Experimental evolution generates novel oncolytic vesicular stomatitis viruses with improved replication in virus-resistant pancreatic cancer cells.

Vesicular stomatitis virus (VSV) based oncolytic viruses are promising agents against various cancers. We have shown that pancreatic ductal adenocarcinoma (PDAC) cell lines exhibit great diversity in susceptibility and permissibility to VSV. Here, using a directed evolution approach with our two previously described oncolytic VSV recombinants, VSV-p53wt and VSV-p53-CC, we generated novel oncolytic VSVs with an improved ability to replicate in virus-resistant PDAC cell lines. VSV-p53wt and VSV-p53-CC encode VSV matrix protein (M) with ΔM51 mutation (M-ΔM51) and one of two versions of a functional human tumor suppressor p53 fused to a far-red fluorescent protein eqFP650. Each virus was serially passaged 32 times (accounts for more than 60 viral replication cycles) on either SUIT-2 (moderately resistant to VSV) or MIA PaCa-2 (highly permissive to VSV) human PDAC cell lines. While no phenotypic changes were observed for MIA PaCa-2-passaged viruses, both SUIT-2-passaged VSV-p53wt and VSV-p53-CC showed improved replication in SUIT-2 and AsPC-1, another human PDAC cell line also moderately resistant to VSV, while remaining highly attenuated in non-malignant cells. Surprisingly, two identical VSV glycoprotein (G) mutations, K174E and E238K, were identified in both SUIT-2-passaged viruses. Additional experiments indicated that the acquired G mutations improved VSV replication at least in part due to improved virus attachment to SUIT-2 cells. Importantly, no mutations were found in the M-ΔM51 protein and no deletions or mutations were found in the p53 or eqFP650 portions of virus-encoded transgenes in any of the passaged viruses, demonstrating long-term genomic stability of complex VSV recombinants encoding large transgenes.IMPORTANCE Vesicular stomatitis virus (VSV) based oncolytic viruses are promising agents against pancreatic ductal adenocarcinoma (PDAC). However, some PDAC cell lines are resistant to VSV. Here, using directed viral evolution approach, we generated novel oncolytic VSVs with an improved ability to replicate in virus-resistant PDAC cell lines, while remaining highly attenuated in non-malignant cells. Two independently evolved VSVs obtained 2 identical VSV glycoprotein mutations, K174E and E238K. Additional experiments indicated that these acquired G mutations improved VSV replication at least in part due to improved virus attachment to SUIT-2 cells. Importantly, no deletions or mutations were found in the virus-encoded transgenes in any of the passaged viruses. Our findings demonstrate long-term genomic stability of complex VSV recombinants encoding large transgenes, and support further clinical development of oncolytic VSV recombinants as safe therapeutics for cancer.


V esicular stomatitis virus (VSV) is a prototypic nonsegmented negative-strand (NNS)
RNA virus (order Mononegavirales, family Rhabdoviridae, genus Vesiculovirus). The 11-kb genome of VSV encodes five proteins: nucleocapsid protein (N), phosphoprotein (P), matrix protein (M), glycoprotein (G), and large polymerase (L) (1). VSV is able to infect and replicate in a wide range of cell types (2) due to the use of ubiquitously expressed cell surface molecules. The low-density lipoprotein receptor (LDLR) and other members of the LDLR family have been shown to serve as VSV receptors (3)(4)(5)(6), and additional studies showed that other cell surface molecules, such as phosphatidylserine (7)(8)(9), sialoglycolipids (10), and heparan sulfate (11) could also play a role in VSV attachment to host cells.
VSV is a promising oncolytic virus (OV) due to its inherent ability to preferentially replicate in cancer cells and because of a lack of preexisting immunity against VSV in the human population (12)(13)(14). The oncoselectivity of most OVs, including VSV, is mainly based on defective or reduced type I interferon (IFN) responses in cancer cells, compared to nonmalignant ("normal") cells (15)(16)(17)(18)(19)(20)(21)(22)(23). These responses are generally unfavorable for tumor development as they are antiproliferative, anti-angiogenic, and proapoptotic (24). Although wild-type (WT) VSV is sensitive to type I IFN-mediated antiviral responses in most normal tissues, WT VSV-M sufficiently inhibits type I IFN responses to allow viral replication in the central nervous system (CNS) (25,26). However, due to a well-established reverse genetics system available for VSV, a large number of safe oncoselective VSV-based oncolytic viruses have been generated and tested in numerous studies (12,14). Some of the most widely used oncolytic VSVs are recombinants carrying a deletion (M-ΔM51) or substitution (M51R) of methionine at amino acid (aa) residue 51 in VSV-M. These mutations attenuate VSV replication in normal cells by preventing WT VSV-M protein from inhibition the nuclear exit of host mRNAs, including transcripts for virus-induced antiviral genes (27)(28)(29). As a result, unlike WT VSV, VSV-M51 mutants have dramatically attenuated neurotoxicity but retain robust oncolytic abilities (16,(30)(31)(32)(33)(34)(35).
Another common approach to generate safe oncolytic VSVs is to introduce a transgene improving oncoselectivity or/and induction of adaptive anti-tumor immune responses (12)(13)(14). For example, previous studies showed that functional human tumor suppressor p53 variants can be successfully integrated into VSV genome (36,37). Our laboratory generated two recombinants, VSV-p53wt and VSV-p53-CC, each expressing M-ΔM51 and a different version of a functional tumor suppressor p53 fused to a near-infrared fluorescent protein, eqFP650 (hereinafter referred to as RFP) (37). VSV-p53wt encodes a human WT p53, while VSV-p53-CC encodes a human p53 with its tetramerization domain substituted for with the coiled-coil (CC) domain of breakpoint cluster region (Bcr) protein (38). The resulting p53-CC protein evades the dominantnegative activities of endogenously expressed mutant p53 (38). Our previous study showed that these VSV-carried p53 transgenes not only enhanced VSV anticancer abilities through the introduction of functional p53 into cancer cells with defective tumor suppression activity, but also through the downregulation of antiviral signaling in cancer cells (37).
As a result of the numerous preclinical studies demonstrating the effectiveness of different VSV recombinants as OVs (12,14,39), VSV-hIFNbeta-NIS, encoding the human cytokine interferon beta (hIFNbeta) and the human thyroidal sodium-iodine symporter (NIS), is currently being tested in the United States in several phase I clinical trials against various malignancies. (For details see ClinicalTrials.gov for trials NCT03647163, NCT02923466, NCT03120624, NCT03865212, and NCT03017820.) Despite these advances, many challenges exist regarding the use of VSV as an oncolytic virus in the clinic. For example, not all tumors are susceptible and/or permissive to VSV (12,14). Our previous studies showed that pancreatic ductal adenocarcinoma (PDAC) cell lines show great diversity in susceptibility and permissibility to VSV-based OVs, such as VSV-ΔM51. We previously identified several mechanisms behind resistance of PDACs to VSV-based therapy, such as abnormal or residual type I IFN antiviral activities (40)(41)(42)(43), inefficient attachment of VSV to some PDACs (44), and resistance of VSV-infected PDAC cells to virus-mediated apoptosis (45).
