Virotherapy with a Semliki Forest Virus–Based Vector Encoding IL12 Synergizes with PD-1/PD-L1 Blockade

Semliki Forest virus (SFV)–based RNA viral vectors have been developed for vaccination and immunotherapy of cancer (1, 2). These propagation-deficient vectors can accommodate expression cassettes and have been engineered to encode cytokines and other proimmunogenic transgenes. A SFV vector encoding the heterodimeric murine interleukin-12 (IL12; SFV-IL12) proteins (3–5) under the control of two independent subgenomic promoters is a powerful antitumor agent when directly injected into experimental tumors (6). The effect is the result of killing tumor cells in an immunogenic fashion, the local effects of the IL12 transgene, and the innate immune activation caused by viral RNA, including a powerful type I IFN response (7). In a previous study, we observed that SFV-IL12 and agonist costimulatory antibodies injected intratumorally (i.t.) and directed against the lymphocyte surface receptor CD137 were markedly synergistic (8).

Cancer immunotherapy is currently being revolutionized by the advent of monoclonal antibodies (mAb) blocking PD-1 and its PD-L1 ligand (9–11). Such antibodies exert antitumor effects in numerous mouse models and have shown frequent durable objective responses in melanoma, renal cell carcinoma, non–small cell lung cancer, and other malignancies (12–14).

In this study, we explored the combination of the local proimmunogenic effects of SFV-IL12 along with the relief of the coinhibitory restraints mediated by PD-1 on the antitumor cytotoxic T-lymphocyte (CTL) response. Potent synergistic effects of the combination were observed with evidence for PD-L1 induction on tumor cells by the IL12–IFNγ axis. Such a train of events results in escape mechanisms that are tackled by anti–PD-1 mAb upon combination.

The BHK-21 cell line (ATCC CCL-10) was cultured as described previously (4). The murine melanoma cell line B16.F10 and the OVA-expressing variant B16-OVA (H-2^b), the MC38 mouse adenocarcinoma cell line and the 4T1 murine breast cancer cell line, were tested to be Mycoplasma free and verified for identity. Five-week-old female C57BL/6 and Balb/c mice were purchased from Harlan Laboratories. Animal studies were approved by the institutional ethics committee for animal experimentation under Spanish regulations (protocol number 071-13).

Plasmids pSFV-enhLacZ and pSFV-enhIL12 have been described previously (4, 15). Plasmid pSFV-NS3 encodes the cytoplasmic Hepatitis C virus NS3 protein (C. Smerdou and colleagues; unpublished results). RNA synthesis from SFV plasmids, transfection into BHK-21 cells by electroporation, and packaging of recombinant RNA into SFV viral particles (vp) were performed as described previously (16, 17). Briefly, BHK-21 cells were coelectroporated with recombinant RNA and two helper RNAs (i.e., SFV-helper-C-S219A and SFV-helper-S2 RNAs), which provided in trans the SFV capsid and spike proteins, respectively (17). SFV vp were harvested and purified by ultracentrifugation as described previously (18). Indirect immunofluorescence was applied to SFV-infected BHK-21 cells to determine the titer of SFV-NS3, SFV-enhLacZ, and SFV-enhIL12 (the last two designated in this article as SFV-LacZ and SFV-IL12, respectively) recombinant virus stocks as described previously (19). Viral vectors were kept frozen at −80°C until use.

Mice were s.c. injected with B16-OVA, B16.F10, MC38, and 4T1 cells (5 × 10⁵ cells per animal) in the right flank. Mice injected with B16-OVA or B16.F10 cells were injected again in the contralateral flank 6 days following the first inoculation in the right flank. Treatments were administered 6 to 10 days following tumor cell inoculation as detailed for each experiment. SFV-LacZ and SFV-IL12 were resuspended in 25 μL of PBS and injected i.t. with 28-G needles. Three doses of 100 μg of anti–PD-1 mAb (clone RMP1-14; Bio X Cell) and/or anti-CD137 mAb (clone 1D8, kindly provided by Maria Jure-Kunkel, BMS) or anti–PD-L1 mAb (10B5, hybridoma) were administered i.p. as detailed for each experiment. The efficacy of treatment was evaluated by measuring the size of tumor nodules every 3 to 4 days.

