Oncolytic vaccinia virus as a vector for therapeutic sodium iodide symporter gene therapy in prostate cancer

Oncolytic strains of vaccinia virus are currently in clinical development with clear evidence of safety and promising signs of efficacy. Addition of therapeutic genes to the viral genome may increase the therapeutic efficacy of vaccinia. We evaluated the therapeutic potential of vaccinia virus expressing the sodium iodide symporter (NIS) in prostate cancer models, combining oncolysis, external beam radiotherapy and NIS-mediated radioiodide therapy. The NIS-expressing vaccinia virus (VV-NIS), GLV-1h153, was tested in in vitro analyzes of viral cell killing, combination with radiotherapy, NIS expression, cellular radioiodide uptake and apoptotic cell death in PC3, DU145, LNCaP and WPMY-1 human prostate cell lines. In vivo experiments were carried out in PC3 xenografts in CD1 nude mice to assess NIS expression and tumor radioiodide uptake. In addition, the therapeutic benefit of radioiodide treatment in combination with viral oncolysis and external beam radiotherapy was measured. In vitro viral cell killing of prostate cancers was dose- and time-dependent and was through apoptotic mechanisms. Importantly, combined virus therapy and iodizing radiation did not adversely affect oncolysis. NIS gene expression in infected cells was functional and mediated uptake of radioiodide both in vitro and in vivo. Therapy experiments with both xenograft and immunocompetent Transgenic Adenocarcinoma of the Mouse Prostate (TRAMP) mouse models showed that the addition of radioiodide to VV-NIS-infected tumors was more effective than each single-agent therapy, restricting tumor growth and increasing survival. In conclusion, VV-NIS is effective in prostate cancer models. This treatment modality would be an attractive complement to existing clinical radiotherapy practice.


INTRODUCTION
The sodium iodide symporter (NIS) membrane protein is responsible for the uptake of iodide by the thyroid tissue. In cases of papillary and follicular thyroid cancer, radioactive Iodine-131 ( 131 I) is used to ablate residual normal (and malignant) thyroid tissue after thyroidectomy, with high rates of efficacy. 1 Targeted delivery of the NIS gene as a mode of cancer gene therapy represents an opportunity to bring the proven therapeutic potential of I 131 to bear against other cancer types.
Oncolytic virotherapy exploits the natural or engineered affinity of certain viral strains selectively to infect and replicate in cancer cells and, in doing so, to kill them. In clinical testing, they have been shown to have excellent safety profiles and promising signs of efficacy. 2,3 The adenovirus H101 has been approved in China for use in head/neck cancer. 4 A positive phase III trial of talimogene laherperepvec (T-Vec, herpes simplex virus type 1) has been reported 5 and this agent received approval by the European Medicines Agency in October 2015 for patients with malignant melanoma. 6 Vaccinia virus belongs to the Poxviridae family and possesses a large linear double-stranded DNA genome consisting of~250 genes, with capacity for insertion of therapeutic transgenes, such as the NIS gene. 7 Vaccinia has been administered widely as the smallpox vaccine and, as such, it has an excellent safety profile. 8 It has also been examined extensively in attenuated forms as an oncolytic agent with comparable safety. 9 The complex life cycle of Vaccinia includes dual mechanisms of infection by separate forms of infectious particles. Intracellular mature virions are the main product of viral lysis and extracellular enveloped virions are actively shed by infected cells. 10 Compared with other agents, Vaccinia offers a number of potential advantages including rapid replication in and lysis of infected cells, the ability to achieve high levels of viral gene expression, the capacity to spread cell-to-cell and the fact that its activity is unhindered by hypoxia 11 and therapeutic irradiation. 12 Genetic modification of Vaccinia to express the NIS gene represents a further refinement of its therapeutic potential by giving it the capacity to drive cellular 131 I uptake for direct killing of infected cells and indirect killing of neighboring cells within the 0.8 mm range of the emitted β particles. 