A Virus-Infected, Reprogrammed Somatic Cell–Derived Tumor Cell (VIReST) Vaccination Regime Can Prevent Initiation and Progression of Pancreatic Cancer

The immune system has long been recognized as critical for maintaining control over nascent tumor cells. Despite this, tumors have an inherent ability to overcome immune constraints and develop into malignant disease (1). Advances in our understanding of the evolutionary mechanisms by which tumors escape control have resulted in groundbreaking new therapeutics, including checkpoint blockade and CAR-T therapies that are redefining survival rates for many patients with cancer. However, there has been little success in applying this understanding to the generation of effective therapeutics for some of the most treatment-resistant tumors such as pancreatic ductal adenocarcinoma (PDAC), which continues to have extremely high mortality rates (2). Our improved understanding of cancer progression in recent years suggests a long period of clinical latency associated with solid tumors (3); thus, a more meaningful approach to control may be to develop prophylactic strategies that can be implemented in this window of opportunity to provoke the immune system into maintaining strict control of the tumor, preventing progression to incurable disease. Theoretically, prophylactic vaccination should be more effective than therapeutic control as nascent tumors lack the complex immune tolerance mechanisms of established tumors; however, although preventative strategies have proven effective for cancers in which the etiologic agents are known (4, 5), for most nonviral cancers, it remains ineffective, largely due to the lack of appropriate tumor antigens and an effective approach to induce robust antitumor immunity against the antigens. Vaccination strategies relying on presentation of known tumor-associated antigens (TAA) have shown little success due to inappropriate choices of TAAs and the inability of these regimes to overcome tumor-induced immune suppression. Autologous whole cancer cell vaccines are more successful in their targeting of tumor cells as these vaccines can present a large number of relevant antigens in the appropriate HLA context, but requirement for sufficient viable biopsy material limits application to few therapeutic and no prophylactic settings. An allogeneic whole-cell vaccine approach offers more scope for preventative use, but does not account for the extensive heterogeneity of tumors and may confer selective pressures that promote tumor divergence (6). These limitations demonstrate a necessity for far more personalized approaches that account for genetic heterogeneity between individuals and are suitably immunogenic to promote robust antitumor immunity for long-lasting prevention of tumor escape.

To address the challenge of ensuring appropriate immunogenicity of prophylactic vaccination, we investigated the use of replicating oncolytic viruses (OV). OV are a class of naturally occurring or genetically modified viruses that selectively replicate in and kill tumor cells, and it is widely recognized that tumor-specific T-cell activation plays a crucial role in OV-mediated therapeutic efficacy. Infected cell vaccines, whereby cancer cells are preinfected with replicating tumor tropic virus prior to delivery as a vaccine, have been shown to provoke high levels of antitumor immunity that was not achieved when cells were delivered without infection (7), suggesting that replicating viruses can provide relevant danger signals required for vaccine immunogenicity. We have recently demonstrated that therapeutic prime–boost delivery of Adenovirus (AdV) and Vaccinia virus (VV), sequentially, is an effective regime for elimination of PDAC in vivo (8). The underlying mechanisms responsible for the efficacy of specific sequential administration remain unclear, but we show here that both viruses can trigger complementary aspects of immunologic cell death that likely promote the strong immunogenicity of the regime. Here, we show that infection of autologous tumor cells with replicating AdV or VV, and delivery of these cells prophylactically in a prime–boost manner using AdV-infected cells as a prime and VV-infected cells as a boost, can induce significant antitumor immune responses and delay disease progression in a transgenic mouse model of PDAC, demonstrating the intrinsic value of OV as vaccine adjuvants. However, as discussed, requirement for biopsy material limits the scope of autologous vaccination regimes to therapeutic and not prophylactic use.

The advent of stem cell technology, especially induced pluripotent stem cells (iPSC), offers novel alternatives for the development of whole-cell–based preventative vaccination protocols, but therapeutic vaccination using iPSCs has historically failed to elicit protective advantages (9). Although a recent report of the use of undifferentiated iPSCs to induce antitumor immune responses in murine cancer models validates the continued exploration of iPSC-based vaccination regimes (10), the associated antitumor efficacy has remained disappointing. We hypothesize that the tumor antigen profile presented by vaccination with undifferentiated iPSCs is insufficient to confer robust protective antitumor immunity.

