Newcastle Disease Virus (NDV) Oncolytic Activity in Human Glioma Tumors Is Dependent on CDKN2A-Type I IFN Gene Cluster Codeletion

Glioblastoma (GBM) is the most aggressive and frequent primary brain tumor in adults with a median overall survival of 15 months. Tumor recurrence and poor prognosis are related to cancer stem cells (CSCs), which drive resistance to therapies. A common characteristic in GBM is CDKN2A gene loss, located close to the cluster of type I IFN genes at Ch9p21. Newcastle disease virus (NDV) is an avian paramyxovirus with oncolytic and immunostimulatory properties that has been proposed for the treatment of GBM. We have analyzed the CDKN2A-IFN I gene cluster in 1018 glioma tumors and evaluated the NDV oncolytic effect in six GBM CSCs ex vivo and in a mouse model. Our results indicate that more than 50% of GBM patients have some IFN deletion. Moreover, GBM susceptibility to NDV is dependent on the loss of the type I IFN. Infection of GBM with an NDV-expressing influenza virus NS1 protein can overcome the resistance to oncolysis by NDV of type I-competent cells. These results highlight the potential of using NDV vectors in antitumor therapies.


Introduction
Glioblastoma (GBM) is the most devastating and least treatable brain tumor, with an incidence of 3.2 cases per 100,000 inhabitants [1]. GBM cells grow quickly and infiltrate the brain parenchyma [2]. The morphology and location of the tumor, as well as the clinical status and age of the patients determine the therapeutic treatment and its efficacy. After surgery, the standard GBM treatment continues with the administration of temozolomide (TMZ) concomitant with radiotherapy [3]. Despite advances in medical image and novel therapies, the median overall survival is 15 months [4], suggesting that other alternatives are needed to find more effective treatment strategies for this heterogenous disease. In this sense, several molecular and genetic alterations have been observed that contribute to tumorigenesis and treatment resistance [5,6]. Some of the most relevant mutations in gliomas are the homozygous and hemizygous deletion at Ch9p21 [7], which involves the genetic loss of CDKN2A (p16 INK4 ) [8,9]. Interestingly, the type I Interferon (IFN) gene cluster share the same chromosome location and the codeletion of both clusters has been further reported [10][11][12][13]. IFN deletion implies the loss of an efficient antiviral response. The absence of the IFN cluster might make these tumors more vulnerable to virus oncolytic therapies.
Newcastle disease virus (NDV) has been proposed as an oncolytic agent for the treatment of different tumors, including GBM [14,15]. It is an avian paramyxovirus whose oncolytic potential derives in part from its capacity to preferentially replicate in tumor cells [16][17][18] and its inherent ability to trigger type I IFN response in humans, mediating the expression of numerous interferon-stimulated genes (ISGs) [19,20]. Many ISGs generate an antiviral state in the target cells blocking the replication of the virus [21]. Different studies have shown that deficiencies in the IFN induction and signaling result in the favorable replication of NDV in tumor cells [22,23]. The oncolytic properties of NDV have been improved using several strategies, including the genetic modification and the insertion of different genes with immunostimulating or immunomodulating properties [24].
In the present work, we have evaluated the genetic loss of the type I IFN locus in 1018 tumors whose copy number alteration (CNA) data are available in the TCGA database [25]. The analysis shows a correlation between CDKN2A deletion and the partial or total loss of the type I IFN gene cluster. GBM CSCs from different patients were evaluated for the presence of CDKN2A and IFNβ1 genes. We also evaluated whether the type I IFN gene cluster determines the ability of NDV virus to replicate and induce CSCs death. Interestingly, the poor NDV oncolytic activity on type I-IFN-competent GBM could be overcome using recombinant NDV expressing a type I IFN antagonistic protein such as the non-structural 1 (NS1) protein of the influenza virus. That viral protein is responsible for inhibiting host immune responses [26,27]. These results highlight the importance of evaluating IFN competence of tumor cells for the choice of a therapy based on virus oncolysis.

