Oncolytic virotherapy with chimeric VSV-NDV synergistically supports RIG-I-dependent checkpoint inhibitor immunotherapy

Abstract

Unraveling the complexities of the tumor microenvironment (TME) and its correlation with responsiveness to immunotherapy has become a main focus in overcoming resistance to such treatments. Targeting tumor-intrinsic retinoic acid-inducible gene-I (RIG-I), a sensor for viral RNA, was shown to transform the TME from an immunogenically “cold” state to an inflamed, “hot” lesion, which we demonstrated previously to be a crucial mediator of the efficacy of immune checkpoint inhibition with anti-cytotoxic T lymphocyte-associated protein 4 (CTLA-4). In this study, we focus on the chimeric oncolytic virus vesicular stomatitis virus (VSV)-Newcastle disease virus (NDV), comprised of genetic components of VSV and NDV, and we investigate its utility to support tumor-intrinsic RIG-I-dependent therapy with anti-CTLA-4. Overall, we demonstrate that treatment with VSV-NDV efficiently delays tumor growth and significantly prolongs survival in a murine model of malignant melanoma, which was further enhanced in combination with anti-CTLA-4. Although the direct oncolytic and pro-inflammatory effects of VSV-NDV therapy were independent of RIG-I activation, the synergism with anti-CTLA-4 therapy and associated activation of tumor-specific T cells was critically dependent on active RIG-I signaling in tumor cells. This work highlights the therapeutic value of utilizing an immune-stimulatory oncolytic virus to sensitize tumors to immune checkpoint inhibition.

Graphical abstract

Altomonte and colleagues explored oncolytic viral therapy with chimeric VSV-NDV to support tumor-intrinsic RIG-I-dependent synergistic effects with anti-CTLA-4 therapy. They demonstrate that the oncolytic effects were independent of RIG-I, while synergism with anti-CTLA4 to prolong survival was only observed in the context of active RIG-I signaling.

Introduction

The US Food and Drug Administration (FDA) approval of ipilimumab, a monoclonal antibody against cytotoxic T lymphocyte-associated protein 4 (CTLA-4), in 2011, revolutionized immune oncology.¹ Despite this breakthrough, multiple factors continue to limit the use of immune checkpoint inhibitors (ICIs), such as immune-mediated side effects and a general resistance of the majority of solid cancers to this type of therapy.^2,3,4,5,6 Therefore, intensive research efforts have focused on optimizing the use of ICIs to overcome these limiting factors by combining them with other drugs to increase sensitivity of resistant tumors and enhance the anti-tumor immune response.⁷ In the past years, our increasing understanding of the relationship between the immune status of the tumor microenvironment (TME) and responsiveness to immune checkpoint blockade has highlighted the importance of turning poorly immune cell-infiltrated tumors into inflamed ones to sensitize tumors to immunotherapy.⁸ Activation of retinoic acid-inducible gene-I (RIG-I), a cytosolic sensor of foreign (e.g., viral) or aberrant endogenous RNA, can affect tumor growth and modulate the TME by increasing immune cell infiltration and type I interferon expression.⁹ In addition, active tumor-intrinsic RIG-I signaling has been shown to result in apoptotic cell death and cross-priming of cytotoxic T cells.^10,11 Application of a specific RIG-I ligand (e.g., in vitro-transcribed 5′-triphosphorylated RNA, 3pRNA) to overactivate the pathway has been used therapeutically to sensitize the TME to immune checkpoint blockade, leading to improved treatment responses.^12,13 Even though such (partly nucleic acid-based) agonists have shown encouraging activity in early clinical trials,^14,15 their widespread clinical use is hindered by limited bioavailability and biodistribution as well as the necessity of often toxic liposome-based delivery vehicles.

In this context, oncolytic viruses (OVs), as a new class of cancer immunotherapeutics, offer the promising possibility to potently synergize with ICIs because of their multiple mechanisms of action. OVs are replication-competent viruses that have an inherent or engineered specificity for tumor cells and exert their effects by direct killing of susceptible tumor cells as well as induction of immune responses against the tumor. Viral oncolysis is associated with release of tumor-associated antigens (TAAs) as well as danger-associated molecular patterns (DAMPs) and results in attraction of immune cells to the TME and their activation.¹⁶ Virus-infected cells are highly effective in delivering antigens to dendritic cells (DCs) for cross-presentation and cross-priming of adaptive immune responses.¹⁷ Furthermore, virus-mediated tumor oncolysis is generally considered to be a form of immunogenic cell death (ICD) that is associated with modulation of immune-suppressive mechanisms in the TME and a shift toward a pro-inflammatory phenotype.^18,19 Reports have demonstrated synergism of OV vectors with ICIs in multiple tumor models, including brain and breast cancer,^20,21 and numerous ongoing clinical trials are investigating this combination approach in the clinical setting.^22,23,24 We have reported previously that ICI therapy with anti-CTLA-4 and its combination with anti-PD-1 is dependent on tumor cell-intrinsic RIG-I activation. Therefore, because RNA viruses are known to be potential agonists of RIG-I signaling, we speculated that oncolytic virotherapy with a potent tumor-specific RNA virus might offer a promising combination approach for synergistic immunotherapy with anti-CTLA-4. We have previously reported a novel chimeric OV construct, vesicular stomatitis virus (VSV)-Newcastle disease virus (NDV), which consists of the VSV backbone where the targeting glycoprotein has been replaced with the envelope proteins of NDV.²⁵ This construct was demonstrated to have superior safety and efficacy compared with its parental viruses and offers enhanced anti-tumor immune stimulation because of its ability to mediate a highly immunogenic mechanism of cell death via syncytium formation.^25,26

Here, we investigate oncolytic VSV-NDV and its dependency on RIG-I to synergize with anti-CTLA-4 therapy in immune-competent mice bearing subcutaneous, syngeneic B16.OVA melanoma lesions. In vitro, we could show that VSV-NDV efficiently infects and kills melanoma cells in a RIG-I-independent manner. Additionally, stimulating DCs with the oncolysate of VSV-NDV-infected tumor cells led to a high degree of activation, again independent of the RIG-I status of the tumor cells. In vivo, VSV-NDV treatment delayed tumor growth not only of directly injected but also of distant untreated lesions. This was associated with a significant increase in tumor antigen-specific T cells in VSV-NDV-treated compared with control-treated mice, especially in mice bearing tumors with an active RIG-I pathway. In contrast, the combination of VSV-NDV with systemic anti-CTLA-4 therapy only generated an additional beneficial effect in delaying tumor growth and prolonging survival in mice bearing wild-type (WT) tumors, while mice bearing RIG-I^−/− tumors displayed no synergistic therapeutic benefit of the combination compared with VSV-NDV monotherapy. Overall, our findings demonstrate that VSV-NDV virotherapy can potently augment the response rate to ICI therapy using anti-CTLA-4. Surprisingly, despite the pro-inflammatory context of VSV-NDV infection, this synergistic effect with ICI immunotherapy was strictly dependent on tumor cell-intrinsic RIG-I activity. Combining oncolytic VSV-NDV with anti-CTLA-4 therapy is an approach that warrants further investigation and development for potential clinical translation.