13 Previous studies have explored the potential of NIS, delivered by a range of oncolytic viruses including measles, HSV and VSV as a therapeutic reporter gene and as a therapeutic agent. [14][15][16][17][18] Oncolytic vaccinia virus has been studied in vitro in a range of tumor types, enabling positron emission tomography and single photon emission computed tomography observation of viral kinetics using a variety of radioiosotopes including 131 I, 124 I and 99m Tc. [19][20][21] This has been shown to be a viable imaging method in a phase I/II trial of measles virus strains encoding NIS in ovarian cancer patients 18 and would be a useful safety-monitoring tool to confirm that viral biodistribution in other human trials is as 1 expected. Furthermore, oncolytic vaccinia enabled NIS therapy has shown additional benefit of 131 I administration in pancreatic and breast cancer models. 21,22 Prostate cancer is the commonest form of male cancer and the second highest cause of cancer death in the United States, with 240 000 new cases and 28 000 deaths annually. 23 Currently, prostate cancer treatment typically involves radical prostatectomy or radiotherapy with good survival outcomes for the 90% of patients whose disease is diagnosed at the local/regional stage. 1 However, the side-effects of such treatments can be significant and include incontinence, bowel complications and erectile dysfunction, with associated long-term detriment to quality of life. The prognosis for those patients who develop castration resistant disease is poor. 24 Despite recent advances in medical therapies, 25-28 men with prostate cancer will ultimately develop treatment-refractory, incurable disease. Therefore, there is a need for novel therapies with improved side-effect profiles in locoregional disease and improved efficacy in metastatic disease. Prostate cancer has been targeted for NIS gene therapy in numerous pre-clinical studies. 29 Using adenovirus as a vector, NIS gene expression in prostate tissue has, in a Phase 1 trial, proven the 99mTc-imaging approach to be both safe and feasible. 30 Further study is ongoing. 31 To date, no human trials have studied the potential of oncolytic viral therapy to additionally enable NIS 131 I therapy.
In this study, we examine the therapeutic potential of the NIS-expressing Vaccinia virus (VV-NIS), GLV-1h153 (VV-NIS), as an oncolytic agent and as a vector for targeting NIS gene therapy to prostate cancer cells in vitro and in vivo. We demonstrate efficient time-and dose-dependent marker gene expression and oncolysis of a panel of human prostate cell lines and confirm that these virally mediated effects are not affected by combination with radiotherapy. Combined GLV-1h153 and radiation therapy is shown to enhance apoptosis of prostate cancer cells. NIS is shown to be both expressed and functional, enabling the prostate cells to concentrate radioiodide. In vivo NIS gene expression and iodide uptake is demonstrated in xenograft tumors of PC3 cells and therapeutic experiments show the combination of virus and 131 I to be significantly more effective against these tumors than either therapy alone. In the immunocompetent Transgenic Adenocarcinoma of the Mouse Prostate (TRAMP) model of prostate cancer this radio-virotherapeutic approach also proved to be effective.

RESULTS
GLV-1h153 cytotoxicity in prostate cancer cell lines A panel of four prostate cancer cell lines, DU145, PC3, LNCaP and WPMY-1, was tested for their susceptibility to the GLV-1h153 virus. The cells were infected with virus at multiplicities of infection (MOI) ranging between 0.0001 and 10 for periods ranging from 24 to 72 h. Reduction in cell proliferation relative to uninfected controls was then assessed by 3-(4,5-dimethylthiazol-2-yl)-2,5diphenyltetrazolium bromide (MTT) assay. The virus was effective against all four cell lines, killing in a dose-and time-dependent manner (two-way analysis of variance (ANOVA) P o0.0001 for both virus and time factors as sources of variation for each cell line individually) ( Figure 1). DU145, LNCaP and WPMY-1 showed a positive interaction between time and dose (two-way ANOVA Po 0.0001), producing increased cell death over time. PC3 were somewhat resistant to this effect, especially at MOI 0.001-0.1.

NIS mRNA expression in GLV-1h153-infected cells
The cell lines DU145, PC3 and WPMY-1 were infected with GLV-1h153 MOIs ranging between 0.001 and 1 for 24 h before levels of NIS mRNA were quantified by quantitative reverse transcriptase PCR. Levels of mRNA detected correlated positively with increasing MOI of the virus (Supplementary Figure 1). There was no NIS mRNA detectable in the uninfected cells.
