Oncolytic virus-delivered GSDME: Unlocking pyroptosis to improve dendritic cell function and amplify cytotoxic T cell response

Abstract

Oncolytic virotherapy is a promising strategy for advanced cancers, but clinical success remains limited. Combining oncolytic viruses with immunogenic cell death modalities offers a potential solution. Pyroptosis, an inflammatory form of programmed cell death mediated by gasdermin proteins, amplifies anti-tumor immune responses through the release of damage-associated molecular patterns. Gasdermin E can switch apoptosis to pyroptosis upon caspase-3 cleavage, but its therapeutic potential is limited due to silenced expression in tumors. Here, we engineered the chimeric Vesicular stomatitis virus (VSV)-GP to express full-length GSDME, enabling pyroptosis induction in GSDME-deficient tumors. VSV-GP-GSDME efficiently delivered GSDME to tumor cells, leading to caspase-3-dependent pyroptotic cell death while retaining replication kinetics and achieving higher therapeutic potency. In vitro, VSV-GP-GSDME shifted cell death from apoptosis to pyroptosis, enhancing tumor cell immunogenicity. In vivo, it improved tumor control, prolonged survival, and promoted dendritic cell activation and tumor-infiltrating CD8 T cell expansion with cytotoxic phenotypes. Thus, VSV-GP-GSDME combines VSV-GP’s oncolytic activity with GSDME-mediated pyroptosis to induce anti-tumor immunity. This dual mechanism broadens applicability across diverse tumors, offering a promising strategy to enhance immunogenicity.

Graphical abstract

Nolden and colleagues demonstrate that the oncolytic virus vector VSV-GP, engineered to deliver GSDME, induces pyroptotic cell death and enhances anti-tumor immunity. This approach sensitizes tumors to immunogenic cell death, overcomes therapy resistance, and promotes dendritic cell activation and cytotoxic CD8⁺ T cell responses, highlighting its potential for advancing oncolytic virotherapy.

Introduction

Oncolytic viruses (OVs) have a long history of preclinical research and are promising treatment options for advanced cancers.^1,2 However, only few OV candidates have made the transition into the clinic and even to approval.³ Combining oncolytic virotherapy (OVT) with other modalities that trigger immunogenic cell death (ICD) has been highlighted as a promising strategy to improve cancer therapy effectiveness.⁴ The synergy may prove advantageous in bypassing tumor immune evasion and increasing therapeutic success.⁵ Pyroptosis, a lytic and inflammatory form of programmed cell death, mediated by gasdermin proteins,^6,7 has emerged as a potential pathway to enhance antitumor immunity.⁸ The gasdermin family comprises six members: Gasdermin A, B, C, D, E, and Pejvakin (PJVK).⁹ During pyroptosis, Gasdermin A-E are activated through proteolytic cleavage, which liberates their N-terminal domains (NTDs).¹⁰ These NTDs subsequently translocate to the plasma membrane, where they oligomerize to form pores.¹¹ The formation of gasdermin pores allows the release of intracellular contents, including damage-associated molecular patterns (DAMPs), which act as danger signals to the immune system.¹² Several DAMPs, such as adenosine triphosphate (ATP)¹³ and high-mobility group box 1 (HMGB1),^14,15 are recognized via specific DAMP receptors by antigen-presenting cells (APCs).¹⁶ Additionally, the release of DAMPs is accompanied by a progressive loss of membrane integrity, thereby enhancing the overall inflammatory response. Inducing gasdermin-mediated pyroptosis has the potential to remodel the immunosuppressive tumor microenvironment (TME) and to further improve the effectiveness of cancer immunotherapies.^17,18 Within the gasdermin family, Gasdermin E (GSDME) has attracted particular attention for its unique role in linking pyroptosis with antitumor immunity.^19,20 In cancer cells expressing GSDME at high levels, the cleavage of GSDME by caspase-3 (CASP3) has been reported to shift the cell death pathway from apoptosis to pyroptosis.^{21,22,23,24,25} This switch is associated with significant therapeutic potential, as several chemotherapeutic agents have been shown to induce GSDME-mediated pyroptosis, which may contribute to enhanced treatment efficacy and improved clinical outcomes.^26,27,28 In addition to chemotherapy, various OVs have demonstrated the ability to trigger GSDME-mediated pyroptosis, further pointing to their therapeutic potential.^29,30,31

However, alterations in gasdermin expression are frequently observed across different human cancer entities. Specifically, GSDME is often epigenetically silenced through hypermethylation of its encoding gene DFNA5,^32,33,34 and may also be inactivated by loss of function mutations.³⁵ These modifications underscore GSDME’s critical role in tumor biology and its potential contribution to immune evasion and tumor progression.³⁶ Overcoming the limitation posed by the absence of GSDME in tumor cells requires the development of strategies for delivering exogenous GSDME. One promising approach involves genetically modified OVs capable of delivering GSDME directly into tumor tissues, thereby restoring or enhancing pyroptotic cell death.^37,38 However, ectopic expression of GSDME by a replicating OV vector is often hampered by the cytotoxic nature of the GSDME protein and might impact viral replication and spread of virus progeny.

The chimeric variant of the Vesicular stomatitis virus (VSV), known as VSV-GP, has shown considerable promise for its oncolytic potential. VSV-GP incorporates the glycoprotein from the lymphocytic choriomeningitis virus (LCMV), enhancing its therapeutic potential by reducing neurotoxicity while maintaining its lytic capabilities against tumor cells.^39,40 Although VSV-GP has demonstrated a broad and potent oncolytic effect across various types of cancer, including melanoma,⁴¹ prostate cancer,⁴² and lung cancer, a study by Schreiber et al. highlighted VSV-GP’s capabilities to directly destroy cancer cells. Here, the lytic activity was a key factor in the therapeutic success in the lung cancer model, without significantly inducing long-term antitumoral immunity.⁴³

In the present study, we modified VSV-GP to express the GSDME protein. We demonstrated that the incorporation of full-length GSDME (FL-GSDME) into VSV-GP achieves (1) efficient delivery of GSDME to tumor cells lacking endogenous GSDME and enables (2) concurrent cleavage of GSDME through VSV-GP mediated CASP3 activation. Importantly, we showed that GSDME does not compromise replication of VSV-GP-GSDME. Furthermore, we could observe that GSDME dominates the cell death phenotype of an infected cancer cell, shifting the virus-induced apoptosis to pyroptosis. In vivo, treatment with VSV-GP-GSDME inhibited tumor growth and prolonged survival in an EMT6 breast cancer mouse model. In the B16-F10 melanoma model, which is known to be resistant to immune checkpoint blockade, VSV-GP-GSDME enhanced the efficacy of anti-PD-1 therapy. We could show that VSV-GP-GSDME treatment led to increased activation and migration of conventional and plasmacytoid dendritic cells (DCs) in tumor-draining lymph nodes (tdLNs). This is accompanied by a cytotoxic CD8⁺ T cell response characterized by elevated expression of effector molecules and activation markers. Using an ex vivo approach with patient-derived colorectal cancer slice cultures, we confirmed GSDME expression and cleavage.

Results

VSV-GP triggers pyroptotic cell death via GSDME activation

Previously, it has been reported that VSV infection activates multiple apoptotic pathways, including the intrinsic mitochondrial pathway, leading to CASP3 activation.⁴⁴ In cells expressing GSDME, VSV infection has been demonstrated to trigger GSDME-mediated pyroptosis,³⁰ a lytic form of programmed cell death.⁷ VSV-GP, a genetically engineered variant of VSV in which the glycoprotein G of VSV is replaced with the glycoprotein of the LCMV, has also been shown to activate CASP3 in patient-derived head and neck squamous cell carcinoma (HNSCC) explants.⁴⁵ Based on these observations, VSV-GP infection may also induce GSDME cleavage and activation, potentially influencing the phenotype of cell death.

We generated two EMT6 murine breast cancer cell variants: GSDME knockout (Gsdme^−/−) and GSDME-overexpressing (GSDME OE) cells, to confirm GSDME cleavage by VSV-GP. Successful knockout or overexpression of GSDME was confirmed via immunoblot analysis (Figure S1C). Both cell lines were infected with VSV-GP at a multiplicity of infection (MOI) of 3 (Figures 1A–1C) or treated with 10 μM raptinal to rapidly activate CASP3 (Figures 1D–1F). Following infection, EMT6 cells overexpressing GSDME exhibited a pronounced pyroptotic phenotype characterized by cell swelling and distension of the plasma membrane, which became evident 24 h after infection (Figure 1A). This phenotype was accompanied by a significant increase in propidium iodide (PI) uptake over time, indicating plasma membrane disintegration and leakage (Figure 1B). In contrast, Gsdme^−/− cells displayed a clear apoptotic phenotype, characterized by membrane blebbing, apoptotic body formation, and a less pronounced PI uptake (Figure 1D). Immunoblot analysis further confirmed GSDME cleavage after infection, as evidenced by the detection of the N-terminal GSDME pore forming domain (GSDME-NT) at the expected molecular weight of approximately 35 kDa (Figure 1C).

Figure 1.

Induction of GSDME cleavage by VSV-GP leads to a pyroptotic phenotype in GSDME overexpressing EMT6 mouse cancer cells

Gsdme knockout (Gsdme^−/−) and GSDME overexpressing (GSDME OE) EMT6 cells were infected with (A–C) VSV-GP (+) at an MOI of 3 for 24 h or left uninfected (−) for control or (D–F) treated with 10 μM raptinal (+) or left untreated (−) for 75 min. (A and D) Phase-contrast images of treated EMT6 cells cultured in the presence of propidium iodide (PI) at 10× magnification; scale bars, 25 μm. (B and E) PI uptake was quantified by red fluorescence signals and is represented as mean red intensity/confluence. Data shown as mean ± SD, n = 3. (C and F) Cell extracts were analyzed 75 min after raptinal treatment and 24 h post-infection with VSV-GP by immunoblotting. GSDME and cleaved GSDME (GSDME NT) were detected (top), viral proteins (VSV-N, -P, and -M) were probed using a polyclonal anti-VSV rabbit serum (middle), and β-actin (ACTB) served as a loading control (bottom).