Results

VSV-NDV efficiently infects B16.OVA cells and causes cytotoxicity independent of RIG I activity

First, we tested susceptibility of B16.OVA melanoma cells to VSV-NDV infection and cytolysis. To address the role of tumor-intrinsic RNA sensing via RIG-I, we used WT and RIG-I-deficient melanoma cells (RIG-I^−/−) (Figure 1A). The protocol for gene targeting of Ddx58 (RIG-I) by CRISPR-Cas9 deletion, phenotypic characterization of the B16.OVA RIG-I^−/− cells, and general activity of the pathway in B16 melanoma cells have been published previously.¹³ WT and RIG-I^−/− melanoma cells were mock infected or infected with recombinant VSV-NDV (rVSV-NDV)-GFP at an MOI of 0.01 to visualize and compare the extent of syncytium formation. Representative photomicrographs were captured 24 h after infection. In both cell lines, the uninfected cells appeared healthy and proliferated as a confluent monolayer, whereas large syncytial formations were prevalent in the infected cells (Figure 1B). Virus-mediated GFP expression facilitated visualization of virus replication, which was primarily located in the syncytium and spread throughout the monolayer independent of RIG-I functionality. In addition, viral growth kinetics were analyzed to determine the amount of released infectious virus particles by 50% tissue culture infectious dose (TCID₅₀) assay of cell supernatants collected at various time points post infection. Unexpectedly, rVSV-NDV-GFP displayed a more efficient and rapid replication rate in B16.OVA WT cells than in B16.OVA RIG-I^−/− (Figure 1C). In B16.OVA WT cells, rVSV-NDV-GFP reached its peak in virus titers at about 1 × 10⁷ TCID₅₀/mL by 24 h post infection, while virus replication was delayed in RIG-I^−/− cells and reached a maximum titer of only 1 × 10⁵ TCID₅₀/mL 48 h post infection. At later time points, the virus titers decreased in both cell lines, which indicates depletion of host tumor cells. To evaluate the cytotoxic effect of virus infection, the same supernatants used for the TCID₅₀ assay were subjected to an lactate dehydrogenase (LDH) detection assay. Despite the differences in virus titers, virus-mediated cytotoxicity in both cell lines was similar at all time points with a maximum of 80% 48 h post infection (Figure 1D). Taken together, although rVSV-NDV replication seemed to be somewhat dependent on the presence of RIG-I, both cell lines were highly susceptible to rVSV-NDV infection and showed comparable levels of cell lysis.

Figure 1.

VSV-NDV readily infects and lyses B16.OVA melanoma cells independent of tumor-intrinsic RIG-I

(A) Representative western blot of B16.OVA cells after CRISPR-Cas9-mediated Ddx58 (RIG-I) deletion in comparison with wild-type (WT) control cells. (B) B16.OVA WT cells (top panels) or RIG-I^−/− cells (bottom panels) were infected with rVSV-NDV-GFP (VN-GFP) at MOI 0.01 or PBS (uninfected, left panels), and images were captured 24 h post infection. Representative images (bright field) demonstrating characteristic syncytium formation upon virus infection and the healthy uninfected monolayer were captured at 200× magnification. Scale bars indicate 100 μm. To localize the virus, GFP in the corresponding field was visualized by fluorescence (right panels). (C) B16.OVA WT or RIG-I^−/− cells were seeded and infected with VN-GFP at MOI 0.01. Virus titers were determined via TCID₅₀ assay from tissue culture supernatants collected 0, 16, 24, 48, and 72 h post infection. (D) LDH release into corresponding supernatant as a surrogate marker for cell lysis was determined and normalized to a maximum release control. Data are pooled from three independent experiments (n = 3), and means + SEM are shown. Statistical significance was determined by Student’s t test (∗∗p ˂ 0.01, ∗∗∗p ˂ 0.001, ∗∗∗∗p ˂ 0.0001).

To characterize the effect of rVSV-NDV infection on RIG-I activation, we investigated transcriptional activity of interferon β (IFN-β) as a surrogate marker (Figure S1A). In contrast to stimulation with a defined RIG-I agonist (3pRNA), rVSV-NDV-induced type I IFN (IFN-I) production was independent of functional tumor-intrinsic RIG-I, possibly because of redundant pathways for detection of viral or endogenous RNA during cellular stress responses. Indeed, gene expression of melanoma differentiation-associated protein 5 (MDA5), an alternative sensor of viral RNA, was found to be even more highly upregulated in RIG-I^−/− cells in response to rVSV-NDV infection compared with the WT (Figure S1B), indicating that increased activity of MDA5 could serve as a compensatory mechanism for deficient RIG-I signaling in this context. Furthermore, increased RIG-I gene expression in response to rVSV-NDV infection in WT cells was confirmed (Figure S1C). In summary, these data show that rVSV-NDV infection in tumor cells engages (potentially multiple) innate RNA receptor pathways and activates an IFN-I response and associated IFN-stimulated genes (ISGs). However, these data do not clarify whether the RIG-I pathway, among others, is directly activated by rVSV-NDV.

VSV-NDV lysis of B16 cells leads to efficient activation of DCs in vitro independent of RIG-I activity in tumor cells

To interrogate the role of tumor-cell-intrinsic RIG-I signaling (engaged by either viral RNA or alternative ligands, such as endogenous tumor RNA) in the reported immunogenicity of rVSV-NDV-based cell lysis in an in vitro setting,²⁵ DCs were generated from murine bone marrow cells. B16.OVA WT and RIG-I^−/− cells were mock infected or infected with rVSV-NDV-GFP at an MOI of 0.01. The resulting cell oncolysate was collected after 24 h and UV irradiated to inactivate the infectious virus. This assay allowed us to investigate the role of virus-mediated RIG-I signaling during ICD of tumor cells without the added confounding factor of a direct infection of DCs, which itself could lead to immune cell activation. DCs were exposed to the UV-irradiated oncolysate for 24 h, and DC activation was then measured using flow cytometry to analyze various activation markers. DCs interacting with the virus-induced oncolysate, but not DCs incubated with a physically induced lysate of uninfected melanoma cells, showed maturation with upregulation of major histocompatibility complex (MHC) class I and MHC class II (Figures 2A–2D) as well as the of co-stimulatory molecules CD80 and CD86 (Figures 2B–2F). Genetic deletion of RIG-I signaling in tumor cells had no impact on virus-induced DC activation. Overall, in vitro lysis of WT and RIG-I^−/− B16.OVA cells by rVSV-NDV-GFP infection led to efficient activation of bystander DCs, indicating that rVSV-NDV-GFP-infected melanoma cells underwent ICD and that this effect was independent of active tumor-intrinsic RIG-I signaling in vitro.

Figure 2.

VSV-NDV induces immunogenic oncolysis with activation of bystander DCs

(A–D) DCs were activated by whole-cell lysates based on B16.OVA WT and B16.OVA RIG-I^−/− created by 10 min of UV irradiation (uninfected) or 24-h infection with rVSV-NDV-GFP at an MOI of 0,01 followed by 10 min of UV irradiation (VN-GFP). Activation is measured by mean fluorescence intensity (MFI) of the DC activation markers MHC class I (A), MHC class II (B), CD80 (C), and CD86 (D), normalized to the MFI of naive DCs (n = 5). Data give values of individual mice (and group means ± SEM are presented as bars). Statistical significance was determined by a matched, one-way ANOVA using the Geisser-Greenhouse correction and Šídák’s multiple-comparisons test (∗p ˂ 0.05, ∗∗p ˂ 0.01). (E and F) Representative histograms of CD80 (E) and CD86 (F) signal after DC activation with the uninfected and VN-GFP lysates based on B16.OVA WT.

