IntroductionOncolytic viruses (OVs) are emerging immunotherapies under active clinical investigation.1 These agents, which can be used in their wild type or engineered forms, preferentially infect, replicate in, and lyse cancer cells, exposing tumor antigens to the host immune system to help facilitate an anti-cancer immune response. Currently, four OVs, and one non-replicating OV, are approved for use clinically, each of which require intratumoral administration.2,3,4,5,6 Additional agents under preclinical investigation are engineered to harbor immune-enhancing and costimulatory payloads, or even suicide cassettes to bolster anti-tumor immunity.7,8,9 In many cases, clinical applications of these agents are limited by their current requirement for local injection. This precludes tumors that are not surgically accessible, or are metastatic. For OVs that are injected systemically, the lack of a “tumor-specific” viral delivery platform limits the choice of immune activating payloads to avoid severe toxicities.8,9,10 Further clinical development of novel viral vectors that have refined specificity for targeting tumors is needed.Vesicular stomatitis virus (VSV) belongs to the Rhabdoviridae family of viruses, having a single-stranded, negative-sense RNA genome of around 11 kb.11,12 VSV mainly infects livestock and causes blister-like lesions in these animals.11,12 While VSV infection in humans is rare, it can cause flu-like symptoms.11 VSV replicates strictly in the host cell’s cytoplasm, does not enter the nucleus, and does not interact with host DNA. This feature, combined with a lack of a DNA phase for VSV, makes it safer compared to DNA viruses (such as Adenovirus) due to the absence of potential insertional mutagenesis to the host.11,12 Owing to its low prevalence (and thus less prior existing immunity against the vector) and mild symptoms in humans, VSV has been studied as a potential vector for cancer immunotherapy.11,13 For instance, the human IFN-β transgene has been cloned into the VSV backbone to further enhance its viral safety profile.14,15,16 Others incorporated interleukin-12 (IL-12) into VSV, and showed that intratumoral injection of these viruses control tumor burden.17,18 VSV harbors its G protein (VSV-G) on the surface of the membrane, which binds to the low-density lipoprotein receptor (LDL-R) on host cells to begin the cell entry process. A recent publication further narrows down the interaction between VSV-G to the CR-2 and CR-3 domains of the LDL-R.19 Due to the wide expression pattern of the LDL-R across different tissues and cell types, the VSV-G protein was used to replace the spike protein of HIV to generate lentivirus that is now commonly used in laboratories.20 However, the broad-tropism of the VSV-G protein also determines the lack of tumor specificity of VSV, limiting its route of administration to intratumoral injection when potent payloads, such as IL-12, are incorporated.Although current immunotherapies have demonstrated clinical success, response rates to these therapies vary dramatically between cancer type, and person.21,22,23,24 For instance, colorectal cancer (CRC) is the third leading cause of cancer-related deaths in the United States,25 and only a small fraction (15%) of CRC that harbor microsatellite instability (MSI) mutations are responsive to immune checkpoint blockade (ICB) therapies such as anti-PD-1.26 Additionally, for advanced CRCs that are microsatellite stable (MSS), immunotherapeutic options are limited and there is a critical need for better treatments. The success of ICB is limited by factors including tumor immunoediting, mutational burden, and the presence and activation state of infiltrating lymphocytes.27 To overcome therapeutic limitations observed with ICB, OVs are used to inflame the tumor microenvironment (TME) and promote T cell responses directed against tumor antigens, as well as increase the expression of checkpoint molecules on immune and tumor cells.28,29,30,31,32,33,34 Currently, various clinical trials (NCT03017820, NCT03647163, NCT03120624, NCT06581406, NCT04735978, NCT03767348, NCT04725331) are actively investigating the efficacy of these therapeutic combinations, as preclinical trials have pointed towards a synergy from dual treatments.29,30,31,32 However, the efficacy of the combination therapy is heavily dependent on several factors not limited to the type of tumor, strain of OV, transgene payload, as well as the dosing scheme, underscoring the importance of understanding