Cancer immunotherapy is joining surgery, chemotherapy, and radiotherapy as a standard treatment modality in the war against cancer. It relies on tipping the cancer–immune set point1; that is, changing the equilibrium between factors that stimulate or inhibit anticancer immunity to restore anticancer effects. These tumor cell-intrinsic or cell-extrinsic factors within the tumor microenvironment (TME) are coordinated at each step of the cancer–immunity cycle.2 3 Immune checkpoint blockade is arguably the most important approach during the past decade to modulate the cancer–immunity cycle for cancer therapy. Such blockades include application of antibodies to target the cytotoxic T-lymphocyte antigen 4 (CTLA-4)/CD284 and programmed cell death 1 (PD-1)/PD-ligand 1 (PD-L1) axes.5 While dramatic, durable, and therapeutic responses are observed after immune checkpoint inhibitor therapy, these successes are still limited to a small percentage of patients.6 This is largely attributed to the lack or paucity of T cell infiltration in most tumors, defining the so-called non-T-cell-inflamed or “cold” tumors.7 Therefore, approaches that can convert non-T-cell-inflamed or “cold” tumors into T-cell-inflamed or “hot” tumors, mainly characterized by an abundance of tumor-infiltrating T cells, are urgently needed.

Oncolytic viruses can selectively replicate in and destroy tumor cells and stromal cells while exposing antigens for cross-presentation and elucidating danger signals, triggering innate and adaptive immunity.8 Oncolytic viruses have positive effects on almost every aspect of the cancer–immunity cycle and can be further armed with chemokines to improve tumor T cell infiltration; with cytokines to reprogram immune cells; or with ligands or antibodies to block immune checkpoints and relieve immunosuppression.9 Thus, oncolytic viruses might be perfect candidates to modulate the TME and convert “cold” tumors into “hot” tumors. Cytokines have gained researchers’ continuous interest for improving the therapeutic effects of oncolytic viruses.10 However, uncontrolled cytokine expression and secretion by a live, replicating virus may lead to toxicity and cytokine release syndrome. Recently, we generated an oncolytic vaccinia virus expressing membrane-bound interleukin 2 (IL-2) to diminish the severe toxicity associated with systemic IL-2 exposure and to modulate the TME. This virus not only reduced toxic side effects but also tipped the cancer–immune set point in the TME and treated several murine tumor models. In combination with PD-1/PD-L1 blockade, this virus cured most of the mice with a high tumor burden.11 IL-12 is another cytokine with great expectations for cancer therapy based on the therapeutic success of IL-12 treatment in a variety of murine tumor models.12 However, these expectations have been greatly diminished because systemic IL-12 in clinical trials evoked severe side effects such as leukopenia and thrombocytopenia.13–15 Several IL-12 modifications have been explored to improve its safety while preserving its activity.16–21 The goal of this study was to investigate whether and how an oncolytic vaccinia virus expressing membrane-bound IL-12 modulates the TME and further improves antitumoral effects alone or in combination with immune checkpoint blockade.