To validate the characteristics of the pyroptotic phenotype induced by VSV-GP-dependent GSDME pathway activation, we treated both cell lines with raptinal as reference treatment. Notably, raptinal induces GSDME-mediated pyroptosis³⁵ independently of inflammasome and BAX/BAK activation.^46,47 Following raptinal treatment, GSDME OE cells exhibited a pyroptotic phenotype, including cell swelling with subsequent membrane distension, disintegration, and increased PI uptake similar to that seen after VSV-GP infection (Figures 1D and 1E). Immunoblot analysis confirmed elevated levels of cleaved GSDME-NT upon raptinal treatment. In contrast, Gsdme^−/− cells showed no evidence of GSDME cleavage, as confirmed by the absence of the GSDME and GSDME-NT signals (Figure 1F). To validate that GSDME cleavage strictly depends on CASP3 activation in the context of virus infections, we generated CASP3 knockout EMT6 cells, which exhibit low endogenous GSDME expression (Figure S1C). The lack of GSDME cleavage in Casp3^−/− cells after raptinal treatment was confirmed by immunoblot analysis (Figure S3B). Neither raptinal treatment nor VSV-GP infection induced a pyroptotic phenotype in GSDME or CASP3-deficient cells (Figures S1A and S1B).

In summary, these results demonstrate that VSV-GP induces GSDME cleavage, thereby triggering GSDME-dependent pyroptosis. A comparable pyroptotic phenotype was observed in GSDME-expressing cells following either VSV-GP infection or chemical induction by raptinal. These findings highlight the critical role of GSDME in mediating the transition from apoptosis to pyroptosis upon CASP3 activation in the context of VSV-GP.

Ectopic expression of GSDME by VSV-GP leads to a pyroptotic phenotype in GSDME negative cancer cells

Many tumors exhibit low or absent GSDME expression,³⁵ which limits their susceptibility to pyroptotic cell death. Thus, we propose using the VSV-GP vector to deliver GSDME and enhance tumor sensitivity to pyroptotic cell death. We inserted the full-length, functional murine GSDME protein into the VSV-GP genome between the LCMV-GP and the VSV polymerase L. We recovered the chimeric virus VSV-GP-GSDME, which expresses full-length GSDME in its autoinhibited form, using established protocols. Additionally, we generated VSV-GP chimeric viruses encoding two variants: a loss of function GSDME mutant (GSDME_F2A) and a non-cleavable GSDME variant (GSDME_D270A). These controls were designed to analyze CASP3 specificity of GSDME cleavage and pore-forming activity in the context of VSV-GP infection (Figure 2A). It is important to note that the active form of GSDME (GSDME-NT) could not be directly expressed by the virus, as attempts to recover a viable virus encoding only this active domain were unsuccessful (data not shown).

Figure 2.

Expression and cleavage of GSDME by the oncolytic VSV-GP vector is timely separated and results in increased PI uptake and reduced cell viability

(A) Schematic representation of GSDME expressing VSV-GP vectors. Mouse GSDME variants, including full-length GSDME, loss of function GSDME_F2A, or non-cleavable GSDME_D270A, were integrated into the VSV-GP genome at position 5 between the glycoprotein GP and VSV polymerase L. (B) Cryo-electron microscopy (CryoEM) analysis of viral particles. Representative micrographs of intact VSV-GP-GSDME particles at indicated magnifications. (C) Individual particle lengths of VSV-GP and VSV-GP-GSDME were measured from the CryoEM images. Each data point represents a single virion, the horizontal line indicates the mean. (D–H) HEK293-F suspension cells were infected in duplicates at an MOI of 0.0005 (D, G) or MOI of 3 (E, F, and H) with the respective VSV-GP-GSDME variants, VSV-GP, or left untreated (mock). At indicated time points, TCID₅₀ and genomic titers were measured in supernatants. Replication kinetics were interpolated using non-linear curve fitting, data points were represented as mean (SD). (G and H) Cell viability was determined after infection at indicated time points. Statistical significance was analyzed by two-way analysis of variance (ANOVA) followed by Tukey’s multiple comparisons test, ∗∗∗∗p < 0.0001 (F and I) Immunoblot analysis of GSDME expression and cleavage at indicated time points in HEK293F (F) and EMT6 Gsdme^−/− (I) cells infected with VSV-GP, VSV-GP-GSDME, or GSDME_D270A at an MOI of 3. (J) Effect of Casp3 knockout on GSDME cleavage in EMT6 Gsdme^−/− cells infected with VSV-GP-GSDME (MOI = 3) at 16 h post-infection. (K) Time-lapse images of EMT6 Gsdme^−/− cells in PI-containing media, infected with either VSV-GP or VSV-GP-GSDME at MOI = 3. The images display phase contrast (phase, top), red fluorescence (PI, middle), and a merged view (bottom), all captured at 10× magnification. Scale bars, 100 μm.

Cryo-electron microscopy confirmed that VSV-GP-GSDME retained the bullet-shaped morphology of VSV (Figure 2B). As insertion of the GSDME open reading frame increases the genome length of parental VSV-GP by 1,569 nucleotides, the VSV-GP-GSDME particles are approximately 19 nm longer than VSV-GP particles (196 ± 7 nm vs. 177 ± 7 nm, Figure 2C). To assess whether the additional genetic information affected viral replication, HEK293-F cells were infected with respective viral variants at an MOI = 0.0005, and viral titers were determined at indicated time points. The GSDME-armed VSV-GP variants replicated comparably to the parental virus, reaching peak titers around 40 h post-infection and plateauing at 48 h (Figure 2D). Notably, VSV-GP-GSDME infection resulted in a decreased cell viability at 48 h compared to VSV-GP, or the loss of function variant VSV-GP-GSDME_F2A (Figure 2G).

Because the timing of detectable GSDME-mediated cytotoxicity depends on infection parameters, we additionally assessed replication and cell killing kinetics under single-step conditions. HEK293F suspension cells were infected at MOI = 3, which resulted in a rapid increase in infectious titers during the first 12 h post-infection (Figure 2E). Under these conditions, longitudinal viability measurements showed that VSV-GP-GSDME began to reduce cell viability compared with VSV-GP between 16 and 20 h post-infection, followed by a progressive time-dependent decline thereafter (Figure S2A), while both viruses reached similar overall reductions in cell viability by 28 h when compared to infections at low MOI (Figures 2G and 2H). In parallel, immunoblot analysis confirmed GSDME expression as early as 8 h post-infection (Figure S2B), whereas the pore-forming cleaved GSDME-NT fragment became detectable at later time points—faintly at 12 h but distinctly at 28 h (Figures 2F and S2B). Together, these data indicate that early GSDME expression precedes detectable cleavage and loss of viability.

To further analyze the effects of virus-encoded GSDME independent of limited virus replication or endogenous GSDME in tumor cells, we generated EMT6 breast cancer cells with a knockout in the type 1 IFN receptor as well as GSDME by CRISPR/Cas9. These EMT6 Ifnar1^−/− Gsdme ^−/− double knockout cells do not express endogenous GSDME but were highly susceptible to VSV-GP infection and oncolysis. To confirm GSDME expression and evaluate cleavage kinetics, EMT6 Ifnar1^−/− Gsdme^−/− cells were infected with VSV-GP-GSDME variants at an MOI of 3. Immunoblot analysis detected full-length GSDME as early as 8 h post-infection, confirming successful transgene expression. Cleaved GSDME signals at 35 kDa were detected by 16 h. No cleavage was observed in cells infected with the VSV-GP-GSDME_D270A variant (Figure 2I). These time-dependent cleavage dynamics were reproducible in 4T1 cells, which express minimal endogenous levels of GSDME (Figure S3C). In the next step, we assessed whether GSDME expression and cleavage were associated with morphological changes using live-cell imaging. EMT6 Ifnar1^−/− Gsdme^−/− were infected with VSV-GP or VSV-GP-GSDME and imaged up to 48 h post-infection. VSV-GP-GSDME infection led to a time-dependent increase in PI uptake, indicative of compromised membrane integrity and cell death. This effect was not observed with VSV-GP alone (Figure 2K). To exclude the possibility that factors other than CASP3 are responsible for GSDME cleavage in the context of virus infection, EMT6 Gsdme^−/− cells additionally lacking CASP3 (Casp3^−/−) were analyzed. Immunoblot analysis demonstrated absence of the N-terminal GSDME domain in Casp3^−/− cells. In contrast, cells positive for CASP3 exhibited a clear GSDME cleavage pattern upon infection (Figure 2J), indicating that CASP3 is essential for GSDME cleavage under these conditions.

In sum, we provide evidence that VSV-GP encoding full-length murine GSDME enhances sensitivity to pyroptotic cell death. VSV-GP-GSDME retained the characteristic morphology and replication kinetics of the parental virus while decreasing cell viability compared to controls lacking functional GSDME. We demonstrate that GSDME cleavage in virus context depends on CASP3 activity and leads to membrane permeabilization, as indicated by increased PI uptake and compromised cell viability. These findings highlight the potential of VSV-GP-GSDME as a therapeutic approach to sensitize tumors, which do not express functional GSDME, to CASP3-mediated pyroptotic cell death.