Bioinformatic Analysis of CNA
DNA copy number of 1018 glioma (astrocytoma n = 169, oligoastrocytoma n = 113, oligodendroglioma n = 172, and GBM n = 564) samples were obtained from the work of Ceccarelli et al. [25]. The amplification/deletion state of a set of IFN genes and CDKN2A, all of them located at chromosome 9p21.3, were rank normalized and then hierarchically clustered with centroid linkage and the usual metric.

gDNA Isolation and PCR
Genomic DNA (gDNA) was isolated using DNeasy Blood and Tissue Kit (QIAGEN) following the manufacturer's recommendations. All PCRs for the amplification of CDKN2A, IFNβ1 and β-actin were performed following the protocol of the Paq5000 enzyme (Agilent Technologies, Las Rozas de Madrid, Madrid, Spain).

Viruses
Recombinant NDV-LaSota-F3aa-GFP (rNDV) was rescued and grown as described previously [29]. Recombinant viruses NDV-LaSota -GFP (rNDV-GFP) and rNDV-LaSota-NS1 viral stocks were generated as previously described [30] and stored at −80 • C. Schematic representation of the viruses genome is provided in Figure S1. Stocks' titration was performed by indirect immunofluorescence using a rabbit polyclonal serum to NDV [31] on Hep2 infected-cells, 48 h after viral inoculation. TCID50 was calculated by the Reed-Muënch method [32]. The multiplicity of infection (MOI), which refers to the number of infectious particles per cell, has been indicated in each experiment.

Infectivity Assay
10,000 GBM cells were plated on a 96-well plate and infected with rNDV-GFP at 0.01 (low) or 1 (high) MOI. To avoid artifacts during cell fixation, 20 µL of DRAQ5 fluorescent probe (Thermofisher) was added to stain cellular DNA after 24-72 h of incubation and 15 min later different fluorescence images were taken.

Replication Assay
Viral-growth kinetics were performed in 6-well plates containing 80,000 cells per well. Cells were inoculated with a MOI of 0.01 in Opti-MEM ® media (Gibco) for 1 h at room temperature (RT). After inoculum removal, cells were maintained in M21 media. Supernatants were collected every 24 h for a total extent of one week and titrated by immunofluorescence as described before. Infected cells were harvested at the given time points, to perform gene expression analysis.

Cell Viability Assays
96-well plates were used to plate 10,000 cells/well in 100 µL of M21. Cells were infected at the indicated MOI using Opti-MEM ® media (Gibco). After one hour of incubation at RT, the inoculum was removed, and the cells were maintained in M21 media.
Cell viability was assessed at the indicated time points for an extent of 7 days by using MTS reagent (Promega, Madison, WI, USA) following manufacturer indications. Absorbance was recorded Cells 2020, 9,1405 4 of 17 at 490 nm and 630 nm using a spectrophotometer plate reader. Values were calculated in reference to the viability of mock-infected cells (negative control).

Gene Expression
RNA was extracted using RNeasy Mini Kit (Qiagen, Germantown, MD, USA) following the manufacturer instructions. After RNA isolation, DNAse treatment was performed using DNAse I (Qiagen) following manufacturer's instructions. RNA concentration and purity assessment were performed by using a Nanodrop (Thermofisher), then samples were stored at −80 • C. cDNA was generated from 1 µg of purified RNA by using high-capacity cDNA reverse transcription Kit (Applied Biosystems, Carlsbad, CA, USA) following manufacturer instructions. Quantitative PCR was performed using SYBR Premix Ex Taq (Takara, Kusatsu, Shiga, Japan) following manufacturer instructions in a 7900HT fast real-time PCR system (Thermofisher) using the primers needed to quantify the expression of the genes under analysis. β-actin and GAPDH were used as housekeeping genes. Primer couples used to quantify the expression of each gene were as follows: β-actinFw: CATCCCCCAAAGTTCACAAT and β-actinRv: ATGGCAAGGGACTTCCTGTA; GAPDH Fw: TCCTCCACCTTTGACGCTG;