Anti-CTLA-4 therapy synergizes with rVSV-NDV virotherapy to delay tumor growth in a RIG-I-dependent manner in murine melanoma

To further investigate a possible role of tumor-intrinsic RIG-I activation through rVSV-NDV infection and therefore its synergistic potential in combination with immune checkpoint blockade using anti-CTLA-4, a survival experiment was performed using a subcutaneous melanoma mouse model. To that end, B16.OVA WT or RIG-I^−/− melanoma cells were inoculated subcutaneously in both flanks of recipient mice. Tumor-bearing animals were treated with PBS, rVSV-NDV-GFP, or the RIG-I agonist 3pRNA by direct intralesional injection into the right-side tumor. Additionally, mice in each treatment group were randomized to receive systemic treatment with a suboptimal dosage of anti-CTLA-4 or an isotype control antibody via intraperitoneal injection (Figure 3A). Tumor growth was monitored daily, and mice were euthanized when either tumor reached a diameter of 15 mm, according local regulations. rVSV-NDV-GFP treatment in B16.OVA WT tumor-bearing mice was associated with delayed tumor growth locally (injected tumor) and in the contralateral tumor (uninjected tumor) compared with PBS with or without anti-CTLA-4 therapy (Figure 3B). The combination of rVSV-NDV-GFP with anti-CTLA-4 showed a trend of enhanced tumor control, especially in the directly injected tumors. However, this was not statistically significant for group mean tumor volumes. In line with our in vitro findings, treatment with rVSV-NDV-GFP delayed tumor growth not only in B16.OVA WT tumors but also in B16.OVA RIG-I^−/− tumors, indicating that virus-induced oncolysis is independent of tumor-intrinsic RIG-I (Figure 3C).

Figure 3.

Anti-CTLA-4 therapy synergizes with rVSV-NDV through activation of RIG-I and delays tumor growth in a melanoma mouse model

(A) Female C57Bl/6J mice were inoculated with 2.4 × 10⁵ (injected tumor) and 1.2 × 10⁵ (uninjected tumor) B16.OVA WT or RIG-I^−/− cells subcutaneously on contralateral flanks. One week after tumor implantation, on day 0, mice were randomly distributed into treatment groups (n = 14) and injected intratumorally with rVSV-NDV-GFP (VN-GFP) (1 × 10⁷) or PBS. In addition, anti-CTLA-4 or isotype control antibodies were administered intraperitoneally on day 3 (100 μg) and day 6 (50 μg). (B–D) Tumor growth (width and length) of WT (B) and RIG-I^−/− vs. WT (C and D) tumors was measured daily. Mean tumor volume was plotted up to day 15 for injected and distant (uninjected) tumors according to the indicated treatment groups. Data are pooled from three independent experiments (n = 14). Error bars indicate SEM. Statistical significance was determined by Student’s t test (∗p ˂ 0.05, ∗∗p ˂ 0.01, ∗∗∗p < 0.001 ∗∗∗∗p ˂ 0.0001).

In line with our previous findings regarding the importance of tumor cell-intrinsic RIG-I activity for anti-CTLA-4 immunotherapy in general,¹³ analysis of tumor volumes over the course of treatment further indicated that anti-CTLA-4 failed to synergize with any of the treatments in controlling the growth of injected and distant RIG-I^−/− tumors (Figure 3D). Even in the presence of the highly immunogenic rVSV-NDV-induced oncolysate, anti-CTLA-4-mediated tumor control was largely diminished in RIG-I^−/− tumors. Taken together, these data indicate that, while rVSV-NDV-mediated oncolysis is independent of tumor cell-intrinsic RIG-I functionality, anti-CTLA-4 therapy and its combination with immunogenic local therapies is only effective in the presence of active tumor-intrinsic RIG-I signaling. Treatment with 3pRNA monotherapy was associated with delayed tumor growth of the injected tumor but not of the uninjected tumor (Figure S2). Its combination with anti-CTLA-4 led to more efficient systemic tumor control, which, however, was not associated with statistically improved overall survival. While it may seem rather unexpected that 3pRNA elicited strong therapeutic effects in the injected tumors of the WT and RIG-I^−/− phenotypes, this effect has been observed previously and was attributed to redundant effects of RIG-I in tumor cells and host stromal cells, which can compensate for each other’s function.¹³

To investigate a possible mechanism for abscopal effects, tumor tissue was subjected to qRT-PCR to quantify viral genomes to rule out virus entering the circulation and causing direct oncolytic effects in the contralateral lesions. Interestingly, no rVSV-NDV genomes could be detected in the distant tumors (Figure S3), indicating that an immune-mediated effect was more likely to be the mechanism. Furthermore, in contrast to the in vitro virus growth rates (Figure 1B), viral genome copy numbers were significantly higher in RIG-I^−/− tumors compared with the WT.

rVSV-NDV virotherapy with anti-CTLA-4 therapy prolongs survival in a RIG-I-dependent manner in murine melanoma

The overall survival of B16.OVA WT and RIG-I^−/− tumor-bearing mice was monitored and plotted for the different treatment groups as a Kaplan-Meier curve (Figures 4A–4C). Monotherapy with rVSV-NDV-GFP significantly increased the overall survival of B16.OVA WT tumor-bearing mice compared with PBS monotherapy, and this effect was further improved in combination with anti-CTLA-4, which resulted in the most beneficial treatment outcome (Figure 4A).

Figure 4.

Anti-CTLA-4 therapy synergizes with rVSV-NDV-GFP through activation of RIG-I and improves survival in a melanoma mouse model

(A–C) Female C57Bl/6J mice were inoculated with 2.4 × 10⁵ (injected tumor) and 1.2 × 10⁵ (uninjected tumor) B16.OVA WT or RIG-I^−/− cells subcutaneously on contralateral flanks. One week after tumor implantation, on day 0, mice were randomly distributed into treatment groups (n = 14) and injected intratumorally with rVSV-NDV-GFP (VN-GFP) (1 × 10⁷) or PBS. In addition, anti-CTLA-4 or isotype control antibodies were administered intraperitoneally on day 3 (100 μg) and day 6 (50 μg). Overall survival of treated B16.OVA WT (A) and RIG-I^−/− (B and C) tumor-bearing mice was monitored and plotted for the different treatment groups as a Kaplan-Meier curve. Statistical significance was determined by log rank test (∗p ˂ 0.05, ∗∗p ˂ 0.01, ∗∗∗p ˂ 0.001, ∗∗∗∗p ˂ 0.0001). (D) Two days after the last treatment, blood was collected from the mice to analyze the systemic expansion of tumor antigen (OVA)-specific T cells by flow cytometry. Data points indicate values of individual mice (and group means ± SEM are presented as bars). Data are pooled from four independent animal experiments (n = 19). Statistical significance was determined by Student’s t test (∗p ˂ 0.05, ∗∗p ˂ 0.01, ∗∗∗p ˂ 0.001, ∗∗∗∗p ˂ 0.0001).