VSV-GP-GSDME suppresses apoptosis and promotes predominantly pyroptotic cell death characteristics in infected cells

To dissect the apoptotic cascade, we assessed the activation of CASP3, its upstream initiator caspase-9 (CASP9), and the downstream substrate poly(ADP-ribose) polymerase-1 (PARP1) after VSV-GP-GSDME treatment. Gsdme^−/− cells were infected with either VSV-GP, VSV-GP-GSDME, or VSV-GP-GSDME_F2A at an MOI of 3 or left uninfected (mock). Cell lysates were analyzed at 8 and 16 h post-infection by immunoblot (Figure 3A). Strikingly, cells infected with VSV-GP-GSDME exhibited strongly reduced levels of cleaved CASP3 compared to VSV-GP and VSV-GP-GSDME_F2A. Importantly, a cleaved CASP3 signal remained detectable only upon enhanced contrast adjustment (Figure 3B). Despite markedly reduced active, cleaved CASP3, GSDME underwent robust proteolytic activation. This indicates that minimal CASP3 activity is sufficient to trigger GSDME cleavage and suggests a preferential allocation of CASP3 catalytic activity toward GSDME over canonical apoptotic substrates.

Figure 3.

VSV-GP-GSDME infection reduces the activation of the intrinsic apoptotic pathway both upstream and downstream of caspase-3

(A) Immunoblot analysis of EMT6 Ifnar1^−/−Gsdme^−/− murine breast cancer cells infected with indicated viruses (MOI = 3) at 8 and 16 h post-infection. The upper image shows GSDME and cleaved GSDME-NT, while the middle image displays cleaved and non-cleaved CASP9, CASP3, and PARP1. For control, the lower image include β-actin (ACTB) and viral proteins VSV-N and -M/P. (B) CASP3 immunoblot as shown in Figure 3A at higher contrast. The faint cleaved CASP3 signal observed after VSV-GP-GSDME infection is marked by a black arrowhead. (C) Representative images of EMT6 Ifnar1^−/−Gsdme^−/− cells infected with VSV-GP or VSV-GP-GSDME, stained with Cytotox-Green and AnnexinV-Red dyes. White arrows indicate nuclear fragmentation. (D) ATP release measured at 8, 12, 16, and 20 h post-infection in EMT6 Ifnar1^−/−Gsdme^−/− cells infected with the indicated virus. (E) LDH release was determined in collected cell supernatants, supernatant from untreated cells (mock) served as negative control. Data are presented as mean (SD) from three biological replicates. Statistical significance was analyzed by two-way ANOVA followed by Tukey’s comparison test, ∗p < 0.05, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

In line with this, cleavage of CASP9 and PARP1 was also strongly reduced in VSV-GP-GSDME-infected cells, indicating attenuation of the intrinsic apoptotic pathway. In contrast, robust cleavage of CASP3, CASP9, and PARP1 was observed in cells infected with VSV-GP or VSV-GP-GSDME_F2A. These findings were further supported by experiments with EMT6 GSDME OE cells treated with either VSV-GP or raptinal (Figures S3A and S3B).

To identify additional differences in cell death phenotype, Gsdme^−/− cells were infected with VSV-GP or VSV-GP-GSDME and analyzed via live-cell imaging with Cytotox and Annexin V stainings. Around 24 h post-infection, VSV-GP-GSDME-infected cells exhibited distinct pyroptotic features, including noticeable membrane swelling. Cytotox staining revealed pronounced nuclear fragmentation in VSV-GP-infected cells, whereas VSV-GP-GSDME-infected cells retained more intact nuclei. Conversely, Annexin V staining was more prominent in VSV-GP-infected cells, which aligns with early apoptosis (Figure 3C). Additionally, lactate dehydrogenase (LDH) release, indicating membrane rupture, was assessed. LDH was detected exclusively in VSV-GP-GSDME-infected cells, whereas VSV-GP or VSV-GP-GSDME_F2A showed no significant LDH release, consistent with more intact membranes of apoptotic cells similar to the uninfected control (Figure 3E).

Another key difference between apoptosis and pyroptosis is the release of DAMPs, such as extracellular ATP (eATP). While apoptosis generally results in minimal ATP leakage,⁴⁸ pyroptosis is associated with substantial ATP release due to the formation of gasdermin pores.⁴⁹ Infection of Gsdme^−/− cells with VSV-GP-GSDME led to a significant increase in eATP levels. By 12 h post-infection, eATP levels were already significantly elevated compared to mock and VSV-GP-infected controls. This increase became even more pronounced at 16 and 20 h post-infection, although a delayed rise in eATP was also observed in VSV-GP-infected cells at later time points (Figure 3D). Similarly, GSDME OE cells treated with raptinal exhibited both ATP and LDH release at early time points (Figures S3D and S3E).

Collectively, these results suggest that VSV-GP-GSDME infection suppresses intrinsic apoptotic events, as evidenced by reduced CASP3 and CASP9 cleavage, while promoting a shift toward GSDME-mediated pyroptosis, as indicated by the absence of nuclear fragmentation and elevated eATP and LDH release.

Human and mouse GSDME delivered by VSV-GP are functionally related and lead to pyroptotic cell death in human and mouse cancer cell lines

Structure-based sequence alignment demonstrates the high conservation of human and mouse GSDME, particularly in the pore-forming N-terminal (GSDME-NT) domain, which exhibits a higher degree of similarity compared to other regions (Figure S4). To enable translational studies, we engineered VSV-GP to express human Gasdermin E (huGSDME) (Figure S5A), leveraging the high similarity between human and murine GSDME to support the feasibility of cross-species functional studies in vivo. Replication kinetics remained unaffected, as evidenced by comparable TCID₅₀, genomic titers between 8 and 48 h (Figure S5B) and reductions in cell viability at 48 h post-infection between VSV-GP-huGSDME and VSV-GP (Figure S5C). To analyze GSDME expression and cleavage, mouse 4T1 and human A549 cells were infected with VSV-GP-huGSDME at an MOI of 3. Immunoblot analysis was performed at 16 h post-infection. The results showed both expression and cleavage of huGSDME (Figures S5D and S6C). Treatment with the pan-CASP inhibitor Emricasan markedly reduced GSDME cleavage, whereas robust cleavage was observed in untreated cells at 16 h post-infection implicating CASP3 activation as a key mediator of GSDME processing independently of the species. Phenotypically, cells infected with VSV-GP-huGSDME exhibited hallmark features of pyroptotic cell death, including cell membrane swelling and permeabilization (Figure S6A). This was confirmed by Annexin V and Cytotox staining, which revealed double-positive signals in VSV-GP-huGSDME-infected cells, indicating simultaneous exposure of phosphatidylserine and loss of membrane integrity. In contrast, cells infected with VSV-GP alone displayed Annexin V⁺/Cytotox⁻ staining at early time points (24 h), consistent with the induction of early apoptosis (Figures S6A and S6B). In addition, a marked release of ATP from 4T1 cells was observed in a dose-dependent manner (Figures S6D and S6E).

These findings show that VSV-GP-mediated delivery of huGSDME induces GSDME-dependent pyroptosis in both human and mouse cancer cell lines, demonstrating its functional equivalence across species. The high conservation between human and murine GSDME allows huGSDME to be processed and activated in mouse cells, highlighting its cross-species applicability. Furthermore, the compatibility of huGSDME expression with VSV-GP replication highlights its potential for translational studies, including in vivo applications in mouse models.

VSV-GP-GSDME enhances tumor control, prolongs survival, and overcomes therapy resistance in EMT6, TC-1, and B16-F10 murine tumor models

Building on the in vitro findings that GSDME shifts VSV-GP-induced apoptotic cell death toward pyroptosis in the context of VSV-GP, we next assessed its therapeutic potential in vivo. Using the EMT6 Ifnar1^−/− breast cancer model, mice received two intratumoral (i.t.) injections of either VSV-GP or VSV-GP-huGSDME (1 × 10⁸ TCID₅₀), while control animals received vehicle treatment. Tumors in the vehicle group progressed rapidly, with two spontaneous remissions observed (Figure 4A). VSV-GP treatment delayed tumor growth. However, it did not result in a higher number of complete remissions compared to vehicle. Notably, VSV-GP-huGSDME treatment led to a marked reduction in tumor growth, with six out of 10 animals achieving complete remission, translating into prolonged survival up to 56 days compared to both control and VSV-GP-treated groups (Figure 4B).

Figure 4.

VSV-GP-GSDME treatment inhibits tumor growth across multiple mouse tumor models

Tumor-bearing mice (n = 10) were treated intratumorally (i.t.) with VSV-GP-huGSDME, VSV-GP, or vehicle control. (A and B) EMT6 Ifnar1^−/− cells were implanted subcutaneously into BALB/c mice, and tumors were treated on days 7 and 10 post-implantation (dotted line) with 1 × 10⁸ TCID₅₀. (A) Individual tumor growth curves and (B) Kaplan-Meier survival analyses are shown, with complete remissions (cr) noted. (C) TC-1 Ifnar1^−/− cells were implanted into the right flank of C57BL/6 mice (n = 8), and tumors were treated on day 8 post-implantation with 1 × 10² TCID₅₀. Tumor growth curves for individual mice are shown. (D) B16-F10 melanoma cells were implanted into the mammary fat pads of C57BL/6 mice (n = 10). Tumors were treated on days 10 and 13 post-implantation with 1 × 10⁸ TCID₅₀, followed by intraperitoneal (i.p.) injections of αPD-1 monoclonal antibody or isotype control starting on day 14 post-implantation. Tumor volumes on day 21 post-implantation are shown as mean (SEM). Statistical significance was determined by ordinary one-way ANOVA,∗p < 0.05. (E and F) GSDME expression in lysates from TC-1 Ifnar1^−/− tumors harvested at day 1 (d1) and day 3 (d3) after intratumoral treatment (1 × 10² TCID₅₀). Representative western blot from one tumor at each time point (E). Quantification of full-length huGSDME and cleaved GSDME-NT (F, left) and VSV-NP (F, right) in tumor lysates (n = 5 per time point), normalized to vinculin (VCN) and shown as mean (SD). (G) Body weight of EMT6 Ifnar1^−/−, TC1 Ifnar1^−/−, and B16-F10 tumor bearing mice was recorded at the indicated time points and normalized to individual baseline values at day of implantation (d0). Data are shown as percent of baseline body weight (mean ± SEM).