IFN Treatment
CSCs were treated with 500 U/mL of recombinant IFN-β (PBL Assay Science, Piscataway, NJ, USA). Pictures were taken on experimental days 0, 2, and 7. Sphere size was measured with CellProfiler 3.1.9 (Broad Institute, Cambridge, MA, USA), objects higher than 1500 pixels were used for the analysis. Total RNA was isolated at 24 h for further gene expression analysis.
When tumors reached a standard size, mice received a single intratumoral dose of 5 × 10 4 iu/mL of rNDV-GFP virus or the corresponding HBSS controls. Images were analyzed using Image J software 1.48q (National Institute of Mental Health, Bethesda, MD, USA). Mice were euthanized when they presented neurological symptoms or a significant weight loss [≥20% of starting body weight].

Evaluation of NDV Toxicity in Mice
Total of ten Nude and five CD1 mice received a single intrathecal dose of 5 × 10 4 IU/mL rNDV-GFP virus as described before. Animals were euthanized when displayed distress symptoms or experienced 20% of body weight loss.

Statistical Analysis
Statistical analysis was performed using a 2-tailed Student t test and two-way ANOVA. Data are presented as means ± standard deviation and were calculated using the software package GraphPad Prism v. 5.0. (GraphPad Software, San Diego, CA, USA). Statistical values of p < 0.05 were considered significant.

Codeletion of CDKN2A and the Type I-IFN Cluster Is Very Common in Glioma
In order to study the prevalence of CDKN2A in brain tumors, we performed an in silico analysis of 1018 glioma tumors. Our results revealed that GBMs have the highest occurrence of CDKN2A gene deletion among gliomas (Figure 1a). 76% of the GBM tumors included in this analysis (n = 428) were CDKN2A deficient, with homozygous and heterozygous deletions being present on the 57% (n = 322) and 19% (n = 106) of them, respectively.
Because of the proximity of the type I IFN gene cluster to CDKN2A at 9p21 (Figure 1b), we decided to investigate whether the coverage of the deletion further included this set of genes ( Figure 1c,d). Gene clustering pointed out that CDKN2A gene state is coordinate with type I IFN genes and our analysis shows that most of the patients with heterodeletion in the gene CDKN2A presented also partial deletion of the type I IFN cluster. Moreover, the CDKN2A homozygous deletion in GBM and astrocytoma entail either partial or complete codeletion of the type I IFN cluster in more than 50% of these tumors (Figure 1d). In contrast, the incidence of type I IFN deficiency for oligoastrocytoma and oligodendroglioma groups remained lower ( Figure 1e).
In light of the results obtained in our in silico analysis, we decided to test if this phenomenon prevailed in a set of primary GBM CSCs derived from primary tumors of glioma patients. Considering that IFNβ1, the gene that codifies for IFN-β, is the one located the farthest from CDKN2A, we assumed that the absence of both genes implies the loss of the entire type I IFN cluster. The genetic analysis of these CSCs showed that three of them (GBM27, GBM128B, and GBM128D) presented a codeletion of the CDKN2A and IFNβ1 genes (type I non-competent CSCs). On the contrary, GBM18, GBM38, and GBM123 (type I-competent CSCs), displayed a preservation of both genes ( Figure 1f or Figure S2). Both sets of CSCs were used to investigate the role of CDKN2A-type-I IFN phenotypes.