In vitro 131 I uptake in GLV-1h153-infected cells Confocal microscopy was used to look directly at WPMY-1, PC3 and DU145 cells during infection to determine the levels of H2Ax foci (DNA double-strand break marker) induced by either viral infection or 131 I uptake (Figure 2a). Cells irradiated with 2 Gy Vaccinia NIS therapy in prostate cancer DC Mansfield et al exhibited H2Ax foci, whereas those incubated with 131 I showed very few. However, when infected cells were incubated with 131 I many more H2Ax foci were observed. These images were then quantified to compare levels of DNA damage in sub-groups of cells ( Figure 2b). First, infected cells, determined by green fluorescent protein (GFP) expression, were compared with noninfected cells. Infected WPMY-1 and PC3 cells were more susceptible to DNA damage by external beam irradiation. DU145 showed a similar profile although this was not statistically significant. In infected DU145 cells incubated with 131 I, noninfected cells had a greater number of H2Ax foci, though this was not seen in WPMY-1 and PC3, where infected cells had a greater proportion of foci. With infected cells accumulating iodide and appearing to cause an increase in DNA damage, it was hypothesized that uninfected cells closer to infected ones should receive a greater dose of radiation than those further away, according to the inverse square law of radiation. We hypothesized that this would account for uninfected DU145 cells receiving a greater proportion of the 131 I-induced DNA damage than the infected cells. To evaluate this, the confocal images were again quantified, with a perimeter of 20 μm around infected cells in which any other cell nuclei would be classed as a 'bystander', cells wholly outside the perimeter were 'non-bystanders' and infected cells were excluded. The bystander DU145 cells exhibited sixfold more H2Ax foci than non-bystanders when exposed to 131 I. This effect was also seen to a lesser extent with external beam radiation, suggesting that infection somehow radiosensitises the non-infected DU145 cells, though this was not statistically significant. WPMY cells did show a significant radio-senitization effect on bystander cells, whereas PC3 did not. No significant bystander effect was observed in the WPMY or PC3 bystander cells exposed to 131 I, although in these cell lines there were few foci in non-infected cells. As another, more general, measure of bystander effect, the images were quantified to compare the total number of H2Ax foci in the field of view to the total number of infected cells (Supplementary Figure 2). Comparison by this method revealed a positive correlation between number of infected cells and H2Ax foci in DU145 cells, and no such correlation in WPMY-1 and PC3, thus supporting the above data.
To verify the functional expression of the NIS transgene, in vitro iodide uptake assays were performed. Cells were irradiated with 5 Gy (or not) and then infected at MOIs of 0.01, 0.1 and 1 for 24 or 48 h. At this time, they were treated with 131 I and radioiodide uptake was measured by gamma emissions from the cells. At 24 h, all four cell lines showed NIS expression and iodide uptake at MOI 1, with a highly significant difference (two-way ANOVA P o 0.001) compared with cells treated with the competitive inhibitor of iodide uptake (potassium perchlorate) (Figure 2c). LNCaP also had significant uptake with MOI 0.1 at 24 h (two-way ANOVA P o 0.01). At 48 h, all four cell lines showed an increase in iodide uptake at MOI 0.1 compared with that seen at 24 h. Cells irradiated with 5 Gy showed little difference in the levels of iodide uptake except in DU145 at 48 h where the level of uptake at an MOI of 0.1 was significantly lower.
GLV-1h153 cytotoxicity in combination with iodizing radiation To confirm that iodizing radiation does not affect the ability of the virus to replicate, express transgenes and cause cell death in prostate cells, external beam irradiation was used as a substitute for 131 I to allow controlled and equal exposure of all cells to iodizing radiation. The range from 1 to 5 Gy was selected to represent radiation doses that would not immediately kill the cells and should allow viral infection to be productive and potentially show synergy as previously published in other models. 12,32 The panel of cell lines was infected with MOIs ranging between 0.01 and 1, either 4 h pre-or post-irradiation with 1, 3 or 5 Gy. Cells were incubated for periods of 24, 48 and 72 h and cell proliferation was assessed by 3-[4,5-dimethylthiazol-2-yl]-2,5 diphenyl tetrazolium bromide (MTT) assay. Cells irradiated prior to infection all showed a significant dose-dependent effect of virus at all time points (two-way ANOVA P o0.0001). Irradiation had a significant dose-dependent effect in LNCaP and WPMY-1 at 24 and 48 h; LNCaP, WPMY-1 and PC3 at 48 h (Supplementary Figure 3); and all four cell lines at 72 h (two-way ANOVA P o0.0001) (Figure 3a). Irradiation never significantly reduced the viral effect. The alternative scheduling, irradiation 4 h post infection, yielded broadly similar results and indicated no benefit of one schedule over the other (Supplementary Figure 4).