Interestingly, survival times of WT versus RIG-I^−/− tumor-bearing mice in response to rVSV-NDV-GFP therapy were not significantly different (Figure 4B). Hence, the therapeutic effects of the monotherapy with rVSV-NDV-GFP appeared to be independent of tumor-cell-intrinsic RIG-I activity, which is consistent with the results of our in vitro infection investigations. In contrast, mice bearing RIG-I^−/− tumors showed significantly reduced benefits in survival prolongation in response to combined treatment of any of the investigated therapies with anti-CTLA-4 (Figures 4C and S2). While we observed prolonged tumor control (>25 days) with the combination therapy only in animals bearing WT and not RIG-I^−/− tumors, there still was a significant increase in overall survival in animals bearing RIG-I^−/− tumors when comparing the combination therapy with anti-CTLA-4 monotherapy. Taken together, these data indicate that, while rVSV-NDV-GFP-mediated oncolysis occurs independent of RIG-I status, anti-CTLA-4 therapy and its combination with immunogenic local therapies is only effective in the presence of active tumor-intrinsic RIG-I signaling.

The frequency of circulating tumor antigen-specific T cells in peripheral blood was analyzed by flow cytometry on day 8 after the first treatment (Figure 4D). rVSV-NDV-GFP resulted in a significant expansion of CD8⁺ T cells with T cell receptor specificity for the tumor model antigen ovalbumin (OVA). This effect was more pronounced in mice bearing B16.OVA WT in comparison with B16.OVA RIG-I^−/−; however, this difference was not statistically significant. The rVSV-NDV-GFP-induced expansion of tumor antigen-specific T cells was further enhanced in combination with anti-CTLA-4 compared with monotherapy. Treatment with the specific RIG-I agonist 3pRNA also resulted in an increase in tumor-specific T cells (Figure S2) but not as efficiently as treatment with rVSV-NDV-GFP. Although an increase in OVA-specific T cells in response to combination therapy compared with rVSV-NDV-GFP monotherapy would have been expected in mice bearing WT tumors, this difference was not statistically significant within the given dataset. Because this effect may be time point dependent, and T cell responses in the periphery (versus within the TME) are only considered surrogate markers in immune oncology, this was not investigated further.

Combination treatment of VSV-NDV with anti-CTLA-4 results in increased T cell infiltration in B16.OVA tumors

To gain further insights into the mechanism of the treatment benefit of combination rVSV-NDV-GFP and anti-CTLA-4 therapy, injected and distant B16.OVA WT and RIG-I^−/− tumors were isolated 9 days after the first treatment, and tumor-infiltrating lymphocytes (TILs) were analyzed by flow cytometry. Monotherapy with rVSV-NDV-GFP resulted in increased infiltration of CD4⁺ T-cells in some of the injected B16.OVA WT tumors compared with PBS (Figure 5A). This effect was also seen after treatment with the RIG-I agonist 3pRNA (Figure S4). Especially rVSV-NDV-GFP in combination with immune checkpoint blockade led to a statistically significant increase in infiltration of CD4⁺ T cells into the injected WT tumor compared with the control treatment groups. In line with our previous finding that tumor cell-intrinsic RIG-I activity strongly modulates the TME,¹³ this effect was completely abrogated in RIG-I^−/− tumors. As described previously,¹³ tumor-infiltrating CD4⁺ T cells were generally much more prevalent in WT tumors compared with RIG-I^−/− tumors. Also, in some distant uninjected WT tumors of rVSV-NDV-GFP-treated mice, we observed increased CD4⁺ T cell infiltration compared with PBS (Figure 5C). In addition, WT tumors showed a higher abundance of CD8⁺ cytotoxic T cells in the TME, not only in the injected but also in the contralateral tumor, after treatment with rVSV-NDV-GFP and its combination with anti-CTLA-4 (Figures 5B and 5D). Again, this effect was largely impaired in RIG-I^−/−-tumors. From these data, we concluded that a combination treatment of VSV-NDV and anti-CTLA-4 results in enhanced tumor T cell infiltration. However, even the strong synergistic immune-stimulatory effects of VSV-NDV and anti-CTLA-4 cannot rescue the previously described immunosurveillance defect in RIG-I^−/− tumors and its diminishing effects on anti-CTLA-4 checkpoint blockade.¹³

Figure 5.

Strong synergistic effects of combination therapy with rVSV-NDV-GFP and anti-CTLA-4 cannot compensate for defective T cell immunosurveillance in the TME of RIG-I^−/− tumors

(A–D) Female C57Bl/6J mice were inoculated subcutaneously with B16.OVA WT or RIG-I^−/− melanoma cells into the right (2.4 10⁵) and left (1.2 × 10⁵) flanks. One week later, mice were randomly distributed into treatment groups (n = 5) and injected intratumorally with rVSV-NDV-GFP (VN-GFP) (1 × 10⁷) or PBS. In addition, anti-CTLA-4 or isotype control antibodies were administered intraperitoneally on day 3 (100 μg) and day 6 (50 μg). Tumors from both flanks were harvested on day 9 after the first treatment. Single-cell suspensions were generated from tumor tissues, and tumor-infiltrating lymphocytes (TILs) were analyzed for CD4⁺ (A and C) or CD8⁺ (B and D) surface expression in directly injected (A and B) and distant uninjected (C and D) tumors. Data points indicate values of individual mice (and group means ± SEM are presented as bars). Statistical significance was determined by Student’s t test (∗p ˂ 0.05, ∗∗p ˂ 0.01).

Combination therapy with VSV-NDV-GFP and anti-CTLA-4 primes cellular anti-tumor cytotoxic function in mice in a tumor-intrinsic, RIG-I-dependent manner

To investigate the functionality of tumor-specific T-cells in treated B16.OVA tumor-bearing mice, splenocytes were isolated from mice on day 9 post treatment start. We have shown previously that loss of tumor cell-intrinsic RIG-I signaling results in defects in priming of tumor antigen-specific T cells by DCs. However, when anti-tumor T cell immunity has formed, the actual killing of target tumor cells by T cells was found to be independent of tumor cell-intrinsic RIG-I.¹³ We now designed an ex vivo experiment to reflect the killing of target tumor cells by T cells primed during previous OV therapy in vivo. Therefore, splenocytes (and included T cells) were exposed to the same tumor cell genotype (in regard to RIG-I functionality) as during their in vivo priming. Following in vivo treatment, bulk splenocytes derived from either WT or RIG-I^−/− tumor-bearing mice were co-cultured with the respective tumor cell line, and the viability of tumor cells was determined after 48 h (see experimental setup shown in Figure S5). Interestingly, splenocytes derived from B16.OVA WT tumor-bearing mice treated with rVSV-NDV-GFP, alone or in combination with anti-CTLA-4, had a significant cytotoxic effect against B16.OVA WT cells ex vivo compared with those from PBS-treated control animals, with approximately 20% of target cells being killed after 48 h of co-culture (Figure 6A). In comparison, splenocytes derived from B16.OVA RIG-I^−/− tumor-bearing mice did not show significant ex vivo cytotoxic function against target cells (Figure 6B).