To reduce the severe toxic side effects caused by the systemic application of IL-12, we used vvDD, a double viral gene-deleted (tk− and vgf−) vaccinia virus, to deliver membrane-bound IL-12 into the tumor bed. This replicating virus is tumor-selective22 and its safety has been demonstrated in clinical trials.23 24 We constructed vvDD-IL-12, vvDD-IL-12-FG, and vvDD-IL-12-RG to express secreted IL-12 or membrane-bound IL-12 after infection in tumor cells, respectively. Membrane association is realized by the glycosylphosphatidylinositol (GPI) anchor form of human CD16b. The difference between vvDD-IL-12-FG and vvDD-IL-12-RG is that between the IL-12p40 subunit and GPI anchor, the former has a flexible linker (G₄S)₃ and the latter has a rigid linker (A(EA₃K)₄AAA)25 (online supplementary figure S1). When MC38-luc, AB12-luc, and B16 cells were infected with these three IL-12-armed viruses or control virus vvDD at a multiplicity of infection (MOI) of 1, viral housekeeping gene (A34R) mRNA levels were similar among the viruses, consistent with similar in vitro infection and replication of the different viruses. In addition, the IL-12 mRNA levels were similar in IL-12-armed viruses, as expected (figure 1A). We also measured IL-12 expression at the protein level using ELISA and flow cytometry. The amount of IL-12 in the supernatant from vvDD-IL-12-infected tumor cells was significantly higher than the other constructs (figure 1B), while cell-surface IL-12 expression was significantly more prevalent in vvDD-IL-12-FG-infected or vvDD-IL-12-RG-infected cells (figure 1C; online supplementary figure S2). This demonstrates the successful realization of membrane association by GPI anchored to one subunit of IL-12. This was further confirmed by measuring the amount of IL-12 cleaved by phosphatidylinositol-specific phospholipase C (PI-PLC) from membrane-associated GPI-anchored IL-12. As expected, membrane-bound IL-12 correlated with virus MOI (figure 1D). We further demonstrated that the GPI-anchored IL-12 is functional in vitro. Con A-activated mouse splenic T cells were co-cultured with mitomycin C (MMC)-inactivated MC38 cells, which were mock-infected or infected with the viruses overnight and washed thoroughly with phosphate-buffered saline (PBS) before co-culture. Only vvDD-IL-12-FG-infected or vvDD-IL-12-RG-infected MC38 cells which had membrane-associated IL-12 could stimulate the proliferation of activated T cells. This stimulation was significantly attenuated by IL-12 antibody neutralization (figure 1E). In a previous report, a single-chain IL-12 fused with GPI anchor from human folate receptor and a flexible linker from pyruvate ferredoxin oxidoreductase was used for membrane association; however, the percentage of IL-12⁺ cells was moderate and a large amount of IL-12 was released into the culture supernatant after adenovirus delivery.17 This indicates that the delivery vector, GPI anchor sequence, and linker sequence might affect the membrane association of GPI-anchored proteins. In fact, we previously showed that a rigid linker between the IL-2 and GPI anchor is better for IL-2 membrane association than a flexible linker,11 whereas here a flexible linker was better for IL-12 membrane association suggesting an effect of the nature of targeted proteins on the design of a GPI-anchored protein. The cytotoxicity of IL-12-armed viruses was tested in four murine tumor cells. The results demonstrated that they had similar in vitro cytotoxicity compared with the parental virus vvDD (figure 1F).

To investigate the toxicity induced by these viruses, we first measured the levels of IL-12 and interferon γ (IFN-γ) in mouse sera and tumor nodules. The IL-12 levels were significantly higher in sera from mice treated with vvDD-IL-12 compared with those treated with membrane-bound forms (online supplementary figure S3A); however, as a main mediator of IL-12-induced antitumor effects,26 IFN-γ levels in sera were similar after the different IL-12-armed virus treatments (online supplementary figure S3B). Therefore, membrane-bound IL-12 might have a similar function with a lower risk of toxicity. The levels of either IL-12 or IFN-γ in tumor nodules receiving vvDD-IL-12-FG treatment were significantly higher compared with other treatments (figure 2A,B). We next investigated the membrane association of IL-12 in vivo using flow cytometry. There were significantly more IL-12⁺ cells in the tumor in mice receiving vvDD-IL-12-FG treatment than in those receiving other treatments (online supplementary figure S3C). This further confirmed that vvDD-IL-12-FG could retain IL-12 at the tumor site, compared with other constructs. We also found that only vvDD-IL-12 treatment (1×10⁹ plaque-forming units (PFU)/mouse) induced lung and kidney edema as evidenced by an increase in water content in lungs and kidneys after treatment (figure 2C,D), and liver toxicity as evidenced by an increase in aspartate transaminase (AST) and alanine transaminase (ALT) in the sera (figure 2E,F). While the mice tolerate doses up to 1×10⁹ PFU and primates have been shown to be more permissive to WR strain vaccinia virus infection,27 we anticipate that a similar dose expressing membrane-tethered IL-12 may be potent in humans. Collectively, our data demonstrate that vvDD-IL-12-FG can effectively maintain IL-12 in the TME and is therefore safer. Hence, we selected vvDD-IL-12-FG for further investigation in the remainder of this study.

