Novel biomarkers of resistance of pancreatic cancer cells to oncolytic vesicular stomatitis virus

Vesicular stomatitis virus (VSV) based recombinant viruses (such as VSV-ΔM51) are effective oncolytic viruses (OVs) against a majority of pancreatic ductal adenocarcinoma (PDAC) cell lines. However, some PDAC cell lines are highly resistant to VSV-ΔM51. We recently showed that treatment of VSV-resistant PDAC cells with ruxolitinib (JAK1/2 inhibitor) or TPCA-1 (IKK-β inhibitor) breaks their resistance to VSV-ΔM51. Here we compared the global effect of ruxolitinib or TPCA-1 treatment on cellular gene expression in PDAC cell lines highly resistant to VSV-ΔM51. Our study identified a distinct subset of 22 interferon-stimulated genes (ISGs) downregulated by both ruxolitinib and TPCA-1. Further RNA and protein analyses demonstrated that 4 of these genes (MX1, EPSTI1, XAF1, and GBP1) are constitutively co-expressed in VSV-resistant, but not in VSV-permissive PDACs, thus serving as potential biomarkers to predict OV therapy success. Moreover, shRNA-mediated knockdown of one of such ISG, MX1, showed a positive effect on VSV-ΔM51 replication in resistant PDAC cells, suggesting that at least some of the identified ISGs contribute to resistance of PDACs to VSV-ΔM51. As certain oncogene and tumor suppressor gene variants are often associated with increased tropism of OVs to cancer cells, we also analyzed genomic DNA in a set of PDAC cell lines for frequently occurring cancer associated mutations. While no clear correlation was found between such mutations and resistance of PDACs to VSV-ΔM51, the analysis generated valuable genotypic data for future studies.


INTRODUCTION
Oncolytic virus (OV) therapy using replicationcompetent viruses has shown preclinical success against many malignancies, and some OVs are now approved for use in the United States and Latvia for melanoma [1,2] and in China for head and neck squamous cell carcinoma [3].Vesicular stomatitis virus (VSV) is efficacious against various cancer types in preclinical studies and is currently in a phase I clinical trial against hepatocellular carcinoma (trial NCT01628640) [4].Our work focuses on VSV-ΔM51, a recombinant VSV with methionine deleted at position 51 in the VSV matrix (M) protein [5].The ΔM51 mutation ablates wild type (WT) M protein's ability to inhibit cellular antiviral gene expression [6][7][8], while still allowing VSV to replicate in and kill cancer cells, as many cancers have defective type I interferon (IFN) antiviral responses.Importantly, the ΔM51 mutation also strongly inhibits neurotoxicity associated with WT VSV, and VSVs with different ΔM51 mutations have been explored extensively [4,9,10].
This work focused on pancreatic cancer, the fourth leading cause of cancer-related deaths worldwide [11].Pancreatic ductal adenocarcinoma (PDAC) is the most common pancreatic neoplasm.Lack of early detection, aggressive local metastases, and limited treatment options means PDAC diagnosis closely mirrors mortality.Surgical removal of tumors is possible in less than 20 percent of patients and current chemo or radiation-based therapies fail to significantly extend life expectancy [12].Various OVs have been tested against PDAC in vitro and in vivo with limited efficacy [13].An understanding of the cellular factors that prevent or allow success is lacking.

Research Paper
The use of VSV-ΔM51 against human PDAC cell lines in vitro and in vivo demonstrated its therapeutic promise [14].However, while VSV-ΔM51 kills a majority of human PDAC cell lines in vitro, resistance of some cell lines to this virus needs to be addressed [14,15].Our previous studies showed that not only resistant but many permissive PDAC cell lines are able to mount type I IFN responses, producing type I IFNs and IFN-stimulated genes (ISGs) in response to VSV-ΔM51 infection [14,15].However, only resistant cell lines showed high-level constitutive expression of the ISGs MX Dynamin-Like GTPase 1 (MX1) and 2'-5'-Oligoadenylate Synthetase 2 (OAS2) [15].We also demonstrated that resistance of PDAC cell lines to VSV-ΔM51 can be overcome by combining virus with IFN signaling inhibitors such as Janus kinase (JAK) inhibitor I and ruxolitinib [15,16].In addition, we showed a similar effect for TPCA-1 [16], which had previously been described as a direct inhibitor of IKK-β [17][18][19].Our study demonstrated [16] pleiotropy for TPCA-1, which inhibited not only IKK-β [17][18][19], but also JAK1 kinase activity [16].
The goal of the current study was to further elucidate the role of ruxolitinib and TPCA-1 in breaking resistance of PDACs to VSV-ΔM51, and to identify gene expression signatures of PDAC resistance to VSV-ΔM51, which could serve as potential biomarkers to predict OV therapy success.The gene expression profiling was the first ever analysis of the global effects of ruxolitinib or TPCA-1 on PDAC transcriptomes, and allowed for further comparison of the molecular mechanisms of action of these drugs.Our study identified a set of 8 ISGs as putative biomarkers of PDAC resistance to VSV-ΔM51, and our data suggest that at least some of the identified ISGs contribute to resistance of PDACs to VSV-ΔM51.Importantly, 4 of these 8 putative biomarkers have never been studied in regard to VSV infection, thus representing potential novel cellular factors restricting VSV replication.Additionally, as certain variants of oncogenes and tumor suppressor genes are often associated with increased tropism of OVs to cancer cells (e.g., by affecting type I IFN signaling regulation), we also conducted a genomic analysis of PDAC cell lines for frequently occurring cancer mutations.