Loss of the IFN Cluster Conditionates NDV Replication and Redefines the Antiviral Response of GBM Cells
In order to study the ability of NDV to replicate in GBM CSCs, we infected the cells at a low MOI (0.01) and monitored the viral growth kinetics as well as the induction of the transcription of IFN-β and MX1, a gene highly expressed exclusively under IFN detection, over a time course of 120 h. Consistent with previous data [23], NDV demonstrated limited replication capacity in type-I-IFN-competent CSCs. In contrast, NDV had higher titers and/or more prolonged replication in GBM27, GBM128B, and GBM128D, which correlated with the induction of IFN-β and MX1 gene expression (Figure 2a).
To characterize the type I IFN response in the different GBM CSCs, we infected them at a high MOI (4) and analyzed the expression of different interferon stimulated genes (ISGs) at 24 h after infection. As shown in Figure 2b, we could differentiate GBM cells into two clusters, according to their ability to express IFN-α, IFN-β, and different ISGs. As expected, both IFN-α and IFN-β genes were induced in type I-competent CSCs upon NDV infection. In comparison, GBM27, GBM128B, and GBM128D did not present a strong response to NDV infection. It can be noticed that GBM128D presents a small but significant stimulation of IFN-α and IFN-β genes ( Figure 2b).
As expected, the analysis of the expression of relevant ISGs involved in the cellular antiviral response to NDV, such as RIG-I, MDA5, STAT1, Mx1, OAS1, ISG54 or ISG15, shows an upregulation of these genes on GBM18, GBM38, and GBM123 cells. In contrast, the level of induction of the same genes in the type I-non-competent CSCs is considerably lower upon NDV infection.
We further investigate other important parameters of the response of cancer cells to NDV, including upregulation of MHC molecules and inflammatory cytokines. Surprisingly, MHC-I mRNA expression could not be detected in GBM18 and low expression values were observed in GBM123 cells. GBM38 showed a basal expression of MHC-I but this gene was not stimulated upon NDV infection. In comparison, MHC-I expression was significantly stimulated upon NDV infection in GBM27, GBM128B, and GBM128D. Nuclear factor kappa B (NF-κB)-dependent pro-inflammatory genes such as interleukin-6 (IL-6) and tumor necrosis factor (TNF-α) were also analyzed. All cell types were able to upregulate the expression of these genes upon NDV infection, except for GBM27. NDV infection can lead to apoptosis through the expression of FAS or TNF-related apoptosis inducing ligand (TRAIL) genes. The expression analysis of the cells showed that all cells were able to stimulate the expression of both genes with the exception of GBM27, that was unable to induce FAS expression, and GBM128B that did not stimulate TRAIL expression upon infection. However, GBM128D induction levels did not reach the basal level of GBM128B ( Figure S3).
Total of ten Nude and five CD1 mice received a single intrathecal dose of 5 × 10 4 IU/mL rNDV-GFP virus as described before. Animals were euthanized when displayed distress symptoms or experienced 20% of body weight loss.

Statistical Analysis
Statistical analysis was performed using a 2-tailed Student t test and two-way ANOVA. Data are presented as means ± standard deviation and were calculated using the software package GraphPad Prism v. 5.0. (GraphPad Software, San Diego, CA, USA). Statistical values of p < 0.05 were considered significant.

Codeletion of CDKN2A and the Type I-IFN Cluster is very Common in Glioma
In order to study the prevalence of CDKN2A in brain tumors, we performed an in silico analysis of 1018 glioma tumors. Our results revealed that GBMs have the highest occurrence of CDKN2A gene deletion among gliomas (Figure 1a). 76% of the GBM tumors included in this analysis (n = 428) were CDKN2A deficient, with homozygous and heterozygous deletions being present on the 57% (n = 322) and 19% (n = 106) of them, respectively.  To characterize the type I IFN response in the different GBM CSCs, we infected them at a high MOI (4) and analyzed the expression of different interferon stimulated genes (ISGs) at 24 h after infection. As shown in Figure 2b, we could differentiate GBM cells into two clusters, according to