Another potential problem is that VSV, as any other RNA virus, can mutate rapidly due to the lack of proofreading activities in virus-encoded RNA-dependent RNA polymerase (RdRp) (46). Such spontaneous mutations could revert attenuated VSV back to a WT phenotype. For example, in the case of VSV-ΔM51 recombinants, secondary mutations in VSV-M could hypothetically restore WT M functions and reduce VSV-ΔM51 oncoselectivity. Also, VSV has a small RNA genome, and the addition of any transgenes typically attenuates viral replication as the added genetic information hinders speed of viral genome replication and attenuates transcription of downstream viral genes (47). A spontaneous loss of a transgene, particularly if the transgene is the attenuating factor, is an undesirable possibility. Another hypothetical complication is single-site mutations in the beneficial transgene, which could completely negate or change its function, resulting in an ineffective or potentially pathogenic function. Thus, while WT p53 is a powerful tumor suppressor, when mutated, p53 can acquire devastating gain-of-function oncogenic activities, promoting cell survival, proliferation, invasion, migration, chemoresistance, tissue remodeling, and chronic inflammation (48,49). Therefore, an important objective of this study was to experimentally examine the stability of VSV-carried transgenes and the retention of the safe oncoselective phenotype.
Here, we used an experimental evolution approach to obtain novel VSVs adapted to better replicate in virus-resistant PDAC cells and to examine the stability of VSV-carried transgenes after virus replication over an extended period of time. Directed evolution and bioselection for more potent oncolytic viruses has been explored in other studies using a variety of oncolytic viruses and cancer types (50)(51)(52)(53)(54)(55)(56). VSV has been a widely used model to study viral evolution for several decades (57) and has been experimentally evolved for various purposes, such as understanding how viruses evade innate immune responses (58) and the generation of novel VSV-G protein variants used to pseudotype retroviral and lentiviral vectors for gene delivery (59) and to produce novel variants of foreign proteins encoded in VSV genome (60). Moreover, several previous studies have successfully used a directed evolution approach to improve VSV's oncolytic abilities (61)(62)(63)(64). Using this approach in our current study, we generated 2 novel oncolytic VSVs with improved replication in VSV-resistant PDAC cell lines. Our findings also demonstrate long-term genomic stability of complex VSV recombinants carrying large transgenes.

RESULTS
Experimental evolution of two oncolytic VSV variants in two human PDAC cell lines. Our experimental evolution experiments employed two different oncolytic VSV recombinants, VSV-p53wt and VSV-p53-CC, previously generated in our laboratory (37) (Fig. 1). Both viruses have a ΔM51 mutation in VSV-M and a transgene inserted between VSV-G and VSV-L and encoding the N terminus of p53 (p53wt or p53-CC) fused to the C terminus of RFP (65) (Fig. 1). Passaging two different VSV recombinants that have the same ΔM51 attenuation and RFP transgene but different p53 variants allowed us to ensure there was no viral cross-contamination over the course of the parallel viral passaging, as the differences between p53wt and p53-CC served as "molecular barcodes" for each recombinant virus while still offering a type of biological repeat since the viruses have shown very similar phenotypes (37).
We passaged 2 founder viruses VSV-p53wt (Founder) and VSV-p53-CC (Founder), which were produced by BHK-21 cells (a highly permissive baby hamster kidney cell line widely used for VSV amplification), in parallel on two different human PDAC cell lines, SUIT-2 and MIA PaCa-2 (Fig. 1). The cell lines SUIT-2 and MIA PaCa-2 were chosen because of their differential permissiveness to VSV and other differences. SUIT-2 cells are more resistant to VSV infection in part because of residual type I IFN responses, yet permissive enough to support sufficient viral replication to produce enough viral progeny for continued viral passaging, while MIA PaCa-2 cells are very permissive to VSV infection in part due to their inactive type I IFN signaling (37,41,66). Also, we showed that SUIT-2 cells showed lower levels of VSV attachment, compared to MIA PaCa-2 cells (44). Each virus has a large transgene (about 17% of the WT VSV genome) encoding a different version of p53 fused to RFP. Regarding p53, a previous study from our lab showed that VSV encoding p53 could stimulate VSV replication in cancer cells with active type I IFN signaling, such as SUIT-2 cells, but had no effect on VSV replication in MIA PaCa-2 cells that are defective in type I IFN signaling (37). The RFP reporter sequences are presumably dispensable for VSV replication in both cell lines, and previous studies suggest that the addition of a reporter transgene to VSV genome slightly attenuates viral replication (47). However, nucleotide substitutions or deletions in the RFP coding region could negatively affect p53 expression or function because the N terminus of p53 (p53wt or p53-CC) is fused to the C terminus of RFP in both viruses (65). In general, while we expected stronger selective pressures in SUIT-2 cells, both viruses could improve viral replication by losing at least some transgenic sequences due to random mutations as it would reduce the time it takes to replicate the viral genome and assemble virions.
For each passage, fresh uninfected cells were infected at a multiplicity of infection (MOI) of 0.1 PFU/ml (here and elsewhere in this study, the MOI was calculated based on titrating viruses on BHK-21 cells) by incubating fresh cells for 1 h with a previous virus passage, washing off any unbound virus, and incubating cells for an additional 23 h. A portion of the supernatant from the infected cells was collected 24 h postinfection (p.i.) to be used for the next passage, while the remaining supernatant was saved and stored at Ϫ80°C. After a final passage (passage 32) on a PDAC cell line, each virus was FIG 1 Scheme of viral passaging. Viruses VSV-p53wt (Founder) and VSV-p53-CC (Founder) ("Passage 0" indicates amplified in BHK-21 cells) were serially passaged independently 32 times on the PDAC cell line MIA PaCa-2 or SUIT-2. VSV-p53wt (Founder) and VSV-p53-CC (Founder) were added at an MOI of 0.1 to fresh cells for each passage. Cells were incubated with virus for 1 h, after which the virus was removed and fresh medium was added. Virus-containing supernatant was collected 24 h p.i., which was used for each subsequent viral passage (MOI of 0.1). Each virus had a final passage on BHK-21 cells, resulting in the following passage 33 viruses used throughout this study: VSV-p53wt (MIA PaCa-2), VSV-p53-CC (MIA PaCa-2), VSV-p53wt (SUIT-2), and VSV-p53-CC (SUIT-2). amplified on BHK-21 to generate the following four passage 33 viruses: VSV-p53wt (SUIT-2), VSV-p53-CC (SUIT-2), VSV-p53wt (MIA PaCa-2), and VSV-p53-CC (MIA PaCa-2). This final amplification on BHK-21 cells was done to generate stocks of virus particles comparable to the founder virus particles that were originally amplified on BHK-21 cells (Fig. 1).