how each unique OV construct impacts the efficacy of the combination therapy.Here, we demonstrate the development of a cancer-specific oncolytic VSV by engineering the endogenous G protein with a “cap-linker” strategy. This modification allows for the safe systemic delivery of potent, immune-stimulating IL-12 directly to the site of the tumor resulting in tumor eradication. We also interrogated the efficacy of the engineered virus in treating metastatic cancers and its ability to mount long-term immune memory. Importantly, we combined tumor-targeted delivery of IL-12 with ICB for enhanced efficacy against established tumors. These efforts pave a new way of applying our novel Pro-VSV technology to safely deliver potent immune activating payloads to human cancers that are not surgically reachable. Further, we generalized our study to include G proteins from two serotypes of VSV and the Maraba virus, to show that the “cap-linker” strategy may be generally applicable for targeted tumor delivery of oncolytic viruses.ResultsRational design of a cancer-specific viral platform using lentivirusVSV-G binds to LDL-R on target cells, resulting in broad tropism in humans and experimental animals.19,35 Specifically, VSV’s clinical use is limited by its potential for neurotoxicity and infectivity of the heart and kidney, which is observed in mice and non-human primates, as well as hepatotoxicity.36,37,38 Lack of tumor specificity by VSV thus limits the choice of payload that the virus may carry, with immune-related adverse events as a concern if potent immune activating payloads (such as IL-12) are used.To improve the cancer-specific targeting of systemically injected VSV, we engineered VSV-GIN (Indiana serotype) to limit its ability to bind its cognate receptor unless present in the tumor microenvironment (TME). We developed a ‘cap-linker’ cancer-specific targeting system where the ‘cap’ acts as an inhibitory domain to VSV-GIN, but is released when the linker connecting the cap to VSV-GIN is cleaved by proteases in the TME. This results in the rescue of VSV-GIN’s ability to bind to the LDL-R and guide viral entry into tumor cells (Fig. 1a).Fig. 1Full size imageCap-linker strategy to modify of VSV-GIN. a Illustration of the ‘cap-linker’ strategy for the development of a cancer-specific oncolytic virus. An inhibitory domain of the VSV-G protein (the “cap”) is engineered to the N’-terminal of VSV-G, spaced by a protease-cleavable linker. Loss of the cap occurs upon viral entry into cancer tissue where the overexpression of proteases allows for the facile cleavage of the amino acid linker sequence that tethers the cap to VSV-G, restoring viral infectivity. b, c Illustration of the domain structure for WT VSV-GIN (b) and Pro-VSV-GIN (c), the latter contains an N’-terminal addition of a FLAG tag with different inhibitory domains linked to mature VSV-GIN by either a 12 or 20 amino acid 3C protease cleavable linker sequence. The mutant CR-2 domain carries W66A, D69A, Q71A, D73A mutations. SP = signal peptide. MBP = Mannose binding protein. d Schematic for the optimization of the inhibitory domain and linker length for Pro-VSV-GIN through the generation of GFP-expressing lentiviruses pseudotyped with WT or Pro-VSV-GIN in HEK-293T cells, with or without proteolytic digestion using 3C protease prior to transduction of HEK-293T cells and flow cytometry analysis. e, f GFP-expressing lentiviruses pseudotyped with WT VSV-GIN or Pro-VSV-GIN with different inhibitory domains and linker lengths were produced by HEK-293T cells and used to transduce HEK-293T cells, followed by flow cytometry analysis. e Representative flow plot of GFP signal in cells following viral infection. In the upper panels of (e), viruses were added onto cells in their native form. In the lower panel of (e), viruses were digested with 3C protease (20 U/ml) for 3 h on ice prior to cell transduction. Quantification of GFP+ cells are shown in (f). g Western blot analysis of VSV-GIN and FLAG expression before and after 3C protease digest of WT and Pro-VSV-GIN lentiviruses to confirm linker cleavage and removal of inhibitory capWe first utilized the lentiviral platform to determine the optimal cap and linker length to mediate viral inhibition. We used WT VSV-GIN (Fig. 1b) and Pro-VSV-G,IN which contains the cap-linker modification (Fig. 1c). For the Pro-VSV-GIN constructs, we selected the cysteine-rich 2 (CR-2) domain (~ 4.5 kDa) from the LDL-R as an inhibitory domain (cap), and cloned it to the