Figure 6.

Tumor cell-intrinsic RIG-I activity modulates the cytolytic potential of VSV-NDV virotherapy-induced antitumor T cell responses

(A and B) B16.OVA WT or RIG-I^−/− melanoma cells were co-cultured with bulk splenocytes derived from genotype-matched tumor-bearing mice following therapy with VSV-NDV with or without anti-CTLA-4. Cytotoxicity of B16.OVA WT (A) or RIG-I^−/− (B) melanoma cells after 48 h was determined based on quantification of ATP. These data were normalized to the values obtained in the respective tumor cell type cultured alone to calculate the percentage of tumor cell killing. (C) Supernatant from co-cultures was collected, and IFN-γ release was determined by ELISA. (D) For some experiments, splenocyte-derived B16.OVA WT tumor-bearing mice, following therapy with VSV-NDV-GFP with or without anti-CTLA-4, were exposed ex vivo to immunogenic peptide epitopes of the model tumor antigen OVA or the endogenous melanoma-associated antigen TRP2. IFN-γ release was determined by ELISA. Data are presented as mean + standard error of the mean, and statistical significance was determined by Student’s t test (∗p ˂ 0.05, ∗∗p ˂ 0.01, ∗∗∗∗p ˂ 0.0001).

To further examine potential differences in T cell functionality in response to the RIG-I genotype of tumor cells, activation of splenocytes after co-culture was also analyzed for IFN-γ production by ELISA (Figure 6C). Especially re-exposure to tumor cells of splenocytes from B16.OVA WT tumor-bearing mice treated with rVSV-NDV-GFP resulted in efficient production of IFN-γ compared with PBS. Compatible with the common understanding that CTLA-4 controls T cell expansion rather than cytolytic function, this effect was not clearly enhanced when donor mice had been treated with the combination of OV and anti-CTLA-4. In line with our previous finding that basal (low-level) RIG-I signaling in tumor cells can modulate tumor immunosurveillance,¹³ co-culturing splenocytes derived from B16.OVA WT tumor-bearing mice led to overall higher production and release of IFN-γ compared with splenocytes from B16.OVA RIG-I^−/− tumor-bearing mice, even in the absence of any tumor treatment. Therefore, it remains to be determined whether the observed increase in IFN-γ production by splenocytes derived from mice bearing WT tumors and treated with rVSV-NDV-GFP, compared to RIG-I^−/− tumors, is specific to the therapy or rather a reflection of an overall suppression in IFN-γ induction in splenocytes of RIG-I^−/− tumor-bearing mice. Finally, splenocytes derived from treated B16.OVA WT tumor-bearing mice were re-stimulated with immunogenic peptide epitopes of tumor antigens ex vivo. Interestingly, splenocytes from mice treated with rVSV-NDV-GFP and its combination with anti-CTLA-4 not only were reactive to re-exposure to the model antigen OVA but also the endogenous melanoma-associated tumor antigen TRP2 (Figure 6D), suggesting that OV therapy with rVSV-NDV-GFP induces a polyclonal antitumor T cell immune response.

Discussion

Immune checkpoint inhibition has become an indispensable approach in immune oncology for a wide range of tumor entities. Nevertheless, its effects as a monotherapy in many solid tumors has proven to be limited. Evidence indicating that activation of tumor-intrinsic RIG-I, a sensor of foreign (e.g., viral) and aberrant endogenous RNA, plays a key role in sensitizing the TME to anti-CTLA-4 therapy,²⁷ led us to the hypothesis that OV vectors, which may serve as potential RIG-I ligands, could act as enablers of anti-CTLA-4 therapy in addition to providing direct oncolytic effects, leading to synergistic treatment responses.

In this study, we focused on the chimeric OV VSV-NDV and its potential to activate RIG-I and therefore mediate synergistic effects in combination with anti-CTLA-4. Not only did we demonstrate that VSV-NDV could efficiently infect and kill melanoma cells, but the dying cells were highly efficient in activating DCs, indicating that VSV-NDV-infected melanoma cells succumbed to an immunogenic form of cell death. This supports previous observations suggesting immune-stimulatory properties of syncytium-based cell lysis.²⁸ However, VSV-NDV-mediated lysis and the subsequent activation of bystander DCs in vitro did not depend on tumor cell-intrinsic RIG-I activation. Interestingly, VSV-NDV titers were even slightly elevated in WT compared with RIG-I^−/− cells, which was somewhat unexpected given the known anti-viral function of RIG-I. This may be due to other viral RNA sensors in the cells, such as MDA5 or OAS1/RNase L, that may overcompensate for the loss of RIG-I signaling.²⁹ In fact, we observed enhanced gene expression of MDA5 in response to rVSV-NDV infection in RIG-I^−/− tumor cells compared with WT cells. Regardless of the mechanism for the slight difference in VSV-NDV replication, cytolytic effects in RIG-I^−/− tumor cells were nearly identical, indicating that the observed alteration in virus replication was probably negligible in the overall oncolytic effect. In general, our findings indicate that active RIG-I signaling in tumor cells is not an important predictor of susceptibility to VSV-NDV infection. Whether this is true for all tumor cells or specific to B16 melanoma cells must be investigated further. It has been well characterized that RNA viruses, such as VSV and NDV, are sensed via the RIG-I-like receptor family.^30,31,32 Nevertheless, the relative permissiveness of a tumor cell to these viruses will ultimately be determined by the functionality of the downstream IFN signaling pathway, which is often impaired in cancer cells.

The immunogenicity of cell death is generally mediated by release of inflammatory factors from dying cells, the so-called DAMPs. In the context of RIG-I-induced ICD, the described roles of IFN-I as a potential DAMP have been conflicting. In vitro studies have suggested a function of IFN-I as a DAMP in response to RIG-I activation in tumor cells.¹⁰ However, while our previous studies suggested a generally important role of tumor-cell-intrinsic activity of the RIG-I pathway in the immunogenicity of therapy-induced programmed tumor cell death (e.g., radiation therapy),³³ we found tumor-cell-derived IFN-I to be dispensable in vivo in this context.¹³ Here we found that infection with VSV-NDV indeed resulted in potent IFN-I transcriptional activity in tumor cells. However, this was not dependent on tumor cell-intrinsic RIG-I, possibly because of activity of redundant anti-viral RNA recognition receptors. It is important to stress that our data do not conclusively determine whether VSV-NDV directly triggers the RIG-I pathway in tumor cells or whether the pathway may be activated by endogenous tumor RNAs leaked into the cytosol during cellular stress reactions.³⁴ However, the RIG-I pathway has branched downstream signaling apart from IFN-I release, such as induction of immunogenic programmed cell death, which has been shown to be important for cancer immunotherapy.¹³ Therefore, our data suggest that signals downstream of RIG-I other than tumor cell-derived IFN-I are responsible for the prolongation of host survival following immunotherapy with checkpoint inhibition and virotherapy with VSV-NDV.