To further investigate the therapeutic potential of VSV-GP-GSDME in the context of a highly permissive tumor model, where oncolysis is the primary driver of antitumor activity, we utilized the TC-1 Ifnar1^−/− model, which supports efficient viral replication and activity. Tumor-bearing mice received a single i.t. injection of VSV-GP or VSV-GP-huGSDME at doses of 10², 10⁴, or 10⁶ TCID₅₀ (Figures S7A–S7C). While VSV-GP delayed tumor growth, no complete remissions were observed at a dose level of 10² TCID₅₀. In contrast, VSV-GP-huGSDME at this low dose-induced tumor remission in four animals by day 28 (Figure 4C), although relapse occurred around day 35, with only one animal remaining tumor-free on day 49 (Figure S7B). At higher doses, both VSV-GP and VSV-GP-huGSDME improved survival in a dose-dependent manner. However, the survival advantage of VSV-GP-huGSDME over VSV-GP observed at the lowest dose was no longer apparent at higher doses, where survival outcomes converged between groups. This pattern is consistent with the notion that in highly permissive Ifnar1^−/− tumors a strong direct oncolytic effect may dominate and mask additional contributions from GSDME-mediated ICD (Figure S7C).

Representative western blot of TC-1 Ifnar1−/− tumor lysates harvested at day 1 (d1) and day 3 (d3) after i.t. treatment with low-dose VSV-GP or VSV-GP-huGSDME (1 × 10² TCID₅₀) shows VSV-NP at d1 and huGSDME expression with appearance of the cleaved GSDME-NT fragment at d3 (Figure 4E). Quantification of full-length huGSDME, cleaved GSDME-NT, and VSV-NP signals was performed for VSV-GP-huGSDME-treated tumors only (n = 5 per time point; normalized to vinculin), providing direct evidence that virally delivered huGSDME can be processed in tumor tissue in vivo (Figure 4F).

Finally, to evaluate whether VSV-GP-huGSDME-induced pyroptosis could overcome resistance to immune checkpoint inhibition, we employed the B16-F10 melanoma model, which typically is refractory to anti-PD-1 therapy under the used conditions. Mice received i.t. injections of VSV-GP-huGSDME, VSV-GP, or vehicle, in combination with intraperitoneal (i.p.) administration of αPD-1 monoclonal antibody (10 mg/kg) or isotype control. In the B16-F10 melanoma model, only VSV-GP-huGSDME combined with αPD-1 reduced tumor volume by day 21, whereas VSV-GP + αPD-1 did not improve tumor volume compared with VSV-GP alone. Overall, virus-treated groups showed delayed tumor growth versus vehicle controls, but improved tumor control was observed only with the VSV-GP-huGSDME + αPD-1 combination (Figure 4D).

Across all tumor models, body weight was largely stable, with only a transient decrease during the first 2 days after treatment and no significant or sustained losses in any virus-treated group. This dip was evident in the EMT6-and TC-1 Ifnar1^−/− models; in the B16-F10 model, body weight remained overall stable but did not show appreciable gain between days 5 and 14 across groups, indicating that this pattern was not related to virus-treatment (Figure 4G). Collectively, these data support the absence of overt VSV-GP-GSDME-associated toxicity compared with VSV-GP under the conditions tested.

In conclusion, these results demonstrate the therapeutic benefit of VSV-GP-huGSDME across diverse tumor models. In EMT6 Ifnar1^−/−, it reduced tumor burden, extended survival, and achieved higher remission rates than VSV-GP. In TC-1 Ifnar1^−/−, VSV-GP-huGSDME induced tumor regression at low-dose, with comparable survival benefits to VSV-GP at higher doses. And finally, in the PD-1 refractory B16-F10 model, VSV-GP-huGSDME synergized with αPD-1, effectively overcoming immune checkpoint inhibitor (ICI) resistance.

VSV-GP-huGSDME enhances DC activation and cytotoxic CD8⁺ T cell responses in mouse tumor models

Mobilizing DCs has emerged as a promising strategy to enhance tumor antigen cross-presentation and improve the efficacy of OVT.⁵⁰ ICD has been demonstrated to increase both the adjuvanticity and antigenicity of dying tumor cells, thereby facilitating DC recruitment, maturation, and activation.⁵¹

To evaluate the impact of VSV-GP-huGSDME-induced pro-inflammatory cell death on DC activation and function, we first utilized the TC-1 Ifnar1^−/− tumor model. Mice were treated intratumorally with either VSV-GP or VSV-GP-huGSDME at low dose (1 × 10² TCID₅₀) or mock treated (vehicle). Analysis of tdLNs revealed that VSV-GP-huGSDME treatment significantly elevated the frequency of CCR7⁺CD86⁺ double-positive conventional dendritic cells (cDCs), reflecting enhanced migratory and antigen-presenting capacities (Figures 5A and 5B). In addition, plasmacytoid dendritic cells (pDCs) exhibited increased CCR7 expression following VSV-GP-huGSDME treatment, suggesting activation across multiple DC subsets (Figures 5C and 5D). Chemokine and cytokine levels were assessed by a multiplex immunoassay in tumor lysates at 24 and 72 h post-treatment and were largely comparable between VSV-GP and VSV-GP-huGSDME, with IFN-β being the only analyte showing a reduction in the VSV-GP-huGSDME group (Figures 4E and S9).

Figure 5.

Low-dose VSV-GP-huGSDME enhances dendritic cell activation in tumor-draining lymph nodes and promotes expansion of activated, cytotoxic tumor-infiltrating CD8 T cells in TC-1 Ifnar1^−/− tumor-bearing mice

TC-1 Ifnar1^−/− cells were implanted subcutaneously into the right flank of C57BL/6 mice. Mice were treated intratumorally (i.t.) with either vehicle (mock), VSV-GP, or VSV-GP-huGSDME with 1 × 10² TCID₅₀. tdLNs were harvested 3 days after the viral treatment and analyzed by flow cytometry (n = 6–7 per group). (A and C) Representative contour plots showing expression of CCR7 and CD86 on cDCs and pDCs in tdLNs from mice treated with VSV-GP (left) or VSV-GP-huGSDME (right). (B and D) Quantification of CCR7⁺, CD86⁺, and double-positive cells among pDCs and cDCs. (E) Analysis of type 1 interferons and IL-6, IL-10, and IL-18 was performed on tumor tissue lysates from five mice per group at 24 and 72 h post-treatment by a multiplex immunoassay. Data are shown as mean (SD) after normalization to total extracted protein. Statistical significance was assessed using ordinary one-way ANOVA, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. For TIL analysis, tumors were harvested 7 days after the viral treatment (n = 6–7 per group). (F and H) Representative contour plots showing expression of granzyme B (GZMB) and perforin (PRF1) and CD25 and the activation-induced glycoform of CD43 (CD43act) in tumors from mice treated with VSV-GP (left) or VSV-GP-huGSDME (right). (G and I) Quantification of granzyme B⁺ and/or perforin⁺ and CD25⁺ and/or the CD43act⁺ of tumor-infiltrating CD8 T cells. Data are presented as mean (SD) from 6 to 7 biological replicates. Statistical significance was analyzed by ordinary one-way or two-way ANOVA, as appropriate, followed by Tukey’s multiple comparison test, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

To further assess the downstream effects on adaptive immunity, tumors were harvested seven days post-treatment for analysis of tumor-infiltrating lymphocytes (TILs). Flow cytometric analysis revealed a significant increase in the frequency of CD8⁺ T cells expressing cytotoxic effector molecules granzyme B and/or perforin in the VSV-GP-huGSDME-treated group compared to vehicle controls (Figures 5F and 5G). Additionally, these CD8⁺ T cells exhibited elevated expression of activation markers CD25 and the activation-induced glycoform of CD43, indicative of functional activation within the TME (Figures 5H and 5I).

These findings were corroborated in the B16-F10 melanoma model, where VSV-GP-huGSDME treatment similarly enhanced the frequency of CCR7⁺CD86⁺ cDCs and CCR7⁺ pDCs in tdLNs (Figures S10A–S10D). Moreover, CD8⁺ T cells in this model exhibited increased expression of granzyme B and PD-1, further supporting the role of VSV-GP-huGSDME in promoting DC activation and functional maturation of cytotoxic CD8⁺ T cells in the TME (Figures S10E and S10F).

These results demonstrate that VSV-GP-huGSDME treatment promotes DC activation and migration into the tdLNs and functional maturation of cytotoxic CD8⁺ T cells within the TME, highlighting its potential to enhance anti-tumor immunity in preclinical mouse models.

Distinct granulocytic and T cell immune profiles define early and late responders to VSV-GP-huGSDME

To investigate how VSV-GP-GSDME influences the i.t. immune landscape during the onset of tumor regression, we analyzed tumors from early and late responders in the TC-1 Ifnar1^−/− model. Mice were treated intratumorally with VSV-GP-GSDME at low dose (1 × 10² TCID₅₀), and tumors were harvested at a fixed experimental endpoint. Notably, a slight, transient post-treatment increase in tumor size in VSV-GP-GSDME-treated mice is consistent with immune-related pseudo-progression (Figure 6A). Based on the timing of tumor regression, animals were classified as early responders (regression within 2–3 days post treatment) or late responders (regression around day 7 post treatment) (Figure 6B). Flow cytometric profiling revealed marked differences in the composition of tumor-infiltrating leukocytes between the two groups. Early responders displayed higher frequencies of CD8⁺ and CD4⁺ T cells, whereas late responders exhibited increased frequencies of granulocytic myeloid cells (CD11b⁺/Ly-6G⁺). In contrast, monocytic myeloid cells, B cells, unclassified CD11b⁺ cells, and other CD45⁺ immune subsets (including regulatory T cells and γδ T cells) were present at comparable levels in both the groups (Figure 6B).