The sulforhodamine B assay was also used to confirm the results observed with the MTT assay. Radiation was delivered 4 h preinfection and cell number was measured at 48 h. These data also showed a significant radiation dose effect (two-way ANOVA P o0.0001) and confirmed that the viral effect is unaffected by irradiation ( Figure 3b).
Clonogenic assays were performed with the two cell lines that were capable of colony formation (LNCaP and PC3) in order to determine the longer-term effects of the virus-radiation combination on cell survival. These data show that the clonogenic potential of cells was reduced by irradiation and by viral infection in dose-dependent fashion (for each agent). The combination of the two treatments reduced the clonogenic capacity of the cells to a greater extent than either treatment alone ( Figure 3c).
Expression of viral transgenes in combination with iodizing radiation Chlorophenolred-β-D-galactopyranoside assay was used to confirm that viral transgene expression was not adversely affected by irradiation. The cell line panel was irradiated with between 1 and 5 Gy 4 h prior to infection with MOIs ranging between 0.016 and 1. Induction of apoptotic cell death by GLV-1h153 and iodizing radiation The panel of cell lines was analyzed for the apoptotic response to virus at MOIs between 0.01 and 1 and irradiation between 1 and 5 Gy. The response varied widely between cell lines ( Figure 5a). By Caspase Glo3/7 assay, DU145 showed strong (up to 5.5-fold) Caspase-3/7 induction in response to virus (two-way ANOVA P o0.0001), but no significant radiation effect was observed. LNCaP also had a small but significant response (up to 1.4-fold) to virus only (two-way ANOVA P o 0.0001). PC3 had no Caspase-3/7 response to either virus or irradiation; and WPMY had a significant response (up to twofold) to irradiation only (two-way ANOVA P = 0.0137).
Western blots for the active cleaved form of Caspase-3 correlated closely with the Caspase-Glo data, showing no functional Caspase-3 in PC3 cells and replicating the patterns of sensitivity observed in the remaining cell lines. It has previously been published that synergy between radiation and vaccinia virus in BRAF mutant melanoma is dependent on signaling via JNK and TNF-α. 32 Phospho-JNK western blots showed that JNK becomes highly active in response to virus in all cell lines, though no clear radiation effect was observed (Figure 5b).  In vivo viral gene expression and 131 I uptake in GLV-1h153infected tumors The PC3 cell line was selected for in vivo experiments, as it had high levels of viral gene expression persisting for 72 h post infection in vitro. Xenografts in CD1 nu/nu mice were allowed to grow to 5-10 mm in diameter before intratumoural injection of virus (1 × 10 6 plaque forming units (PFU)) or control injection. After 48 h, 131 I was administered intraperitoneally as described. A further 48 h later, tumor iodide uptake and GFP expression were measured (Figures 6a and b). Tumors injected with virus had gamma emissions 2.7-fold higher than control tumors (t-test P o0.05). There was no significant difference in emissions from thyroid tissue between the two treatment groups. In tumors injected with virus, GFP expression was observed at low levels.

Combination of viral infection and irradiation
Bioluminescence generated by viral Renilla luciferase indicated that viral replication and gene expression is limited to the tumor  tissue only (Figure 6d). Uptake of 131 I in non-NIS-expressing tissue was minimal, with tumor tissue uptake being twice that in the blood (Supplementary Figure 8).