Activation of bystander DCs by VSV-NDV-induced oncolysates of RIG-I-deficient melanoma cells suggests redundant mechanisms in these tumor cells that trigger the release of DAMPs. This highlights a possible role of RIG-I independent mechanisms, such as sensing of viral RNA by MDA5 in infected tumor cells (which would also induce IFN-I release)³⁵ as well as mechanisms of DCs directly sensing various aspects of viral infection and/or the resulting programmed tumor cell death response. These mechanisms include direct recognition of viral RNA via Toll-like receptor 3 (TLR3), TLR7, and TLR9 and sensing of alternative DAMPs,^36,37,38 such as HSP90, HSP70, and HMGB1,^39,40 all of which are released during VSV-NDV cell lysis.²⁵ At least in this in vitro model, these mechanisms seem to be more central to bystander DC activation than RIG-I-induced IFN-I release from tumor cells, which we initially predicted based on earlier reports.²⁷ However, we cannot fully exclude that UV irradiation of oncolysates, methodologically necessary to prevent spillover of viable virus to DC cultures, artificially inactivated specific DAMPs released after tumor-intrinsic RIG-I activation.

In vivo, we could demonstrate that localized rVSV-NDV-GFP treatment of mice bearing bilateral WT tumors prolonged overall survival, associated with delayed tumor growth of the injected tumor, but also resulted in abscopal effects in distant lesions. This emphasizes formation of a systemic anti-tumor immune response in which the effects in the contralateral tumor are not mediated through direct oncolysis by VSV-NDV-GFP (as evidenced by the lack of detectible virus genomes in the distant tumor), consistent with previous data reported by us and others.^41,42 Furthermore, VSV-NDV-GFP treatment resulted in enhanced expansion of circulating tumor antigen OVA-specific T cells, infiltration of tumors by T cells, and increased anti-tumor cytolytic function. The direct oncolytic and distant abscopal effects of monotherapy with VSV-NDV-GFP were independent of tumor-intrinsic RIG-I activity. This was contrary to the finding that lack of tumor-intrinsic RIG-I signaling was associated with impaired T cell infiltration in the TME with reduced lytic function. We have shown previously that tumor-intrinsic RIG-I activity governs the cross-priming and activation of T cells but not the intrinsic susceptibility of tumor cells to T cell-mediated killing.¹³ These data suggest that VSV-NDV-mediated antitumor immune responses may comprise several redundant cellular components (possibly including innate immune mechanisms such as natural killer [NK] cells) and do not necessarily rely solely on cytotoxic T cells.

In contrast, the very strong synergistic anti-tumor effects of VSV-NDV in combination with anti-CTLA-4 checkpoint inhibition associated with particularly high T cell infiltration in tumors with potent cytolytic activity were only evident in mice bearing tumors with functional RIG-I signaling. In general, treatment with rVSV-NDV-GFP in combination with anti-CTLA-4 increased the abundance of these TILs, especially in the injected tumor, which supports the hypothesis that OV therapy leads to transformation of the TME from that of an immune desert to an inflamed tumor.⁸ However, despite its plethora of immunostimulatory functions, including release of DAMPs, oncolysis via VSV-NDV could, surprisingly, not completely compensate for the previously described defect in tumor immunosurveillance in RIG-I^−/− tumors.¹³ We have shown that basal (low-level) RIG-I signaling in tumor cells modulates the immunogenicity of programmed tumor cell death and therefore is a prerequisite for the success of immune checkpoint inhibition with anti-CTLA-4 with or without anti-PD-1. The ligands that trigger tumor cell-intrinsic RIG-I pathway activity in this context remain to be identified and might comprise endogenous as well as (retro)viral RNAs. Here we show that the necessity of tumor cell-intrinsic RIG-I signaling for the efficacy of checkpoint inhibitors holds true, even with simultaneous virotherapy with VSV-NDV. As mentioned above, whether direct or indirect VSV-NDV-mediated RIG-I signaling contributed to the synergistic effects with anti-CTLA-4 animals bearing RIG-I WT tumors was not definitively clarified in our study. In addition to tumor cell-intrinsic RIG-I, RIG-I signaling in the stromal compartment (including infiltrating innate immune cells) also plays an important role in tumor immunosurveillance and therapy.^13,27 Therefore, we would expect that RIG-I activation in stromal cells also contributes to the therapeutic effects of oncolytic virotherapy with VSV-NDV. It should be noted, however, that only the B16 melanoma cell line was investigated in this study. Therefore, it remains to be seen whether these findings in the context of virotherapy are applicable in the broader tumor cell context and, importantly, in the human setting.

Interestingly, we were able to detect a significant increase in T cells specific not only for the model antigen OVA but also for the endogenous melanoma antigen TRP2, indicating that the therapeutic approach with VSV-NDV was also effective in stimulating polyclonal T cell responses against naturally occurring tumor antigens. The suggested role of tumor-cell-intrinsic RIG-I in generation of antigen-specific T cells in response to OV treatment was not clearly reflected in the frequency of circulating OVA-specific T cells. This seeming discrepancy might be time point dependent or an artifact because of the strong immunogenicity of the model antigen OVA because T cells in these mice showed expansion in peripheral blood but failure to accumulate in the TME.

In summary, here we demonstrate the potent therapeutic effects of oncolytic VSV-NDV treatment with immune checkpoint inhibition using anti-CTLA-4 while identifying RIG-I sensing in the tumor as a necessary step to elicit systemic anti-tumor T cell immunity and the potent synergistic therapeutic benefit of the combination. We suggest that VSV-NDV infection of melanoma cells occurs independent of tumor-cell-intrinsic RIG-I signaling. Local tumor control following intralesional OV injection seems to be mainly mediated by direct tumor cell lysis and, accordingly, occurred also in the absence of active RIG-I signaling. In contrast, efficient expansion and intratumoral accumulation of T cells with high cytotoxic capacity in response to VSV-NDV infection, and thus its synergism with anti-CTLA-4 ICI, was dependent on tumor-cell-intrinsic activity of the RNA receptor RIG-I. This would indicate that VSV-NDV RNA transcripts and/or potentially endogenous tumor cell RNAs released as a consequence of oncolysis act as tumor-intrinsic RIG-I agonists, much in the same way as shown previously for therapeutic 3pRNA, to mediate downstream signaling and subsequent modulation of the TME to enable efficacy of ICI therapy with anti-CTLA-4.

The results presented in this study indicate that VSV-NDV in combination with anti-CTLA-4 therapy could provide profound systemic therapeutic effects in patients harboring tumors with functional RIG-I signaling pathways. This finding, therefore, has high translational relevance and could support use of intratumoral RIG-I activity as a predictive biomarker for identifying patients who would benefit most from the combination of OV therapy and anti-CTLA-4. Furthermore, this work supports continued clinical investigation of OV and ICI-based combination approaches in solid cancers.

Materials and methods

Tumor cell lines and virus

B16.OVA and B16.OVA RIG-I^−/− cells were cultured in complete Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% heat-inactivated fetal calf serum (FCS), penicillin (50 U/mL), streptomycin (50 μg/mL), minimum essential medium (MEM) non-essential amino acids, and sodium pyruvate.^11,13 rVSV-NDV-GFP was generated as described previously.⁴¹ Virus stocks used for the experiments were produced in adherent AGE1.CR.pIX cells (ProBioGen, Berlin, Germany) and purified by ultracentrifugation over a sucrose gradient. For animal experiments, the virus stocks were resuspended in PBS after an additional ultracentrifugation step.