Figure 6.

VSV-GP-GSDME treatment reveals distinct granulocytic and T cell activation states in early and late responders in the TC-1 Ifnar1^−/− tumor model

TC-1 Ifnar1^−/− cells were implanted subcutaneously into the right flank of C57BL/6 mice. Mice were treated intratumorally with VSV-GP-GSDME at 1 × 10² PFU or mock treated (vehicle). Tumor growth was monitored over time. Tumors were harvested at day 18 post implantation and analyzed by flow cytometry (n = 5–9 per group). (A) Tumor growth curves comparing mock-treated and VSV-GP-GSDME-treated mice (left), and tumor growth curves of VSV-GP-GSDME-treated mice stratified into early responders (regression within 2–3 days after treatment) and late responders (regression around day 7) (right). (B) A 10 × 10 grid plot was used to visualize frequencies of major tumor-infiltrating leukocyte populations (% of CD45⁺ cells) in early and late responders. (C) Quantification of MHCII⁺, CD86⁺, and Arginase-1⁺ granulocytic myeloid cells per gram tumor in early and late responders. (D) Quantification of activated CD8⁺ T cells (CD25⁺ and/or CD43act⁺) per gram tumor in early and late responders. (E) Quantification of granzyme-B–positive CD4⁺ T cells per gram tumor in early and late responders. Statistical significance was analyzed by unpaired two-tailed Mann-Whitney U tests, ∗p < 0.05, ∗∗p < 0.01.

Further phenotypic analysis of granulocytic myeloid cells showed that late responders had higher numbers of MHCII- and CD86-expressing cells, indicating a more pronounced pro-inflammatory activation state at the time of tumor harvest. Late responders also exhibited increased numbers of arginase-1-expressing granulocytic cells, whereas early responders showed markedly reduced arginase-1 expression within this compartment (Figure 6C). In contrast, early responders displayed higher numbers of activated CD8⁺ T cells, defined by expression of CD25 and/or the activation-induced glycoform of CD43 (Figure 6D), as well as increased numbers of granzyme-B-positive CD4⁺ T cells (Figure 6E).

Together, these data demonstrate that early and late responders differ in both the abundance and activation state of granulocytic myeloid cells and in the magnitude of i.t. T cell activation at the time of analysis.

VSV-GP-GSDME effectively targets patient-derived colorectal cancer slice cultures, demonstrating tumor-specific permissivity and GSDME cleavage

To further explore the therapeutic applicability of VSV-GP-GSDME, we turned to patient-derived colorectal cancer slice cultures. While these ex vivo human systems do not fully recapitulate the immune dynamics observed in mice, they provide a valuable platform to investigate the fundamental tumor-intrinsic mechanisms of virus permissivity, ectopic GSDME expression, and cleavage in a human tumor ecosystem.

For quality control, baseline colorectal cancer (CRC) slices were characterized using hematoxylin and eosin (H&E) staining, revealing that TME cells were the most abundant (>50%), while tumor cells varied significantly (7.87%–54.85%). Necrotic areas were minimal (<5%), except for CRC-044, which showed higher proportions of dead cells. (supplemental information and Table S7).

Tissue slices were treated with VSV-GP-Katushka or VSV-GP-huGSDME-Katushka at 1 × 10⁷ TCID₅₀/slice, and infection was assessed by the Katushka protein signal over 72 h. A semi-quantitative scoring system was used ranging from negative (no signal) to +++++ (entire slice encompasses positive signal) to evaluate the permissivity of CRC samples. Permissivity varied across biospecimens, with CRC-043 showing strongest Katushka signals (Figure 7A and Table S10) and CRC-044 displaying more variability, which might be attributed to a higher number of necrotic cells at baseline (supplemental information; Tables S7 and S11). Due to the high tumor heterogeneity observed across CRC slices, the overall means are moderate and do not fully reflect the variability in Katushka fluorescence between individual slices (Figure 7B, supplemental information and Tables S8–S11). Albeit the mentioned intra- and inter-patient heterogeneity, VSV-GP-Katushka, and VSV-GP-huGSDME-Katushka displayed comparable virus spread and propagation from 24 to 72 h indicating similar permissivity in human CRCs ex vivo. Of note, CRC-042 remained non-permissive for both indicated viruses (Table S9). Tumor slices of highly permissive tumor specimens were harvested at 72 h post-infection and subsequent immunoblot analysis confirmed GSDME expression and cleavage in VSV-GP-huGSDME-Katushka-infected slices, correlating with fluorescence signals observed (Figure 7C). Elevated GSDME expression and cleavage were detected in two of three slices derived from CRC-043, while one slice showed levels comparable to endogenous GSDME, potentially due to variability in tumor tissue composition and infection. These findings were further supported by quantification of GSDME and GSDME-NT levels normalized to VSV-N protein (Figure 7D).

Figure 7.

VSV-GP-GSDME infects patient-derived colorectal tumor specimens in an ex vivo slice culture system, resulting in elevated GSDME levels and subsequent GSDME cleavage

(A) Tissue slices of human colorectal cancer (CRC) specimens were treated ex vivo with 1 × 10⁷ TCID₅₀ of VSV-GP-Katushka or VSV-GP-GSDME-Katushka. Katushka fluorescence intensity in the slices was scored using fluorescence microscopy (negative to +++++) over a period of 72 h. Representative images of permissive CRC samples are shown; scale bars, 500 μm. (B) Slice cultures from four different CRC cancers were scored at the indicated time points post-infection. Data are presented as mean (SEM) of n ≥ 5 slices per CRC specimen and condition (supplemental information and Tables S8–S11). (C) Individual slices (n = 3) from specimen #43 (CRC-43) with Katushka reporter-positive (permissive) signals were analyzed for GSDME expression and cleavage by immunoblot. The upper image shows GSDME and cleaved GSDME-NT, the middle image shows VSV-N, and the lower image shows β-actin (ACTB) as a loading control. (D) GSDME and cleaved GSDME-NT levels were quantified relative to VSV-N signal (not shown) using Compass for Simple Western 6.1.0. Data are presented as mean (SD).

In summary, VSV-GP-GSDME-Katushka demonstrated the ability to infect patient-derived CRC slice cultures similar to VSV-GP-Katushka, leading to elevated GSDME expression and cleavage. The observed variability in viral permissivity and GSDME activation highlights the heterogeneity of CRC tumors and underscores the importance of personalized approaches in therapeutic development. This ex vivo system provides a robust platform for evaluating oncolytic viral therapies in the context of human tumor biology.

Discussion

OVT is a promising cancer treatment strategy, combining direct tumor lysis with immune activation. However, the therapeutic efficacy of these viruses can be further improved by integrating mechanisms that enhance their capacity to induce ICD, such as pyroptotic cell death. Recent studies suggest that GSDME-mediated pyroptosis enhances tumor lysis and immune stimulation.³⁵

In our study, we demonstrate that engineering the oncolytic virus VSV-GP to express full-length GSDME effectively redirects the apoptotic pathway toward pyroptosis, thereby amplifying anti-tumor immune responses. Initial experiments confirmed that VSV-GP alone activates CASP3 and induces GSDME-mediated pyroptosis, comparable to the chemically induced pyroptosis observed with raptinal treatment. Morphological and biochemical analyses revealed hallmark features of pyroptosis, including membrane swelling and PI uptake, supporting the induction of lytic cell death. Given that only the NTD of GSDME is responsible for pore formation, direct delivery of GSDME NTD has been explored in other systems to bypass the need for upstream caspase activation.^52,53 However, attempts to incorporate the active NT domain into VSV-GP were unsuccessful, likely due to its high intrinsic cytotoxicity, which impaired viral viability. Notably, to our knowledge, only one other group has successfully incorporated GSDME NTD into a replicating viral vector.⁵⁴ To circumvent high toxicity, we inserted the autoinhibited full-length GSDME (GSDME-FL) into the viral genome. Importantly, the incorporation of GSDME-FL did not compromise viral morphology or replication kinetics, as confirmed by cryo-electron microscopy and viral replication assay. Infected cells exhibited a clear pyroptotic phenotype, and immunoblotting confirmed GSDME cleavage. Interestingly, further analysis of the apoptotic cascade revealed reduced cleavage of CASP3, CASP9, and PARP in cells infected with VSV-GP-GSDME. In contrast, robust activation of these apoptotic markers was observed in cells treated with the parental VSV-GP or the loss of function mutant VSV-GP-GSDME_F2A. This shift in signaling is consistent with previous reports showing that in GSDME-high tumor cells, chemotherapeutic agents preferentially induce CASP3-mediated cleavage of GSDME rather than PARP cleavage, resulting in attenuated apoptotic signaling.^24,25,26,55 Furthermore, studies on gasdermin D (GSDMD) have shown that its NTD causes severe and rapid mitochondrial damage due to its strong cardiolipin-binding affinity, occurring before plasma membrane damage.^56,57 A similar mechanism has been reported to permeabilize mitochondrial membranes, contributing to cell death and causing neurite loss.^58,59 While earlier reports suggest that mitochondrial damage should lead to cytochrome c release and apoptosome formation, thereby augmenting CASP3 activation and GSDME cleavage,⁵⁸ we speculate that in our setting, high levels of GSDME, either stably overexpressed or delivered by VSV-GP-GSDME, interfere with the progression of the apoptotic cascade. Under these conditions, initial CASP3 activation, triggered by either raptinal or VSV-GP infection, may be sufficient for rapid GSDME cleavage. The resulting GSDME NT fragment disrupted mitochondrial membranes, leading to early mitochondrial damage and amplifying pyroptotic signaling. Notably, a substantial pool of cytosolic GSDME remained available for further cleavage, suggesting that minimal CASP3 activity is sufficient to initiate and sustain GSDME-mediated pyroptosis and is dominant to apoptotic pathway progression. Redirecting CASP3 activity to GSDME cleavage promotes pore formation and cell lysis.