Effect of ruxolitinib and TPCA-1 on transcriptomes of PDAC cell lines
Our previous studies showed that while most of the tested human PDAC cell lines are permissive to VSV-ΔM51, some are highly resistant to this virus [14,15,20].The current study is focused on two permissive PDACs, MIA PaCa-2 and Capan-1, and two resistant PDACs, HPAF-II and Hs766T.As cancer cell can be genotypically and phenotypically unstable, we reexamined permissiveness of these 4 PDAC cell lines to VSV-ΔM51.
MIA PaCa-2, Capan-1, HPAF-II, and Hs766T were infected with VSV-ΔM51 at a range of MOIs (calculated based on VSV-ΔM51 titer on BHK-21, a reference cell line highly permissive to VSV), and monitored for GFP expression to measure virus replication kinetics (Figure 1A), and for virus-mediated oncolysis using MTT cell viability analysis (Figure 1B).Consistent with previous observations, Hs766T and HPAF-II showed strong resistance to VSV as extremely limited GFP was detected at all time points (Figure 1A) and practically no cell death occurred even at the highest tested MOI (Figure 1B).In contrast, MIA PaCa-2 and Capan-1 cell lines were permissive to VSV-ΔM51 as GFP was readily detectable at most time points (Figure 1A), and all cells were dead by the endpoint at all tested MOIs (Figure 1B).
By examining expression of select antiviral ISGs by RT-PCR, we previously showed that only resistant cell lines showed high-level constitutive expression of MX1 and OAS2 [15].Furthermore, inhibition of type I IFN signaling by ruxolitinib or TPCA-1 dramatically improved VSV-ΔM51 replication and PDAC killing while decreasing expression of these ISGs in both infected and uninfected PDACs [16].Despite having very similar effects on VSV-ΔM51 in PDAC cells, these two drugs differ in target specificity, with ruxolitinib targeting both JAK1 and JAK2 [16,21], and TPCA-1 targeting IKK-β [17][18][19], JAK1 [16], and possibly other JAKs.As these drugs are capable of inhibiting pathways other than just type I IFN signaling, determining the global impact of these drugs on PDAC cells is a key step in identifying the mechanism(s) of resistance to VSV-ΔM51 and the role of these drugs in breaking that resistance.To compare the global effects of ruxolitinib and TPCA-1 on cellular gene expression, microarray analysis was performed on both untreated and ruxolitinib or TPCA-1 treated HPAF-II and Hs766T, the two cell lines with the strongest resistance to VSV-ΔM51.Cells were treated with 2.5 μM ruxolitinib or 8 μM TCPA-1 as these doses show the greatest effect on VSV replication without significant drug-mediated toxicity [16].Total RNA was collected at 6 h posttreatment as in the absence of treatment VSV replication is already severely impaired by 6 h p.i., suggesting that the relevant targets are downregulated rapidly [14,16].Furthermore, by choosing a relatively early time point we hoped to primarily capture changes in gene expression directly affected by the drug treatment.Total RNA was collected and reverse transcribed for use with Affymetrix Human Genome U133+ Plus PM array strips, which report on expression levels of more than 47,000 transcripts and variants selected from GenBank, dbEST, and RefSeq.The threshold for our initial analysis was set so that only transcripts with a 2-fold expression level change or greater at least in one sample were included.For transcripts identified at the 2-fold or greater level at least in one sample, expression at the 1.5-fold level was also determined for the remaining samples.
Ruxolitinib treatment of either HPAF-II or Hs766T affected the expression of a surprisingly small number of RNAs with 26 probes representing 20 genes and 32 probes representing 28 genes respectively changing by 2-fold or more upon treatment (Table 1; full listing of probes in Supplementary Table S3).Nearly all identified genes were downregulated and the majority are documented ISGs as based on the Interferome Database v2.01 (see Methods).This list includes MX1, which was identified in our previous studies as constitutively expressed in VSV resistant cell lines [15], but also a number of ISGs where expression was not previously examined.There is a high degree of overlap in the genes differentially expressed in both HPAF-II and Hs766T with 9 of 38 genes with at least a 2-fold difference in expression being shared.For the remaining genes with a minimum of a 2-fold difference in only one cell line, the change in expression in the other cell line was determined to be at least 1.5-fold for most ISGs, but not necessarily for the other gene classifications (Table 1; complete listing of genes changed at the 1.5fold threshold is given in Supplementary Table S4).This suggests that downregulation of a relatively small and consistent set of ISGs is associated with ruxolitinib treatment of two different VSV resistant PDACs.These data are consistent with ruxolitinib as a specific JAK1/ JAK2 inhibitor.
As expected for a drug that targets not only type I IFN signaling but also NF-kB signaling, TPCA-1 affected a much larger number of genes with 260 probes representing 226 genes for HPAF-II and 422 probes representing 348 genes for Hs766T at the 2-fold level (Supplementary Table S3).A number of these genes are known ISGs and many of these are the same genes differentially expressed in response to ruxolitinib treatment, with eight being shared in common between all four cell-treatment combinations (CMPK2, IFI44, IFIT1, IFIT3, PARP9, USP18, XAF1) and an additional eight between three of the treatments at the 2-fold level (Table 1).Expression of these genes, individually or together, may serve as biomarkers for resistance of PDAC cells to VSV-ΔM51.Several additional ISGs were also identified as being changed at least 2-fold in response to TPCA-1 treatment, but that did not reach that threshold with ruxolitinib (Table 2), although several of these genes were affected at least 1.5-fold in response to ruxolitinib treatment (Table 2 and Supplementary Table S4).
In addition to impacting ISGs, treatment with TPCA-1 also affected expression of a number of other genes, including downregulation of several genes associated with the NF-kB pathway including IL8 (Supplementary Tables S3 and S4).This supports our recent report that TPCA-1 acts as an NF-kB pathway inhibitor as well as a JAK1 inhibitor [16].Other differentially expressed genes affected by TPCA-1 could not be obviously identified as associated with either the type I IFN or NF-kB pathways (Supplementary Tables S3 and S4).