Type-I Deficient GBM CSCs Are More Susceptible to NDV-Derived Oncolysis
A major feature of NDV's oncolytic activity is its capacity to induce cell death in tumor cells. In order to determine if the loss of the IFN cluster could affect the susceptibility of GBM cells to viral-induced cytopathic effect (CPE), we performed a MTS cell viability assay on CSCs infected at different MOIs. As observed in Figure 3a, all the GBM cells were susceptible to be infected by rNDV-GFP virus. However, a clear dependence on MOI was noted along all groups, in reference to CPE or loss of viability (Figure 3b).
GBM18 and GBM123 (type I-competent) CSCs displayed the highest level of resistance to NDV-induced cell death, showing significant loss of cell viability (≥50%) at a MOI higher than 1. In the case of GBM38, this effect was reached with a MOI of 0.1 (Figure 3b). In comparison, a MOI of 0.01 or lower is sufficient to reduce the cell viability below 50% in type-I-IFN non-competent CSCs. Noteworthy, GBM27 cells showed the highest susceptibility to NDV-induced cell death, with a reduction on cell viability higher than 50% at the lowest MOI used in this study (MOI of 0.001) (Figure 3b).
To better characterize the CPE induced by NDV, GBM CSCs were infected at low and high MOIs (0.01 and 1 respectively) and loss of viability was monitored every 24 h, over a time course of 120 h. Type-I-IFN-competent CSCs displayed a significant drop in viability (≥50%) at 48 h post-infection at a MOI of 1 (Figure 3c). However, and confirming our previous results, this type I-competent CSCs presented high resistance to virus when they were infected at a low MOI ( Figure S4). In contrast, GBM27, GBM128B, and GBM128D non-competent cells were more susceptible to the virus infection at a MOI of 0.01 than the IFN competent cells over the extent of the study. This effect was observed after 24 h of NDV infection in GBM128B and GBM128D. Instead, GBM27 presented a delayed susceptibility to the virus that is significative only after 48 h of NDV infection.
In order to confirm the lytic effect of NDV in vivo, we choose GBM18 and GBM27 luciferase-expressing cells, as type-I-IFN competent and non-competent CSCs respectively, to develop GBM xenotransplant tumors in mice. Tumor growth was monitored using an IVIS luminometer using the luciferase intensity of these tumors as a surrogate measure of the tumor size. When the tumors reached an average radiance of 6.77 × 10 3 [p/s/cm 2 /sr], they were injected with 5 × 10 4 IU/mL of rNDV-GFP. Seven days after intratumoral NDV inoculation, we observed a significant decrease in the luciferase expression only in the IFN deficient GBM27 xenotransplanted tumors (p < 0.001) (Figure 3d).

GBM Cells Are Able to Respond to Recombinant IFN-β Treatment
The classification of our GBM CSCs as type-I IFN competent or non-competent was based on the presence of IFNB deletion and, therefore, in their individual capacity to produce IFN-β. We have already described the autocrine effect of the IFN-β produced by these cells (Figure 1; Supplementary  Figure 1). However, whether or not these cells could respond differently to a stronger exogenous IFN-β stimulation needed to be stated. To investigate if the CSCs were able to respond to exogenous interferon, GBM cells were treated with 500 U/mL of recombinant IFN-β. As shown in Figure 4a, no morphological changes were noted between treated and untreated samples. However, an anti-proliferative effect of the addition of exogenus IFN-β was noticed for all the cell lines studied. This biological effect was statistically significant in GBM18 (p < 0.05) and GBM27 (p < 0.01). When the sphere diameters were measured, we observed a dramatical decrease in GBM27 (p < 0.01) and in the relative expression of stemness-associated genes (BMI1 and CD44). The addition of recombinant IFN-β also triggered a significant increase in the expression of MX1, OAS1, and ISG15 genes (Figure 4b). No dramatical changes were observed in the expression of pro-inflammatory cytokine IL-6 or chemokine CXCL1, except for the cell lines GBM123 and GBM27, in which CXCL1 levels only increased after addition of IFN-β.

