Generation and Functional In Vitro Analysis of Semliki Forest Virus Vectors Encoding TNF-α and IFN-γ

Cytokine gene delivery by viral vectors is a promising novel strategy for cancer immunotherapy. Semliki Forest virus (SFV) has many advantages as a delivery vector, including the ability to (i) induce p53-independent killing of tumor cells via apoptosis, (ii) elicit a type-I interferon (IFN) response, and (iii) express high levels of the transgene. SFV vectors encoding cytokines such as interleukin (IL)-12 have shown promising therapeutic responses in experimental tumor models. Here, we developed two new recombinant SFV vectors encoding either murine tumor necrosis factor-α (TNF-α) or murine interferon-γ (IFN-γ), two cytokines with documented immunostimulatory and antitumor activity. The SFV vector showed high infection rate and cytotoxicity in mouse and human lung carcinoma cells in vitro. By contrast, mouse and human macrophages were resistant to infection with SFV. The recombinant SFV vectors directly inhibited mouse lung carcinoma cell growth in vitro, while exploiting the cancer cells for production of SFV vector-encoded cytokines. The functionality of SFV vector-derived TNF-α was confirmed through successful induction of cell death in TNF-α-sensitive fibroblasts in a concentration-dependent manner. SFV vector-derived IFN-γ activated macrophages toward a tumoricidal phenotype leading to suppressed Lewis lung carcinoma cell growth in vitro in a concentration-dependent manner. The ability of SFV to provide functional cytokines and infect tumor cells but not macrophages suggests that SFV may be very useful for cancer immunotherapy employing tumor-infiltrating macrophages.

Two well-known cytokines that may be potentially useful for cancer immunotherapy are IFN-γ and tumor necrosis factor-α (TNF-α). The pleiotropic cytokine IFN-γ is a tumor-suppressive factor that protects against carcinogen-induced and spontaneous tumor development (26,27), inhibits tumor angiogenesis (28)(29)(30)(31) and may improve tumor immunogenicity by upregulating major histocompatibility complex class I molecules (32,33). Furthermore, it has been shown that IFN-γ enhances macrophage activation toward a tumoricidal phenotype in vitro (34)(35)(36) and in vivo (31,(37)(38)(39). TNF-α was discovered in 1975 as a serum factor inducing haemorrhagic necrosis in tumors (40) and, therefore, this cytokine was proposed as a potential anti-cancer agent. TNF-α has been shown not only to selectively destroy tumor vasculature (41,42), but also increase tumor vessel permeability, thus, improving drug penetration into tumors (43)(44)(45). Moreover, low doses of TNF-α have been shown to promote antitumor immune responses by enhancing T-cell infiltration and by activating macrophages toward a tumor-suppressive phenotype (46). Notably, a synergistic action of IFN-γ and TNF-α was reported in early studies showing tumor growth inhibition in mice (47,48) and tumor disappearance in patients after local limb perfusions with IFN-γ and TNF-α in combination with a chemotherapeutic agent (49). The antitumor effects were likely due to decreased endothelial cell adhesion and survival in response to TNF-α and IFN-γ leading to destruction of tumor vasculature (50). The synergism may also be explained by the fact that IFN-γ enhances TNF-α receptor expression in malignant cells (51,52), thus improving TNF-α treatment. Another synergistic action of IFN-γ and TNF-α has been shown on macrophage activation toward a tumoricidal phenotype in vitro (53). However, the clinical usefulness of TNF-α and IFN-γ is limited by their systemic toxicity (54,55) and short in vivo half-lives (56,57).
To the best of our knowledge, no previous studies have reported using rSFV vectors that encode the cytokine IFN-γ or TNF-α. To provide new tools for cancer immunotherapy, we developed two rSFV vectors that encoded either murine TNF-α or IFN-γ and tested the functionality of the resulting rSFV-encoded cytokines in vitro. Here, we show that the rSFV vectors were capable of infecting Lewis lung carcinoma (LLC) cells and exploiting them to produce rSFV-encoded TNF-α and IFN-γ. Moreover, in this study, we show that rSFV-encoded IFN-γ effectively enhanced macrophage activation toward a cancer-suppressive phenotype in vitro.