Here, we sought to utilize iPSC technology to generate whole tumor cell vaccines in which the unique characteristics of an individual's tumor were preserved. The development of lineage-specific cells from somatic cells can now be achieved consistently in vitro (11), and transformation of iPS-differentiated pancreatic cells into tumor cells by lentiviral transduction of KRas^G12D and p53^R172H, driver mutations common to a spectrum of cancers, has been reported (12), although this regimen demonstrated a low transduction efficiency and random insertion of the driver mutation genes. We have developed an alternative technological platform for derivation of pancreatic cancer cells from healthy somatic cells by in situ gene editing of iPSCs using stable knock-in of inactive KRas^G12D and p53^R172H prior to lineage differentiation and transformation. The derived murine tumor cells displayed high antigenic similarities to PDAC cell lines derived from the related KPC or KP transgenic mouse model of pancreatic cancer, demonstrating that this process models neoantigen accrual, based on the genetic and epigenetic profile of autologous stem cells, during the earliest stages of tumorigenesis. These cells can supply a large number of relevant neoantigens and TAAs for induction of specific antitumor immunity. As such we can generate an effective alternative to autologous whole-cell vaccines that is antigenically compatible with each patient, but suitable for provision in a preventative setting that can, via immune surveillance mechanisms, detect initiation of malignancies and prevent their development.

By combining OV with the iPSC technology platform, we created a Virus-Infected Reprogrammed Somatic cell-derived Tumor cell (VIReST) vaccination regime. Prophylactic prime vaccination with AdV-infected reprogrammed iPSC tumor cells followed by booster vaccination with the VV-infected cells was extremely effective for prevention of PDAC development and progression in a robust transgenic model that faithfully recapitulates the disease progression and the associated complex immune-suppressive environment. This platform provides a realistic prospect for cancer prevention in at-risk individuals, which may be the most constructive approach to affect survival statistics for this disease.

Tail-tip fibroblasts from LSL-KRas^G12D/+; Trp53^R172H/+; Pdx-1-Cre (KPC) or wild-type (WT) littermate mice and murine embryonic fibroblasts (MEF) from E13.5 WT mice were cultured using DMEM supplemented with 10% FBS. Embryogenic stem cells or iPSCs were cultured using either mES medium (DMEM supplemented with 15% FBS; GEMINI), leukemia inhibitory factor (GEMINI; 1,000 U/mL), β-mercaptoethanol (0.45 μmol/L), 1× nonessential amino acids, 1× Glutamax, 1× sodium pyruvate, or KSR medium (mES medium with 15% Knockout Serum Replacement replacing 15% FBS). MEF feeder cells were inactivated by mitomycin C (MMC) treatment (Melonepharma; 20 mg/mL for 2.5 hours).

The murine PDAC cell line DT6606 was established from LSL-KRas^G12D/+; Pdx-1-Cre mice that had developed PDAC. The PDAC cell lines TB11381, TB32043, and TB32047 were cultured from LSL-KRas^G12D/+; Trp53^R172H/+; Pdx-Cre mice that had developed PDAC (13). These were kindly provided by David Tuveson (Cancer Research UK Cambridge Institute, now at Cold Spring Harbor Laboratory) and maintained in DMEM supplemented with 5% FBS. CV1 (African monkey kidney) cells and JH293 cells were obtained from the ATCC and maintained in DMEM containing 5% FBS. WT-KP, KPC, and pancreatic progenitor like cells (PPLC) were maintained in DMEM containing 10% FBS. All cell lines were routinely tested for the presence of Mycoplasma.

The thymidine kinase (TK)–deleted Lister strain VV (VVL15) was described previously (14). VVL15 was further modified by XhoI/EcoRI-mediated removal of the LacZ open-reading frame from the VV TK shuttle vector pSC65 (GenBank: HC193923.1), which was replaced with red fluorescent protein (RFP) derived by NheI/AflII digestion of pCMV-dsRED-Express (Clontech). The virus was produced as described previously (15). Ad-Cre nonreplicative virus was purchased from Vector Biolabs and propagated in our laboratory. WT Adenovirus serotype 5 (Ad5) was described previously (8).