NDV Lytic Activity Could Be Restored in IFN-Competent GBM Cells by Coexpressing A Type I IFN Antagonist
Based on our previous data, NDV infection has a stronger lytic effect and therapeutic potential on IFN-deficient GBM. Therefore, we decided to test whether a recombinant NDV (rNDV-NS1)

NDV Lytic Activity Could Be Restored in IFN-Competent GBM Cells by Coexpressing a Type I IFN Antagonist
Based on our previous data, NDV infection has a stronger lytic effect and therapeutic potential on IFN-deficient GBM. Therefore, we decided to test whether a recombinant NDV (rNDV-NS1) bearing the NS1 gene from influenza A virus, a strong antagonist of the IFN response, could exert higher oncolytic activity in these IFN-competent CSCs. The insertion of the NS1 into the NDV genome has previously probed to enhance the viral replication and lytic capacity of NDV in other human cancer cells [31].
For this purpose, we used GBM18 and GBM38 as IFN competent CSCs and GBM27 and GBM128D as deficient CSCs ( Figure 5 and Figure S6). As shown in Figure 5a, rNDV-NS1 viral titer was 1000-fold higher than rNDV-GFP virus in GBM18 and were more effective at high MOI reducing the number of viable cells below 20% at 120 h. The rNDV-NS1 decreased the viability in GBM38 drastically compared with rNDV-GFP at both MOIs. However, the growth kinetics of rNDV-GFP and rNDV-NS1 viruses did not change in GBM27 and GBM128D and similar viability effects were observed when these CSCs were infected with MOI 0.01 and 1 (Figure 5b or Figure S6). We further confirmed the IFN-β inhibition by mRNA expression and observed that in competent CSCs model GBM18 the expression decreased drastically when rNDV-NS1 was used in the infection (p < 0.001). As expected, no changes were observed in the non-competent CSCs model. bearing the NS1 gene from influenza A virus, a strong antagonist of the IFN response, could exert higher oncolytic activity in these IFN-competent CSCs. The insertion of the NS1 into the NDV genome has previously probed to enhance the viral replication and lytic capacity of NDV in other human cancer cells [31].
For this purpose, we used GBM18 and GBM38 as IFN competent CSCs and GBM27 and GBM128D as deficient CSCs ( Figure 5 and Figure S6). As shown in Figure 5a, rNDV-NS1 viral titer was 1000fold higher than rNDV-GFP virus in GBM18 and were more effective at high MOI reducing the number of viable cells below 20% at 120 h. The rNDV-NS1 decreased the viability in GBM38 drastically compared with rNDV-GFP at both MOIs. However, the growth kinetics of rNDV-GFP and rNDV-NS1 viruses did not change in GBM27 and GBM128D and similar viability effects were observed when these CSCs were infected with MOI 0.01 and 1 (Figure 5b or Figure S6). We further confirmed the IFN-β inhibition by mRNA expression and observed that in competent CSCs model GBM18 the expression decreased drastically when rNDV-NS1 was used in the infection (p < 0.001).
As expected, no changes were observed in the non-competent CSCs model. Figure 5. NDV-NS1 restored oncolytic activity in competent CSCs model. (a) rNDV-GFP and rNDV-NS1 viral growth curves were monitored for 120 h in GBM18 and GBM27 (b) GBM18 and GBM27 were exposed to NDV-GFP and NDV-NS1 MOI 1 and 0.01. Cell viability was measured every 24 h during 120 h (c) 100,000 CSCs per well were seeded in a 12 well plate and infected with NDV-GFP and NDV-NSI MOI 0.01. mRNA was isolated at 72 h. Error bar represents standard deviation. * p < 0.05, ** p < 0.01, *** p < 0.001.