Viral genome sequence analysis of passaged viruses. To examine if any mutations within coding (viral or transgenic) or noncoding regions of viral genomes took place over the course of the 33 passages, the genomes of each founder virus and passage 33 virus were fully sequenced using Sanger sequencing. Despite the advantages of next-generation sequencing techniques, Sanger sequencing allowed us to focus on major mutations that would become fixed or at least highly prevalent in viral populations by passage 33. For sequencing, supernatants containing viral particles were used to isolate viral genomic RNA that was reversed transcribed into cDNA using random hexamer primers. The generated cDNA was then PCR amplified to generate overlapping DNA products covering the entire viral genomes. (Primer sequences are not shown.) To examine the stability of VSV-carried transgenes, we amplified, analyzed by size, and sequenced a portion of viral genome containing transgene sequences between VSV-G and VSV-L coding regions ( Fig. 2A). As controls, we used a plasmid containing a full-length cDNA copy of the viral genome of VSV-p53wt (the VSV-p53wt plasmid in Fig.  2A) and cDNA generated from a VSV-eq-FP650 virus that carries a shorter transgene (RFP only, no p53 sequences) (VSV-eq-FP650 in Fig. 2A). If passaged viruses lost any significant portions of their transgenes, we expected to see shorter PCR fragments. In addition, all these PCR fragments were sequenced to detect any nucleotide changes in this region. We did not detect any deletions in the transgene regions in any of the passage 33 viruses ( Fig. 2A) (data not shown). Moreover, we detected no nucleotide deletions, additions, or substitutions in the transgenes by Sanger sequencing. To independently address the issue of potential transgene loss, we also examined virus titers for founder and passage 33 viruses by comparing the numbers of PFU and fluorescent focus units (FFU). PFU would account for all infectious viruses (with and without RFP expression), while FFU would account only for viruses retaining their functional RFP transgene. We did not observe any significant changes in the FFU/PFU ratios for founder and passage 33 viruses indicating that the passage 33 viruses have not lost RFP transgene sequences (Fig. 2B). Together, our data demonstrate long-term stability of VSV recombinants carrying RFP-p53 transgenes after extended replication of tested viruses in either permissive or moderately resistant PDAC cell lines. Figure 2C summarizes all genome alterations in viruses detected by Sanger sequencing. No mutations were detected in the VSV regions of N, M, p53, or RFP or any intergenic regions of the viral genome. The absence of any novel mutations in VSV-M after 33 passages is particularly important, indicating the stability of M-ΔM51 as an oncolytic virus attenuator. Of the passage 33 viruses that were passaged on the cell line MIA PaCa-2, one missense mutation, E860D, only partially present in passage 33 viral population (data not shown), was detected in the L protein coding region of VSV-p53wt (MIA PaCa-2). This mutation was not present in any other virus. As we expected, SUIT-2-passaged viruses acquired more mutations than the MIA PaCa-2-passaged viruses, likely because of the stronger selective pressures in SUIT-2 cells. VSV-p53wt (SUIT-2) had a total of 3 nucleotide (nt) substitutions: 2 missense mutations in VSV-G and one silent mutation in VSV-L. VSV-p53-CC (SUIT-2) had a total of 5 nt substitutions: 3 missense mutations in VSV-G, 1 silent mutation in VSV-P, and 1 silent mutation in VSV-L (Fig. 2C).
Surprisingly, both of the SUIT-2-passaged viruses acquired 2 identical missense mutations in VSV-G at aa positions 174 (K174E, A¡G substitution) and 238 (E238K, G¡A substitution) (Fig. 2C). To see at what point these mutations occurred during viral passaging, we sequenced VSV-G of each virus at intermittent passages. Figure 3 shows that in both VSV-p53wt (SUIT-2) and VSV-p53-CC (SUIT-2), E238K appeared first around FIG 2 (A) Analysis of transgene-containing sequence between VSV-G and VSV-L to examine the stability of VSV recombinants carrying transgenes. Supernatants containing viral particles for the founder and passaged viruses were used to isolate viral genomic RNA that was reversed transcribed into cDNA using random hexamers. This cDNA was then amplified by PCR. Red arrows show an approximate position of annealing sites for VSV-specific primers located in the VSV-G and VSV-L regions. As controls, we used a plasmid containing a full-length copy of viral genome of VSV-p53wt and cDNA generated from VSV-eq-FP650 virus, which carries a shorter transgene (coding for RFP only, with no p53 sequences). (B) Comparison of the ratios of viral titers calculated by FFU divided by PFU indicates no loss of RFP transgene sequences from the viral genome after 33 passages. Results shown are representative of 2 independent repeats. Data shown represent the means and standard errors of the means (SEM). Results were analyzed to determine significance using one-way analysis of variance (ANOVA) with a Tukey posttest at a 95% confidence interval for comparison between each condition. All conditions tested were statistically insignificant, with no P value of Ͻ0.05. (C) The entire genomes for all founder and passage 33 viruses were sequenced using Sanger sequencing. Supernatants containing viral particles for the founder and passaged viruses were used to isolate viral genomic RNA, which was reversed transcribed into cDNA using random hexamers. This cDNA was then amplified by PCR. All identified mutations are listed in the table above. Silent mutations are denoted in black font whereas missense mutations are denoted in boldface black font and highlighted in gray if only present in one virus or highlighted in yellow if present in two viruses. The region of the viral genome where the mutations were identified is located at the top of the

FIG 3
The chronological order of the appearance of VSV-G mutations E238K and K174E during passaging of VSV-p53wt (A) and VSV-p53-CC (B) on SUIT-2 cells. Supernatants containing viral particles for the shown passages were used to isolate viral genomic RNA, which was reversed transcribed into cDNA using random hexamers, and cDNA was PCR amplified and sequenced using VSV-specific primers. The nucleotide substitutions are highlighted in orange boxes, and the presence of either mutation is indicated by a red arrow above the sequences. The amino acid numbering starts from the first amino acid of the mature VSV-G and does not include the 16-aa N-terminal signal peptide.