GLV-1h153 and 131 I combined treatment in PC3 xenografts PC3 xenografts in CD1 nu/nu mice were allowed to grow to 5-10 mm in diameter. Mice were divided into four groups, two of which received intratumoural injection of virus. After 48 h, one treated group and one non-treated group received 131 I. Tumors treated with 131 I only continued to grow at the same rate as the controls (Log-Rank P = 0.41). Virus alone slowed growth marginally and significantly improved survival to 60% at study termination, 60 days post treatment (Log-Rank P o0.05). The virus and 131 I treatment combination stabilised tumor growth and improved survival at study termination to 80% versus 0% in the control group (Log-Rank P = 0.004), with all surviving mice having stable or completely regressed tumors (Figure 6c).
GLV-1h153 biodistribution from PC3 xenografts PC3 xenografts in CD1 nu/nu mice were allowed to grow tõ 10 mm in diameter and then received intratumoural injection of virus (1 × 10 6 PFU). At 1 h and at 4, 6 and 9 days post virus administration 100 μg coelenterazine, the substrate for the viral encoded Renilla luciferase, was administered to the animals to allow bioluminescent detection of areas of viral gene expression (Figure 6d and Supplementary Figure 9). These images show that viral replication was limited to the tumor tissue only and did not disseminate to other tissues. Low levels of gene expression were detected in 75% of the animals at the 1 h time point, and the level of gene expression increased markedly over the following time points.
GLV-1h153 in TRAMP-C cell lines Following the positive results from the PC3 model in immunodeficient CD1 nu/nu mice, we considered that a fully immunocompetent model would more closely reflect the challenges faced by the virus in a clinical setting. The TRAMP transgenic model of murine prostate cancer was selected.
To test the species specificity of GLV-1h153, three cell lines derived from TRAMP tumors (TRAMP-C1, -C2, -C3) were challenged with the virus. The TRAMP-C cell lines were susceptible to infection and viral GFP expression was observed 24 h post infection (Figure 7a, Supplementary Figure 10). DNA damage due to radioiodide uptake was also seen in both infected cells and bystander cells.  In cell proliferation assays (Figure 7b), the TRAMP-C1 line showed some resistance to viral infection and a slower death as compared to the data for the human cell lines at 48 h post infection. Tramp-C2 and -C3 showed improved viral cell kill. Infection following exposure to iodizing radiation appeared to produce an additive effect on cell death in most cases.
GLV-1h153 and 131 I combined treatment in TRAMP mice The TRAMP mouse strain was used to evaluate the efficacy of the virus in an immunocompetent mouse model of prostate cancer. Pilot experiments showed that TRAMP mice aged between 29 and 31 weeks showed clear histological signs of either a tumor or extensive neoplasia in the prostate and associated tissues, that is, the seminal vesicles as has been described. 33 It was, therefore, selected as the optimal time point for intervention. Animals Histological analysis revealed extensive hyperplasia of the prostate (Figure 7d) and seminal vesicles, with metastatic disease present in the lung, liver and lymph nodes (Supplementary figure  12-14). Immunohistochemical staining for the proliferation marker Ki67 confirmed prostate and seminal vesicle tissues to be highly proliferative, as well as all metastatic deposits (Supplementary figure 15-17). Immunohistochemical staining for probasin showed that probasin expression is lost as the tumor tissues become less differentiated and metastatic deposits usually did not express probasin (Supplementary figure 18-20). Normal liver tissue appears to stain positive for probasin, though this can be explained as false positive owing to the presence of probasinrelated antigen in rodent liver cells. 34 The prostate tissues treated with GLV-1h153 appear to have retained more differentiation than the control tissues and more so in the tissues treated with the combination of GLV-1h153 and 131 I.