Discussion
Despite the development of new drugs, GBM remains the most lethal brain tumor, mainly due to its resistance to conventional treatments, high heterogeneity and the presence of CSCs [33]. Although GBMs can be classified based on histology and molecular characteristics [34][35][36], there are no specific genetic biomarkers or signatures for treatment stratification and the survival rates have remained unchanged during the past years [37].
Our group has isolated and characterized several primary cultures of CSCs which nowadays are featured as the best GBM in vitro models [28]. In the present work we propose that NDV oncolytic therapy could be a good approach for the treatment of a specific subgroup of primary brain tumors. NDV-based antineoplastic therapy has demonstrated to elicit a robust long-term anti-tumor immunity in different preclinical tumor models [38,39]. In the present work, we present a new glioma stratification system based on the type I-IFN cluster genomic expression, that could classify patients as potential responders and non-responders to NDV therapy and possibly other oncolytic viruses. Figure 5. NDV-NS1 restored oncolytic activity in competent CSCs model. (a) rNDV-GFP and rNDV-NS1 viral growth curves were monitored for 120 h in GBM18 and GBM27 (b) GBM18 and GBM27 were exposed to NDV-GFP and NDV-NS1 MOI 1 and 0.01. Cell viability was measured every 24 h during 120 h (c) 100,000 CSCs per well were seeded in a 12 well plate and infected with NDV-GFP and NDV-NSI MOI 0.01. mRNA was isolated at 72 h. Error bar represents standard deviation. * p < 0.05, ** p < 0.01, *** p < 0.001.