Another important hallmark to distinguish cell death modalities is the release of DAMPs, such as eATP.⁶⁰ VSV-GP-GSDME-infected cells showed elevated eATP and LDH levels, indicating gasdermin pore formation and membrane rupture. Recent studies have shown that gasdermin pores facilitate ATP efflux, while LDH release is mediated by Ninjurin-1 (NINJ1)-dependent membrane rupture.⁴⁹ In GSDME OE cells treated with raptinal, ATP, and LDH were released simultaneously. However, ATP levels released from GSDME-deficient cells were significantly lower, suggesting that secondary necrosis may occur in the absence of functional GSDME. The abundant availability of GSDME in overexpression models likely accelerates membrane rupture, resulting in the concurrent release of both ATP and LDH.

To evaluate the therapeutic potential of VSV-GP-huGSDME, we tested its efficacy in multiple murine tumor models possessing varying degrees of viral permissivity and resistance to OVT.⁶¹ Treatment of EMT-6 Ifnar1^−/− tumors with VSV-GP-huGSDME delayed tumor growth and improved survival but did not lead to complete tumor control in all mice.

In the highly permissive TC-1 Ifnar1^−/− tumor mouse model,⁶² low-dose treatment with VSV-GP-huGSDME resulted in effective tumor control and remission, while higher doses of VSV-GP-huGSDME led to pronounced oncolytic activity and comparable tumor growth delay of VSV-GP and VSV-GP-huGSDME-treated groups. This may suggest that lower viral loads preferentially engage immune-mediated mechanisms rather than predominantly utilizing direct oncolysis, indicating a dose-dependent shift in therapeutic mode of action. While the Ifnar1^−/− tumor mouse models do not fully capture an intact antitumor immune landscape, their use was necessary in this study because commonly used murine tumor lines such as TC-1 and EMT6 are IFN-competent and therefore poorly permissive to VSV-GP infection. Employing Ifnar1^−/− tumor cells enabled productive viral replication and mechanistic evaluation of therapeutic activity in an IFN-competent host. Importantly, this approach remains translationally relevant, as defects in type 1 interferon signaling are observed in a substantial fraction of human tumors and cancer cell lines (≈60%), including head and neck cancers, which display broad permissivity to VSV-GP-based OVs.⁴⁵

Effective anti-tumor immunity depends on the priming of T cells, which requires efficient antigen cross-presentation by DCs. Enhancing DC mobilization, antigen uptake, and activation is a promising strategy to potentiate T cell-based therapies.⁵⁰ To further investigate the immune response at low viral doses, we analyzed DCs in the tdLNs and TILs. VSV-GP-huGSDME treatment led to enhanced activation of cDCs in the tdLNs, as evidenced by upregulated expression of CCR7 and CD86. CCR7 facilitates DC migration to secondary lymphoid organs, promoting antigen presentation and T cell priming,⁶³ while CD86 provides essential costimulatory signals for T cell activation.⁶⁴ Importantly, cDCs are particularly known for their ability to efficiently cross-present antigens to CD8⁺ T cells, a critical function for initiating cytotoxic immune responses against tumors.^65,66 This aligns with our observation that cDCs exhibited a higher frequency of CCR7⁺CD86⁺ double-positive cells, indicating a mature and migratory phenotype. These findings suggest that VSV-GP-huGSDME promotes cDC maturation and trafficking, thereby enhancing cross-presentation and contributing to the initiation of effective antitumor immunity. In parallel, CD8⁺ T cells within the TME exhibited increased expression of activation markers CD25 and a specific glycoform of CD43, along with effector molecules granzyme B and perforin. This profile reflects enhanced cytotoxic T lymphocyte (CTL) activation and functional capacity. The upregulation of cytolytic molecules further supports the ability of these T cells to mediate direct tumor cell lysis. Compared to VSV-GP, VSV-GP-huGSDME induced significantly stronger activation of both cDCs and CD8⁺ T cells. This enhanced immunogenicity is likely driven by the shift from apoptosis to GSDME-mediated pyroptosis. This nature of cell death is inflammatory, promoting the release of DAMPs, which are known to stimulate DC maturation, antigen presentation, and CTL activation.^67,68 These findings were further supported in the B16-F10 melanoma model, which is known for its resistance to ICIs and its immunologically “cold” TME.⁶¹ While VSV-GP-huGSDME treatment delayed tumor growth, complete tumor control was not achieved. However, combination with anti-PD-1 therapy significantly improved outcomes, consistent with previous reports showing that artificial activation of the GSDME pyroptosis pathway can sensitize tumors to ICI therapy.⁶⁹ Notably, we observed upregulation of granzyme B and PD-1 in CD8⁺ T cells within the tdLNs, suggesting that pyroptosis-induced inflammation may prime the TME for checkpoint blockade responsiveness. While our data demonstrates early DC activation, cytotoxic CD8 T cell responses and increases ICI responsiveness; longer follow-up of responses in applied mouse models and rechallenge experiments will be required to directly confirm durable, tumor-specific immune memory. Additionally, the biological significance of chemokines and cytokines within the TME between d1 and d3 post treatment remains unclear. Only, the selective reduction in IFN-β may suggest that GSDME-driven, accelerated tumor cell killing modulates local IFN-β outputs—potentially dampening antiviral signaling—without broadly altering the inflammatory cytokine milieu, warranting dedicated mechanistic investigations.

To place these findings within the broader dynamics of the antitumor response, it is informative to consider how the i.t. immune landscape diverges between animals that initiate tumor regression rapidly and those that respond only after a delay. Because all tumors were collected at a uniform endpoint, the immune profiles of early and late responders effectively capture distinct temporal stages of the evolving antitumor response. Early responders, having initiated tumor regression several days earlier, displayed features of a more advanced immune phase, including increased T cell infiltration and higher numbers of activated CD8⁺ T cells and granzyme-B-positive CD4⁺ T cells. Late responders, in contrast, showed characteristics of an earlier stage marked by higher frequencies of granulocytic myeloid cells. An important methodological consideration is that responder status could not be prospectively assigned at the time of tumor harvest. Because tumors were collected at a fixed endpoint, the immune profiles reflect discrete temporal states of the antitumor response rather than predictors of eventual therapeutic outcome, positioning the observed differences as sequential mechanistic stages rather than stable responder-versus-non-responder identities. Granulocytic cells in late responders exhibited elevated MHCII and CD86 expression, consistent with a proinflammatory, T cell-activating phenotype.^70,71 In early responders, this activation phase appears to have transitioned further, as indicated by lower MHCII/CD86 expression and the presence of fully activated CD8⁺ and cytotoxic CD4⁺ T cell populations. Late responders also showed higher arginase-1 expression, a feature associated with suppressive granulocytic states in other tumor models,^72,73 suggesting that such programs may still be emerging at the time of analysis. These patterns imply that granulocytic activation and subsequent acquisition of suppressive features unfold along a temporal continuum, raising the possibility that arginase-1-expressing granulocytes could influence both, relapse dynamics or the durability of antitumor immunity.

Finally, to bridge the translational gap between murine models and human clinical relevance, we employed an ex vivo system using patient-derived tumor slices. This platform enables the assessment of viral permissivity, GSDME expression, and CASP3-dependent cleavage in an intact human TME. Building on previous work in HNSCC,⁴⁵ we here extended this approach to CRC, a distinct tumor entity and the third leading cause of cancer-related death worldwide.⁷⁴ Notably, GSDME is frequently silenced in CRC due to promoter hypermethylation, while its expression remains intact in adjacent non-tumor intestinal epithelial cells.^32,75 This differential expression has been linked to radio resistance, with studies showing that restoring GSDME sensitizes CRC cells to radiotherapy.⁷⁶ In our ex vivo system, we confirmed that GSDME can be ectopically expressed and cleaved in cultured human CRC slices, demonstrating the model’s feasibility for mechanistic studies.

In conclusion, VSV-GP-GSDME represents a promising strategy to induce pyroptosis in GSDME-deficient tumors. Leveraging the CASP3-GSDME axis enhances the therapeutic potential and broadens the applicability of VSV-GP virotherapy.

Materials and methods

Cell lines and cell culture conditions

HEK293-T cells (Eufets, #ZM-0003-N-1308) were maintained in Dulbecco’s modified Eagle’s medium (DMEM; Gibco, #11960044) supplemented with 10% fetal calf serum (FCS; Gibco, #10100147) and 2 mM CTS-GlutaMax-I (Gibco, #A1286001), 1% MEM non-essential amino acids solution (100×, Gibco, #11190-035), and 1% sodium pyruvate (Gibco, #11360-070). Here, 4T1 cells (ATCC CRL-2539), were cultured in RPMI 1640 (Gibco, #21875-034) supplemented with 10% fetal bovine serum (FBS; Gibco, #10100147). A549 cells (ATCC, #CCL-185TM) were cultured in F12K Nut Mix, (Gibco, #21127022) supplemented with 10% FBS (Gibco, #10100147).

TC-1 Ifnar1^−/− cells, kindly provided by Dr. Guido Wollmann (Tyrolean Cancer Research Institute, Innsbruck, Austria), were cultured in RPMI 1640 media (Gibco, #21875-034) supplemented with 10% FBS (Gibco, #10100147), 1% sodium pyruvate (Gibco, #11360-070), 1% MEM non-essential amino acids solution (Gibco, #11190-035) and 0.4 mg/mL G418 (Gibco, #10131027). EMT6 cells were cultured in Waymouth’s MB752/1 (Gibco, #31220-023) supplemented with 15% FBS (Gibco, #10100147). B16F10 cells (ATCC, CRL-6475) were maintained in (DMEM; Gibco, #11960044) supplemented with 10% FCS (Gibco, #10100147) and 2 mM CTS-GlutaMax-I (Gibco, #A1286001). All adherent cell lines were cultured in humidified incubators at 37°C and 5% CO₂.