Experimental Evolution of VSV in Cancer Cells
Journal of Virology passage 10, followed by K174E that first appeared around passage 26 in VSV-p53wt (SUIT-2) and passage 27 in VSV-p53-CC (SUIT-2). Interestingly, only after K174E became dominant in both viruses (around passage 30), E238K quickly reached fixation (complete sweep) (Fig. 3). Also, while the E238K mutation was slowly replacing the WT position between passages 10 and 33, the K174E change reached fixation (complete sweep) surprisingly quickly, in just several passages after appearing first around passage 27. Compared to VSV-p53wt (SUIT-2), VSV-p53-CC (SUIT-2) obtained another mutation in VSV-G, M184I (G¡A substitution) ( Fig. 2C and Fig. 4). Although M184I has never completely replaced the WT position in the viral population of passage 33 VSV-p53-CC (SUIT-2), it was fixing in the population surprisingly quickly, first appearing at passage 28 and becoming prevalent by passage 33 (Fig. 4).

SUIT-2-passaged viruses show an improved replication in PDAC cells, while retaining their oncoselectivity.
To determine whether the mutations in passaged viruses altered VSV abilities to replicate in PDAC or nonmalignant cells, virus replication kinetic assays were conducted to compare the founder viruses to the passage 33 viruses. In addition to MIA PaCa-2 and SUIT-2, we tested another human PDAC cell line, AsPC-1, which has a similar phenotype to SUIT-2 in terms of moderate resistance to VSV and inducible type I IFN signaling (37,41). In addition, we tested the viruses in BHK-21 cells, which are highly permissive to VSV and many other viruses, at least in part due to their defective antiviral responses (67,68). To examine the possible loss of oncoselectivity of the passaged viruses as a result of the acquired mutations, we also compared the viruses in the nonmalignant human pancreatic duct epithelial (HPDE) cell line (69)

VSV-p53wt (SUIT-2) G: M184
VSV-p53-CC (SUIT-2) G: M184I The chronological order of the appearance of VSV-G mutation M184I, which was found in VSV-p53-CC (SUIT-2), but not in VSV-p53wt (SUIT-2). Supernatants containing viral particles for the shown passages were used to isolate viral genomic RNA, which was reversed transcribed into cDNA using random hexamers, and cDNA was PCR amplified and sequenced using VSV-specific primers. The nucleotide substitutions are highlighted in orange boxes, and the presence of the M184I mutation is indicated by a red arrow above the sequences. and 72 h p.i. (Fig. 5A). As shown in previous studies, due to its downstream position between VSV-G and VSV-L, virus-encoded reporter expression can be used to measure virus replication levels as it could be detected only if the virus genome is replicated (70). The experiment showed that while all tested viruses showed similar levels of replication in BHK-21 cells, both VSV-p53wt (SUIT-2) and VSV-p53-CC (SUIT-2) replicated better in SUIT-2 cells, compared to founder viruses, especially at 21 h p.i. Importantly, both VSV-p53wt (SUIT-2) and VSV-p53-CC (SUIT-2) replicated better not only in SUIT-2 cells, but also in AsPC-1 cells, and they retained the abilities of founder viruses to replicate in MIA PaCa-2 cells, indicating that experimental evolution of viruses in SUIT-2 cells widened the range of PDAC cells permissive to VSV.
While SUIT-2-passaged viruses show an improved ability to replicate in the SUIT-2 and AsPC-1 cell lines, they also show a retention of oncoselectivity, as none of the tested viruses showed detectable replication in HPDE cells or either of the tested primary human fibroblast cell lines (Fig. 5A). In agreement with virus replication kinetics assay (Fig. 5A), crystal violet cell cytotoxicity assay showed improved cell killing for SUIT-2-passaged viruses in SUIT-2 cells, but no cell killing in HPDE cells (Fig. 5B).
K174E is required for improved replication of VSV experimentally evolved in SUIT-2 cells. As both VSV-p53wt (SUIT-2) and VSV-p53-CC (SUIT-2) obtained the same 2 mutations in VSV-G (first E238K, then later K174E) ( Fig. 2 and 3), we wanted to determine whether E238K alone or both mutations were required for the observed similar phenotypes of these SUIT-2-passaged viruses. To address this question, SUIT-2 passage 20 of VSV-p53wt (to isolate viruses with no mutations or only E238K) and SUIT-2 passage 33 of VSV-p53wt (to isolate viruses with both E238K and K174E mutations) were serially diluted until only 1 FFU was microscopically observed in a tissue culture well, and then each virus originated from a single FFU was amplified in BHK-21 cells, and the viral genome was sequenced to verify VSV-G sequence. Using this approach, we obtained 12 independent VSV-p53wt-based viruses, each of which originated from a single FFU (hereinafter referred to as "independent virus clones"), 4 with no mutations in G (WT VSV-G), 4 with only the E238K mutation ("single mutant"), and 4 with both E238K and K174E mutations ("double mutant") ( Fig. 6). No virus clones with only the K174E mutation were evaluated in this study as that mutation was only present together with the E238K mutation. Based on the determined virus titers on BHK-21 cells, BHK-21 and SUIT-2 cells were infected at an MOI of 0.1, and replication of these "independent virus clones" was examined by Western blotting by analyzing accumulation of viral proteins at 8, 13, 18, and 24 h p.i. As shown in Fig. 6, no stimulation of viral replication was observed for any of the 4 single mutants (G-E238K) in either BHK-21 or SUIT-2 cells, while clear improvement in viral replication can be seen for all double mutants at all tested time points in SUIT-2 cells and at earlier time points (especially at 8 h p.i.) in BHK-21 cells. These data indicate that the second mutation, K174E, was required for improved replication of VSV experimentally evolved in SUIT-2 cells.