Discussion
Despite the development of new drugs, GBM remains the most lethal brain tumor, mainly due to its resistance to conventional treatments, high heterogeneity and the presence of CSCs [33]. Although GBMs can be classified based on histology and molecular characteristics [34][35][36], there are no specific genetic biomarkers or signatures for treatment stratification and the survival rates have remained unchanged during the past years [37].
Our group has isolated and characterized several primary cultures of CSCs which nowadays are featured as the best GBM in vitro models [28]. In the present work we propose that NDV oncolytic therapy could be a good approach for the treatment of a specific subgroup of primary brain tumors. NDV-based antineoplastic therapy has demonstrated to elicit a robust long-term anti-tumor immunity in different preclinical tumor models [38,39]. In the present work, we present a new glioma stratification system based on the type I-IFN cluster genomic expression, that could classify patients as potential responders and non-responders to NDV therapy and possibly other oncolytic viruses.
As more than 50% of GBM patients present a loss of heterozygosity (LOH) at chromosome 9p21 which harbors IFN-I cluster [40,41], the six CSCs models used in this study are representative of the equivalent human phenotype.
The results of the study indicate that there are two clusters of GBM CSCs with different degrees of susceptibility to NDV infection. One in which the NDV replication process occurs completely and another cluster in which the replication is abrogate. These differences could be explained by the induction of type I IFNs. The secretion of these type I IFN cytokines can trigger an antiviral immune response which activates type I IFN signaling and the induction of ISGs that blocks viral replication [42]. As it is known that the IFN-I signaling regulates the expression of TNF-α, FAS ligand and TRAIL [43], that effect is related with the high levels of those pro-apoptotic genes expressed by the cluster composed by GBM18, GBM38, and GBM123 upon NDV infection. Moreover, GBM tumors with high levels of IFN-β are more sensitive to TMZ and have better overall survival [44]. Thus, for the poor prognosis group composed by the type I IFN non-competent or deficient CSCs, NDV therapy could serve as a potential therapeutic option.
In line with previous researchers that demonstrate the NDV oncolytic effect in orthotopic GL261 mouse glioma model with a meaningful impact on overall survival [15], we observed a clear effect in tumor size reduction in type I IFN-deficient cell line (GBM27) model as compared to the type I IFN-competent tumors. Unfortunately, high toxicity levels were developed in these immunocompromised nude mice and we were unable to monitor survival rates. To demonstrate that the toxicity detected was due to the lack of cell-mediated immunity in that mice strain, we treated CD1 mice with the same virus concentration and no harm was detected ( Figure S5). However, additional research is required in immunocompetent murine models to further confirm these data.
The use of NDV oncolytic virotherapy in several clinical trials has been presented as a safety strategy for treating solid tumors [45], such as GBM [46,47], melanoma [48], renal cell carcinoma [49], and ovarian cancer [50], among others. This therapy seems to have a high response heterogeneity in patients. Based on the data presented here, prior characterization of the type of tumor with regard to the presence or absence of the type I IFN cluster prior to the treatment with NDV (and possibly with other oncolytic tumors) is strongly suggested in order to better predict the patients' response.
Other oncolytic viruses have shown efficacy in treating GBM and pediatric brain tumors mediated by direct oncolysis followed by an immune response [51][52][53]. Notably, the use of virotherapy together with radio and chemotherapy has produced a synergistic effect [54].
Tumors that lack type I IFN induction response will be more receptive to replicate the virus and get a better oncolytic activity as compared to those that are type I IFN-competent. In the case of the type I IFN-competent tumors, the viral replication would be limited, and the oncolytic effect less effective. Still, the induction of the immune response against the infected cells may trigger the activation of immune cells against the tumor and show oncolytic effect.
Type I IFN induction has important roles in tumor immunogenicity such as antigen presentation, maturation of dendritic cells (DCs), activation and survival of memory cytotoxic T lymphocytes, NK cell activation, or neutrophil recruitment [55,56]. Loss of type I IFN cluster by cancer cells partially disables the ability of viral infection to trigger tumor interaction with the adaptive immune response [57]. In this scenario, production of type I IFN relies on non-tumor cells such as tumor infiltrated macrophages and DC or other cells present at the tumor microenvironment that do not have the genetic alterations of the cancer cells.
The tumor microenvironment plays a key role in participating in the response and secretion of type I IFN to the tumor stroma [58]. As we have seen, there are GBM CSCs with CDKN2A-Type I IFN gene cluster codeletion, that could activate the signaling pathway in response to exogenous IFN-β and could be potentially activated by other type I IFNs. In the type I IFN-deficient tumors, NDV infection could trigger type I IFN production by other tumor-infiltrating immune cells. IFN production in this context will induce type I IFN signaling that can trigger ISG expression and induce a program in the tumor microenvironment that could control the virus replication and also generate an antitumor immune response. As a consequence, IFN produces anti-proliferative effects, described since 1962 [59], and decreases stem properties as it has been reported in other pathologies [60]. These antiproliferative effects are not the same in all our GBM CSCs, which opens another factor of heterogenicity in the oncolytic response to the virus. As mentioned above, NDV therapy reduced CSCs self-renewing capacity and increased their ability to differentiate. Targeting these mechanisms could be used in differentiation-based cancer therapy, triggering synergistic antitumor effects [61].
We have confirmed that IFN I responses underlie GBM CSCs selective replication of NDV. Based on this GBM classification, it is possible to employ alternative strategies that could bypass the IFN-competent GBMs. We have proved that type I IFN-competent GBM cells refractory to NDV can become susceptible to the virus if the type I IFN system is inhibited in those cells by other means. In this case, we have used a recombinant NDV expressing the influenza NS1 gene that blocks IFN induction (rNDV-NS1) [62]. This recombinant virus has an oncolytic effect in both type I IFN-competent or deficient CSCs. rNDV-NS1 has been used before, showing improved oncolytic efficacy as a therapeutic and safe agent as compared to a NDV control [24,63]. Increasing the oncolytic potential of NDV by inserting type I IFN antagonistic proteins will need further characterization in GBM mouse models as the immune privileged brain environment response to these modified viruses is uncertain.
In conclusion, our data highlight the importance of a glioma patient stratification based on the CDKN2A-IFN gene cluster deletion. That new classification could predict the patients that might potentially respond to NDV oncolytic therapy and possibly use it as a sensitivity biomarker to other oncolytic viruses [64]. For optimal patient management in GBM clinical practice, accurate CDKN2A CNA analysis prior to treatment is recommended to identify potential NDV non-responder patients and exclude them. The combination of conventional therapies with NDV oncolytic agent may provide a new promising approach for reducing GBM recurrence and improving the overall survival rate.