Inducing tail-tip fibroblasts to pluripotency was carried out according to the protocol of Dr. Duanqing Pei's laboratory (16). Retroviral vectors (pMX) containing the murine cDNAs of Oct4, Sox2, and Klf4 were purchased from Addgene. These plasmids were transfected into PlatE cells (provided by Dr. Jiekai Chen, Guangzhou Institute of Biomedicine and Health, Chinese Academy of Science, cultured using DMEM supplemented with 10% FBS) using calcium phosphate transfection for packaging before WT male mouse tail-tip fibroblasts and KPC male mouse tail-tip fibroblasts were infected with the retroviruses. After 48 hours, cells were cultured in iCD1 media as detailed previously (16), and this day was indicated as d0. At d10, high-quality iPSC clones were picked under a microscope according to their morphology and passaged.

pBlueScript-PGK-LSL-K-Ras^G12D and pBlueScript-PGK-LSL-p53^R172H plasmids (kind gifts from Dr. David Tuveson, Cold Spring Harbor Laboratory) containing LoxP-flanked transcription stop signals were used to modify WT iPSCs. We modified the puromycin-resistant gene of pBlueScript-PGK-LSL-P53^R172H to neomycin resistance. For cell modification, the KRas plasmid was NotI-linearized and electroporated into iPSCs. The cells were selected using puromycin (2 μg/mL) and genotyped by PCR using primers flanking the LoxP-Stop-LoxP (LSL) cassette and the genome of right homologous arm. KRas_forward (F): 5′-CCATGGCTTGAGTAAGTCTGC-3′; KRas_reverse (R): 5′-TGACTGCTCTCTTTC-3′ (PCR band 4,100 bp); the mutant site was amplified and sequenced using the following primers: amplification primer: F: 5′-GCCTGCTGAAAATGACTGAGTAT-3′; R: 5′-CTCTATCGTAGGGTCATAC-3′ (PCR band 180 bp). Sequencing primer: 5′-CTCTATCGTAGGGTCATAC-3′. The PCR fragment was sequenced in GENWIZ, Inc. The p53 plasmid was linearized using NotI and electroporated into the KRas^G12D/+ iPSCs with TALENS targeting the p53 homologous site. TALENS target sequences were 5′-GGAAGGCCAGCCCTGGTTG-3′ and 5′-GTAAGAAAATGTTGGCTGGG-3′. The cells were selected by G418 (400 μg/mL) and genotyped by PCR using primers flanking the genome of the left homologous arm and the LSL cassette. P53_F: 5′-CACGCTTCTCCGAAGACTGG-3′; P53_R: 5′-TATGCTATACGA AGTTATGTCG-3′ (PCR band 1,900 bp). The mutant site was amplified and sequenced using the following primers: amplification primer: F: 5′-TCTCTTCCAGT ACTCTCCTCC-3′; R: 5′-AATTACAGACCTCGGGTGGCT-3′ (PCR band 500 bp); sequencing primer: 5′-AAGCTATTCTGCCAGCTGGCG-3′. The PCR fragment was sequenced at GENWIZ, lnc.

iPSCs were cultured in mES medium on an MMC-treated MEF feeder layer. For differentiation, iPSCs were cultured for two passages without feeder cells, dissociated into single cells using 0.25% trypsin before culture in Medium 1 (Med1; 25% Nutrient mixture F12, 75% Iscove modified Dulbecco's media, 1× N-2 supplement, 1× Glutamax, 0.05% BSA, and 0.45 μmol/L β-mercaptoethanol) at 5 × 10⁵ cells/mL in low adhesion plates to form embryoid bodies (EB). Forty-eight hours later, EBs were cultured using 0.1% gelatin-coated plates in Med1 supplemented with 100 ng/mL Activin A and 3 μmol/L Chir99021 for 24 hours. Medium was replaced with Med1 supplemented with 100 ng/mL Activin A. Cells were cultured for 48 hours to form definitive endoderm (DE). DE were cultured in Med2 (DMEM, 1× B27 supplement, and 1× Glutamax) with 2 μmol/L retinoic acid (Sigma); 1 μmol/L A83-01 (Stemgent); 0.5 μmol/L Cyclopamine (Tocris); 50 ng/mL Noggin (R&D Systems); and 280 μmol/L Vitamin C (Vc; Sigma) for 48 hours. Medium was replaced with Med2 containing 1 μmol/L A83-01; 0.5 μmol/L cyclopamine; 50 ng/mL Noggin; and 280 μmol/L Vc (Sigma) for 48 hours to induce differentiation into PPLC (17, 18).