Splenocytes from naïve C57BL/6 mice were stained with 5 μmol/L of Violet Proliferation Dye 450 (BD Biosciences) and then divided into three different samples. Two of these samples were pulsed with 10 μg/mL of the OVA_257–264 peptide (SIINFEKL, NeoMPS) or the TRP-2_180–188 peptide (SVYDFFVWL, NeoMPS) for 30 minutes at 37°C in 5% CO₂ and extensively washed. The third sample received no peptide as an additional control. OVA-pulsed, TRP-2–pulsed, and nonpulsed splenocytes were then labeled with 1 μmol/L (CFSE^hi), 100 nmol/L (CFSE^med), or 10 nmol/L (CFSE^low) of CFSE (Sigma-Aldrich) and with Violet Proliferation Dye 450 (BD Horizon) at 10 μmol/L in all cases. The three samples were mixed at the same ratio and injected i.v. (3 × 10⁶ cells of each population) into B16-OVA–bearing or naïve mice. Twenty hours later, spleens were harvested, and the transferred target cells were identified and gated using Violet Proliferation Dye 450. Specific cytotoxicity was analyzed by flow cytometry and calculated as follows: 100 − [100 × (% CFSE^hi or CFSE^med tumor–bearing mice/% CFSE^low tumor–bearing mice)/(%CFSE^hi or CFSE^med naïve mice/% CFSE^low naïve mice)].

B16-OVA tumors were generated as described above and inoculated with 10⁸ vp of SFV-LacZ, or SFV-IL12, or an equivalent volume of saline. SFV-IL12–injected mice also received 1 mg of a blocking anti-IFNγ mAb (clone HB170, hybridoma) or an equivalent volume of saline i.p. After 3 days, tumors were harvested and treated with 400 U/mL collagenase D and 50 μg/mL DNase I (Roche Diagnostics). After mechanical tissue dissociation, cells were passed through a 70-μm nylon mesh filter (BD Falcon, BD Bioscience) and washed. Cells were also treated with ACK lysing buffer and washed before further use. Finally, cell suspensions were immunostained to assess H2-K^b and PD-L1 expression on gated CD45-negative malignant cells by flow cytometry.

B16-OVA cells were plated in triplicate in 12-well cell culture plates at a seeding density of 5 × 10⁴ cells per well. When cells had attached to the wells, 3 hours later, fresh splenocytes from C57BL/6 mice were added or not at three different B16-OVA/splenocytes ratios: 1:1, 1:5, and 1:10. Supernatants from SFV-enhIL12– or SFV-NS3–infected BHK cells were added to the cell cocultures, in the presence or absence of a blocking anti-IFNγ mAb (clone HB170, hybridoma) at 10 μg/mL. As an additional negative control, B16-OVA cells without splenocytes were similarly treated and analyzed. As a positive control, B16-OVA cells were cultured in the presence of 10³ U/mL of IFNγ (Miltenyi Biotec). Three days later, B16-OVA cells were collected, washed with PBS, and analyzed for H2-K^b and PD-L1 surface expression on gated CD45-negative populations.

Single-cell suspensions obtained from in vivo growing tumors or in vitro experiments were pretreated with anti-CD16/32 mAbs (clone 2.4G2; BD Biosciences Pharmingen) to reduce nonspecific binding to Fc receptors. Then, cells were stained as needed with the following fluorochrome-conjugated antibodies: CD45.2 (100), PD-L1 (10F.9G2), and H2-K^b (AF6-88.5) purchased from Biolegend. Dead cells were discarded during the analysis by staining with Sytox Green or 7-AAD (Molecular Probes). A FACS-Canto II (BD-Biosciences) was used for cell acquisition, and data analysis was carried out using FlowJo software (TreeStar Inc.).

All error terms are expressed as SD. Prism software (GraphPad Software Inc.) was used for statistical analysis. Survival of tumor-bearing animals was represented by Kaplan–Meier curves and analyzed by log-rank tests. Data were analyzed first by the Kolmogorov–Smirnov normality test. The Kruskal–Wallis tests followed by Dunn multiple comparison tests were used to permit comparisons in experiments with four experimental groups. P values of <0.05 were considered to be statistically significant. The synergy index (S) was studied using the formula described by Mazat and colleagues (20).