HEK293-F suspension cells obtained from Thermo Fisher Scientific (#11625019) were maintained in serum-free BalanCD (FUJIFILM Irvine Scientific, #91165), supplemented with 4 mM CTS-GlutaMax-I (Gibco, #35050061). Suspension cells were cultured in humidified incubators at 37°C, 5% CO₂ and 120 rpm in shaking flasks.

Cell counting and viability were determined using the Via1-Cassette (chemometic, #941-0012) with the NucleoCounter NC-200 instrument. Measurements were performed according to manufacturer’s instruction.

Generation of knockout and knockin cell lines

EMT6 Gsdme^−/−, EMT6 Casp3^−/−, EMT6 Ifnar1^−/− cells (kindly provided by M. Strubel, Boehringer Ingelheim, Germany) and EMT6 Ifnar1^−/− Gsdme^−/− (Horizon Discovery Ltd. Cambridge, UK) were generated using the CRISPR-Cas9 system. Sequence information of the gene-specific gRNA used is provided in Table S12. To generate the TC-1 Ifnar1^−/− cell line, the plasmid pSpCas9(BB)-2AGFP (PX458) (Addgene, plasmid #48138; Watertown, MA, USA), designed to target three regions of Ifnar1 located in exons 2 and 3, was transfected into TC-1 cells using standard protocols. Single-cell sorting was then performed as previously described.⁶²

EMT6 GSDME stably GSDME OE was established via lentiviral transduction using a pSFFV-GFP-Puro-GSDME plasmid. Puromycin selection was applied to isolate GSDME OE cells. Successful modification was confirmed by sequencing and immunoblot.

Viruses

VSV-GP was described previously.⁷⁷ Variants of VSV-GP expressing GSDME, the red fluorescent protein Katushka or a T2A linked combination of GSDME and Katushka were generated using the Gibson NEBuilder HiFi DNA Assembly kit (New England Biolabs, #E5520S), adhering to the manufacturer’s recommendations. In brief, the coding sequences for GSDME, Katushka or GSDME-Katushka were amplified by PCR. The resulting transgene was added into VSV-GP vector between the LCMV-GP and the VSV-L open reading frame. GSDME_F2A and GSDME_D270A were generated by PCR-mediated mutagenesis of the VSV-GP-GSDME genomic vector. Recombinant virus variants were subsequently recovered using a helper virus-free rescue protocol as described previously.⁷⁸ HEK293T cells were co-transfected with helper plasmids using the calcium phosphate method. Virus progeny was then plaque-purified twice on adherent HEK293-F cells.

For virus amplification, HEK293-F suspension cells were seeded into Erlenmeyer shake flasks. About 48 h of post-seeding, the cells were infected with the stock virus at a MOI of 0.0005. At 32–34 h post-infection, 0.2 M NaCl was added to the culture to facilitate virus particle release into the supernatant. Subsequently, the suspension was clarified by centrifugation at 2,000× g for 5 min and filtered through a 0.22 μm syringe filter. This material was either aliquoted, stored at −80°C and analyzed or further purified, via centrifugation or chromatographic methods as previously described.⁷⁹

Determination of viral genomic copies

Viral RNA was extracted using the MagMAX Viral RNA Isolation Kit (Ambion, #AMB1836) following the manufacturer’s instructions. Quantification of genomic VSV-N RNA copies was performed via quantitative real-time PCR (real-time qPCR) using 2 μL of extracted RNA and the iTaq Universal Probes One-Step Kit (Bio-Rad, #1725141). The reaction utilized VSV-N-specific primers (forward: 5′-AGTACCGGAGGATTGACGACTAAT-3′; reverse: 5′-TCAAACCATCCGAGCCATTC-3′) and a 6-carboxyfluorescein (FAM)-labeled probe (5′-FAM-ACCGCCACAAGGCAGAGATGTGGT-BHQ1-3′). Absolute quantitation was achieved by generating standard curves with serial log-fold dilutions ranging from 1 × 10⁸ to 10 genome copies/mL. Standards were measured in technical duplicates, while all samples and no-template controls (NTC) were measured in technical triplicates. Amplification conditions were as follows: 50°C for 10 min, 95°C for 2 min, followed by 40 cycles of 95°C for 15 s and 60°C for 30 s. Data were processed and exported using BioRad CFX Manager software.

TCID₅₀ assay

For determination of infectious titers, a TCID₅₀ assay was performed as described previously.⁸⁰ In brief, 1 × 10⁴ BHK-21[cl13] cells (CLS Cell Lines Service GmbH, #800125) maintained in Glasgow minimum essential medium (GMEM; Gibco, #21710082) supplemented with 10% FCS (Gibco, #10100147), 5% tryptose phosphate broth (Gibco, #18050039) and 1% penicillin-streptomycin solution (Gibco, #5070063), were seeded into a 96-well plate in 100 μL supplemented GMEM per well. At 24 h post-seeding, cells were infected with half-log10 serial dilutions of the virus in replicates (n = 6 per dilution). At 3 days post-infection, the confluence of each well was measured using an automated plate reader (Tecan, Spark 3M). A confluency threshold of ≤95% was set to identify cytopathic effects (CPEs) in wells resulting from viral infection. The TCID₅₀/mL value was subsequently calculated using the Spearman-Kärber method.^81,82

Immunoblot

A total of 1–2 × 10⁶ cells were seeded into 6-well plate. After 24 h, cells were infected at an MOI of 3. At indicated time points, cells were harvested and pelleted by centrifugation at 5,000 rpm. Cells were then lysed in pre-cooled RIPA buffer (Thermo Fisher Scientific, #89901) containing a protease inhibitor cocktail (Sigma-Aldrich, #11836170001) and incubated at 4°C for 30 min, with agitation (800 rpm). Cell debris were removed by centrifugation at 12,000 rpm for 20 min at 4°C. Cell extracts were stored at −80°C until further use. Total protein concentration of lysates was determined using a Pierce BCA assay kit (Thermo Fisher Scientific, #23225) and subsequently analyzed on a Jess Simple Western System (BioTechne). Protein analysis was performed according to the manufacturer’s protocol and the following reagents were used: 12−230 kDa Jess separation module (BioTechne, #SM-W001) 8 × 25 capillary cartridges (BioTechne, #SM-W00X). For detection, the following antibodies and reagents were used: primary antibodies: anti-DFNA5/GSDME (Abcam, #ab215191), anti-CASP9 (Cell Signaling Technology, #9504), anti-cleaved CASP9 (Cell Signaling Technology, #9509), anti-CASP3 (Cell Signaling Technology, #9662), anti-PARP (Cell Signaling Technology, #9532), anti-β-Actin (Cell Signaling Technology, #3700), VSV polyclonal rabbit serum (kind gift from Prof. Stefan Finke, Friedrich-Loeffler-Institute, Isle of Riems, Germany). For antibody detection, the anti-Rabbit detection or anti-Mouse detection modules (BioTechne, #DM-001, #DM-002, #DM-009) were used. Immunoblots were quantitated using the software Compass for Simple Western (v.6.1.0).

PI uptake and time-lapse microscopy

For virus-induced cell death, 2 × 10³ cells per well were seeded into a 96-well plate with media supplemented with 0.625 μM PI (Abcam, #ab14083). About 24 h post-seeding, cells were infected at a MOI of 3. Imaging was performed in an Incucyte S3 Live Cell Analysis System (Sartorius) over a 72 h period at 10× magnification, with images recorded every 2 h. For raptinal-induced cell death, cells were treated with 10 μM raptinal and PI uptake was imaged at 10× magnification. Images were recorded every 10 min, over a period of 75 min. PI uptake was quantified using the Incucyte software v.2022B Rev2.

To quantify early apoptotic signals, 2 × 10³ cells were stained with IncuCyte AnnexinV Red Reagent (Sartorius, #4641) and IncuCyte Cytotox Green Reagent for dead cells (Sartorius, #4632 or #4633) according to the manufacturer’s recommendations before infection. Images at 10× magnification were recorded every 2 h for 72 h and analyzed using the Cell-by-Cell analysis module of the Incucyte Software. Quantification of AnnexinV-Red and/or Cytotox-Green positive cells was performed in triplicate (n = 3) and expressed as a percentage of total cells at the indicated time points post-infection.

LDH release

Cell supernatants were collected at indicated time points and LDH release was measured using the Bio LDH kit with the Cedex Bio Analyzer (Roche) according to manufacturer’s instruction. Briefly, 8 × 10⁴ cells per well were seeded into a 12-well plate. Around 24 h post-seeding, cells were either treated with raptinal or infected with virus at an MOI of 10. Cell culture supernatants were clarified by centrifugation at 300 g for 5 min before LDH measurement.

ATP release

eATP in the cell culture supernatant was quantified using a bioluminescence-based RealTime-Glo Extracellular ATP Assay (Promega, #GA5010) according to the supplier’s recommendations. In brief, 2,000 cells were seeded per well on a 96-well plate in cell culture media supplemented with 10 mM HEPES. After 24 h, cells were infected at an MOI of 10 and the 4×-reconstituted RealTime-Glo Extracellular ATP Assay reagent was added to the virus-infected cells. For control, cells were left untreated. Luminescence was recorded for 24 h in a multimode plate reader (GloMax Explorer, Promega) at 37°C. Relative light units (RLU) were analyzed and plotted as n = 3 replicates ±SD using GraphPad Prism software (v.10.4.1).