Acquired G mutations do not evade antiviral responses, but stimulate VSV replication at least in part due to improved virus attachment to SUIT-2 cells. Our previous analysis of permissive and resistant PDAC cell lines identified at least 2 mechanisms responsible for resistance of SUIT-2 cells to VSV. First, SUIT-2 cells are able to induce a functional type I IFN response to VSV (37,41,66). Second, we observed that, compared to MIA PaCa-2 and some other tested PDAC cell lines, VSV attaches less efficiently to SUIT-2 cells (44). Therefore, we hypothesized that extensive passaging of VSV on SUIT-2 cells selected for spontaneous VSV mutants via the following mechanisms: (i) an improved ability to evade type I IFN signaling and/or (ii) improved attachment to SUIT-2 cells. If the first hypothesis is correct, we would expect to see an increase in VSV replication accompanied by a decrease in antiviral signaling when SUIT-2 cells are infected with the evolved viruses. To test this hypothesis, we infected SUIT-2 and other cell lines shown in Fig. 7 with different viruses at an MOI of 0.1, and total protein was isolated at 13 h p.i. and analyzed by Western blotting for accumulation of VSV-encoded proteins as well as total STAT1 and phosphorylated STAT1 Supernatants containing viral particles for the shown passages were used to isolate viral genomic RNA, which was reversed transcribed into cDNA using random hexamers. This cDNA was PCR amplified using VSV-specific primers, and the (Continued on next page) Experimental Evolution of VSV in Cancer Cells Journal of Virology (STAT1-P) levels as a marker of type I IFN signaling induction (Fig. 7). In agreement with our data in Fig. 5 and 6, VSV-p53wt (SUIT-2) and VSV-p53-CC (SUIT-2) showed an increased ability to replicate on SUIT-2 cells (Fig. 7). Interestingly, these 2 viruses also showed a slightly increased ability to replicate in MIA PaCa-2 cells, which are defective in type I IFN signaling. We did not detect significant differences in viral replication levels in BHK-21 cells at 13 h p.i., which is consistent with the data shown in Fig. 6, where the differences in BHK-21 cells were seen mainly at an earlier time point (8 h p.i.). No significant viral replication was detected in the nonmalignant HPDE pancreatic ductal cell line HPDE and human primary fibroblast cell lines AG0159 and AG08498, confirming retained oncoselectivity of the evolved viruses (Fig. 7). Importantly, although viral infections did not significantly alter total STAT1 levels, for both of the SUIT-2-passaged viruses in the SUIT-2 and HPDE cell lines, as well as both of the fibroblast cell lines, there was an increase, rather than decrease, in STAT1 phosphorylation in VSV-p53-CC (SUIT-2)-and VSV-p53wt (SUIT-2)-infected cells (Fig. 7). This result shows that the improved replication of the evolved viruses was not due to their acquired abilities to evade innate antiviral responses. (In such a case, we would see inhibition rather than stimulation of STAT1 phosphorylation.) The observed increased STAT1 phosphorylation was likely a result of the higher number of cells infected with SUIT-2-passaged viruses, which produce collectively a stronger antiviral response.
Since both VSV-p53wt (SUIT-2) and VSV-p53-CC (SUIT-2) mutations were located in VSV-G, a region that plays an important role in viral attachment, and because our previous studies have shown that VSV does not attach to SUIT-2 cells as well as to some other PDAC cell lines, we also compared the ability of the founder and SUIT-2-passaged viruses to attach to SUIT-2 cells. To examine virus attachment, BHK-21 or SUIT-2 cells were incubated at various MOIs with purified VSV-p53wt (Founder) or VSV-p53wt (SUIT-2) at 4°C for 1 h, and the cells were extensively washed to remove any unbound virus and analyzed by Western blotting for virus proteins bound to cells (attachment assay in Fig. 8). At 4°C, the viral particles can only attach to the outside of cells and not enter them. A duplicate set of cells was treated the same way (incubated with virus at 4°C for 1 h and then extensively washed), but then incubated for 7 more h at 37°C before protein was isolated to examine virus replication (replication assay in Fig. 8). In BHK-21 cells, there was only a minor difference in virus attachment or replication between VSV-p53wt (Founder) or VSV-p53wt (SUIT-2) viruses. However, in SUIT-2 cells, VSV-p53wt (SUIT-2) was able to attach much more efficiently (about 3-fold better based on serial dilutions of viruses) than VSV-p53wt (Founder), and VSV-p53wt (SUIT-2) replication of the passage 33 virus was also higher in SUIT-2 cells. Interestingly, we observed about 3-fold improvement in virus attachment, but about 10-fold enhancement in virus replication for the SUIT-2-passaged viruses in SUIT-2 cells. While it is possible that the 3-fold improvement in virus attachment was solely responsible for even stronger enhancement in virus replication due to the exponential growth rate of virus replication, we cannot rule out that E238K and K174E VSV-G mutations also improve other steps of the virus replication cycle. Future studies will test these possibilities.
These data suggest that the SUIT-2-passaged viruses were selected to attach to SUIT-2 cells more efficiently, which could improve new infection efficacy and explain at least in part the observed improvement in viral replication of SUIT-2-passaged viruses. To test whether SUIT-2-passaged viruses could initiate infections more efficiently, compared to founder viruses, titers of serial dilutions of each virus were determined on   Fig. 8B, VSV-p53wt (SUIT-2) and VSV-p53-CC (SUIT-2) improved their abilities to initiate infections on SUIT-2 cells by about 3-fold, which is consistent with our data on relative attachment efficiency of founder and passaged viruses (Fig. 8A). LDLR and LDLR family members have been shown to serve as receptors for VSV (3-6). As VSV-G is responsible for VSV attachment to host cells and we observed an improved attachment of SUIT-2-passaged viruses to SUIT-2 cells (Fig. 8A), we wanted to examine the abilities of the evolved viruses to attach to LDLR. Mutations in VSV-G could improve VSV's ability to interact with LDLR or, rather, utilize an alternative receptor. The affinity of VSV for LDLR was examined using soluble LDLR (sLDLR), which neutralizes VSV virions and inhibits viral infectivity. To test if sLDLR could inhibit infectivity of "independent virus clones" (WT-G, single mutant, or double mutant) of VSV-p53wt, the same number of infectious particles corresponding to an infection at an MOI of 0.1 on BHK-21 cells were incubated with or without sLDLR in situ, and the SUIT-2 cells were incubated with these VSV Ϯ sLDLR combinations for additional 30 min, washed to remove any unattached virus and sLDLR, and then incubated for an additional 12 h. Cells were then trypsinized and analyzed for the percentage of VSV-infected cells (RFP-positive cells). In agreement with Fig. 6 data, the presence of both VSV-G mutations E238K and K174E in double mutants resulted in a dramatic increase in the percentage of VSV-infected cells in the absence of sLDLR (Fig. 9A). In the presence of sLDLR (Fig. 9B), we observed a significantly lower percentage of VSV-infected cells for all tested viruses (Fig. 9B). However, we did not observe statistically significant differences between viruses with regard to the inhibiting effect of sLDLR on viral infectivity when we compared the ratios of VSV-infected cells in the absence and presence of sLDLR, which were close to 2 under our experimental conditions (Fig. 9C). These data suggest that VSV is still able to attach to and infect SUIT-2 cells through an interaction with LDLR.