To test the combined effects of i.t. SFV-IL12 and anti–PD-1 mAb blockade, we chose the B16-OVA bilateral melanoma model. In this setting, mice were s.c. inoculated twice with tumor cells in each flank with a 6-day interval between injections. On day 10, SFV-LacZ or SFV-IL12 was given i.t. in the nodule firstly induced, along with systemic anti–PD-1 mAb or saline given on days 10, 13, and 17. Figure 1A shows the progression of individual tumors and demonstrates that combined treatment was superior to achieve control of the bilateral experimental disease, resulting in an evident survival advantage (Fig. 1B) with over 80% of mice remaining completely tumor free at the end of the experiment. According to the formula of the synergy index (S) described by Mazat and colleagues (20), we calculated an S value of +9.11, clearly superior to zero. Therefore, we can conclude that SFV-IL12 + anti–PD-1 synergized to reject treated tumors completely.

Such synergistic effects were also observed upon treatment of MC38-derived tumors that were injected with a suboptimal dose of SFV-IL12 that by itself only cured 1 of 6 mice (Fig. 1C). However, upon treatment with the combined regimen, most of the mice were completely cured in an experimental setting in which anti–PD-1 mAb as a single agent only resulted in a marginal retardation of tumor growth with no survival benefit (Fig. 1D). In this experiment, a positive S (+3.88) value was computed, thus demonstrating a synergistic rather than a merely additive effect of the combined treatment. Again, a majority of mice treated with the combination were tumor free at the end of the experiment.

The PD-1/PD-L1 pathway is targeted in the clinic with agents directed against either one of these counter receptors. In our combination regimen, we tested in parallel both options (anti– PD-1 or anti–PD-L1 mAb) and obtained similar results on B16-OVA–derived tumors (Supplementary Fig. S1).

These experiments were extended to the untransfected bilateral B16.F10 tumor model and the 4T1 breast cancer model. In both instances, a less evident but reproducible therapeutic effect was observed with the combined treatment (Supplementary Fig. S2A and S2B). In the case of 4T1-derived breast carcinomas, a group of mice was treated with a triple combination of SFV-IL12 i.t. + anti–PD-1 + anti-CD137 mAbs with a trend suggestive of better survival (Supplementary Fig. S2B). Our data strongly support synergistic therapeutic effects of i.t. injection of SFV-IL12 with PD-1 blockade that were exerted both against the SFV-IL12–injected tumor and distantly implanted lesions that were not injected with the viral vector.

Both SFV-IL12 and anti–PD-1 mAbs rely on amplified CTL antitumor responses to exert therapeutic effects (4, 21). Hence, an alternative explanation for the combined effects could be the induction or amplification of a more robust antitumor CTL response. In the same therapeutic setting of Fig. 1A and B, we performed in vivo killing experiments testing CTLs against the OVA-derived peptide SIINFEKL and an endogenous melanosomal peptide antigen (TRP-2) also presented by H2-K^b. The results indicated only a marginal improvement in cytotoxicity with the combination regimen on the CTL response in the case of the TRP-2 antigen but none in the case of the OVA-derived antigen (Fig. 2). In the case of the OVA epitope, SFV-IL12 as a single agent induced a nearly optimal response, which is difficult to be improved upon by the PD-1 blockade. Of note, PD-1 as a single agent induced a very modest increase in CTL responses, at least in this difficult-to-treat tumor model. Thus, we reasoned that the cooperative mechanism for synergy ought to be different and hypothesized that the virally encoded IL12 could act as an inducer of PD-L1 on tumor cells via IFNγ (21), thereby resulting in a tumor-protective factor against CTLs as elicited by SFV-IL12.

Indeed, as shown in Fig. 3A and B, SFV-IL12 i.t. injection induced bright expression of PD-L1 on CD45-negative tumor cells analyzed by flow cytometry. This was paralleled by a concomitant increase in the surface expression of MHC class I molecules. Induction of both MHC-I and PD-L1 only took place with the vector encoding IL12 but not with a similar SFV vector encoding β-galactosidase as a control transgene. Moreover, IFNγ blockade in vivo with a neutralizing mAb abolished the induction of both PD-L1 and MHC-I. Similar results were observed in MC38-derived tumors (data not shown). Therefore, IL12 induction of IFNγ in the tissue microenvironment is responsible for the induction of PD-L1 and MHC-I. Although PD-L1 would be detrimental for CTL tumor destruction (22), MHC I would probably enhance tumor-antigen presentation.