N’-terminus of the mature VSV-GIN after the endogenous signal peptide and connected it to VSV-GIN by a 12- or 20-amino acid (aa) protease cleavable linker (Fig. 1c).19 We also generated a CR-2 cap which harbors four different point mutations at the amino acids that have been shown to be critical for mediating VSV-G’s binding to the CR-2 domain (Fig. 1c).19 Additionally, we investigated the use of a mannose binding protein (MBP) (~43 kDa) cap, in place of the CR-2 domain, and connected it to VSV-GIN with a 20 aa linker (Fig. 1c). Finally, we used a 3C protease cleavable linker to avoid inadvertent activation of Pro-VSV-GIN pseudotyped viruses by proteases endogenous to the HEK-293T cells used for viral production and screening.39To probe the infectious activity of Pro-VSV-GIN pseudotyped viruses, we transduced HEK-293T cells with WT VSV-GIN or Pro-VSV-GIN pseudotyped, GFP-encoding lentiviruses prior to or directly following 3C protease digest and analyzed the number of infected cells by GFP signal (Fig. 1d). Compared to lentivirus harboring WT VSV-GIN, the infectious activity of Pro-VSV-GIN pseudotyped viruses, prior to 3C protease digest, was completely abolished with all constructs, except Pro-VSV-GIN harboring the mutated CR-2 cap (Fig. 1e, upper panels). We next treated viruses with 3C protease and examined the degree of restoration of lentivirus infectivity (Fig. 1e, lower panels). While only a marginal rescue of the viral titer was observed with the removal of the MBP or mutant CR-2 caps, the removal of the CR-2 cap resulted in complete restoration of lentivirus infection, with the 20 aa linker length proving to be robust beyond that of the WT VSV-GIN control (Fig. 1e–g). The Pro-VSV-GIN pseudotyped lentivirus with a CR-2 cap and 20 aa linker showed a near 700-fold activation compared to its undigested counterpart, demonstrating the versatility of the cap-linker strategy in VSV-G inhibition and protease-dependent activation. The higher signal of the Pro-VSV-GIN pseudotyped lentivirus (CR-2 cap, 20 aa linker) may be attributed to the inability of the virus to infect host cells during viral production, resulting in enhanced viral accumulation in the supernatant.Optimization of a cancer-specific virus using viral spike protein cognate binding domainsWith the success of the CR-2 domain as an effective cap in our engineered Pro-VSV-GIN (Fig. 1), we next compared the efficacy of the additional six CR domains of the LDL-R for their ability to inhibit VSV-GIN function. We included the New Jersey (NJ) serotype of VSV-G (VSV-GNJ), which is more pathogenic than the Indiana serotype,40,41 and investigated the Maraba virus glycoprotein which uses the LDL-R for cell entry and is more resistant to serum neutralization.42We covalently linked the CR-1 through CR-7 domains of the LDL-R to the N’-terminus of VSV-GIN, VSV-GNJ, or Maraba-G, using the 20 aa 3 C protease cleavable linker (Fig. 2a). We transduced HEK-293T cells and quantified viral titer by assessing GFP+ cells 3 days post infection. The CR-2 domain provided the greatest fold of activation for Pro-VSV-GIN and Pro-VSV-GNJ (Fig. 2b–e; supplementary Figs. 1 and 2); however, the higher basal level of activity of Pro-VSV-GNJ in its native form reduced the extent of activation of the virus compared to the Indiana serotype (Fig. 2c, e). Interestingly, unlike both VSV-G serotypes, Pro-Maraba-G pseudotyped lentivirus demonstrated the greatest fold of activation when utilizing the CR-4 domain as a cap, although to a much lesser extent than Pro-VSV-GIN (Fig. 2f, g; supplementary Fig. 3).Fig. 2Full size imageOptimization of the cap-linker strategy for VSV-GIN, VSV-GNJ, and Maraba-G. a Schematic of the investigation testing the use of LDL-R cysteine rich (CR) domains 1–7 as inhibitory domains for Pro-VSV-GIN, Pro-VSV-GNJ, or Pro-Maraba-G constructs. b, d, f Flow cytometry analysis of HEK-293T cells infected with GFP-expressing lentiviruses pseudotyped with (b) WT VSV-GIN and Pro-VSV-GIN, (d) WT VSV-GNJ and Pro-VSV-GNJ, and (f) WT Maraba-G and Pro-Maraba-G. c, e, g Fold of activation as calculated by analyzing GFP+ HEK-293T cells 72 h post infection, and dividing the viral titer of each strain with 3C protease digested vs undigested viruses for (c) VSV-GIN, (e) VSV-GNJ (g) Maraba-G, n = 3. All data represent consistent findings in three independent experimentsThus, the CR-2 domain as an inhibitory cap connected to VSV-GIN with a 20 aa cleavable linker provides the cleanest signal-to-noise ratio across all