BHK-21 and SUIT-2 cells, and the ratios of virus titers on SUIT-2 cells to those on BHK-21 cells were calculated. As shown in
We have previously shown that infectivity of VSV-ΔM51 in several resistant PDAC cell lines, including SUIT-2, was dramatically improved when cells were treated with the polycations DEAE-dextran or Polybrene (44). Although the exact mechanism of polycation-mediated improvement of virion attachment is not clear and several alternative mechanisms, including charge shielding and virus aggregation, have been proposed (71), it is believed that polycations help the initial nonspecific anchoring of virus particles to cell surface, which facilitates their further association with specific receptors (such as LDLR and LDLR family members for VSV) (72)(73)(74)(75)(76). As our data suggest that the VSV-G mutations E238K and K174E did not dramatically change VSV-G affinity for LDLR, we decided to test a hypothesis that E238K and K174E mutations improve the efficacy of this initial nonspecific VSV binding to target cells. In such a case, the efficient infection of evolved viruses would be less dependent on polycation treatment, compared to the founder viruses. To test this hypothesis, we infected BHK-21, MIA PaCa-2, and SUIT-2 cells with the founder viruses and the SUIT-2-passaged viruses at various MOIs in the presence or absence of DEAE-dextran and analyzed virus replication kinetics by measuring VSV-encoded RFP fluorescence over time. In agreement with our previous study using VSV-ΔM51 (44), we did not observe any significant positive effect of DEAE-dextran on replication of any tested viruses in the highly permissive BHK-21 and MIA PaCa-2 cell lines (Fig. 10). On the other hand, in agreement with the same study (44), DEAE-dextran treatment strongly improved infectivity of VSV-p53 (Founder) at all tested MOIs in SUIT-2 cells (Fig. 10). In contrast, DEAE-dextran treatment had a rather small positive effect on VSV-p53 (SUIT-2) infection at only the lowest MOI tested (MOI of 0.01), and no effect was observed at an MOI of 0.1. Moreover, DEAE-dextran treatment actually inhibited VSV-p53 (SUIT-2) at an MOI of 1 (Fig. 10). In general, these data indicate that VSV-G mutations K174E and E238K make the evolved viruses less dependent on polycations for efficient infection of resistant cell lines, such as SUIT-2, suggesting that the evolved viruses have an improved nonspecific attachment to target cells.

DISCUSSION
In this study, using a directed viral evolution approach, we generated novel oncolytic VSVs that show improved replication in virus-resistant PDAC cells. Both SUIT-2passaged VSV-p53wt and VSV-p53-CC showed improved replication in the human PDAC cell lines SUIT-2 and AsPC-1, which are moderately resistant to VSV, and they retained the abilities of founder viruses to replicate in MIA PaCa-2 cells, indicating that experimental evolution of viruses in SUIT-2 cells widened the range of PDAC cells permissive to VSV. Importantly, both evolved viruses remained highly attenuated in the nonmalignant human pancreatic duct epithelial cell line HPDE and the primary human fibroblast cell lines AG0159 and AG08498. Moreover, no mutations were found in M-ΔM51, and no deletions or mutations were found in the p53 or eqFP650 portions of virus-carried transgenes in any of the passaged viruses, demonstrating long-term genomic stability of complex VSV recombinants carrying large eqFP650-p53 transgenes.
Interestingly, both of the SUIT-2-adapted viruses passaged in parallel, VSV-p53wt (SUIT-2) and VSV-p53-CC (SUIT-2), acquired two identical missense mutations in VSV-G, E238K and K174E. The fact that both mutations (complete sweep) were acquired independently in two different viruses passaged in parallel suggests a strong fitness benefit to VSV in SUIT-2 cells from the introduction of these mutations. Interestingly, both viruses first obtained the E238K mutation before acquiring the K174E mutation. It is likely that the first mutation, E238K, provided a minor fitness benefit to both VSV-p53wt and VSV-p53-CC, founder viruses, but the observed major shift in virus replication required both mutations. It should be noted that the second mutation, K174E, was never identified alone in any of the viral passages and thus was not individually evaluated. VSV-p53-CC (SUIT-2) also showed a third missense mutation in The data points and error bars shown represent the means and SD of the means, respectively. Results were analyzed to determine significance using two-way ANOVA with a Tukey's posttest at a 95% confidence interval for comparison between each condition. *, P Ͻ 0.05. If no error bars appear, the error is too small to appear on the graph.
VSV-G, M184I, but only in a portion of the passage 33 VSV-p53-CC population (not a complete sweep). Although it will be interesting to examine the role of this mutation in VSV replication in the future studies, it is unlikely that this particular mutation plays a critical role in the improved ability of this virus to replicate in SUIT-2 cells, as VSV-p53wt (SUIT-2) and VSV-p53-CC (SUIT-2) exhibited very similar phenotypes. Although the VSV-G E238K K174E double mutant had never been described in the past, a previous study described VSV-G E238G (VSV mutants "G 6 " and "G 6R ") and E238Q (VSV mutants "G 5 " and "G 5R ") mutants (63). Most of these previously described VSV mutants had additional mutations in VSV-G, although one of the mutants, VSV-G 6R , had only the single amino acid substitution E238G in VSV-G (63). The study showed that VSV-G 6R infection of L929 cells (mouse fibroblasts) produced higher levels of IFN-␤ compared to WT VSV, and those levels were similar to or even higher than those with the VSV-M51R mutant. Based on that result, the authors proposed that E238G mutation in VSV-G enhances type I interferon secretion and responses via some unclear mechanism not involving VSV-M (63). However, we propose another explanation for their observations: that E238G mutation may have resulted in an improved replication of VSV-G 6R virus in L929 cells, and the observed overall increase in IFN-␤ was due to the increase in the number of infected cells rather than the increased IFN-␤ production by each infected cell. In agreement with this hypothesis, no attenuation was observed for VSV-G 6R replication in L929 cells, which would be expected for a mutant in which replication enhances antiviral response (63). Our data show the increased STAT1 phosphorylation in cells infected with the SUIT-2-adapted viruses carrying the E238K K174E VSV-G double mutation, but we think that was due to their higher levels of replication, which we observed throughout this study, suggesting that these evolved viruses are not modulating antiviral signaling in each cell, but instead the higher number of infected cells (Fig. 9) produces collectively higher STAT1 phosphorylation levels.
Although we cannot exclude additional mechanisms of the improved VSV replication in the presence of E238K and K174E mutations in VSV-G, our data show that these mutations result in improved VSV attachment to SUIT-2 cells. We envision that during VSV passaging, when a virus passage was incubated with fresh cells for a 1-h period (after which the incubation medium containing unbound virus was aspirated and cells were washed with phosphate-buffered saline [PBS]), there was a selective pressure for VSV mutants capable of more efficient attachment to SUIT-2 cells.