This train of phenomena could be mimicked in cell cultures, for example, the coculturing B16-OVA melanoma cells with syngeneic splenocytes as a putative source of IFNγ in response to IL12. Indeed, when such cocultures were supplemented with virus-free and IL12-containing supernatants from BHK cells that had been infected with SFV-IL12, tumor cells increased their surface expression of MHC I (data not shown) and PD-L1, an effect that was also abolished with the addition of an anti-IFNγ mAb to neutralize this cytokine as produced by the cocultured lymphocytes (Fig. 3C and D).

This study provides strong preclinical evidence to combine virotherapy and checkpoint blockade for cancer treatment. Previous results have pointed to the synergistic effects of tumoral infection with Newcastle disease virus and blockade with anti–CTLA-4 mAbs (23). Immunostimulatory mAbs directed against CD137 to enforce T-cell costimulation are also strongly synergistic with intratumoral Vaccinia virus–derived vectors (24) and SFV-IL12 (8).

Viral vectors, even if not replicative, mimic an acute pathogenic viral infection, sending a compendium of alarm signals that call for a CTL-mediated immune response (25). We hypothesize that mimicking viral infection of the tumor should trigger or sustain the proper type of cytotoxic inflammation in the tumor microenvironment. With this rationale in mind, checkpoint inhibitors operating in the malignant tissue can unleash the efficacy of CTL response that had resulted from tumor infection or from the expression of proimmunogenic transgenes, such as IL12.

In this study, an undesirable loop is elicited by SFV-IL12, which, in an IFNγ-mediated fashion, induces PD-L1 on tumor cells. Hypoxia via HIF1α may also contribute to PD-L1 overexpression (21) because IL12 is known for its antiangiogenic effects (3). We have observed that blockade of such adaptive resistance mechanisms results in powerful synergistic effects that can control distant tumor lesions not treated with the viral SFV vector.

Combinations of local virotherapy and systemic checkpoint blockade are suitable for testing first in metastatic melanoma with accessible cutaneous metastasis, in which treatment is provided to one or some of the lesions in an attempt to turn transduced lesions into an in situ cancer vaccine (26). Combined strategies of this kind, based on knowledge of the immunodynamic interaction of the viral vector and checkpoint blocking antibody, hold much promise for clinical translation because both types of agents are currently undergoing clinical trials.

I. Melero reports receiving a commercial research grant from Pfizer and is a consultant/advisory board member for Bristol-Myers Squibb, Roche-Genentech, and AstraZeneca. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.I. Quetglas, I. Melero

Development of methodology: J.I. Quetglas, C. Smerdou, I. Melero

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.I. Quetglas, M.Á. Aznar, E. Bolaños, A. Azpilikueta, E. Casales

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.I. Quetglas, V. Segura, I. Melero

Writing, review, and/or revision of the manuscript: J.I. Quetglas, S. Labiano, M.Á. Aznar, A.R. Sánchez-Paulete, V. Segura, C. Smerdou, I. Melero

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.I. Quetglas, S. Labiano, M.Á. Aznar, E. Bolaños, A. Azpilikueta, I. Rodriguez, E. Casales, A.R. Sánchez-Paulete

Study supervision: J.I. Quetglas, I. Melero

The authors thank Drs. J.L. Perez-Gracia, J. Prieto, S. Hervas-Stubbs, J.J. Lasarte, and S. Martín-Algarra for scientific discussion and long-term collaborations in the development of combined immunotherapies. The authors also thank Dr. Diego Alignani for flow cytometry support and Eneko Elizalde for excellent animal facility care.

This study was financially supported by MICINN (SAF2008-03294 and SAF2011-22831, to I. Melero) and FIS (PI11/02190 and PI14/01442, to C. Smerdou). I. Melero was also funded by the Departamento de Salud del Gobierno de Navarra, Redes temáticas de investigación cooperativa RETICC, and European Commission VII Framework Program (ENCITE and IACT).