investigated modified glycoproteins.Engineering of Pro-VSV-GIN pseudotyped lentivirus for tumor-specific protease activationAmong different proteases, the matrix metalloproteinase (MMP) family has been shown to be upregulated across different human cancers.39,43,44,45,46,47 Additionally, MMP activity is known to promote tumor development by aiding angiogenesis and metastasis.48 The abundance of MMPs across multiple tumor types makes them ideal for use as an activating signal for the liberation of Pro-VSV-GIN.39,43,44,45,46,47 Therefore, we engineered our 20-aa linker to contain a universal MMP cleavage site (Pro-VSV-GIN-MMP, Fig. 3a).Fig. 3Full size imageMMP-dependent activation of Pro-VSV-GIN-MMP. a Illustration of the design for Pro-VSV-GIN-MMP which harbors a CR-2 inhibitory domain and a 20 aa flexible, pan-MMP cleavable linker fused to the N’-terminus of VSV-GIN. b–e WT VSV-GIN and Pro-VSV-GIN-MMP containing plasmids were co-transfected with lentiviral helper plasmids to HEK-293T cells to generate lentiviruses coated with WT or Pro-VSV-GIN-MMP spike proteins, respectively. Cell supernatant containing lentivirus was used to transduce HEK-293T cells in 35 mm dishes. During viral production and transduction, indicated concentrations of broad-spectrum MMP inhibitor, marimastat, were added to cell culture medium. Additionally, once generated, Pro-VSV-GIN-MMP coated lentivirus was digested with 0.3 U/ml of MMP-7 for 1 h prior to infecting HEK-293T cells and analyzed 72 h following viral transduction by b Flow cytometry, n = 3, c fluorescent cell imaging, and d western blotting using viral supernatants for the presence of VSV-GIN and FLAG across sample conditions. e Representative flow plots and percentage of virally infected, GFP+ HEK-293T cellsThe incorporation of an MMP-cleavable linker resulted in a viral activity that was only marginally lower than the WT virus (Fig. 3b–e), which was due to HEK-293T cells producing MMPs. Addition of marimastat, a broad-spectrum MMP inhibitor that blocks MMPs 1, 2, 3, 7, 8, 9, and 14, resulted in a dose-dependent decrease in the viral titer of Pro-VSV-GIN-MMP lentivirus (Fig. 3b–e).49 Pro-VSV-GIN-MMP lentivirus pre-digested with recombinant MMP-7 prior to infection, resulted in a significant increase in the percent of virally infected cells compared to the marimastat-treated counterparts (Fig. 3b–e). Taken together, these data demonstrate the MMP-dependent activity of Pro-VSV-GIN-MMP in guiding viral entry into the cell.CRISPR knockout of MMPs in HEK-293T cells allows for efficient production of Pro-VSV-GIN pseudotyped lentivirus in its native formWe suspected that the remnant activity of Pro-VSV-GIN-MMP pseudotyped lentivirus was due to the expression of MMPs in HEK-293T cells not targeted by marimastat. RT-qPCR analysis of all MMP mRNAs show that MMPs 2, 11, 15, 16, and 17 are expressed at relatively high levels in HEK-293T cells (supplementary Fig. 4a), which prompted us to engineer HEK-293T cells using CRISPR/Cas9 technology to knockout all five MMPs (supplementary Fig. 4b, MMP-KO HEK-293T cells). Successful knockout of these genes was confirmed by Sanger sequencing of genomic DNA from single MMP-KO HEK-293T cell clones, and we selected the best clone C2-11 for subsequent study (supplementary Fig. 4b, c). We next performed a functional test to investigate the ability of our MMP-KO HEK-293 T (clone C2-11) to produce lentivirus (supplementary Fig. 4d) and showed that the MMP-KO HEK-293T cells were as efficient as WT HEK-293T cells in producing lentiviruses (supplementary Fig. 4e). The activity of Pro-VSV-GIN-MMP lentivirus produced in MMP-KO HEK-293T cells is abolished when used to infect MMP-KO HEK-293T cells and is completely rescued when recombinant MMP-7 is used to pre-digest the virus prior to infection of these cells (supplementary Fig. 4f, g). Thus, the MMP-KO HEK-293T cell line serves as an ideal platform to produce MMP-dependent viruses by preventing premature MMP cleavage on the viral particles.Systemic injection of Pro-VSV-GIN-MMP pseudotyped vesicular stomatitis virus exhibits a tumor-restricted tropismGiven the many attractive qualities of VSV, we replaced the VSV-GIN spike protein on an EGFP-expressing VSVIN (VSVIN-GFP) with Pro-VSV-GIN-MMP and designated this recombinant VSV as Pro-VSVIN-GFP (Fig. 4a, b). As anticipated, the structures of WT VSVIN-GFP and Pro-VSVIN-GFP under transmission electron microscopy (TEM) resembles the classic bullet-shape for VSV, confirming that the genetic