Plasmids and construction of expression Vectors
The pSFV1-DsRed vector, which carries a gene for Discosoma sp. red fluorescent protein (DsRed), was generated in our lab as previously described (61) using the pSFV1 vector (20). The pSFV1 and pSFV-Helper1 plasmids (20) were generously provided by Garoff (Karolinska Institute, Stockholm, Sweden).
To generate the pSFV1-Tnfa-Flag plasmid, we used the pSFV1-NruI vector, which is a derivate of pSFV1 that was generated in our lab by deleting a 527 bp StuI-HindIII fragment from pSFV1 (bp 7603-8130) and changing a unique SpeI site to a NruI site (62). The murine TNF-α-coding gene (Tnfa) (GenBank No. NM_013693.2) fused to a Flag-tag sequence was inserted into the pSFV1-NruI vector under the control of the SFV 26S subgenomic promoter. The Tnfa-Flag fragment was amplified from the template plasmid pCMV3-mTNF-Flag (Cat. No. MG50349-CF; Sino Biological Inc., Beijing, China) using PCR with the following primers (synthesized by Microsynth AG, Balgach, Switzerland): 5′-GCGGATCCATGAGCACAGAAAGCATGATC-3′ (forward) and 5′-TGCCCGGGTTTACTTATCGTCGTCATCCTTG-3′ (reverse). BamHI and SmaI restriction sites (shown here in bold) were introduced to facilitate the ligation of the PCR fragment into the pSFV1-NruI vector.
The pSFV1/Enh-Luc plasmid was generously provided by A. Merits (Institute of Technology, University of Tartu, Estonia) and has been previously described (63). The pSFV1/Enh-Ifng vector was generated by replacing the XmaI-ApaI fragment (bp 7572-9268), which contained the firefly luciferase gene (Luc) of pSFV1/ Enh-Luc, with the murine IFN-γ-coding gene (Ifng) ( (20) for encapsidation of the rRNA. The electroporated BHK-21 cells were resuspended in 15 mL of BHK-21 cultivation medium containing 1% FBS (Sigma-Aldrich) and then incubated at 33°C for 48 h. The cell growth medium containing the infectious rSFV particles was then harvested, rapidly frozen in liquid nitrogen, and subsequently used as a source of rSFV particles. The titers of the rSFV particles were determined based on the "one virus particle-one infected cell" hypothesis regarding the replication-deficient virus particles. BHK-21 cells were infected with serial dilutions of the rSFV particles. Infected cells expressing DsRed were counted in 10 viewfields using fluorescence microscopy at 24 h post-infection. The cells that were infected with either SFV-Tnfa or SFV-Ifng were stained with rabbit polyclonal antibodies that were specific for the nsp1 subunit of SFV replicase (generously provided by A. Merits, Institute of Technology, University of Tartu, Estonia) and with a fluorochrome-conjugated secondary antibody (Cat. No. A11034; Thermo Fisher Scientific). The stained cell monolayers were observed using EVOSfl digital inverted fluorescence microscope (Thermo Fisher Scientific), and infection-positive cells were counted in 10 viewfields using fluorescence microscopy.