DISCUSSION
We have shown that oncolytic NIS-expressing vaccinia virus exerted significant activity against prostate cancer as a singleagent, in combination with external beam radiotherapy, and with therapeutic radioiodide. VV-NIS had potent antitumour efficacy against prostate cancer cells in vitro, which was most apparent at later time points, presumably as a result of ongoing viral replication (Figure 1). The presence of the NIS gene gave the virus the ability to accumulate radioiodide within infected cells and, importantly, we showed that resulted in DNA damage in uninfected bystander cells in close proximity (Figure 2). The observation that the bystander DU145 cells exhibited a greater number of γH2Ax foci could be explained by the cell cycle of the infected cells being arrested, compared with the bystander cells that are actively cycling and experiencing replication stress that would degrade single-strand breaks into double-strand breaks and thus result in more γH2Ax foci. The same effect was not seen in the other cell lines, perhaps simply because of biological variations in the rates of cell cycle or DNA repair mechanisms. This effect, which was seen in a two-dimensional in vitro assay, was due to emission of β-particles by 131 I and is likely to be a significant underestimate of what would occur in vivo within the densely  packed three-dimensional tumor environment. In combination with external-beam radiotherapy, viral oncolysis was not inhibited and the combination effectively reduced the clonogenic potential of irradiated cells (Figures 3 and 4). The mechanism of cell death involved signaling through apoptotic pathways. However, in contrast to our previous observations in melanoma cells that showed that vaccinia virus and radiation acted synergistically by modulating JNK/ TNFα signaling, 32 we did not see evidence of this effect in prostate cancer cell lines ( Figure 5). This highlights the fact that different tumor types are likely to show different combinatorial effects with virus plus radiotherapy and/or chemotherapy. We also provided in vivo evidence that intratumourally delivered VV-NIS was capable of driving iodide uptake in xenograft tumors and that this translated to a powerful antitumour effect. Indeed, we were able to demonstrate that the combination of VV-NIS and radioiodide was capable of delaying tumor growth significantly and of achieving long-term control in PC3 xenograft tumors ( Figure 6). These data were consistent with in vivo evidence of gene expression-measured by GFP imaging and radioiodide uptake assay ( Figure 6). To evaluate off-target viral effects, viral luciferase was used to produce bioluminescent imaging of areas of viral replication. Viral replication was high in the tumor and undetectable in other organs, indicating little or no capacity for the virus to leak from the tumor to establish infection at other sites, even in immunedeficient animals. Therefore, there is no evidence to suggest that there would be any significant offtarget accumulation of iodide in any organs other than perhaps the thyroid, which as previously mentioned has not caused adverse effect in these studies. Having demonstrated this in an immunodeficient animal model, we were keen to test the therapeutic effect in immunocompetent TRAMP mice. In these experiments, we were able to confirm that intravenously delivered virus significantly extended life-span, even as a single agent. Importantly, when combined with radioiodide long-term survival was further improved, quadrupling survival compared with control animals and with 50% of the combination group surviving at the day 200 termination of the experiment (Figure 7). The 131 I dose used in these in vivo studies is scaled down from the typical human dose and is in the range used in similar studies. 35,36 It is notable that those animals that received 131 I showed no sign of late adverse effects of the treatment, indicating that any undetected off-target effects were minimal. This is reflected in clinical radiotherapy where 131 I is used in patients with thyroid gland in situ on a regular basis without any long-term complications. Given that the TRAMP-C cell lines appeared relatively resistant to viral oncolysis, we hypothesize that the efficacy observed in the TRAMP model may be related to immunemediated effects rather than direct oncolysis. The positive role of the immune system in the efficacy of oncolytic viruses has been reported in several recent studies [37][38][39][40][41] and certainly is an avenue requiring more research, in which immunocompetent models such as the TRAMP mice and novel immune-checkpoint inhibitors will be valuable.
It had been our intention to test VV-NIS-driven radioiodide therapy combined with external-beam radiotherapy, but the efficacy of the individual component parts meant that this was not feasible. However, based on the studies reported here, we believe that viral-mediated radioiodide therapy of prostate cancer has the potential to contribute significantly to the radiation dose delivered by external-beam radiotherapy. This would theoretically improve patient outcomes (for example, progression-free and overall survival) by improving the ability of radiotherapy to secure locoregional control of prostate cancer. In addition, the selective replication-competence of VV-NIS would ensure that normal tissue toxicity would be minimized.
In summary, the data presented here not only contribute further evidence of single-agent efficacy of oncolytic vaccinia, but also point the way to further pre-clinical and clinical studies of oncolytic viruses in radio-virotherapeutic approaches. These further studies would build on the promising, but numerically limited, early-phase clinical studies utilizing adenovirus and measles which, to date, have not explored the potential additional benefit of 131 I therapy, but have successfully demonstrated functional NIS transgene expression using non-therapeutic imaging isotopes. 18,30 In addition, the growing realization that oncolytic virotherapy represents a form of 'oncolytic immunotherapy' means that further translational studies of VV-NIS, radiotherapy and other immune therapies (for example, checkpoint inhibitors) should be a priority for researchers in this field. Prostate cancer will be an excellent model system to investigate these therapies, given its accessibility for direct intratumoural injection and needle-core biopsy.