ISGs constitutively expressed in resistant but not permissive PDAC cell lines
Our microarray data show that a number of ISGs are downregulated by both ruxolitinib and TPCA-1, suggesting they are constitutively expressed in uninfected HPAF-II and Hs766T.To confirm these findings and determine if they are specific to VSV-resistant HPAF-II and Hs766T cells but not to VSV-permissive MIA PaCa-2 and Capan-1 cells, and to determine whether these genes are inducible by VSV infection, these 4 cell lines were pretreated (or mock-treated) with ruxolitinib for 24 h, then infected (or mock-infected) with VSV-ΔM51 (at MOI 10) for 12 h in the presence or absence of ruxolitinib.Total RNA was isolated, reverse transcribed, and analyzed by semi-quantitative PCR.For this analysis, we selected the genes which were downregulated in both HPAF-II and Hs766T at least 1.5-fold with both ruxolitinib and TPCA-1 and at least 2-fold with at least one treatment.All 22 genes meeting these criteria were ISGs (based on the Interferome Database v2.01) and are included in Table 1.
In agreement with our microarray data, all selected ISGs were constitutively expressed in VSV-resistant HPAF-II and Hs766T cells, with ruxolitinib significantly inhibiting their expression (Figure 2).Importantly, expression was generally not increased in response to VSV-ΔM51 expression (Figure 2).This profile is not seen in the 2 permissive cell lines, MIA PaCa-2 and Capan-1.Out of 22 tested genes, expression of 14 genes was detectable in at least one VSV-permissive cell line even in the absence of VSV-ΔM51.Expression of eight of these genes (IFIH1, OAS2, IFIT1, IFIT2, IFIT3, IFIT5, USP18 and DDX58F) was significantly elevated in response to VSV-ΔM51 infection in permissive cells, but not in resistant PDACs.Expression of the remaining four genes (PARP9, SP110, SP100 and PRIC285) was unchanged by infection.Importantly, no clear effect of ruxolitinib could be seen in VSV-permissive MIA PaCa-2 or Capan-1 cells.Despite these key differences, these genes are not optimal biomarkers of resistance due to their detectable expression in at least some permissive PDACs.
On the other hand, 8 genes were constitutively expressed in resistant cells but showed no detectable expression in permissive MIA PaCa-2 or Capan-1 cells.These 8 genes, MX1, EPSTI1, XAF1, IFI44L, GBP1, SAMD9, SAMD9L, and CMPK2 (Figure 2), do represent candidate biomarkers of resistance of PDAC cells to VSV-ΔM51.To confirm this result at the protein level, a similar experiment was conducted but with infection for 16 h rather than 12 h to detect changes in protein accumulation.Total protein from VSV-resistant and VSV-permissive PDAC cell lines was analyzed for expression of these 8 potential biomarkers (Figure 3).Western blot analysis confirmed the transcriptome and RT-PCR analyses for six potential biomarkers of resistance: MX1, EPSTI1, XAF1, GBP1, SAMD9 and SAMD9L (Figure 3).However, compared to mRNA analyses by microarray and RT-PCR (Figure 2), we observed smaller effect of ruxolitib on SAMD9 protein by Western blot.It is very likely that the observed differences are due to the stability of the preexisting pool of SAMD9 protein that may cause To further investigate whether the selected potential biomarkers are expressed only in resistant PDAC cell lines, we compared constitutive levels of the 6 putative biomarkers for which good antibodies were available in  S1) were used to amplify cDNA.PCR products were run on a 2% agarose gel.All reactions were run in duplicate/triplicate, and one representative sample is shown for each condition.
3 resistant cell lines (HPAF-II, Hs766T, and CFPAC-1) and 7 permissive cell lines (Figure 4).Note that in this experiment HPAC, which in previous publications behaved as a moderately resistant cell line [14,15], now behave as a permissive cell line while all other cell lines behave in a manner consistent with previous observations (Figure 4A).Four of the six potential biomarkers (MX1, EPSTI1, XAF1, and GBP1) were expressed only in resistant PDAC cell lines, with the exception of the T3M4 cell line, which is relatively permissive to VSV, but still shows detectable levels of MX1, EPSTI1, and GBP1 (Figure 4B).This result is consistent with our previous observation that T3M4 was the only VSV-permissive PDAC cell line that constitutively expressed MX1 even in the absence of viral replication [15].While T3M4 can be completely killed in vitro by low MOI VSV infection (our criterion for being "permissive"), T3M4 is less susceptible to VSV infection and killing than other permissive cell lines [14].Therefore, it is not surprising that it shares some features in common with the highly resistant cell lines.In agreement with Figure 3, the other 2 putative biomarkers, SAMD9 and SAMD9L, were expressed in highly resistant HPAF-II and Hs766T cells, but not in the highly permissive MIA PaCa-2 and Capan-1 cells.However, we also observed expression of SAMD9 and one of the bands for SAMD9L (the upper band in Figure 4B that could be associated with a posttranslational modification of the same protein or an alternative mRNA vartiant) in some VSV-permissive cell lines, albeit frequently at lower levels, indicating that these genes cannot be used individually as biomarkers of PDAC  resistance to VSV.Together, the analysis of the 10 PDAC cell lines demonstrates that the 4 genes (MX1, EPSTI1, XAF1, and GBP1) are constitutively co-expressed in VSVresistant, but not in VSV-permissive PDACs, thus serving as potential biomarkers to predict OV therapy success.