modification made to Pro-VSVIN-GFP did not compromise the overall capsid structure of the virus (Fig. 4c–e). We next probed the conservation of the MMP-dependent activation of Pro-VSVIN-GFP and confirmed the ability of the CR-2 cap to be removed (Fig. 4f). Importantly, when added to MMP-KO HEK-293T cells, we found that Pro-VSVIN-GFP showed a marked decrease in viral spreading, as observed by flow cytometry at timepoints across 12 h, compared to cells transduced with WT VSVIN-GFP (Fig. 4g). However, MMP-7 pre-digestion of Pro-VSVIN-GFP restored the ability of the virus to infect cells at 6 h, although viral spreading was slower at 12 h compared to WT VSVIN-GFP because the newly produced Pro-VSVIN-GFP was again capped (Fig. 4g).Fig. 4Full size imageEngineering VSV with Pro-VSV-GIN-MMP enhances safety and tumor specificity. a Schematic for the design of EGFP transgene expressing WT VSVIN (WT VSVIN-GFP). b Schematic for the design of an EGFP transgene expressing Pro-VSVIN (Pro-VSVIN-GFP) which possesses a Pro-VSV-GIN-MMP spike protein harboring an N’-terminal FLAG tag, and CR-2 inhibitory domain connected to VSV-GIN by a 20 aa flexible, MMP cleavable linker (MPLGLRA). c, d TEM at a 50,000x magnification of WT VSVIN-GFP (c) and Pro-VSVIN-GFP (d). e Quantification of TEM viral particle size analysis of WT VSVIN-GFP and Pro-VSVIN-GFP, n = 3. f Western blot detection of viral proteins (VSV-G, VSV-N, and VSV-M), as well as Pro-VSVIN-GFP FLAG tag from purified viruses. Recombinant MMP-7 protease was used to digest Pro-VSVIN-GFP, to remove the FLAG and CR-2 inhibitory domain prior to Western blot analysis (third lane), resulting in a downward shift of the VSV-GIN detection band. g Quantification by flow cytometry of the number of GFP+ cells at 0, 3, 6, and 12 h following viral transduction of C2-11 MMP-KO HEK-293T cells with VSVIN-GFP, Pro-VSVIN-GFP, and MMP-7 protease pre-digested Pro-VSVIN-GFP viruses at an MOI of 0.1, n = 3. Uninfected cells were used as controls. h Observed maximum tolerated dose (MTD) for WT VSVIN-GFP and Pro-VSVIN-GFP injected i.v. into C57BL/6J mice. Mice were observed for up to 2 months post injection. The MTD was determined to be the highest dose with no observed treatment related paralysis or death, n = 5. i–k NSG-c-Kit mice were grafted with 5 × 106 CACO2 human colon cancer cells subcutaneously on one flank. One month later, when tumor sizes reached 200 to 300 mm3, mice received i.v. injection of 1 × 107 pfu of either WT VSVIN-GFP or Pro-VSVIN-GFP (Day 0). Mice were sacrificed 8 days post treatment, followed by collection of tumor tissue and indicated organs which were examined for (j) green fluorescence by microscopy, and (k) the level of GFP mRNA by RT-qPCR, n = 5. Scale bars = 100 nm for (c) and (d), and 100 μm for (j). l Primary human brain microvascular endothelial cells (pBECs) were transduced with Pro-VSVIN-GFP, or Pro-VSVIN-GFP digested with 0.3 U/ml of MMP-7 for 1 h prior to infection at an MOI of 0.25. WT HEK-293T cells were also transduced with Pro-VSVIN-GFP at an MOI of 0.25. Cells were analyzed by flow cytometry for GFP expression at 8 h post infection, n = 3. m, n Representative flow cytometry analysis of MC38 (m) or 4T1 (n) transduced with Pro-VSVIN-GFP, or Pro-VSVIN-GFP digested with 0.3 U/ml of MMP-7 for 1 h prior to infection at an MOI of 0.25. Cells were analyzed by flow cytometry for GFP expression 8 h later. Uninfected cells were used as controlsWe examined the safety profile of the Pro-VSVIN-GFP compared to WT VSVIN-GFP. Using non-tumor bearing, immune competent C57BL/6J mice, we probed the safety profile by dose escalation to determine the maximum tolerated dose (MTD) of each virus. Pro-VSVIN-GFP proved to have a nearly 4-times higher MTD than WT VSVIN-GFP (Fig. 4h).We then performed a biodistribution analysis over eight days using CACO2 human colon tumor bearing NSG-c-KIT mice intravenously (i.v.) injected with 1 × 107 pfu of Pro-VSVIN-GFP. Pro-VSVIN-GFP preferentially expanded in the tumor environment (Fig. 4i–k, supplementary Fig. 5a). Compared to all normal tissues examined, 500 times the amount of GFP RNA was detected in the tumor of mice treated with Pro-VSVIN-GFP (Fig. 4i–k). The level of Pro-VSVIN-GFP quickly elevated in the tumor through the 8-day period, suggesting robust replication of the virus in the tumor environment (supplementary Fig. 5a). Additionally, when compared to systemic injection of WT VSVIN-GFP into CACO2 bearing NSG-c-Kit mice, Pro-VSVIN-GFP showed much greater accumulation in