from WT bone marrow cells (Fig. 5E), as shown in Fig. 4E. The results described earlier indicate that the reovirus dsRNA genome is recognized by TLR3 in MDSCs and that TLR3 signaling, not IPS-1–dependent signaling, is largely involved in the reovirus-mediated inhibition of immunosuppressive functions of MDSCs.

To further examine the mechanism of TLR3-dependent suppression of the immunosuppressive activity of MDSCs by reovirus, we first examined the expression levels of type I IFNs, IL-6, and IL-12β in in vitro–generated MDSCs after treatment with reovirus (Fig. 6). Reovirus significantly upregulated the mRNA levels of all of the cytokines examined in in vitro–generated MDSCs. In particular, the expression levels of IFN-α and IFN-β were largely increased by reovirus. The mRNA levels of type I IFNs were significantly (∼2-fold) lower in the TLR3 KO MDSCs, compared with WT MDSCs, although there were no significant differences in the mRNA levels of IL-6 or IL-12β between WT MDSCs and TLR3 KO MDSCs after reovirus treatment. Expression levels of all of the cytokines were greatly lower in IPS-1 KO MDSCs than in WT MDSCs. When comparing IPS-1 KO MDSCs with TLR3 KO MDSCs, the cytokine mRNA levels in IPS-1 KO MDSCs were significantly lower than those in TLR3 KO MDSCs. These data indicate that the reovirus genome is recognized by both TLR3 and RIG-I and/or MDA5 in MDSCs, although RIG-I and/or MDA5 play greater roles in reovirus-mediated cytokine induction in MDSCs compared with TLR3.

Next, the expression levels of several cytokines in the splenic MDSCs were evaluated after systemic administration of reovirus. mRNA levels of several cytokines, including type I IFNs, were significantly upregulated in the splenic MDSCs of tumor-bearing WT mice 48 h after i.v. administration of reovirus (Supplemental Fig. 3B). IPS-1 KO MDSCs exhibited a significant reduction in the mRNA levels of type I IFNs and IL-10, compared with WT MDSCs, after reovirus administration. In contrast, the mRNA levels of the cytokines examined were comparable between TLR3 KO MDSCs and WT MDSCs after reovirus administration, although the mRNA levels of type I IFNs in in vitro–generated MDSCs of TLR3 KO mice were significantly lower than those of WT mice following reovirus treatment. This is probably because cytokines and/or soluble factors secreted from other cells in a TLR3-independent manner induced type I IFN expression in splenic MDSCs following reovirus administration. We also determined the mRNA levels of cellular factors that have been reported to be involved in the immunosuppressive activity of MDSCs. mRNA levels of inducible NO synthase (iNOS) and arginase-I tended to be elevated in WT MDSCs following reovirus administration (Supplemental Fig. 3C). The mRNA levels of iNOS and arginase-I were significantly lower in IPS-1 KO MDSCs, compared with WT MDSCs, following reovirus treatment, whereas the levels of iNOS and arginase-I mRNAs were comparable between the WT and TLR3 KO MDSCs. There were no apparent differences in the expression levels of the other genes, including IL-1β or cyclooxygenase 2, between WT MDSCs and IPS-1 or TLR3 KO MDSCs recovered from tumor-bearing mice following reovirus administration. Next, to examine whether reovirus induced differentiation of MDSCs to immunostimulatory cells in a TLR3-dependent manner, the expression levels of CD86 and MHC II, which are immune activation markers, on the MDSCs were analyzed following reovirus injection. Significant upregulation of CD86 and MHC II expression was observed on the WT, IPS-1 KO, and TLR3 KO MDSCs (Supplemental Fig. 3A). The DC marker CD11c was also upregulated on MDSCs in all types of mice receiving reovirus, although there was no statistically significant difference between PBS- and reovirus-infected WT mice when WT and TLR3 KO MDSCs were directly compared. The macrophage differentiation marker F4/80 was not significantly increased on MDSCs after reovirus administration. Comparable levels of these surface markers were found between the WT and TLR3 KO MDSCs, whereas the expression levels of CD86 and MHC II on IPS-1 KO MDSCs were lower than those on WT MDSCs following reovirus treatment. These results suggest that an unknown mechanism dependent on TLR3 signaling is involved in reovirus-mediated suppression of the immunosuppressive activity of MDSCs.