Role of MX1 in resistance to VSV
Our data identified 8 cellular genes constitutively expressed in VSV-resistant, but not in VSV-permissive PDACs, thus serving as potential biomarkers of PDAC resistance to VSV-ΔM51 and possibly other OVs.While we have shown previously that MX1 is constitutively expressed in resistant cell lines [15], the remaining 7 genes are a novel finding regarding PDAC resistance to OV therapy.While a biomarker of resistance need not be a causative factor of resistance, it is likely that at least some of these ISGs contribute to resistance of PDACs to VSV-ΔM51, because ruxolitinib and TPCA-1 treatment, which downregulated these genes, strongly enhanced VSV-ΔM51 replication in resistant PDAC cell lines [16].The role for each of these putative biomarkers in antiviral activity against VSV is beyond the scope of this project, especially as some of these 8 genes have been already demonstrated to have antiviral activities in other cell types.Here we focused on the potential contribution of MX1 to the resistance of HPAF-II to VSV-ΔM51.MX1 was the first potential biomarker of PDAC cell resistance identified by our group [15] and it is a known inhibitor of VSV in other systems [25,26].However, our earlier works only showed a correlation between MX1 expression and resistance to VSV, without examining causation, and the antiviral role of MX1 against VSV has never been examined in PDAC cells.Using a lentivector system carrying shRNA expression cassettes, we generated two HPAF-II cell lines with stable shRNA-mediated knockdown (KD) of MX1 expression (Figure 5A), MX1-1B with partial downregulation of MX1 and MX1-3C with complete (undetectable MX1 levels) downregulation of MX1.
To examine effect of MX1 downregulation on VSV-ΔM51 replication, we analyzed viral replication in MX1-1B and MX1-3C KD cell lines, and compared it to the original HPAF-II cell line.Cells were infected with VSV-ΔM51 at MOI 10 (based on VSV-ΔM51 titer on BHK-21 cells) or mock-infected, in the presence or absence of ruxolitinib.As shown in Figure 5B, in the original HPAF-II VSV replication could be detected only in the presence of ruxolitinib, whereas VSV bands could be seen in MX1-1B and MX1-3C clones even in the absence of ruxolitinib.VSV replication levels were higher in the MX1-3C clone than in the MX1-1B, likely because MX1 was completely downregulated in MX1-3C and only partially downregulated in MX1-1B.Although MX1 downregulation strongly stimulated VSV replication, this effect was markedly lower than the effect of ruxolitinib in both MX1-1B and MX1-3C clones, suggesting that that MX1 expression is only partially responsible for resistance of HPAF-II cells to VSV-ΔM51.
In a separate experiment, cells were mock treated or treated with ruxolitinib, then infected with VSV-ΔM51 at 3 different HPAF-II specific MOIs (Figure 5C).Treatment with ruxolitinib resulted in almost complete killing of all HPAF-II clones at all MOIs.In the absence of ruxolitinib, knockdown of MX1 significantly improved killing compared to the original HPAF-II and scramble-shRNA control clone.As in the Figure 5B experiment, the effect of MX1 KD alone was lower than the effect of ruxolitinib alone (Figure 5C).This is not surprising considering that a number of ISG are downregulated by this JAK1/2 inhibitor, many of which have potential antiviral activity.Therefore, it is likely that a combination of the identified ISGs collectively contributes to resistance.