The modified WT-KP PPLCs were infected with nonreplicating Ad5-Cre (50 pfu/cell) to remove the LSL cassette at day 9 and cultured in DMEM containing 10% FBS for 10 passages to facilitate their transformation to pancreatic tumor cells. The KPC PPLCs were directly cultured in DMEM containing 10% FBS for 10 passages to allow natural transformation to tumor cells. p53 and KRas were genotyped at passage 3 by PCR using primers to detect a WT band of 290 bp (P53) and 285 bp (KRas) or the mutant band of 330 bp (P53) and 325 bp (KRas). P53g_F-5′-AGCCTGCCTAGCTTCCTCAGG-3′; P53g_R-5′-CTTGGAGACATAGCCACACTG-3′. KRasg_F-5′- GGGTAGGTGTTGGGATAGCTG-3′; KRasg_R-5′-TCCGAATTCAGTGACTACAGATGTACAGAG-3′.

Additional methods can be found in the Supplementary Methods.

It has been demonstrated that OV infection of whole tumor cell vaccines significantly enhances their immunogenicity (7), and we and others have shown that the administration of two antigenically distinct viruses sequentially provides superior therapeutic control over tumor growth compared with use of one virus alone (8, 19). To determine whether OV infection of induced tumor cells provides a beneficial adjuvant in a prophylactic setting, we investigated the underlying mechanisms for the improved efficacy associated with in situ administration of AdV and VV.

As defined by Kroemer and colleagues, features of immunogenic cell death (ICD), known to be key for effective induction of antitumor immunity, comprise HMGB1 release, calreticulin (CRT) exposure, and ATP release (20). Multiple studies have confirmed that OVs are potent inducers of ICD (21); hence, we investigated the induction of each of these features upon infection with AdV and VV. Encouragingly, we found that both viruses were able to induce ATP release 48 hours after infection of pancreatic tumor cells (Fig. 1A), although only AdV was able to elevate these levels significantly compared with mock treatment. Only VVL15-RFP was able to induce HMGB1 release and CRT exposure on pancreatic tumor DT6606 cells (Fig. 1B and C). Thus, it appears that both viruses can trigger certain features of ICD, which may complement each other to improve immunogenicity when viruses are used sequentially as opposed to individually.

Next, we tested whether infecting autologous cells and delivering them using a Virus-Infected Cancer Cell Vaccination (VICCV) regime could result in tumor-specific immunity. KPC transgenic mice were immunized with OV-infected DT6606 cells at weeks 0 and 4. Ex vivo splenocyte reactivation on exposure to growth-arrested DT6606 (Fig. 1D) or unrelated CMT93 cells (Fig. 1E) demonstrated that only a VICCV regime comprising delivery of AdV-infected tumor cells at week 0 followed by VV-infected tumor cells delivered at week 4 induced significant IFNγ responses to syngeneic, but not unrelated tumor cells. As proof of principle of the efficacy of this regime, we vaccinated immunocompetent C57/Bl6 mice using the VICCV regime described above. Subsequent challenge 2 weeks post boost with DT6606 cells indicated that delivering Ad5-infected tumor cells as a prime and VV-infected tumor cells as a boost was the best regime to induce tumor-specific immunity and prevent challenge tumor growth (Fig. 1F). We then investigated the use of syngeneic tumor cells in the VICCV regime to protect against disease progression in a more complex KPC transgenic mouse model. This demonstrated that vaccination of KPC mice prior to disease emergence using AdV-infected TB11381 tumor cells, derived from the KPC mouse, followed by booster vaccination with VV-infected TB11381 tumor cells, was able to significantly delay disease progression (Fig. 1G), demonstrating the principle of using preinfected tumor cells, delivered in a prime–boost manner, as a prophylactic approach for anticancer vaccination.

For development of a prophylactic vaccine strategy, it is necessary to generate antigenically compatible, lineage-specific tumor cells using healthy somatic cells derived from an individual. We first confirmed that iPSCs containing inactive tumor driver mutations (KRas^G12D and p53^R172H) remained pluripotent. Fibroblasts from KPC transgenic mice with endogenous, inactivated KRas^G12D and p53^R172H mutations were induced following the strategy illustrated in Fig. 2A by infection with retroviruses expressing Oct4, Sox2, and Klf4 reprogramming factors (16, 22) to create KPC iPSCs (Supplementary Table S1). Their pluripotency was confirmed using colony formation assays (Fig. 2B) and detection of iPSC markers by qPCR, which demonstrated reactivation of endogenous Oct4, Sox2, and Klf4, with the concurrent loss of expression of exo-reprogramming factors (Fig. 2C).