Previous studies have demonstrated that reovirus activated antitumor immunity after i.v. administration into tumor-bearing host, resulting in tumor regression (9–13). To examine whether TLR3 was involved in the reovirus-mediated activation of immune cells, we assessed the activation status of immune cells, including NK cells and T cells, on the basis of CD69 expression levels. Reovirus induced significant upregulation of CD69 expression on NK cells, CD4⁺ T cells, and CD8⁺ T cells of WT mice in the spleen as previously shown; in contrast, the CD69 expression levels on these immune cells in TLR3 KO mice were lower than those in WT mice following the reovirus administration (Fig. 7A, 7B). Moreover, UV-Reo was i.v. injected into tumor-bearing WT and TLR3 KO mice to evaluate the antitumor effects of reovirus-mediated immune activation in a TLR3-dependent manner. UV-Reo was used to remove the contribution of reovirus-mediated lysis of tumor cells to tumor regression in this experiment. UV-Reo significantly suppressed tumor progression in WT mice, but not in TLR3 KO mice (Fig. 7C). These results suggest that reovirus activates the immune cells, at least partly in a TLR3-dependent manner, which contributes to tumor regression.

MDSCs are a crucial target for cancer therapy due to their strong induction of immunosuppressive environments via multiple mechanisms, including suppression of CTL activities and induction of Tregs (17). Various approaches have been used to inhibit the immunosuppressive activities of MDSCs and to induce death of MDSCs. The numbers of Gr-1⁺/CD11b⁺ cells, a cell population that includes MDSCs, have been dramatically reduced in tumor-bearing hosts by the administration of gemcitabine and 5-FU (22, 23). In another approach, administration of paclitaxel and CpG oligodeoxynucleotides has been shown to induce differentiation of MDSCs into mature DCs and macrophages, respectively (25, 43). Thus, in addition to exerting direct cytotoxic effects, these anticancer agents also inhibit MDSCs, which further contributes to their efficient antitumor effects.

In a similar manner, the data in this study indicate that reovirus efficiently inhibits the immunosuppressive activities of MDSCs after administration in tumor-bearing mice. Two previous studies have reported the effects of reovirus treatments on the frequencies and immunosuppressive functions of MDSCs following administration in tumor-bearing mice (44, 45). Clements et al. (45) reported that the numbers of Gr-1⁺/CD11b⁺ cells in the tumor increased 3 d after i.p. administration of reovirus. They also demonstrated that the mRNA levels of inflammatory cytokines, including IL-6 and TNF-α, which had been reported to be involved in induction of immunosuppressive activities of MDSCs (46–48), were significantly upregulated in reovirus-treated mice, compared with control mice, although immunosuppressive activities of MDSCs were not examined in that study. We also found that the mRNA levels of these inflammatory cytokines were significantly elevated in the MDSCs of reovirus-treated mice (Supplemental Fig. 3); however, we did not observe significant increases in the frequency of MDSCs (Gr-1⁺/CD11b⁺ cells) in the spleen of tumor-bearing mice, compared with PBS-treated mice, following i.v. administration (Fig. 2A, 2B). In addition, the i.v. administration of reovirus resulted in a significant reduction in the suppressive activities of MDSCs against T cell proliferation (Fig. 2E). These differences in the effects of reovirus on MDSCs between our present study and the previous reports may have been attributable to the different administration routes or the different tumor models, because the characteristics of MDSCs have been shown to differ by tumor types (19). This suggests that we should pay attention to the type of MDSCs in cancer patients, even though the characteristics and classification of MDSC types remain to be clarified.