Genomic biomarkers of resistance to VSV
Genomic mutations of oncogenes and tumor suppressor genes have been shown to affect viral tropism.For example, mutation to Cyclin-Dependent Kinase Inhibitor 2A (CDKN2A) facilitates herpes simplex virus (HSV) replication in cancer cells [27].Conversely, knockdown of the retinoblastoma (Rb) tumor suppressor reduced replication efficiency of human cytomegalovirus (HCMV) [28].Important to our study, mutation of tumor suppressor gene p53 can disrupt the antiviral type I IFN signaling pathway [29,30].Other examples with different oncolytic viruses also link oncogenic KRAS mutations, which are initiating mutations in PDAC, to downregulation of ISGs in multiple cancer types [31][32][33].To determine if a particular oncogene or tumor suppressor profile is associated with resistance or permissiveness of PDACs to VSV, we isolated genomic DNA from the resistant (HPAF-II and Hs766T) and permissive (Mia PaCa-2 and Capan-1) cells lines, and searched for more than 2,800 mutations in the Catalogue of Somatic Mutations in Cancer (COSMIC) mutations of 50 oncogenes and tumor suppressors using Ion Ampliseq Hotspot analysis (Table 3; complete listing is given in Supplementary Table S5).Importantly, this assay did not allow for entire genome analysis and focused only on 50 common tumor suppressors or oncogenes.Table 3 shows the predicted damaging genomic mutations found in all 4 cell lines.Consistent with the previous studies, all PDAC cell lines had common KRAS mutations, a known initiating mutation of PDAC.Additionally, HPAF-II, MIA PaCa-2, and Capan-1 have TP53 mutations.As shown in our previous study, Hs766T has a TP53 deletion that Ion Ampliseq reports as "wild-type" [20].Interestingly, different TP53 mutations were detected in the resistant HPAF-II cell line compared to the permissive cell lines MIA PaCa-2 and Capan-1 mutations (Table 3).In addition, the MET Proto-Oncogene, Receptor Tyrosine Kinase (MET) p.S178S mutation was found only in permissive cells, and the Neurogenic Locus Notch Homolog Protein 1 (NOTCH1) p.Q2459K mutation was found in resistant cells versus the p.L2457V found in permissive cells.To examine if these genomic signatures indeed correlated with PDAC phenotype in regard to VSV, we expanded this analysis to a larger set of 7 additional PDAC cell lines as well as 2 non-malignant pancreatic ductal cell lines.Such analysis of a larger set of PDAC cell lines showed that these putative correlations do not hold up (Supplementary Table S5).In general, no clear correlation was found between the assayed genomic mutations and resistance of PDACs to VSV-ΔM51, although the analysis generated valuable genotypic data for future studies.As this study was limited to the most common cancer associated mutations, future genomic analysis may identify changes to other genes (e.g., ISGs) that could be correlated with resistance of PDACs to VSV-ΔM51.