Cell lines
The following cell lines were used in these studies: Viral plaque assays Viral titers were determined by viral plaque assay. Monolayers of CV1 cells in 24-well plates were treated with serial dilutions of the viral solution to be titred and incubated for 24 h. Cells were fixed with 2% formaldehyde (Sigma-Aldrich, Dorset, UK)/0.2% glutaraldehyde (Sigma-Aldrich) in phosphate-buffered saline (PBS, Sigma-Aldrich) and stained for 4 h with 5 mM potassium-hexa-cyanoferrat III (Sigma-Aldrich), 5 mM potassiumhexa-cyanoferrat II-tri-hydrate (Sigma-Aldrich), 2 mM magnesium chloridehexahydrate (Sigma-Aldrich) and 0.6 mg ml − 1 5-Bromo-4-chloro-3-indolyl β-D-galactopyranoside (X-Gal, CalBioChem, Watford, UK) in PBS. Cells were then washed with ultrafiltered water and dried. Macroscopic X-gal-stained viral plaques were counted manually. Sulforhodamine B cytotoxicity assay Cell viability was quantified by fixing cells with 10% trichloroacetic acid (Sigma-Aldrich) and staining with 0.05% sulforhodamine B in 1% acetic acid (Sigma-Aldrich). The sulforhodamine B bound to cells was dissolved with 10 mM TRIS and absorbance was measured at 510 nm on a SpectraMax M5 plate reader.

Clonogenic assay
Cells were plated at 5 × 10 5 in a T25 flask (Corning, Glendale, AZ, USA). The next day, cells were irradiated at 2 Gy and 6 h later infected with GLV-1h153 at MOI of 0.01. After 48 h of treatment, cells were washed in PBS, trypsinized and counted using a haemocytometer. Cells were then plated into six-well dishes at 400-800 cells per well. After 10-14 days, plates were stained with 0.2% crystal violet (Sigma-Aldrich) in 7% ethanol and the colonies consisting of~50 cells or greater were counted manually.
Caspase-3/7 luminometry assay Relative levels of Caspase-3/7 activation 48 h post infection were measured by Caspase-Glo3/7 Assay (Promega, Southampton, UK), in which a proluminescent caspase-3/7 DEVD-aminoluciferin substrate is cleaved by active caspases 3/7, producing free aminoluciferin as a luciferase substrate. The detected levels were normalized according to the proportion of surviving cells, as determined by a tandem MTT assay carried out simultaneously.

Western blotting
Radioiodide uptake assay Cells were plated in 24-well plates (Nunc) at 1 × 10 5 cells per well in 500 μl DMEM and incubated overnight before infection with virus. At the experimental end point, 1 μCi (0.37 MBq) of 131 I (Perkin Elmer, Waltham, MA, USA) was added to each well for 1 h. Duplicate wells received 131 I premixed with 50 μM potassium perchlorate (Sigma-Aldrich). All wells were washed with copious amounts of PBS before detaching cells with 100 μl 1 M sodium hydroxide (Sigma-Aldrich) for 20 min. Gamma emissions from each sample were measured with a Wallac 1470 Wizard automatic gamma counter (Perkin Elmer).

Confocal microscopy
Cells were seeded into glass bottomed 35 mm dishes (MatTek Corporation, Ashland, MA, USA), and infected at an MOI of 0.01 24 h later. Following a 24 h infection period cells were then irradiated or the media was spiked with 5 μCi (1.85 MBq) of 131 I. Cells were then incubated for 1 h to allow H2Ax foci to form at the site of any DNA double-strand breaks induced, before being fixed in 4% paraformaldehyde. Permeabilised with 0.2% Triton X-100 in PBS for 20 min. H2aX foci were detected by a 1:200 dilution of rabbit anti-γH2aX primary antibody (Cell signaling) incubated at 4°C overnight and Alexafluor-546 conjugated secondary antibody (Life Technologies) diluted 1:1000 at room temperature for 90 min. DNA was stained with 1:10 000 dilution of DAPI (Life technologies). Quantification of the confocal images was done using CellProfiler 2.0 (Broad Institute, Cambridge, MA, USA) software configured to recognize cell nuclei, GFP-expressing cells and γH2Ax foci. The bystander cells were classified as cell nuclei lying wholly or partially within a 20 μm perimeter of GFPpositive cells.