Immunoblotting

Western blots were performed as described previously.¹³ In short, tumor cell proteins were extracted using radioimmunoprecipitation (RIPA) buffer (Invitrogen), including complete protease inhibitor cocktail (Roche, Basel, Switzerland). Protein yield was measured by bicinchronic acid (BCA) assay (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s protocol. Protein separation by SDS-PAGE was done on 10%–12% polyacrylamide gels for 90 min at 80 V. Proteins were blotted to a nitrocellulose blotting membrane (GE Healthcare) for 90 min at 0.3 A, and membranes were subsequently blocked for 90 min in 5% bovine serum albumin (BSA) or 5% milk in 1× TBST (Tris-buffered saline and Tween 20). After incubation with primary antibody (SS1A, Enzo Life Sciences) in blocking buffer overnight at 4°C and three additional washing steps with 1× TBST (each lasting for at least 10 min), secondary antibodies coupled with horseradish peroxidase (HRP) were incubated for 1–2 h at room temperature. After three additional washing steps, signals were visualized using Pierce enhanced chemiluminescence (ECL) western blot substrate according to the manufacturer’s protocol on the INTAS Science imaging system.

Virus growth curves

To generate growth curves of the virus, B16.OVA WT or RIG-I^−/− cells were seeded in a 24-well plate at a density of 1 × 10⁵ cells/well the day before infection. Each cell line was infected with rVSV-NDV-GFP at the calculated MOI of 0.01 or treated with PBS containing Mg²⁺ and Ca²⁺ as a negative control. After 1 h of incubation at 37°C, the cells were washed 3 times with PBS to remove remaining virus particles, and fresh culture medium was added to each well. At various time points (0, 16, 24, 48, and 72 h) after infection, samples of the supernatant were collected for the TCID₅₀ assay in AGE1.CR.pIX cells to determine virus titers.

Cytotoxicity assay

To demonstrate cytotoxicity in response to virus infection with rVSV-NDV-GFP, release of LDH was measured in cell culture supernatant. Cells were seeded, infected, and washed as described for the growth curves. After 16, 24, 48, and 72 h, samples of the supernatant were collected, and LDH-release was determined using the CytoTox96 Non-radioactive Cytotoxicity Assay (Promega, Madison, WI, USA) protocol. Additional wells of uninfected cells were treated with the supplied lysis buffer for 30 min to define maximum release. Absorption at 450-nm wavelength was detected with an absorbance microplate reader (Tecan, Austria). All data are normalized to the maximum release and presented as percent cytotoxicity.

Microscopy

Virus infection was visualized by microscopy, and representative images (bright field and fluorescence) at 200× magnification were captured using an Axiovert 40 CFL microscope (Carl Zeiss, Oberkochen, Germany) with an AxioCam ICm1 attached to the microscope.

In vitro pulsing of DCs with B16 lysates

Murine bone marrow cells were isolated from the femur and tibia of male C57Bl/6J mice. For 7 days, they were cultured in Gibco Roswell Park Memorial Institute (RPMI) 1640 medium with GlutaMAX supplement (Thermo Fisher Scientific), complemented with 10% heat-inactivated FCS, penicillin (50 U/mL), and streptomycin (50 μg/mL) (DC medium). Murine granulocyte-macrophage colony-stimulating factor (GM-CSF) (20 ng/mL) was added to facilitate differentiation into DCs. B16.OVA WT or RIG-I^−/− cells were seeded in a 96-well flat-bottom plate with a density of 1 × 104 cells/well the day before infection. On the day of infection, the medium was changed to DC medium, and each cell line was infected with rVSV-NDV-GFP at the calculated MOI of 0.01 or treated with PBS as a negative control. After 24 h, cells (infected and uninfected) were treated with UV light (302 nm) to inactivate the virus, and 100 μL DC medium containing 1 × 105 DCs was added to the tumor cell lysate. Successful virus inactivation was confirmed by TCID₅₀ assay, which demonstrated that no infectious virus remained after UV treatment. After 24 h of co-culture, the lysate-DC co-cultures were analyzed using flow cytometry.

Animal studies

All animal studies were approved by the institute’s commission for preclinical animal research and the regional government commission for animal protection (Regierung von Oberbayern, Munich, Germany). Female C57Bl/6J mice 6–8 weeks of age were purchased from Janvier Labs (France) and maintained under specific pathogen-free conditions.

A subcutaneous bilateral tumor model was used for survival analysis. Mice were shaved and then injected subcutaneously with B16.OVA WT or B16.OVA RIG-I^−/− cells into the right (2.4 × 10⁵ cells) and left (1.2 × 10⁵ cells) flanks in a volume of 50 μL of PBS. One week later (day 0), when tumors were visible (tumor volume approximately 20–50 mm³), mice were treated by intratumoral injection of rVSV-NDV-GFP (1 × 10⁷/50 μL), 3pRNA (25 μg/50 μL), or PBS (50 μL) into the tumor located on the right lateral flank (injected tumor). The left tumor (uninjected tumor) was not directly treated to allow observation of an abscopal effect on tumor growth. In addition, the monoclonal antibody anti-CTLA-4 (anti-mouse CTLA-4 [CD152]; Bio X Cell, West Lebanon, NH, USA) or an isotype control (polyclonal Syrian hamster immunoglobulin G [IgG], Bio X Cell) was administered intraperitoneally on day 3 (100 μg) and day 6 (50 μg). Approximately 50 μL of blood was collected 2 days after the last treatment from the vena facialis of mice using blood lancets (Supra Megro) and collected in EDTA tubes (Sarstedt, Newton, NC, USA). Blood samples were lysed twice with red blood cell lysis buffer for 5 min and stopped with PBS. The samples were then used for characterization of circulating tumor-specific T cells by flow cytometry. Throughout the experiment, mice were monitored daily, and tumor sizes were measured with a caliper and recorded. Tumor volumes were then calculated using the formula for a sphere: 4/3 × π × ([width + length] / 4)³. According to the requirements of local regulatory agencies, mice were routinely euthanized when the maximum tumor diameter exceeded 15 mm or tumor rupture occurred. At the end of the experiment, survival times with respect to the first injection of treatment were plotted in a Kaplan-Meier survival curve, and median survival times were calculated.