the tumor (Fig. 4j, k). VSV infects liver cells,50 and H/E staining of liver showed that WT VSVIN-GFP, but not Pro-VSVIN-GFP, caused marked damage (supplementary Fig. 5b). Pro-VSVIN-GFP readily infects HEK-293T cells, but not primary human brain microvascular endothelial cells (pBEC) (Fig. 4l). Pre-digestion of Pro-VSVIN-GFP with MMP rescued its defective infection in the non-cancerous endothelial cells (Fig. 4l). In addition, Pro-VSVIN-GFP also infects two mouse cancer cell lines, MC38 and 4T1, and its viral activity is further enhanced when the virus is pre-digested with MMP (Fig. 4m, n). Collectively, our data demonstrate that Pro-VSVIN-GFP is tumor-specific and possess an excellent safety profile.Systemic administration of Pro-VSVIN expressing a single-chain interleukin-12 transgene eradicates MC38 murine colorectal cancer tumorsInterleukin (IL)-12 (IL-12) possesses potent anti-tumor activity,51,52 but systemic delivery of recombinant IL-12 resulted in severe toxicity and even death in clinical trials.53,54,55 To avoid the toxicity of IL-12, but harness its anti-cancer properties, OVs have been investigated to improve the safety profile of the cytokine and express the cytokine locally in the tumor.17,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72 Given the improved safety profile and enhanced tumor-specificity observed with Pro-VSVIN-GFP compared to that of WT VSVIN-GFP (Fig. 4), we hypothesized that upon systemic injection, Pro-VSVIN engineered to express a single-chain (sc) IL-12 (sc-IL-12) would be safe and effective.Both WT VSVIN and Pro-VSVIN were engineered to carry scIL-12 as a payload (WT VSVIN-IL-12 and Pro-VSVIN-IL-12, respectively) (Fig. 5a, b). TEM of both showed the classic VSV bullet shape and size (~ 200 nm) (Fig. 5c, d; supplementary Fig. 6a–c). Viral GIN, N, and M protein expression, as well as the FLAG-tagged CR-2 domain connected to VSV-GIN in the Pro-VSVIN-IL-12 virus were confirmed by immunoblot analysis (Fig. 5e). Additionally, to assess scIL-12 secretion, BHK-21 cells were transduced with either WT VSVIN-IL-12 or Pro-VSVIN-IL-12, and cell supernatant was tested in immunoblot analysis with an IL-12 p35-specific antibody (Fig. 5e). Further, the plaque diameter formed by WT VSVIN-IL-12 and Pro-VSVIN-IL-12 was decreased compared to their GFP expressing counterparts, indicating slower replication kinetics due to the IL-12 payload, which is larger in size than GFP (supplementary Fig. 6d).Fig. 5Full size imageSystemic injection of Pro-VSVIN-IL-12 eradicates tumors. a Schematic for scIL-12 expressing WT VSVIN (WT VSVIN-IL-12). b Schematic for the design of scIL-12 expressing Pro-VSVIN (Pro-VSVIN-IL-12). c, d TEM at a 50,000x magnification of (c) WT VSVIN-IL-12 and (d) Pro-VSVIN-IL-12. e Western blot detection of viral proteins (VSV-G, VSV-N, and VSV-M), and Pro-VSVIN-GFP FLAG tag from purified viruses. Detection of scIL-12 performed by harvesting viral supernatant from infected BHK-21 cells. f Increasing doses of WT VSVIN-IL-12 and Pro-VSVIN-IL-12 were injected i.v. into C57BL/6J mice. Mice were observed for up to 15 days post injection. MTDs were determined to be the dose where there was no treatment-related paralysis or death, n = 3. g, h NSG-c-Kit mice were grafted with 5 × 106 CACO2 human colon cancer cells subcutaneously on one flank. One month later, when tumor sizes reached 200 to 300 mm3, mice received i.v. injection of 1 × 107 pfu of either WT VSVIN-IL-12 or Pro-VSVIN-IL-12 (Day 0). Mice were sacrificed 8 days post treatment, followed by collection of tumor tissue and indicated organs which were examined for the level of scIL-12 mRNA by RT-qPCR (h), n = 5. i, j C57BL/6J mice were injected i.v. with 200 μl of PBS, 1 × 107 pfu of WT VSVIN-IL-12, or 1 × 107 pfu of Pro-VSVIN-IL-12 twice, three days apart. Mouse serum was collected one day following the last injection and analyzed for IL-12p70 (I) and IFN-γ (J) by multiplex ELISA. k–r 5 × 105 MC38 cells were grafted subcutaneously to one flank of C57BL/6J mice. On days 7, 10, 13, and 16 post tumor graft, 5 × 106 pfu of WT VSVIN-GFP, 1.875 × 107 pfu of Pro-VSVIN-GFP, 5 × 106 pfu of WT VSVIN-IL-12, 5 × 107 pfu of Pro-VSVIN-IL-12, or PBS were injected i.v., as indicated by the pink arrows. Average tumor volumes (k), and individual tumor volumes over time for mice treated with PBS (l), WT VSVIN-GFP (m), Pro-VSVIN-GFP (n), WT VSVIN-IL-12 (o), and Pro-VSVIN-IL-12 (p). q Respective cure rates of each treatment determined by the number of mice achieving tumor