This study demonstrated that the reovirus dsRNA genome was recognized by TLR3 after internalization of reovirus into MDSCs, leading to inhibition of the immunosuppressive activity of MDSCs. In addition, UV-Reo–treated TLR3 KO MDSCs did not suppress the tumor growth following transfer into the tumor-bearing mice. However, UV-Reo–treated WT MDSCs did significantly delay the tumor growth. These results indicate that TLR3 plays a major role in the reovirus-mediated reduction of the immunosuppressive activity of MDSCs. TLR3 is mainly localized on the endosomes (49). We consider that the reovirus dsRNA genome is recognized by TLR3 on the endosomal membrane following endocytosis of reovirus in MDSCs after i.v. administration. We showed that in vitro–generated TLR3 KO MDSCs expressed lower levels of type I IFN mRNAs than in vitro–generated WT MDSCs following incubation with reovirus, and that there was no apparent reduction in the viability of MDSCs following incubation with reovirus (Figs. 4D, 6). In addition, the previous reports showed that the tumor and virus Ags, including the virus genome, were released into the tumor microenvironment because of virus-infected tumor cell death following oncolytic virus administration (50, 51). Therefore, MDSCs not only took up reovirus, but also might take up the reovirus dsRNA genome released from dying tumor cells and/or reovirus-infected cells following reovirus infection, leading to recognition of the reovirus dsRNA genome by TLR3 in the endosomes.

In contrast, reovirus-mediated inhibition of MDSCs was found in IPS-1 KO mice, similarly to WT mice, indicating that neither RIG-I nor MDA5 is involved in the reovirus-mediated inhibition of MDSCs. Several groups, including ours, reported that the reovirus dsRNA genome is recognized by RIG-I and/or MDA5 postinfection, leading to activation of innate immunity (41, 52, 53). In this study, the mRNA levels of several cytokines, including type I IFNs, were significantly lower in IPS-1 KO MDSCs than in WT MDSCs following reovirus injection (Supplemental Fig. 3). Rather, the mRNA levels of IFN-β were more largely reduced by IPS-1 KO than by TLR3 KO. These data indicate that the RIG-I/MDA5-IPS-1 pathway plays a greater role in reovirus-induced innate immunity in MDSCs than the TLR3 pathway.

Nonetheless, at the present time it remains unclear why the signaling pathway of TLR3, not RIG-I or MDA5, is involved in the reduction in the immunosuppressive activity of MDSCs by reovirus, although both the RIG-I/MDA5 and TLR3 signaling pathways activate IRF3 and NF-κB. An unknown signaling pathway specific for TLR3 might play an important role in the reovirus-mediated suppression of the immunosuppressive activity of MDSCs. The TLR3 signaling pathway would be important for the inhibition of immunosuppressive cells or differentiation of bone marrow–derived immune cells. Shime et al. (54) showed that tumor-supportive macrophages were converted to tumoricidal macrophages via TLR3 signaling following treatment with a dsRNA analog, polyI:C, which is a TLR3 ligand. However, we found that the cell surface markers, including a DC marker CD11c, on splenic Gr-1⁺/CD11b⁺ cells of WT mice were not altered by TLR3-dependent signaling following reovirus injection (Supplemental Fig. 3A). The expression profiles of several cytokines and enzymes that are known to be involved in the immunosuppressive activity of MDSCs were comparable between the WT and TLR3 KO MDSCs following reovirus injection. Further experiments are now ongoing to elucidate the mechanisms of reovirus-mediated inhibition of MDSCs via TLR3.