Cryo-EM imaging

Data collection was performed by NanoImaging Services (Woburn, MA, USA). Cryo EM sample preparation was performed as described previously.⁸³ The virion lengths of randomly selected intact bullet-shaped virus particles were measured using ImageJ.⁸⁴

In vivo studies

Female BALB/cByJ mice (6–8 weeks old) were obtained from Charles River Laboratories (Wilmington, MA, USA), and C57BL/6JRj mice (6–8 weeks old) were sourced from Janvier Labs. For subcutaneous (s.c.) tumor induction, 2.5 × 10⁵ EMT6 Ifnar1^−/− cells were injected into the right flank of BALB/c mice, while 1 × 10⁵ TC1 Ifnar1^−/− cells were injected into the right flank of C57BL/6J mice. To induce tumors in the mammary fat pad (mfp), 5 × 10⁴ B16-F10 cells were injected directly into the mfp tissue of C57BL/6J mice. Mice were anesthetized with isoflurane, and a 5 mm skin incision was made over the lateral thorax to expose the second mammary fat pad for cell injection. Tumor size was measured three times a week with a caliper and volume was calculated using the formula: length × width² × 0.5. Animals were randomized based on their individual tumor volume. Randomization in preparation of treatment was performed when the mean tumor volume reached a size of 80–150 mm³. Animal behavior was observed daily, while body weight and tumor volume were measured a minimum of three times a week. Mice were sacrificed when their tumor volume reached 1,500 mm³ or tumors showed signs of ulcerations.⁸⁵ Animals were euthanized by overdose on isoflurane anesthesia followed by cervical dislocation. Tissue samples were transferred individually to a sterile vial filled with T-store (Life Science Group Ltd.) and stored at 4°C overnight for analysis the next days.

Preparation of cell lysates from mouse tumors

Tumors were collected and lysed in Cell Lysis Buffer (10×) (Cell Signaling Technology Inc., USA) supplemented with 1 mM PMSF protease inhibitor (Cell Signaling Technology Inc., USA) immediately prior to use. For each 100 mg of tumor tissue, 500 μL of ice-cold lysis buffer were added to gentleMACS M Tubes (Miltenyi Biotec, Germany) maintained on ice. Samples were mechanically dissociated using a GentleMACS Dissociator (Miltenyi Biotec, Germany) with the Protein_01 program. Following dissociation, M tubes were centrifuged at 500× g for 1 min at 4°C to collect the sample material, which was subsequently transferred to microcentrifuge tubes. Lysates were clarified by centrifugation at 16,000× g for 10 min at 4°C, and supernatants were collected, aliquoted and stored at −80°C until further processing. Protein concentrations were determined using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, USA) according to the manufacturer’s instructions and were used to equalize protein input for Jess automated western blotting (ProteinSimple, USA) and to normalize cytokine concentrations to total protein content.

Multiplex immunoassay

IFN-α, -β and -γ, IL-1β, -6, -10, -18, CCL-2, CXCL-1, and TNF-α concentrations were quantified in murine serum samples and tumor lysates using Meso Scale Discovery (MSD) U PLEX Custom Biomarker Group 1 (mouse) assays (Cat. #K15069M-2). Samples were diluted 2-fold in assay diluent, measured in duplicates and processed according to the manufacturer’s instructions using MSD 96 well 10 spot plates. Signal detection and data acquisition were performed on a QuickPlex SQ 120 instrument (MSD). Analyte concentrations were calculated using the MSD Discovery Workbench software. Measurements below the limit of detection were assigned a value of zero.

Flow cytometric analysis

Mouse tumors and tdLNs were harvested three and seven days after virus treatment. Single cell suspensions were prepared from tdLNs by passing them through a 70 μm cell strainer using the rubber stamp of a 3 mL plastic syringe (mfpB16F10 study) or using the spleen dissociation kit (Miltenyi Biotec) in combination with the gentleMACS Octo Dissociator with heaters (TC-1 Ifnar1^−/− study) using program 37C_m_SDK_1. For dissociation of tumors, a combination of enzymatic digestion with murine tumor dissociation kit (Miltenyi Biotec) and mechanical dissociation on an OctoDissociator (program 37C_mTDK_1) according to the manufacturer’s instruction was used. Cells were then washed with PBS, cell pellet was suspended in a defined volume of PBS and stained with a live/dead discrimination dye. After blocking of Fc-receptors with an Fc-receptor blocking reagent (BioLegend) in combination with blocking with True Stain Monocyte BlockerTM (BioLegend) to prevent non-specific binding of tandem fluorophores to myeloid cells. Cell suspension from tdLN was stained with antibodies characterizing DC populations and differentiation, whereas tumor single cell suspensions were stained with antibodies characterizing the composition of tumor-infiltration leukocytes as well as activation and killing capacity of T cells. Antibody panels are listed in Tables S1–S6. Counting beads (Invitrogen, #C36950) were added in the TC1 Ifnar1^−/− study to the final suspension and cells were analyzed on a MacsQuant16 (Miltenyi Biotec) conventional flow cytometer or on a ID7000 Spectral Cell Analyzer (Sony Biotechnology). Data were analyzed using FlowJo software and visualized using GraphPad Prism v.10.4.1.

Ex vivo tumor slice culture system

Four freshly resected patient-derived CRC samples were sectioned into 200 μm thin tissue slices using a Leica VT1000S Vibratome (Leica Biosystems). Individual slices were sequentially distributed into wells of a 48-well cell culture plate and cultivated as free-floating tissue cultures in DMEM (Gibco, #21063029) supplemented with 7% FCS (Gibco, #10100147), 0.1% β-Mercaptoethanol (Gibco, #21985023), 1% penicillin-streptomycin (Gibco, #5070063), 1% Amphotericin B (Gibco, #15290026), and 1% Protease Inhibitor Cocktail (Sigma-Aldrich, #P1860). To ensure quality control at baseline, tumor content, TME composition, and viability were characterized by fixing, paraffin-embedding, and staining every seventh tumor slice with H&E for histopathology assessment, as described in the supplemental information.

Cultivated CRC slices were treated with 1  ×  10⁷ TCID₅₀ of VSV-GP-Katushka, VSV-GP huGSDME-Katushka or left untreated. Samples were incubated at 37°C, 5% CO₂ in a humidified incubator for up to 72 h. Tissue integrity and presence of the Katushka signal in each individual slice was monitored at 24, 48, and 72 h using an inverted fluorescence microscope (Cell Discoverer 7, Zeiss) at low magnification (Plan-Apochromat 5×/0.35 objective and 0.5× Tubelens optovar). A semi-quantitative scoring system ranging from negative to +++++ was utilized to assess permissivity as described in Runge et al.⁴⁵ At 72 h post-infection, slices were recovered from individual wells of the cell culture plate and protein lysates were prepared and analyzed via immunoblotting as described previously.

Ethical statement

Human CRC samples were obtained from the Department of Visceral, Transplant, and Thoracic Surgery at the Medical University of Innsbruck, Austria. Tumor samples were randomly selected from four patients who underwent surgery between July and October 2023. Written informed consent was obtained from all patients. The study was approved by the Ethics Committee of the Medical University of Innsbruck, Austria (vote no. 1043/2020).

Animal experiments were conducted at Oncodesign (Dijon, France). Animal housing and experimental procedures adhered to French and European regulations as well as the NRC Guide for Care and Use of Laboratory Animals. The animal facility is authorized by the French authorities (agreement no. C 21 231 011 EA). All animal procedures were reviewed and approved by the Institutional Animal Care and Use Committee of Oncodesign (Oncomet) and authorized by French authorities (CNREEA agreement N° 91).

Data analysis and statistics

Graphs and figures were generated using GraphPad Prism v.10.4.1. and Adobe Illustrator v.29.7.1. Statistical analyses were performed using GraphPad Prism v.10.4.1, with appropriate statistical tests selected based on the data.

Data availability

Data are available within the manuscript and upon reasonable request from the corresponding author.

Acknowledgments

We thank Simone Bartl (ViraTherapeutics GmbH) for excellent technical support. We thank Michaela Smolle and the ViraTherapeutics Bioprocess Innovation Team as well as our colleagues in the Quality Control Department for intensive support for virus amplification and analytics. We are grateful to André Volland (ViraTherapeutics GmbH) for experimental advice. We would like to thank Patrizia Stoizner (Medical University Innsbruck), Guido Wollman, and Sieghart Sopper (Tyrolean Cancer Research Institute) for scientific advice. Funding was provided by ViraTherapeutics, Boehringer Ingelheim and the Life Science Program of the Austrian Research Promotion Agency (Österreichische Forschungsförderungsgesellschaft, FO999912222) “Rewiring the Tumor Microenvironment with Oncolytic Virotherapy and Cancer Vaccines.” The graphical abstract was created in BioRender.

Author contributions

S. Schneider, and T.S. designed and performed experiments, analyzed, and compiled data; C.M., M.M., M.P., J.K., I.B., E.P., I.A., and M.G. performed experiments and analyzed data; C.B.K.-T. assisted with data and material acquisition; B.S., K.D., and C.B.K.-T. analyzed data; M.P. and D.X. assisted with method establishment and reviewed the manuscript; S. Schneeberger and S. Scheidl gained institutional approval, were involved in formal study and methods design, material, and data acquisition; T.N., T.S., B.S., and M.P. conceived and supervised the project; S. Schneider, T.S., and T.N. wrote the manuscript; T.S., T.N., B.S., C.B.K.-T., and K.D. revised the manuscript; K.E. and S.C. gave advice and revised the manuscript.

Declaration of interests

ViraTherapeutics is a subsidiary company of Boehringer Ingelheim International GmbH. VSV-GP-GSDME and variants thereof described in this work are included in a patent application filed by Boehringer Ingelheim International GmbH.