DISCUSSION
This is the first study to conduct global transcriptome analysis PDAC cell lines treated with ruxolitinib and TPCA-1, two drugs that dramatically increase  Conducted on resistant (HPAF-II and Hs766T) and permissive (MIA Paca-2 and Capan-1) cells lines.Translation impact signifies an amino acid difference from wild type while Gene Region denotes where the mutation was detected: E=exonic, I=intronic, U=3'UTR; Pathogenicity: B=benign, P=pathogenic, LP=likely pathogenic.
permissiveness to VSV-ΔM51 infection and oncolysis in otherwise resistant cell lines [16].Ruxolitinib (a specific JAK1/JAK2 inhibitor) was highlighted here based on its use in current phase 1, 2, and 3 clinical trials against pancreatic cancer (trials NCT01423604, NCT02117479, NCT02119663, and NCT01822756).We were also interested in comparing effect of ruxolitinib to TPCA-1, which potentially could be used therapeutically as a dual inhibitor of IKK-β and JAK1 [16].As an inhibitor of both JAK1 and JAK2, key signaling molecules in multiple cellular pathways, ruxolitinib has the potential to impact the expression of thousands of genes.Instead it had a surprisingly modest effect, changing the expression of only 38 genes in both cell lines combined, most of which were ISGs.As expected for a drug targeting additional pathways, TPCA-1 affected expression of a much larger number of genes, but downregulated many of the same ISGs as ruxolitinib, suggesting that these common targets are behind the shared enhancement of viral replication by both ruxolitinib and TPCA-1 [16].Additionally, our study allowed us to identify a set of putative biomarkers of cellular resistance to VSV, all of which are ISGs.In particular, 8 genes (Mx1, EPSTI1, XAF1, IFI44L, GBP1, SAMD9, SAMD9L, and CMPK2) were expressed in the highly resistant cell lines HPAF-II and Hs766T but not the highly permissive cell lines MIA PaCa-2 and Capan-1 in the absence of virus infection.Six of these genes (MX1, EPSTI1, XAF1, GBP1, SAMD9, and SAMD9L) were confirmed in these 4 cell lines not only on the RNA level, but also on the protein level.When we expanded our analysis to 10 human PDAC cell lines, most of these genes are overexpressed only in highly resistant HPAF-II and Hs766T cell lines.However, we also observed expression of SAMD9 and SAMD9L in VSVpermissive AsPC-1, Capan-2, HPAC, and T3M4 cell lines, indicating that these genes cannot be used individually as biomarkers of PDAC resistance to VSV.Nevertherless, the analysis of the 10 PDAC cell lines demontsrate the constitutive co-expression of all 6 potential biomarkers (MX1, EPSTI1, XAF1, GBP1, SAMD9, and SAMD9L) in VSV-resistant, but not in VSV-permissive PDACs.
While we have shown previously that MX1 is constitutively expressed in resistant cell lines [15], the remaining 7 genes are a novel finding regarding PDAC resistance to OV therapy.All of these genes have been shown previously to have antiviral activity against various viruses: MX1 [34], Epithelial Stromal Interaction Protein 1 (EPSTI1) [35], Sterile Alpha Motif Domain Containing 9 (SAMD9) [36], Sterile Alpha Motif Domain Containing 9-Like (SAMD9L) [37], Guanylate Binding Protein 1 (GBP1) [38] and Cytidine Monophosphate (UMP-CMP) Kinase 2 (CMPK2) [39], Interferon-Induced Protein 44-Like (IFI44L) [40], and XIAP Associated Factor 1 (XAF1) [41].Importantly, 3 of these genes (MX1, GBP1, and CMPK2) have specifically been shown to play a role in resistance of cells to VSV [15,38,39].While MX1 is known to interfere with VSV transcription initiation [25], the role for the other proteins in anti-VSV activity is not as clear.GBP1 has been shown to promote IFN production in response to VSV infection [38], and CMPK2 has been found in purified VSV preparations and is thought to have activity associated with the VSV large (L) or nucleocapsid (N) proteins [39].The remaining genes, to the best of our knowledge, have not been specifically linked to VSV infection.However, XAF1 is involved in the apoptotic pathway, which may be important for the VSV life cycle [42].The role for each of these putative biomarkers in antiviral activity against VSV is beyond the scope of this study, but further investigation into how they influence the virus life cycle would be important in regard to basic VSV biology and clinical applications.
The known antiviral effects of these ISGs along with the fact that ruxolitinib effectively breaks resistance to VSV-ΔM51 while affecting very few non-ISGs suggests that the constitutive expression of these genes is likely a causative rather than merely correlative factor.Furthermore, we have previously shown that WT VSV, which is capable of inhibiting host antiviral responses but cannot be used clinically due to its neurotoxicity, more effectively kills resistant cell lines than VSV-ΔM51 [14], also consistent with causation.To determine if KD of the identified genes would indeed facilitate virus replication, we conducted a proof of principle study using shRNAmediated KD of MX1.Our results show a partial reversal of resistance.This finding demonstrates that these putative biomarkers may play a synergistic role in resistance and that a combinational KD approach may be required to achieve optimal efficacy.Future studies will examine a combinational KD on virus replication.
Intriguingly, treatment with ruxolitinib or TPCA-1 resulted in downregulation of a distinct subset of the hundreds of known ISGs.It is likely that only some ISGs are constitutively expressed in VSV-ΔM51 resistant PDAC cell lines rather than ruxolitinib and TPCA-1 specifically target certain ISGs for downregulation [15].Cancer cells selected for radiation damage resistance were shown to constitutively express select ISGs in a manner associated with STAT1 overexpression [43].This set overlaps with that seen in our present study.Further, upregulation of ISGs like STAT1, STAT2, or IRF9 is known to promote cancer resistance to chemotherapy and radiotherapy as demonstrated in human head and neck squamous cell carcinoma and breast cancer [43][44][45][46].Chronic exposure to type I IFN has also been shown to cause expression of only a subset of ISGs (again with significant overlap with that seen here) and to lead to resistance to DNA damaging agents as well as virus infection [47].It has been suggested that chronic type I IFN exposure may occur naturally in cancer cells as a result of environmental insults, mutations leading to dysregulation of interferon production, or continual stimulation of the type I IFN pathway by damage associated molecular patterns caused by the cancerous state [47].Overall this suggests a possible mechanism for constitutive ISG expression and VSV-ΔM51 resistance in PDAC cell lines although this remains to be experimentally determined in future studies.Our work extends the observations made in regard to chemotherapy and radiation resistance to OV therapy and suggests there may be a common mechanism and set of biomarkers.It will be interesting to expand this study to other cancer types and viruses to determine similarities or differences in the ISG profiles of resistance cells.

Virus replication and cell viability assays
Cells were seeded in 96-well plate so that they reached approximately 80% confluence after 24 hours (h).Cells were mock infected or infected with VSV-ΔM51 in DMEM without FBS at the specified multiplicity of infection (MOI, based on virus titer on BHK-21 cells).Virus-containing media was aspirated after 1 h absorption period, and replaced with growth media containing 5% FBS and either mock-treated or treated with inhibitor (here and elsewhere mock treatment and inhibitor treatment contained 0.3% DMSO).After infection, virus-driven GFP fluorescence was measured at regular intervals (CytoFluor Series 4000, excitation filter of 485/20 nm, emission 530/25 nm, gain=63; Applied Biosystems).Cell viability was analyzed by a 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) cell viability assay (Sigma-Aldrich).