For mechanistic analysis, B16.OVA WT or RIG-I^−/− melanoma cells were implanted into the right and left flanks as described above. One week after tumor implantation, mice were treated as for the survival analysis. On day 9 after the first treatment, mice were euthanized, and tissue samples (blood, spleen, tumors) were collected. For flow cytometry analysis of tumor-infiltrating immune cells, the tumor was mashed through a 40-μm filter after a 30-min incubation in Liberase (Roche) at a concentration of 20 μg/mL. Immune cells were then concentrated via gradient centrifugation (20 min, 1,025 × g) in LymphSep (Biowest, Nuaillé, France) and used for staining of infiltrating antigen-specific T cells, followed by flow cytometry. The measurement time was limited to 60 s to normalize each cell count to the size of the tumor before euthanasia and to correlate immune cell infiltration with tumor size. Blood was collected in EDTA-microvettes (Sarstedt) and centrifuged (10 min, 1,000 × g) and used for flow cytometry staining of circulating tumor-specific T cells. Harvested spleens were mashed through a 40-μm filter and incubated with red cell lysis buffer for 2 min. After washing with mouse T cell medium and a second filtration step, the single-cell suspension of splenocytes was frozen in 10% DMSO in FCS at −80°C until thawing for the co-culture assay and peptide activation assay.

qRT-PCR

RNA was isolated from tumor cell pellets after Ficoll gradient centrifugation using the innuPREP RNA Mini Kit 2.0 (Analytik, Jena, Germany). The presence of VSV-NDV genomes was quantified by qRT-PCR using the GoTaq 1-Step qRT-PCR kit (Promega) and oligonucleotides designed to amplify the transgenic region spanning the VSV N and P genes (Forward (Fwd): 5′-CATGTCACTGCAAGGCCTAAG-3ʹ; Reverse (Rev): 5′-GCCTGATCCAGACGAGAATAG-3′). The reaction was performed using a LightCycler 480 (Roche) and quantified using the integrated data analysis software. A serially diluted RNA standard (the corresponding intragenic region of 200 bp produced by Eurofins Genomics, Germany) was used to generate a standard curve of known genome copy numbers that allowed absolute quantification of genome copies per microgram of RNA of the tumor samples.

For gene expression analysis of IFN-β, B16.OVA WT or RIG-I^−/− cells were seeded in a 6-well plate at a density of 3 × 10⁵ cells/well. The following day, each cell line was mock treated, infected with rVSV-NDV-GFP at the calculated MOI of 1, or transfected with 3pRNA (3 μg/mL, using Lipofectamine 3000 according to the manufacturer’s protocol) as a positive control. After 12 h of incubation at 37°C, cells were lysed, and RNA was isolated from the cell pellet using the innuPREP RNA Mini Kit 2.0 (Analytik). After the RNA was reverse transcribed to cDNA using the ProtoScript II First Strand cDNA Synthesis Kit (New England Biolabs), 2 ng of total cDNA was used for qRT-PCR analysis. Transcript amplification of the murine target genes was performed with Power Up SYBR Green (Thermo Fisher Scientific) using the QuantStudio 3 (Thermo Fisher Scientific) real-time PCR system. The relative transcript level of each gene was calculated according to the 2−ΔΔCt method with normalization to β-actin. Primer sequences were as follows: Ifnb1 Fwd, 5ʹ-GCCTTTGCCATCCAAGAGATGC-3; Ifnb1 Rev, 5ʹ-ACACTGTGCTGGTGGAGTTC-3ʹ; ddx58 (RIG-I) Fwd, 5ʹ-AGCCAAGGATGTCTCCGAGGAA-3ʹ; ddx58 (RIG-I) Rev, 5ʹ-ACACTGAGCACGCTTTGTGGAC-3ʹ; Ifih (MDA5) Fwd, 5ʹ-GCCTGGAACGTAGACGACAT-3ʹ; Ifih (MDA5) Rev, 5ʹ-TGGTTGGGCCACTTCCATTT-3ʹ; Actb Fwd, 5ʹ-CATTGCTGACAGGATGCAGAAGG-3ʹ; Actb Rev, 5ʹ-TGCTGGAAGGTGGACAGTGAGG-3ʹ.

Functional co-culture assay and peptide activation assay

For the functional co-culture assay, frozen splenocytes from treated tumor-bearing mice were thawed and cultured overnight at 37°C in RPMI 1640 medium complemented with 10% heat-inactivated FCS, MEM non-essential amino acids, sodium pyruvate, 1% β-mercaptoethanol, and 102 IU/mL of interleukin-2 (IL-2). The next morning, B16.OVA WT or RIG-I^−/− cells were plated into a 96-well plate (1 × 10⁴ cells/well). The splenocytes were counted and then added at a 10:1 ratio to the B16.OVA cells. The remaining splenocytes were used for the peptide activation assay. After 48 h of co-culture, splenocytes were transferred into a new 96-well plate and spun down (8 min, 150 × g). The supernatant was harvested for IFN-γ quantification by ELISA. Viability of the B16.OVA cells after co-culturing with splenocytes was determined by CellTiter-Glo Luminescent Cell Viability Assay (Promega) with respect to the values obtained from the tumor cells cultured alone. For the peptide activation assay, a 96-well plate was prepared with peptide solution at a final concentration of 1 μg/mL OVA or TRP2 peptide, and 2 × 10⁶ splenocytes from treated B16.OVA WT tumor-bearing mice were added to each well and incubated overnight at 37°C. The next day, the supernatant was harvested for IFN-γ ELISA.

IFN-γ ELISA

To quantify production of IFN-γ by splenocytes co-cultured with B16.OVA or from the peptide activation assay, culture supernatant of the samples was thawed and diluted at a 1:1 or 1:100 ratio with diluent reagent, and the IFN-γ concentration was determined using the IFN-γ Mouse ELISA Kit (Invitrogen, Thermo Fisher Scientific) according to the manufacturer’s protocol. Absorption at 450-nm wavelength was detected with an absorbance microplate reader (Tecan).

Flow cytometry

Flow cytometric measurements were performed using the CytoFLEX S platform (Beckman Coulter Genomics, Brea, CA, USA). For extracellular staining of T cells, isolated splenocytes were incubated for 25 min with FcR blocking reagent (Miltenyi Biotec, Bergisch Gladbach, Germany); washed in PBS; incubated for 1 h at room temperature with CD3, CD4, and CD8 (Miltenyi Biotec) staining antibodies or H-2kb OVA Tetramer-SIINFEKL (MBL International, Woburn, MA, USA); washed again; and processed according to the manufacturer’s protocols.

Invitro-generated and activated DCs were washed in PBS, incubated for 30 min at room temperature with the 7-AAD viability dye (Sigma Life Sciences, Merck, Darmstadt, Germany), CD11c, CD80 (BioLegend, San Diego, CA, USA), CD86 (BD Biosciences, Franklin Lakes, NJ, USA), MHC class I (Invitrogen, Thermo Fisher Scientific), MHC class II (Miltenyi Biotec) DC lineage and activation marker antibodies, washed again, and processed according to the manufacturer’s protocols. The compensation was performed based on staining results from UltraComp beads (Thermo Fisher Scientific). All flow cytometry results were analyzed using Flow Jo (Ashland, OR, USA) software.

Statistical analysis

All data were plotted and analyzed using GraphPad Prism 7.0 (GraphPad, San Diego, CA, USA). Individual data points were compared for statistical significance using an unpaired Student’s t test, and p values of less than 0.05 were considered to be statistically significant (∗p ˂ 0.05, ∗∗p ˂ 0.01, ∗∗∗p ˂ 0.001, ∗∗∗∗p ˂ 0.0001). Survival data were plotted as Kaplan-Meier curves, and statistical significance was calculated by log rank test. For in vitro DC activation, DCs were generated from five separate mice in five separate experiments. Different conditions in each experiment were applied in triplicates, and the mean of each triplicate was used for further analysis. Data gathered in one experiment were considered part of one experimental block. Statistical significance was determined by a matched, one-way ANOVA using the Geisser-Greenhouse correction and Šídák’s multiple-comparisons test (∗p ˂ 0.05, ∗∗p ˂ 0.01).

Data and code availability

All raw and analyzed data used in the preparation of this manuscript are available upon request.