To evaluate the antitumoral efficacy of vvDD-IL-12-FG, we intraperitoneally injected virus at the dose of 2×10⁸ PFU/mouse to treat B6 mice bearing 5-day-old peritoneal murine colon cancer (MC38-luc). Survival results demonstrated that vvDD-IL-12-FG and vvDD-IL-12 elicited potent antitumoral effects compared with PBS or vvDD treatment (figure 2G). Impressively, vvDD-IL-12-FG treatment cured all mice that received the treatment, though there was no significant difference in survival between the different IL-12-armed virus treatments. All the mice bearing peritoneal MC38-luc cured by vvDD-IL-12-FG treatment received a subcutaneous re-challenge of either MC38 or an irrelevant tumor control, Lewis lung cancer (LLC). MC38 tumor growth was retarded in the cured mice (figure 2H), but LLC growth was not (figure 2I), compared with naïve mouse control, suggesting that a systemic tumor-specific antitumor immunity was elicited. We explored the therapeutic efficacy of vvDD-IL-12-FG using 2×10⁸ or 1×10⁸ PFU/mouse to treat BalB/c mice bearing 5-day-old peritoneal murine colon cancer (CT26-luc; online supplementary figure S3D) or murine mesothelioma (AB12-luc; online supplementary figure S3E) with similar results, respectively. We also evaluated the antitumoral efficacy of vvDD-IL-12-FG in 9-day tumor-bearing mouse models, which are more akin to metastatic human tumors, characterized by heavier tumor burden and increased immunosuppressive factor expression in the TME (PD-1, PD-L1, CTLA-4, transforminggrowth factor β (TGF-β), CD105, and vascular endothelial growth factor (VEGF)).11 Both IL-12-armed virus treatments significantly improved survival compared with PBS or vvDD treatment in this 9-day MC38 model (figure 2J). The splenocytes from 9-day MC38-bearing mice receiving IL-12-armed virus treatment secreted significantly more IFN-γ after restimulation with MMC-inactivated MC38, suggesting systemic antitumor immunity in the late-stage tumor model (figure 2K). Similar survival results were obtained using a 9-day AB12-bearing mouse model (figure 2L). Occasionally, a few mice that received vvDD-IL-12 treatment, but not vvDD-IL-12-FG treatment, died earlier than those that received PBS treatment (figure 2L), consistent with the IL-12-induced toxicity.