RNA microarray analysis
HPAF-II and Hs766T cells (in triplicate) were seeded in 6-well plates so that they reached approximately 80% confluence at 24 h.Cells were washed with PBS, and then mock treated (with serum-free DMSO) or treated with 2.5 μM ruxolitinib or 8 μM TCPA-1 in serum-free DMSO for 6 h.Cellular RNA was extracted with TRIzol (Life Technologies) per the manufacturer protocol with slight modification.In brief, following the first phase separation, the aqueous layer was transferred to a new tube.Then, 500 μl of TRIzol and 100 μl of chloroform were added and phase separation was repeated.Isolated RNA was run on a Bioanalyzer 2100 (Agilent) to check for purity.RNA integrity number (RIN) values were ≥ 7. Samples were reverse transcribed, amplified and labeled using the 3' IVT Express Kit (Affymetrix).The resultant labeled complementary RNA (cRNA) was purified and fragmented as per vendor's instructions.The cRNA samples together with probe array controls were hybridized onto Affymetrix Human Genome U133+ Plus PM array strips, which cover more than 47,000 transcripts and variants selected from GenBank, dbEST, and RefSeq.Hybridization controls were spiked into the cRNA samples to monitor and troubleshoot the hybridization process.Probes for housekeeping genes were used to assess sample integrity.Hybridization, washing, staining and scanning were performed using Affymetrix GeneChip system instruments.Affymetrix GeneAtlas instrument control software version 1.0.5.267 was used to analyze microarray image data and to compute intensity values.Affymetrix .CEL files containing raw, probe-level signal intensities were analyzed using Partek Genomics Suite version 6.6.12.0713 (Partek).Robust multichip averaging (RMA) was used for background correction, quantile normalization and probe set summarization with median polish (195).Statistical difference was calculated by two-way ANOVA analysis with a false discovery rate (FDR) of 0.05.Based on the Interferome Database v2.01, genes with at least a 10-fold increase in expression in human tissue upon type I interferon stimulation under at least 2 experimental conditions were considered to be Type I Interferon stimulated ISGs.Microarray data were deposited to the ArrayExpress database (accession E-MTAB-4576, "Global effects of ruxolitinib and TPCA-1 on cellular gene expression").

RNA RT-PCR analysis
Cells were seeded in a 6-well plate as described above and treated with the specified inhibitor for 24 h before infection.Cells were then mock infected or infected with VSV-ΔM51 in DMEM without FBS at a MOI of 10 (based on VSV-ΔM51 titer on BHK-21 cells).Virus-containing media was aspirated after 1 h absorption period, and replaced with growth media containing 5% FBS and the same treatment as prior to infection.Cells were harvested 12 h post-infection (p.i.) and total RNA was extracted using TRizol as per manufacturer instructions (Ambion), and reverse transcribed using SMART-Scribe reverse transcriptase (Clontech Laboratories, Inc.) as per manufacturer's protocol.PCR products were electrophoresed on a 2% agarose gel with ethidium bromide and photographed using a GelDoc-It imager (UVP Imaging).Primers used for PCR are shown in Supplementary Table S1.

Production of stable MX1-shRNA HPAF-II cells
For stable knockdown of MX1 expression, a pLKO.1-puroplasmid-based shRNA targeting the sequence CCTCTATTACTGAATGGAGAT or GCTTTGTGAATTACAGGACAT was employed (MX1-shRNA1 and MX1-shRNA3, respectively).Additionally, a scramble-shRNA plasmid CAACAAGAT GAAGAGCACCAA was used as a control.A primer containing the target sequence along with a stem loop followed by the reverse target sequence was annealed to a complimentary primer and inserted into the EcoRI and AgeI sites of the pLKO.1-puroplasmid, which was a gift from David Root (Addgene #10878) [52].Lentiviral particles were produced via TRANSIT-TKO (Mirus)mediated triple transfection of AD293 cells with MX1-shRNA1, MX1-shRNA3, or the scramble-shRNA plasmid along with the lentiviral envelope plasmid pMD2.G and the lentiviral packaging plasmid psPAX2.Both pMD2.G (Addgene #12259) and psPAX2 (Addgene #12260) were a gift from Didier Trono.HPAF-II cells were transduced with MX1-shRNA1, MX1-shRNA3, or scramble-shRNA containing lentiviral particles and stable clones were selected using 6 μg/mL puromycin.

Figure 1 :
Figure 1: Phenotypes of VSV-permissive and VSV-resistant PDAC cell lines.VSV-ΔM51 replication A. and VSV-ΔM51mediated oncolysis B. in 4 different human PDAC cell lines.Cells were infected with VSV-ΔM51 at 3 different MOIs (0.001, 1, or 10) or mock-treated, and (A) virus replication driven GFP fluorescence was monitored through 90 h p.i. and (B) cell viability was analyzed by MTT assay at 90 h p.i. and is plotted as percentage of the uninfected control.The assays were done in triplicate and data represent the mean±SEM.Statistical analysis was performed using GraphPad Prism Software, using one-way ANOVA with Bonferroni post-test for comparison to the control.(*) P<0.01; (***) P<0.0001 (*) indicate statistical significance between infected and uninfected cells within the same cell line.