It is unclear how the E238K and K174E mutations in VSV-G improve VSV attachment to PDAC cells. We analyzed the positions of E238K and K174E mutations using the crystallographic structures of prefusion conformations of VSV-G with and without LDLR (PDB code no. 5oyl, 5oy9, and 5i2s) and low-pH, postfusion conformations of VSV-G without LDLR (PDB code no. 2cmz) (6,77,78) (Fig. 11). As shown in Fig. 11, both mutations are located away from the interaction interface between VSV-G and the CR2 or CR3 (cysteine-rich) domains of LDLR. Both mutations are also located on the side of the protein opposite to the intermonomer interface in the postfusion trimer. Therefore, it is unlikely that they affect the trimerization of the glycoprotein G or its interaction with LDLR, at least with the CR2 or CR3 domains for which X-ray structures have been solved. In agreement with that, we did not detect any significant differences between WT and mutant G viruses with regard to the inhibiting effect of sLDLR on viral infectivity (Fig. 9C). Also, our previous study showed that, despite lower levels of VSV attachment to SUIT-2 cells, SUIT-2 expressed high levels of LDLR, suggesting no limitation of the surface receptor for VSV in this cell line (44). Together, these data suggest that VSV-G mutations did not dramatically alter the abilities of mutant VSV-G proteins to attach to and infect SUIT-2 cells through an interaction with LDLR.
Interestingly, our data indicate that the VSV-G mutations E238K and K174E make the evolved viruses less dependent on polycations for efficient infection of SUIT-2. Previous studies have suggested that efficient cell attachment of virions of many viruses, including VSV and VSV-G-pseudotyped viruses (79,80), requires an initial nonspecific binding of virus particles to cell surface, followed by attachment of virus particles to their specific receptors, which is required for virus internalization (72)(73)(74)(75)(76). The initial nonspecific binding step can be dramatically enhanced by treating cells with polycations, such as DEAE-dextran or Polybrene (81)(82)(83). Importantly, we have previously shown that infectivity of VSV-ΔM51 in several resistant PDAC cell lines, including SUIT-2, was dramatically improved when cells were treated with the polycation DEAE-dextran or Polybrene (44). Although the exact mechanism of polycation-mediated improvement of virion attachment is not clear and several alternative mechanisms, including charge shielding and virus aggregation, have been proposed (71), the most widely accepted hypothesis is that polycations decrease electrostatic repulsion between negatively charged molecules on the surface of cells and many viruses, including VSV, and thus facilitate nonspecific binding of virus particles to cell surface (82)(83)(84). Our data indicate that the VSV-G mutations E238K and K174E make the evolved viruses less dependent on polycations-possibly by decreasing dissociation of virions at the initial step of attachment due to electrostatic repulsion between cell surface molecules and VSV. One possibility is that these mutations change the overall structure of VSV-G, which could significantly change charge distribution on the surface of VSV particles, possibly reducing the repulsion. These mutations could also reduce repulsion by posttranslational modifications of VSV-G: for example, VSV-G N-glycosylation that occurs through the N163 and N320 positions (85)(86)(87)(88)(89)(90)(91)(92). The VGV-G N-glycosylation can dramatically affect viral infectivity, although the effect strongly depends on the target cell type and specific mode of VSV-G N-glycosylation (89,90,93). Moreover, it is believed that one of the major mechanisms of DEAE-dextran-mediated improvement of VSV infection is removal of repulsion between negatively charged molecules on the cell surface (such as anionic phospholipid sialic acid residues) and terminal sialic acid residues associated with those N-glycosylated VSV-G (89). Therefore, it is possible that E238K or K174E mutations could alter VSV-G N-glycosylation pattern or juxtaposition of sialic acid residues and thus decrease repulsion, thus making mutant viruses less dependent on DEAE-dextran for virus infection.
We cannot rule out that E238K and K174E VSV-G mutations could also improve other steps of the virus replication cycle. VSV-G is responsible for both viral attachment and entry into host cells, and it is possible one or both of these mutations improved not only attachment, but also virus entry into the infected cells. Interestingly, E238K and K174E mutations are located near the region that undergoes the major conformational rearrangement during the transition from the prefusion to postfusion state (6,77,78). E238 and K174 are about 24 Å apart in the prefusion conformation, but only about 10 Å apart in the postfusion conformation (Fig. 11). It is possible that these mutations play a role in this transition. Future studies will test these possibilities.
Each of the passaged viruses was fully sequenced and compared to its corresponding founder virus. This analysis revealed overall impressive resistance of complex oncolytic VSV recombinants to mutations as none of the passaged virus had any nucleotide changes in the noncoding regions or amino acid substitutions in the N, P, or M region. Importantly, no viruses had any secondary mutations in M-ΔM51. We have demonstrated that not only can VSV's efficacy as on OV be improved, it can do so while retaining oncoselectivity. This retention of oncoselectivity was demonstrated on 3 separate nonmalignant cells lines: HPDE (nonmalignant human pancreatic ductal cells) and the primary human fibroblast lines AG0159 and AG08498.
Over the course of 33 viral passages, all viruses retained both the p53 (p53wt or p53-CC) and RFP sequences of the virus-carried transgenes. No transgenes in any of the viruses obtained any mutations, demonstrating long-term genomic stability of complex VSV recombinants carrying large transgenes, even after replication over an extended period of time (more than 768 h of continuous viral replication). To our knowledge, our study is the first to passage VSV carrying a transgene for 33 passages and exhibiting complete transgene retention and absence of any mutations in the transgenes. This result is surprising considering that neither p53 nor RFP expression presumably benefits VSV-ΔM51 replication in MIA PaCa-2 cells, although p53 expression stimulates VSV-ΔM51 replication in SUIT-2 cells via inhibition of antiviral signaling (37). RNA viruses are known to have high mutation rates due to a lack of proofreading activities by viral RNA-dependent RNA polymerase (46,94). This lack of proofreading can result in the introduction of mutations in the viral genome or transgenic regions. Such mutations can have a detrimental effect on the expression of viral or recombinant proteins. For instance, in a study where VSV expressed recombinant CD4 protein, a single nucleotide deletion resulting in a frameshift mutation caused the loss of expression of the transgene that was seemingly stable after 26 passages (95). In the same study, another VSV recombinant carrying measles virus F protein lost the transgene after 1 passage.