Mouse MDSCs are divided into the two subsets, granulocytic MDSCs and monocytic MDSCs. Monocytic MDSCs have a greater potential for immunosuppressive activity than granulocytic MDSCs (20). In this study, similar levels of granulocytic MDSCs were observed in the spleen of reovirus-treated and PBS-treated mice (Fig. 2C, 2D). In contrast, the percentages of monocytic MDSCs in the spleen were significantly decreased following reovirus administration. The decrease of monocytic MDSCs might partly explain the reovirus-mediated reduction in the suppressive activity of MDSCs.

iNOS and arginase-I are well-known as the important molecules for the immunosuppressive activities of MDSCs; however, reovirus treatment significantly elevated the mRNA levels of iNOS and arginase-I in the splenic MDSCs, although reovirus reduced the immunosuppressive activity of MDSCs. Previous studies reported that iNOS and arginase-I were upregulated by stimulation of innate immunity (55, 56). Reovirus strongly activates innate immunity, resulting in significant upregulation of iNOS and arginase-I. Unknown factors involved in the immunosuppressive activities of MDSCs would be downregulated by reovirus treatment.

We mainly examined the effects of reovirus on splenic MDSCs after i.v. administration in tumor-bearing mice. It remains unclear whether immunosuppressive activities of tumor-infiltrating MDSCs were downregulated following reovirus administration, although tumor-infiltrating NK and T cells were activated (Fig. 1C, 1D). Tumor microenvironment factors, including hypoxia and inflammation, contribute to the immunosuppressive functions of MDSCs (57). Previous studies reported that monocytic MDSCs were more prominent than granulocytic MDSCs in the tumor (58, 59) and that tumor-infiltrating MDSCs have more potent immune suppressive activity than splenic MDSCs (60, 61). In contrast, spleen functions as a reservoir of myeloid cells, including MDSCs (62, 63). Cortez-Retamozo et al. (63) demonstrated that splenic MDSCs physically relocated from the spleen to the tumor. Reovirus-mediated inhibition in the immunosuppressive activity of splenic MDSCs would contribute to the activation of immune cells in the tumor and tumor regression.

Several oncolytic viruses, including vaccinia virus (VV), vesicular stomatitis virus (VSV), HSV, and adenovirus (Ad), which are widely studied as oncolytic viruses, have been reported to increase or decrease the numbers of Gr-1⁺/CD11b⁺ cells (MDSCs) after injection in tumor-bearing hosts (64–67); however, there have been few studies reporting the functions of MDSCs after administration of these oncolytic viruses. In this study, we revealed that oncolytic reovirus did not appear to increase the numbers of Gr-1⁺/CD11b⁺ cells in the spleen after administration, and that the immunosuppressive activity of MDSCs recovered from reovirus-treated mice was significantly suppressed in a TLR3-dependent manner, indicating that TLR3 signaling is crucial for the reduction in the immunosuppressive activity of MDSCs. Reovirus possesses a dsRNA genome, whereas the genomes of VV, HSV, and Ad are dsDNAs. The VSV genome consists of ssRNA. TLR3 signaling would not be efficiently activated by VV, VSV, HSV, or Ad. Reovirus, which has a dsRNA genome, would be a unique oncolytic virus that can efficiently inhibit the immunosuppressive functions of MDSCs.

In summary, we demonstrated that reovirus inhibits the immunosuppressive activity of MDSCs in a TLR3-dependent manner, but not an RIG-I/IPS-1–dependent manner, after systemic administration in tumor-bearing mice. This study indicates that reovirus not only kills cancer cells directly, but also overcomes the immunosuppressive environment in tumor-bearing hosts, leading to efficient activation of antitumor immune responses. Reovirus shows effective antitumor effects through the various mechanisms described earlier and could become a promising antitumor agent.

We thank Sayuri Okamoto and Eri Hosoyamada (Osaka University, Osaka, Japan) for their help, and Dr. Naoki Okada and Kento Fujiwara (Osaka University, Osaka, Japan) for their technical assistance. Reovirus was kindly provided by Dr. Akira Nishizono and Dr. Tsuyoshi Eto (Oita University, Oita, Japan).

The authors have no financial conflicts of interest.