To create a platform that can be applied prophylactically in addition to therapeutically, induction of tumor cells from healthy individuals via iPSC reprogramming is required. We next isolated somatic cells from WT mice (129J/Bl6 KPC littermates) that do not contain endogenous genomic mutations. These cells were induced to pluripotency as described and demonstrated similar pheno- and genotypes in colony formation assays and qPCR, respectively (Fig. 2B and C).

Induced cells then underwent homologous recombination to introduce LSL-controlled common tumor driver mutations, creating KRas^G12D/p53^R172H heterozygote subclones (WT-KP iPSCs; Supplementary Table S1; Supplementary Fig. S1A–S1C). Neither endogenous nor in situ insertion of the two silent mutations affected pluripotency of the iPSCs as determined by expression of both Nanog and Oct4 (Fig. 2D), generation of viable chimeric mice (Supplementary Fig. S1D), and teratoma formation after implantation into immunocompromised mice (Fig. 2E). Next, WT iPSCs (derived from WT mice), KPC iPSCs (derived from KPC transgenic mice), or WT-KP iPSCs (derived from WT mice iPSCs engineered to contain in situ LSL-controlled KRas^G12D and p53^R172H mutations) were effectively differentiated into pancreatic progenitor cells using the protocol indicated in Supplementary Fig. S2A and lineage differentiation confirmed using immunofluorescence analysis (Fig. 3A) or qPCR analysis (Supplementary Fig. S2B) of standard differentiation markers.

KRas^G12D and p53^R172H mutations were activated spontaneously in differentiated KPC pancreatic progenitor cells due to pancreas-specific expression of Pdx-1–activating Cre recombinase, resulting in transformed KPC PDAC cells (Supplementary Table S1). LSL-controlled KRas^G12D/+ and p53^R172H/+ mutations in WT-KP pancreatic progenitor cells were activated by infection with nonreplicating Ad5 vector expressing Cre (AdCre) to create KP-AC (PDAC) cells (Supplementary Table S1; Supplementary Fig. S2C). KPC and KP-AC cells were tumorigenic when inoculated into the flanks of nude (Fig. 3B) or immunocompetent syngeneic mice (Fig. 3C), producing characteristically stroma-rich tumors with similar cell morphology to those resulting from implantation of DT6606 or TB11381 cells derived from the KC or KPC PDAC transgenic mouse models (Fig. 3D; Supplementary Table S1).

Most importantly, transcriptome analysis of reprogrammed tumor cell lines demonstrated highly similar gene expression profiles to transgenic mouse-derived tumor cell lines (Fig. 3E). Concordance between transformed and nontransformed iPSC cells (KP-AC or KPC vs. WT-iPSCs or WT-KP iPSCs) was low, and gene expression concordance never exceeded 30%, suggesting that untransformed iPSCs were unsuitable as vaccination material as the antigen spectrum presented would not confer a suitable level of protection. Interestingly, the concordance between the KP-AC or KPC tumor cells and the DT6606 tumor cell line, derived from KRAS driver mutation alone, was high. These results demonstrate that the generation of neoantigens upon tumorigenesis is not dependent on the driver mutation, but the epigenetics within each lineage. Furthermore, transformed lineage-specific cells provide the highest number of relevant neoantigens or TAAs and thus represent the most rational mechanism for creating effective whole cancer cell vaccination regimes.

Prophylaxis using autologous cells is possible using mouse models due to the inbred nature of mouse colonies preventing genetic discrepancies that would restrict their use in a human prophylactic setting. For translation to the clinical setting, cells that are matched to each individual are required. Moreover, these cells must be obtained prior to disease emergence to be of prophylactic use. Genotypically matched cells are currently only available via biopsy and thus limited to therapeutic use. In this regard, iPSC-derived tumor cells, which can be developed from at-risk individuals, may be an effective alternative given the high levels of similarities between transcriptomes of derived and endogenous tumor cells. To assess the potential of this approach, a VIReST immunization protocol was developed in which mice were immunized first with iPSC-derived tumor cells (KPC or KP-AC) preinfected with AdV, followed at an interval of 4 weeks with iPSC-derived tumor cells preinfected with VV (Fig. 4A). These virus-preinfected tumor cells were treated with MMC before immunization to prevent ongoing virus replication and tumor cell growth. Both AdV and VV were cytotoxic to (Supplementary Fig. S3A and S3B) and replicated efficiently in (Supplementary Fig. S3C–S3F) KPC and KP-AC tumor cell lines. Importantly, MMC treatment of cells inhibited ongoing viral replication and prevented tumor cell proliferation (Supplementary Fig. S3G and S3H), but viral proteins remained detectable in cell lysates 72 hours after infection (Supplementary Fig. S3I), demonstrating that this approach to vaccination is intrinsically safe with no compromise of vaccine immunogenicity.