eradication. r Kaplan–Meir survival curves of mice treated with recombinant viruses, n = 10–11. s, t MC38 tumor-bearing mice were treated with i.v. injected Pro-VSVIN-IL-12 to achieve tumor eradication. Around 1 to 1.5 months after the last treatment, mice were challenged with a subcutaneous injection of 5 × 105 MC38 cells, and tumor growth was followed for 30 days (s), or sacrificed and 1 × 106 splenocytes were harvested and co-cultured with 5 × 104 MC38 tumor cells and supernatants were assayed 48 h later for IFN-γ and compared against splenocytes isolated from naïve control mice (t)To interrogate the safety profile of the IL-12 containing viruses, a dose escalation study in C57BL/6J hosts was performed. Pro-VSVIN-IL-12 is safe at a dose 10-fold higher than WT VSVIN-IL-12, since no treatment related paralysis, or death was observed following systemic injection of Pro-VSVIN-IL-12 at titers up to 2 × 108 pfu per injection, compared to 2 × 107 pfu for WT VSVIN-IL-12 (Fig. 5f). A biodistribution study over eight days using CACO2 tumor bearing NSG-c-KIT mice showed the preference of Pro-VSVIN-IL-12 to accumulate in tumor, compared to that of WT VSVIN-IL-12, which also accumulated in liver (Fig. 5g, h). When injected i.v. to immune competent C57BL/6J mice, WT VSVIN-IL-12 led to significant elevation in IL-12 and IFN-γ protein levels in the serum (Fig. 5i, j). This effect was largely abolished in the case of Pro-VSVIN-IL-12 (Fig. 5i, j). It is known that IL-12 is toxic when systemically exposed, and the reduction in IL-12 and IFN-γ exposure in the serum by Pro-VSVIN-IL-12 agrees with its enhanced safety profile.Next, to explore the potential therapeutic benefit of Pro-VSVIN-IL-12 as a cancer treatment, we established a syngeneic, ectopic model of colon adenocarcinoma by grafting MC38 cells subcutaneously to the flanks of C57BL/6J mice (Fig. 5k–r). The tumors were palpable 7 days following the tumor graft, at which time the mice were treated with four doses every three days by i.v. injection of PBS or one-quarter of the determined MTD of the transgenic viruses. While little or no tumor control was observed for mice treated with PBS (Fig. 5l, supplementary Fig. 6e), WT VSVIN-GFP (Fig. 5m, supplementary Fig. 6f), as well as Pro-VSVIN-GFP (Fig. 5n, supplementary Fig. 6g), the addition of the potent immune-stimulating scIL-12 in both WT VSVIN-IL-12 (Fig. 5o, supplementary Fig. 6h) and Pro-VSVIN-IL-12 (Fig. 5p, supplementary Fig. 6i) resulted in delayed tumor growth. However, unlike WT VSVIN-IL-12, Pro-VSVIN-IL-12 proved to be the superior treatment, eradicating tumors in nine of ten mice, and significantly increasing the survival of mice compared to all other treatment groups (Fig. 5p–r). Importantly, no treatment-related deaths or signs of toxicity were observed in the Pro-VSVIN-IL-12 treated cohort. Additionally, to test if treatment with Pro-VSVIN-IL-12 conferred long-term immune memory, Pro-VSVIN-IL-12 treated mice were rechallenged with MC38 tumor cells 40–60 days after the initial tumor graft. None of the 9 mice grew tumors following MC38 subcutaneous rechallenge (Fig. 5s). Immune memory was further confirmed by isolating splenocytes from mice with tumors eradicated following treatment with Pro-VSVIN-IL-12, co-culturing with MC38 tumor cells, and analyzing for the expression of IFN-γ (Fig. 5t).To explore the scope of anti-tumor efficacy of Pro-VSVIN-IL-12, we treated tumor-bearing mice with four doses of PBS or Pro-VSVIN-IL-12 by intraperitoneal (i.p.), instead of i.v., injection every three days (supplementary Fig. 6j–o). Treatment with Pro-VSVIN-IL-12 resulted in a significant decrease in the growth rate of tumors compared to PBS treated controls, and eradicated tumors in seven of ten mice (supplementary Fig. 6j–l). Likewise, Pro-VSVIN-IL-12 significantly extended the lives of mice compared to PBS treated controls (supplementary Fig. 6m). As observed with i.v. administration of Pro-VSVIN-IL-12, i.p. administration likewise conferred immune memory (supplementary Fig. 6p).Finally, to test the scope of Pro-VSVIN-IL-12 efficacy, 4T1 breast cancer cells were grafted into BALB/c mice via tail vein injection to allow for the formation of lung metastasis (Fig. 6a). 