In vivo xenograft studies
Xenograft tumors (PC3) were established by injecting 3 × 10 6 cells in 100 μl PBS subcutaneously in to the right flank of CD1 nude mice (Charles River Ltd., Harlow, UK). Tumors were allowed to grow to 5-8 mm in diameter and randomly allocated to treatment groups before beginning therapy. All animal work was carried out in compliance with UK Home Office regulations and under the scrutiny of the Institutional Review Board.
Virus at the relevant concentration was administered intratumourally in 100 μl PBS. Control animals received intratumoural injections of PBS alone. A single cutaneous puncture site and multiple intratumoural injection tracks were used to improve the distribution of injectate within the tumor. After 48 h, 131 I was administered by intraperitoneal injection (1 mCi (37 MBq) in 100 μl Hanks balanced salt solution.
Prior to administration of 131 I, cage water was supplemented with 5% Lugol's iodine solution (Sigma-Aldrich) for a minimum of 72 h to attempt to saturate the thyroid gland with non-radioactive iodide. This was replaced with normal drinking water 24 h before 131 I administration.
Tumors were measured twice weekly in two dimensions using Vernier callipers and the volume was estimated using the formula: (width × length 2 )/2.
For some experiments, tumor iodide uptake was measured by excision of the whole tumor and other tissues, and detection of gamma emissions using a Wallac 1470 Wizard automatic gamma counter (Perkin Elmer). Results were normalized to the weight of the tumor/thyroid gland. GFP imaging was performed under inhalational anesthesia (Vetflourane, Virbac Animal Health, Bury St Edmunds, UK) with IVIS Lumina II imaging system (Caliper Life Sciences, Waltham, MA, USA). GFP fluorescence was detected using an excitation wavelength of 465 nm, a GFP-specific emission filter and an exposure time of 0.2 s. Bioluminescence was measured following intraperitoneal administration of 100 μg h-Coelenterazine-SOL (Nanolight Technologies, Pinetop, AZ, USA) immediately prior to imaging.

TRAMP mouse in vivo studies
The TRAMP mouse model was acquired from the Jackson Laboratory (Bar Harbor, ME, USA). A breeding colony was established in the C57BL/6 background to produce heterozygous TRAMP males. Characterization studies were carried out to observe the rate of tumor growth and select a time point for intervention based on the level of disease progression. Animals were culled at a range of time points and the prostate gland and seminal vesicles were dissected and observed histologically. Animals were found to develop prostate adenocarcinoma, as well as seminal vesicle epithelial stromal tumors that resemble phyllodes tumors of the human breast, as has been previously described in this model. 33 Therapeutic experiments were carried out on heterozygous TRAMP males aged 29-31 weeks. Animals received 5 × 10 7 PFU of GLV-1h153 in 100 μl PBS by tail vein injection. Five days later, animals received 1 mCi 131 I in 100 μl Hanks balanced salt solution by intraperitoneal injection. All groups were given Lugol's iodine supplement as described previously. The animals were observed long-term for disease progression and culled if showing signs of discomfort. Animals with abdominal tumors not causing any obvious discomfort were culled if the tumor was palpable and over 1 cm in diameter. Post mortem dissection of all mice was done to enable histological analysis of the extent of the tumor progression and metastasis. Tissues were paraffin embedded and sectioned before Shandon Harris Hematoxylin (Thermo Fisher Scientific) and eosin (Leica Biosystems, Wetzlar, Germany) staining, and immunohistochemical staining with Ki67 (Dako UK Ltd., Ely, UK) and probasin (Santa-Cruz Biotechnology Inc., Dallas, TX, USA) antibodies with Shandon Gill Hematoxylin counterstain (Thermo Fisher Scientific).