Using the KPC model of PDAC, both KPC and KP-AC–based VIReST applied prior to tumor lesion development could enhance splenocyte IFNγ production after ex vivo stimulation with different tumor cell lines derived from this model (Fig. 4B). Elevated CD8⁺ and CD4⁺ T-cell infiltrates were shown in the pancreas of VIReST-immunized animals after, but not during, VIReST treatment (Fig. 4C and D; Supplementary Fig. S4A). Analysis of the spleen and lymph node compartments indicated that there were increased levels of activated CD8⁺ and CD4⁺ T cells at both 1 and 3 weeks post-VIReST (Fig. 4E; Supplementary Fig. S4B), with a progression from increased effector memory to central memory populations, that can rapidly expand following rechallenge, over time. Treatment was unable to alter the levels of regulatory T-cell (Treg) infiltration into the tumor (Supplementary Fig. S4C), and T-cell infiltration into the tumor was lost by 3 months after treatment (Supplementary Fig. S4D), suggesting avenues for improvement of this regime. Of note, virus-infected, nontransformed iPSCs did not induce significant tumor-specific immunity against PDAC (Supplementary Fig. S4E and S4F).

We next investigated the effect of VIReST-based prophylaxis on mortality. Four-week-old KPC transgenic mice, who do not have tumor lesions at this age, were vaccinated as previously. Volumetric analysis of pancreatic tumor burden revealed significantly delayed tumor development in VIReST-vaccinated animals (Fig. 5A and B). At 6 weeks, tumor was detected in 50% of PBS-treated mice versus 25% of VIReST-treated animals. Progression in VIReST-treated animals was delayed compared with PBS treatment. This translated into a significant survival advantage, with a median survival time of untreated mice (129 days) being increased to 195 days or 169 days (a 51% or 31% extension of lifespan) using KPC (Fig. 5C) or KP-AC (Fig. 5D) VIReST regimes, respectively. This demonstrates that the current VIReST regimen can significantly postpone disease development and progression in these complex models of cancer, a conclusion supported by histologic analysis demonstrating delayed pancreatic intraepithelial neoplasia progression during VIReST treatment (Supplementary Fig. S5A–S5C). Of note, the use of iPSC-derived tumor cells compared with syngeneic tumor cells was more or equally effective at prolonging survival in this model (Supplementary Fig. S6A), although only iPSC-derived tumor cells have the potential for both therapeutic and prophylactic uses, although syngeneic therapy is limited to treatment of established disease. This model was also used to demonstrate the importance of virus infection of iPSC-derived tumor cells as statistical treatment efficacy, and induction of antitumor immunity was lost when virus infection was omitted from the regime (Supplementary Fig. S6B and S6C). The importance of early infiltration and activation of both CD8⁺ and CD4⁺ T-cell populations was confirmed in vivo by complete abrogation of the survival advantage afforded by VIReST in CD8⁺-depleted (Fig. 5E) or CD4⁺-depleted (Fig. 5F) PDAC transgenic animals. Importantly, immunization did not result in evident signs of colitis or ileitis associated with autoimmune disorders, no weight loss was noted, there was no tumor growth detected at the immunization site, and there was no difference in the amount of circulating anti-nuclear antibodies (ANA) detected (Fig. 5G). PDAC is characteristically unresponsive to immune checkpoint therapy; however, given the influx of T cells into the tumor following VIReST therapy, and the reliance of this therapy on CD8⁺ and CD4⁺ T-cell subsets, we investigated that addition of α-PD1 to the regime. Surprisingly, this addition had no impact on the efficacy of treatment, suggesting that eventual tumor escape from control occurs by mechanisms independent of PD1-PD-L1 checkpoint activation (Fig. 5H).