We think several different factors contributed to the surprising transgenic stability of our tested viruses. First, while NNS RNA viruses, like any RNA viruses, are associated with a high mutation rate (46,94), they show lower incidence of genetic recombination caused by polymerase slippage compared to positive-strand RNA viruses, because the viral genome in NNS RNA viruses is encapsidated at all times, as opposed to naked RNA genomes of positive-strand RNA viruses, where polymerase can disassociate with one strand of RNA and reassociate on another strand (96,97). Second, the helical nucleo-capsid and overall bullet shape of VSV can accommodate extra transgenes by adding length to the virus rather than being geometrically limited in icosahedral virions of some other viruses (98). Third, our transgene sequences were not fused to viral proteins, which could increase selective pressures to lose the transgene. Previously, an in-frame fusion of green fluorescent protein (GFP) to VSV-P, an essential viral protein, resulted in the loss of transgene expression within 20 passages, assumedly because the reporter gene resulted in reduced viral fitness (99). Fourth, our viruses had transgene sequences located between VSV-G and VSV-L genes. Although placing a transgene closer to the 3= end of the viral genome can increase its expression, it also results in greater virus attenuation (47,97,100,101), which increases the selective pressures to lose the transgene sequences. On the other hand, several studies have successfully inserted transgenes between the G and L proteins of both VSV and rabies virus (RABV) without diminishing viral replication or activity (47,(102)(103)(104). Finally, the observed stability of the RFP-p53 transgenes is likely due to the selective pressures associated with beneficial effect of p53 expression on VSV replication in PDAC cells. The fact that p53 transgene was stable in both cell lines after 33 passages suggests that p53 has at least a minor positive effect on VSV replication even in MIA PaCa-2 cells.
While this study was focused on the adaptation of oncolytic VSV recombinants to PDAC cells and VSV transgene stability after extended virus passaging, future experiments will compare the efficacy and safety of the founder and SUIT-2-passaged VSV-p53 viruses in vivo. With our current findings describing improved VSV-p53 viruses and demonstrating long-term genomic stability of complex VSV recombinants encoding p53 transgenes, all of these data support further clinical development of VSV-p53 OVs as safe therapeutics for cancer.

MATERIALS AND METHODS
Viruses and cell lines. The recombinant viruses VSV-p53wt and VSV-p53-CC were previously engineered using the VSV-ΔM51 backbone and were previously described in detail (37). Purified virus was obtained exactly as described previously (105). Plaque-isolated viruses were obtained by isolating individual viral plaques, which were then amplified on BHK-21 cells. The baby hamster kidney BHK-21 fibroblast cell line (ATCC CCL-10) was used to grow viruses and to determine their titers. Viral titers for both viruses were determined by standard plaque assay on BHK-21 or SUIT-2 cells using an agar overlay and then calculated as PFU/ml or FFU/ml. To calculate PFU/ml, cells were fixed and stained with crystal violet, whereas to calculate FFU/ml, VSV-encoded RFP fluorescent foci were counted using fluorescence microscopy. The following human PDAC cell lines were used in this study: SUIT-2 (106), MIA PaCa-2 (ATCC CRL-1420), and AsPC-1 (ATCC CRL-1682). The human origin of all these PDAC cell lines was confirmed by partial sequencing of KRAS and actin, as well as genomic mutation profiling using Cancer Hotspot Panel v2 (Life Technologies) to analyze for 2800 Catalogue of Somatic Mutations in Cancer (COSMIC) mutations of 50 oncogenes and tumor suppressor genes (43). As expected, all PDAC cell lines had a mutation in KRAS, as it is typical for PDACs (42,43). A nonmalignant human pancreatic duct epithelial (HPDE) cell line was previously generated by introduction of the E6 and E7 genes of human papillomavirus 16 into normal adult pancreas epithelium. HPDE retains a genotype similar to pancreatic duct epithelium, is nontumorigenic in nude mice, and has no cancer-associated mutations (69). HPDE was grown in keratinocyte-SFM (K-SFM; Gibco, 17005042) without serum. Normal untransformed human fibroblast AG08498 and AG01519 cells from foreskin of healthy donors were obtained from the Coriell Institute. MIA PaCa-2, SUIT-2, and AsPC-1cells were maintained in Dulbecco's modified Eagle's medium (DMEM [Cellgro, 10-013-CV]), while BHK-21, AG08498, and AG01519 cells were grown in modified Eagle's medium (MEM [Cellgro, 10-010-CV]). All cell growth media (except for K-SFM, which was supplemented with manufacturer-provided human recombinant epidermal growth factor 1-53 [EGF 1-53] and bovine pituitary extract [BPE]) were supplemented with 10% fetal bovine serum (FBS [Gibco]), 4 mM L-glutamine, 900 U/ml penicillin, 900 g/ml streptomycin, and 1% nonessential amino acids. MEM was additionally supplemented with 0.3% glucose (wt/vol). Cells were kept in a 5% CO 2 atmosphere at 37°C. For all experiments, all cell lines were passaged no more than 15 times. Viral passaging. MIA PaCa-2 and SUIT-2 cells were seeded into 35-mm-diameter dishes to be approximately 95% confluent in 24 h. Cells were washed once with PBS and incubated with viruses at an MOI of 0.1 PFU/ml (calculated based on virus titration on BHK-21 cells) in DMEM without fetal bovine serum (FBS) for 1 h at 37°C. After 1 h of incubation, medium containing unbound virus was aspirated, cells were washed with PBS, and fresh DMEM medium containing 5% FBS was added to the cells. After 24 h of incubation, dishes were checked under a microscope (Olympus IX70 fluorescence microscope) to ensure all cells were infected, which was detected by the presence of RFP signal, and that all cells were detached from the 35-mm dishes. The entire supernatant was collected at 24 h p.i. and centrifuged at 4,000 rpm at 4°C for 10 min to pellet cellular material. The virus containing supernatant was transferred to new tubes and stored at -80°C. Each collected viral passage was used for the subsequent viral passage.
Mg 2ϩ and Ca 2ϩ . Prior to infection, cells were pretreated with MEM without FBS (mock) or with 10 g/ml DEAE-dextran in MEM without FBS for 30 min. After the 30-min pretreatment, VSV-p53wt (Founder) or VSV-p53wt (SUIT-2) in MEM with 0% FBS was directly added to cells at an MOI of either 1, 0.1, or 0.01 (MOI calculated based on virus titration on BHK-21) for 1 h at 37°C. Virus plus DEAE-dextran-containing medium was removed after the 1-h infection period, and fresh DMEM containing 5% FBS was added to cells. Cells were maintained at 37°C, and the level of VSV-encoded RFP fluorescence was measured over the course of 68 h p.i. using a fluorescence multiwell plate reader at the wavelength 590/645 nm.