5 days later, at which time the lung metastatic nodules were already established (Fig. 6b, left panel), mice were treated with Pro-VSVIN-IL-12 or PBS. Results showed that systemic Pro-VSVIN-IL-12 injection significantly reduced metastatic lung nodules with no signs of toxicity (Fig. 6c, d). Further, H/E staining showed that the remaining metastatic nodules in Pro-VSVIN-IL-12 treated group were significantly smaller in sizes compared to the control group (Fig. 6b, middle and right panels, e).Fig. 6Full size imageSystemic Pro-VSVIN-IL-12 is effective against metastatic breast cancer. a–e 3 × 105 4T1 cells were grafted to lungs of BALB/c mice by tail vein injection (Day 0). 5 days after tumor cell injection, mouse lungs were fixed and embedded in paraffin, followed by sectioning, H/E staining, and analyzed to confirm the presence of tumor nodules prior to treatment with virus (b, left panel). On days 5, 9, and 13 post tumor cell injection, 5 × 107 pfu of Pro-VSVIN-IL-12 or PBS (control) were injected i.v. to the treatment groups (a). On day 15, mice were sacrificed and their lungs were analyzed for the presence of tumor nodules by H/E staining on sectioned lungs (b, right two panels), visual examination and quantification of metastatic nodules on the surface of the lungs (c, d), and quantification on the sizes of metastatic nodules on H/E-stained lung sections (e)Systemic Pro-VSVIN-IL-12 activates tumor-infiltrating T cellsTo elucidate the mechanism of action of Pro-VSVIN-IL-12, we investigated the tumor-infiltrating lymphocyte (TIL) populations impacted by systemic treatment with Pro-VSVIN-IL-12. To do so, we implanted MC38 tumor cells into the flanks of mice, and treated mice with three doses of PBS, WT VSVIN-GFP, Pro-VSVIN-GFP, WT VSVIN-IL-12, or Pro-VSVIN-IL-12 (Fig. 7a). One day after the last injection, TILs were isolated by Ficoll-Paque density centrifugation and analyzed by single-cell RNA and TCR sequencing, as well as flow cytometry (Fig. 7b–j, supplementary Fig. 7). Across all treatment conditions, we observed a remodeling of the CD4+ T cell compartment, with the greatest change observed in Treg and Th1 cell populations (Fig. 7c, h–j). Specifically, treatment with Pro-VSVIN-IL-12 resulted in the greatest increase in the expression of Th1-associated genes, and a downregulation of genes associated with Tregs (Fig. 7d). The shift towards Th1 signature genes was a result of the increased proportion of Th1 cells, and decreased proportion of Tregs, in the tumor (supplementary Fig. 8a, b). Th1 cells within MC38 tumors of the Pro-VSVIN-IL-12 treated mice further show enhanced expression of IFN-γ and T-bet (supplementary Fig. 8c, d). Consistent with prior literature on the effects of IL-12 on Treg stability and suppressive function,73,74 Tregs in Pro-VSVIN-IL-12 treated tumors showed increased levels of T-bet (supplementary Fig. 8e, f). These Treg cells also show markedly reduced level of CD103 (encoded by Itgae) expression, suggesting a deficit in Treg migration into the tumor that could also contribute to the observed changes in Treg number (supplementary Fig. 8g). The reduction in Tregs and increase in Th1 cells following Pro-VSVIN-IL-12 treatment was further validated by flow cytometry (Fig. 7h–j).Fig. 7Full size imageSystemic treatment with Pro-VSVIN-IL-12 depletes Tregs and bolsters Th1s in the TME. a Schematic for the analysis of tumor infiltrating lymphocytes (TILs) by flow cytometry or single-cell RNA or TCR sequencing. 5 × 105 MC38 cells were grafted subcutaneously to the flanks of C57BL/6J mice. When tumors reached a size of 200–300 mm3, mice received i.v. injection of 5 × 106 pfu of WT VSVIN-GFP, 1.875 × 107 pfu of Pro-VSVIN-GFP, 5 × 106 pfu of WT VSVIN-IL-12, and 5 × 107 pfu of Pro-VSVIN-IL-12, or PBS on days 0. Two additional treatments were administered on days 3 and 6, followed by tumor dissociation, and analysis of TILs by flow cytometry or sequencing on day 7. b Merged UMAP of cell types captured by sequencing. Cell type assignments were determined based on expression of markers highlighted in supplementary Fig. 7. c CD4+ T cell subsets (CD4+ T conventional cells plus CD4+ Tregs) were reclustered and analyzed for expression of CD4 T cell subset markers. FeaturePlots showing module scores for each Treg subset. Module scores were created using the R package Seurat (see “Materials and methods” for details). d Heatmap showing the module scores for Treg and Th1 CD4+ T cell subsets across each experimental condition. e Boxplot shows the distribution of clonal frequency (number of matching clones for each unique TCR) for CD8+ T cells from 10x Genomics TCR (VDJ) sequencing. In each plot, the box represents the interquartile range, the horizontal line inside the box indicates the median. f Proportion of expanded (0.01