To explore the mechanism by which vvDD-IL-12-FG treatment elicits antitumoral immune activity in the profoundly immunosuppressive advanced tumor model, we investigated the immune cell profile in the TME using the late-stage tumor model. The percentages of CD4⁺Foxp3⁻ and CD8⁺ T cells from tumors receiving IL-12-armed virus treatment were increased compared with those treated with PBS or vvDD (figure 3A,B) and the IFN-γ secretion from both CD8⁺ and CD4⁺ T cells was increased (figure 3C,D). The results also showed that the exhausted PD1⁺CD8⁺ T cells, more severely exhausted PD1⁺Tim-3⁺CD8⁺, PD1⁺TIGIT⁺CD8⁺, and PD1⁺LAG-3⁺CD8⁺ T cells in the tumor-infiltrating CD8⁺ T-cell population decreased after IL-12-armed virus treatment (figure 3E–H). We examined myeloid-derived suppressor cells (MDSCs) in tumors after virus treatment. As reported by a previous study,28 we found that granulocytic MDSCs (G-MDSCs, CD11b⁺Ly6G⁺Ly6C^low) were increased after vvDD treatment; however, they were decreased after IL-12-armed virus treatment (figure 3I). We also examined regulatory T cells (Tregs, CD4⁺Foxp3⁺) and found that the percentage of CD4⁺Foxp3⁺ T cells in tumor-infiltrating CD4⁺ T cells was decreased after IL-12-armed virus treatment (figure 3J), implying suppression of tumor-induced Treg proliferation by the IL-12/IFN-γ axis,29 owing to the significantly higher IL-12 and IFN-γ levels in tumors that received IL-12-armed virus treatments (figure 2A,B; online supplementary figure S4A). However, the increase of IFN-γ in the TME did not upgrade the expression of PD-1 on T cells (figure 3E; online supplementary figure S4B-C) or PD-L1 on CD45⁻ or CD11b⁺ cells (online supplementary figure S4D-F), suggesting that IL-12 did not reinforce adaptive immune resistance in vvDD-related therapy.30 We next found a significant decrease in the expression of pro-cancer factors, including TGF-β, cyclooxygenase-2 (COX-2), and angiogenesis markers (VEGF and CD105) in tumors after IL-12-armed virus treatment compared with other treatments (figure 3K–M; online supplementary figure S4G). It is interesting that although the IL-12-armed virus treatment did not decrease the TGF-β expression in CD11b⁺ cells (online supplementary figure S4H), it did decrease the TGF-β expression in Tregs cells (figure 3N) and CD45⁻ cells (figure 3O), which suggests a potential synergy with PD-1/PD-L1 blockade since the reduction of TGF-β signaling in stromal cells can enhance the antitumor immunity of anti-PD-1 antibody treatment.31 We further depleted IFN-γ, NK1.1⁺ cells, and CD4⁺ and CD8⁺ T cells by antibodies post vvDD-IL-12-FG treatment (figure 3P; online supplementary figure S5) and found that the antitumor effect elicited by vvDD-IL-12-FG treatment was IFN-γ-dependent and CD8⁺ T-cell-dependent, but not CD4⁺ T-cell-dependent or NK1.1⁺ cell-dependent (figure 3Q). Collectively, these data demonstrated that vvDD-IL-12-FG treatment, as well as vvDD-IL-12 treatment, tipped the cancer–immune set point in tumor-bearing mice and turned “cold” tumors to “hot” tumors, which significantly extended the survival of mice receiving IL-12-armed virus treatment. Therefore, we hypothesized that the combination with immune checkpoint blockade would enhance the antitumor effects. We previously reported that the anti-PD-1 antibody either alone or combined with vvDD could not elicit an effective antitumor response in the late-stage tumor model.11 We tested whether the combination of vvDD-IL-12-FG and anti-PD-1 antibody could improve the therapeutic effects using the late-stage tumor model. The survival results showed that in combination with anti-PD-1 antibody, vvDD-IL-12-FG cured all the advanced tumor-bearing mice (figure 3R). Thus, the safe delivery of tethered IL-12 into the TME via oncolytic virus, combined with immune checkpoint blockade, makes it possible to revisit IL-12 as a cancer immunotherapy agent.32

Our data demonstrate that vvDD-IL-12-FG treatment can deliver IL-12 to the tumor bed and tether IL-12 on cell membranes, which is safe and effective in modifying the cancer–immune set point and producing an immune-favorable microenvironment that leads to improved efficacy as a monotherapy. Impressively, in profoundly immunosuppressive, advanced stage disease, vvDD-IL-12-FG synergizes with anti-PD-1 antibody therapy, leading to the cure of all late-stage MC38 tumors. Our data suggest that vvDD-IL-12-FG as a new form of IL-12 immunotherapy represents a treatment for cancers that have historically been unresponsive to immune checkpoint blockade-based immunotherapy.