Lack of success in vaccination strategies to control tumor development can be defined by two challenges. The first is derivation of large pools of relevant, immunogenic antigens that are not subject to tolerance mechanisms. In the case of prophylactic strategies, immune priming must occur prior to disease development to aid effective elimination of nascent tumor cells before their escape from immune control via direct immunosuppression or HLA pathway loss. It is clear that immune prevention at the earliest stages can overcome the imprinting of the immune system by previous, nonimmunogenic encounters that promote immune tolerance. Our vaccine, using patient-specific transformed iPSCs, potentiated using viruses to enhance the immunogenicity, is designed to create immune responses against the neoantigens that accrue upon initial development of cancer such that the immune system can prevent development at the earliest stages. Fundamental to the success of the regime is the use of patient-matched iPSC technology, which provides the key to accessing unique neoantigens by modeling specific epigenetic changes that arise early in disease genesis to drive stepwise progression of carcinogenesis via patient-specific accrual of passenger mutations (23, 24) in a way that allogenic vaccination regimes cannot. Using genotypically matched cells ensures adaptive immunity is appropriately raised against neoantigens, and not irrelevant histocompatibility antigens and the epigenetic abnormalities specific to each individual play a seminal role in the earliest steps of cancer initiation (25), affecting both the initiation of the disease (26, 27) and the progression of the disease, for example the patient-specific accrual of passenger nonsynonymous mutations during tumorigenesis (23). From our data, it is clear that driving tumorigenesis with these two driver mutations results in an accurate modeling of the neoantigen profile found in spontaneous PDAC from transgenic mice. Most importantly, we have demonstrated a high level of similarity between transcriptomes of our KRas/P53-driven iPSC-based cells and DT6606 PDAC cells that were isolated from a 129J/C57Bl6 transgenic model driven solely by KRas. This provides further confirmation that the epigenetics and not the driver per se account for the specific pattern of neoantigens expressed. Clearly, the situation in human patients is far more complex than reflected in the mouse model used in this study. Driving tumorigenesis by introducing further driver mutations is possible and the focus of our ongoing work. Identification of genetic predisposition to PDAC can enable us to tailor iPSCs even more accurately, for example, BRCA2 mutations, common in inherited familial atypical multiple mole melanoma (FAMMM) syndrome and LKB1/STK11 mutations common in inherited Peutz–Jeghers syndrome, both of which increase the risk of PDAC development, provide options to tailor treatments to family history data that may increase the already high antigenic relevance of the vaccine to the patient. However, the data presented elegantly support this platform as suitable for development of antigenically relevant vaccines for at-risk individuals.

The second challenge concerns antigen delivery, which must potentiate vaccine immunogenicity. OV is known to exert its most potent effects via activation of antitumor immune responses (7). The oncolytic process provides critical danger signals as a consequence of virus-induced ICD mechanisms, which initiate potent antitumor immune responses (28). Indeed, we found infection of tumor cells with both AdV and VV resulted in production of ICD signals. Interestingly, the order of virus-infected cell administration was critical for treatment efficacy, and this phenomenon was also observed during in vivo therapeutic use against established tumors (8). The underlying mechanism is not yet known; however, AdV is known to be extremely effective at TLR activation, resulting in early improvement of T-cell activation and Treg suppression (29, 30). VV expresses a wide range of immune-modulatory proteins that may either boost activated T-cell responses (31) or alternatively downregulate the induction of effective immune responses when administered before AdV.

Here, we demonstrate that sequential use of two distinct OV as part of a whole tumor cell vaccination regime is required to provide effective prophylaxis against cancer progression in the absence of detectable autoimmunity.

Assessment of the efficacy of prophylaxis was carried out using a transgenic mouse model of PDAC that preserves the relationship between developing tumors, the immune system, and surrounding tissues. Using these models, we showed that immunization using the induced PDAC (KPC or KP-AC) cells alone was unable to prevent disease development, despite antigenic compatibility. However, when these cells were preinfected with OV and used in a VIReST regime, a significant delay in mortality was achieved. This correlated with a delay in progression of PanINs to invasive disease and demonstrated both a preventative and therapeutic effect of the vaccine in vivo.