Figure 2 :
Figure 2: RT-PCR analysis of gene expression of putative biomarkers of resistance in PDAC cell lines.Resistant cells (HPAF-II and Hs766T) and permissive cells (MIA PaCa-2 and Capan-1) were mock treated or treated with ruxolitinib (2.5 μM) for 24 h prior to mock treatment or infection with VSV-ΔM51 at MOI 10 (based on virus titer on BHK-21 cells).Virus was aspirated after 1 hour absorption and replaced with growth media containing 5% FBS.Total RNA was extracted 12 h p.i. and reverse transcribed.Gene specific primers (Supplementary TableS1) were used to amplify cDNA.PCR products were run on a 2% agarose gel.All reactions were run in duplicate/triplicate, and one representative sample is shown for each condition.

Figure 3 :
Figure 3: Western blot analysis of putative biomarkers of resistance in PDAC cell lines.Protein expression following ruxolitinib treatment and VSV-ΔM51 infection.Cells were mock treated or treated with ruxolitinib (2.5 μM) for 24 h prior to mock treatment or infection with VSV-ΔM51 at an MOI of 10 (based on virus titer on BHK-21 cells).At 16 h p.i., cell lysates were prepared and analyzed by Western blot for the indicated protein.Protein sizes (kDa) are indicated on the right.Actin protein levels and Coomassie Blue staining of total protein demonstrate equal loading of protein.

Figure 4 :
Figure 4: Western blot analysis of putative biomarkers of resistance in 10 PDAC cell lines.A. Protein expression following mock-treatment or VSV-ΔM51 infection.Cells were mock treated or infected with VSV-ΔM51 at an MOI of 5 (based on virus titer on Suit-2 cells, which have an average permissiveness to VSV-ΔM51).At 8 h p.i., cell lysates were prepared and analyzed by Western blot for the indicated protein."NL" -not loaded, the well #18 was skipped.B. Cells were seeded and protein was isolated 24 h later.Protein sizes (kDa) are indicated on the right.Actin protein levels and Coomassie Blue staining of total protein demonstrate equal loading of protein.

Figure 5 :
Figure 5: Effect of MX1 knockdown on VSV-ΔM51 replication and oncolysis.A. HPAF-II based cell line clones were generated using a lentivector system with the vector genomes carrying MX1-shRNA1 or MX1-shRNA3 sequences.Multiple cell clones for each shRNA construct were puromycin selected and cell lysates were prepared for Western blot analysis of MX1 expression.B. Protein expression following ruxolitinib treatment and/or VSV-ΔM51 infection.Cells were mock treated or treated with VSV-ΔM51 at an MOI of 10 (based on virus titer on BHK-21 cells) and/or ruxolitinib (2.5 μM).At 48 h p.i., cell lysates were prepared and analyzed by Western blot for the indicated protein.Protein sizes (kDa) are indicated on the right.C. HPAF-II and HPAF-II based clones MX1-1b, MX1-3c and SCRA (scramble shRNA) were mock treated or treated with ruxolitinib (2.5 μM) for 24 h prior to infection with VSV-ΔM51 at an MOI of 0.001, 0.05, or 1 (based on virus titer on HPAF-II cells; 1 MOI HPAF equates 1500 MOI BHK-21 ).MTT cell viability analysis was conducted 96 h p.i..The MTT assay was done in triplicate and data represent the mean±SEM.Statistical analysis was performed using GraphPad Prism Software, using multiple t-tests for comparison to uninfected control.(*) indicates statistical significance (p<0.05) between infected and uninfected cells within the same cell line.

Table 2 : ISGs with at least a 2-fold change in expression upon TPCA but not Ruxolitinib treatment Entrez Gene
Matches more than one transcript variant; see Supplementary TableS3for full list of transcripts b Multiple values indicate multiple probes; see Supplementary TableS3for list of probes c For treatments where the change is expression was less than 2-fold: changes greater than 1.5-fold are indicated in italics (see Supplementary Table S4 for full listing); changes less than 1.5-fold are indicated by "-" Changes in expression of those same genes upon TPCA treatment is also indicated.aProbesetIDgivenfor non-mRNAs b Matches more than one transcript variant; see Supplementary TableS3for full list of transcripts c Multiple values indicate multiple probes; see Supplementary TableS3for list of probes a d For treatments where the change is expression was less than 2-fold: changes greater than 1.5-fold are indicated in italics (see Supplementary TableS4for full listing); changes less than 1.5-fold are indicated by "-" significant amounts of this protein to still be present 16 h after treatment.Unfortunately, we were not able to confirm our RT-PCR results for IFI44L and CMPK2 by Western blot.Two commercial antibodies were tested for each protein (see Materials and Methods for details).For IFI44L there was no detectable signal with either antibody.For CMPK2 protein, two antibodies showed different patterns of bands of different sizes and intensities, and both patterns were inconsistent with our RNA data (data not shown).