Despite the potential of this regime, the protocol failed to afford complete protection from disease development in these models. This is likely due to progressive failure of tumor-specific immunity over time as we have observed that the increased infiltration of CD8⁺ T cells within tumors is lost 3 months after the VV-infected tumor cells boost, and the inability to reduce the prolific Treg populations within the tumor (32). Interestingly, and perhaps surprisingly given the influx of T cells prompted by VIReST and the reliance of the regime on these adaptive immune cells, we found no improved impact when combining the VIReST regime with α-PD1. This may suggest that eventual tumor escape from control occurs independently of PD1-PD-L1 checkpoint activation, or may purely reflect the need for further optimization of the delivery of checkpoint inhibitors such as α-PD1 or other checkpoint pathways. Previous reports regarding clinical efficacy of the allogeneic pancreatic tumor vaccine GVAX and Listeria vaccine against early-stage pancreatic intraepithelial neoplasms demonstrated that efficacy was far superior when inhibition of Tregs is induced in parallel with vaccine delivery (32, 33). Therefore, combination with immune checkpoint inhibition to mitigate immune suppression initiated early in tumorigenesis or coadministration with CD25 antagonists may further improve the efficacy of VIReST (34, 35), although the timing of delivery of multiple agents will need to be clearly defined for maximal activity. A further consideration in this regard is the myeloid-derived suppressor cell (MDSC) populations in premalignant or malignant lesions. Previous studies have reported that MDSC levels in premalignancy correlates negatively with the development of antigen specific humoral and adaptive immune responses after therapeutic vaccination using a MUC1-directed vaccine (36). Mechanisms to reduce these populations specifically would benefit VIReST therapy. In addition, an awareness of MDSC levels in at-risk populations may direct treatment suitability. Of note, as genome sequencing provides more information about the genetic evolution of cancers, potential immunogenic, and driver mutations, it will be possible to modify iPSCs further, incorporating further initiating, tissue specific truncal mutations to increase the antigenic relevance of the vaccine to patients or tailor treatments to family history (37). An assessment of all these factors may vastly improve the development of adequate levels of T-cell memory responses that confer long-term protection to vulnerable populations.

For clinical translation, the determination of vaccine eligibility still remains challenging. This technology has potential for therapeutic use, after tumor resection where adequate viable autologous material is not recovered, as a mechanism to prevent tumor recurrence. Determination of patients for primary prophylactic use is more complex; however, with the current lack of effective approaches for early diagnosis and screening, there has been a recent report of individuals with germline mutations or familial risk factors opting for radical pancreatectomy to mitigate their risk (38, 39), and these would be obvious populations to benefit from the option of noninvasive prophylaxis. The European registry of hereditary pancreatitis and familial pancreatic cancer exists to stratify patients with pancreatic cancer for the purpose of clinical intervention, and similar registries exist worldwide that stratify patients with pancreatic cancer by risk. Our expanding knowledge of the progression of genomic alterations and the inflammatory microenvironment that drive premalignancy are providing unprecedented possibilities to identify at-risk individuals for early intervention with cancer prevention strategies (40), and ambitious efforts are underway to detect early cancers via liquid biopsies and analysis of circulating tumor DNA (41) in addition to a coordinated effort to determine effective serum or urine biomarkers to detect PDAC early in disease (42, 43), which will provide candidates for early intervention strategies such as VIReST to prevent malignant progression of nascent cancer.

No potential conflicts of interest were disclosed.

Conception and design: S. Lu, Y. Wang

Development of methodology: S. Lu, P. Du, W. Yan, M. El Khouri, Y. Wang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Lu, Z. Zhang, P. Du, M. El Khouri, Y. Wang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Lu, Z. Zhang, P. Du, L.S. Chard, M. El Khouri, Z. Wang, A. Nagano, J. Wang, C. Chelala, J. Chen, Y. Dong, N.R. Lemoine, Y. Wang

Writing, review, and/or revision of the manuscript: S. Lu, L.S. Chard, J. Wang, J. Liu, Y. Dong, S. Wang, J. Dong, N.R. Lemoine, Y. Wang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Yan, Z. Zhang, Y. Chu, D. Gao, Q. Zhang, J. Liu, S. Wang, X. Li, Y. Wang

Study supervision: L. Zhang, J. Liu, P. Liu, D. Pei, Y. Wang

This project is supported by the National Key R&D program of China (2016YFE0200800), the Nature Sciences Foundation of China (U1704282, 81771776, and 31301007), and the core funding for development of the Cell and Gene Therapy Program by Zhengzhou University. L.S. Chard is funded by The MRC (MR/M015696/1). The authors are grateful to Professor David Tuveson for providing materials for this study.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.