Production of sFV Vector-encoded cytokines
To produce rSFV-encoded TNF-α and IFN-γ, BHK-21 cells were cultivated in 75 cm 2 cell culture flasks (Cat. No. CLS430641; Sigma-Aldrich) until 80% confluency, washed once with PBS+/+ and then incubated at 37°C for 80 min with 5 mL of a suspension containing 1.92 × 10 8 IFU of SFV-Tnfa or 1.21 × 10 8 IFU of SFV-Ifng in PBS+/+. The virus-containing suspensions were then replaced with BHK-21 cultivation medium containing 1% FBS. After 36 h, the cell growth medium was harvested and centrifuged at 300 g for 5 min to remove the cells before the supernatants were transferred to a new tube, which was then centrifuged at 11,000 g for 8 min. The remaining supernatants were stored at −80°C and used as a source of vector-derived (vd) TNF-α or vdIFN-γ. The concentrations within the supernatants of the cytokines vdTNF-α or vdIFN-γ were 332.28 and 795.92 ng/mL, respectively. The levels of cytokines within the supernatants were quantified using Luminex bead-based assay with Bio-Plex Mouse Cytokine Group I IFN-γ or TNF-α sets (Cat. No. 171G5017M and 171G5023M; Bio-Rad, Hercules, CA, USA) according to the manufacturer's instructions.
secretion of sFV Vector-encoded cytokines from cancer cells Baby hamster kidney and LLC cells were plated in 24-well plates (Cat. No. 142475; Thermo Fisher Scientific) and cultivated until the next day when the cells had reached 100 and 50% confluency, respectively. The cells were washed once with PBS+/+, and 600 µL of infectious suspension containing either SFV-Tnfa or SFV-Ifng particles at a MOI = 1, MOI = 10 or MOI = 40 (calculated according to the titration in BHK-21 cells) in PBS+/+ was added per well. After cells were incubated with the infectious suspension for 80 min at 37°C, the infectious suspension was discarded, and 500 µL of fresh cell growth medium was added per well. The cells were cultured for 24 h before the growth medium was collected and centrifuged for 6 min at 300 g. The supernatant was then transferred to a new tube, which was centrifuged at 11,000 g for 10 min. The supernatants were stored at −80°C, and the levels of secreted vdTNF-α and vdIFN-γ were determined using Luminex bead-based assay with Bio-Plex Mouse Cytokine Group I IFN-γ and TNF-α sets (Cat. No. 171G5017M and 171G5023M; Bio-Rad) according to manufacturer's recommended protocols.
induction of cell Death in l929 Fibroblasts and llc cancer cells by TnF-α and iFn-γ L929 fibroblasts or LLC cells were plated in 24-well plates (Cat. No. 142475; Thermo Fisher Scientific) and cultivated in RPMI 1640 containing 10% FBS (Sigma-Aldrich) for approximately 24 h until the cells reached 50-90% confluency. To verify the functional activity of vdTNF-α, it was added to L929 cells in parallel with recombinant (r)TNF-α (Cat. No. 315-01A; PeproTech, Rocky Hill, NJ, USA) to reach a final concentration of 6.6, 20, or 60 ng/mL, and the cells were then cultivated for 24 h. L929 cells that were treated with 1 µM staurosporine (Cat. No. S4400; Sigma-Aldrich) for 24 h were used as the positive control for apoptosis induction (65), and untreated cells were used as the negative control. LLC cells were treated with either 50 ng/mL rTNF-α or 100 ng/mL rIFN-γ for 48 h to test cytotoxicity of cytokines to LLC cancer cells. LLC cells were also treated with 50 ng/mL rTNF-α in combination with 100 ng/mL rIFN-γ for 24 h or 48 h. LLC cells were pretreated with 100 ng/mL rIFN-γ for 24 h followed by 24 h treatment with 50 ng/mL rTNF-α. Untreated cells were used as the negative control, whereas L929 cells that were treated with 50 ng/mL of rTNF-α were used as the positive control for apoptosis induction. After treatments L929 or LLC cells were detached using 0.25% trypsin-EDTA and centrifuged before they were labeled with annexin V-FITC and PI using an Apoptosis Detection Kit (Cat. No. 556547; BD Biosciences) according to the manufacturer's protocol. Briefly, the cell pellet was resuspended in 100 µL 1 x annexin V binding buffer and stained using 1 µg/ mL annexin V-FITC and 5 µg/mL PI by incubating the cells at RT for 15-20 min. Then, 400 µL of additional annexin V binding buffer was added. The cell suspension was immediately analyzed using flow cytometry in a BD LSRFortessa instrument (BD Biosciences). The data were analyzed using FlowJo V10 software.

resUlTs construction of sFV-Based Vectors and Production of recombinant Viral Particles
We constructed two new replication-deficient SFV-based vectors that encoded either mTNF-α or mIFN-γ ( Figure 1A). To express mTNF-α in the SFV vector, a sequence containing the Tnfa gene followed by a Flag-tag-encoding sequence at the 3-prime end was subcloned into the pSFV1-NruI vector ( Figure 1A). The pSFV-Ifng vector was made by inserting the mIFN-γ gene (Ifng) into the pSFV1/Enh vector downstream and in-frame with the SFV capsid translation enhancer sequence (Enh) that was fused to the 2A auto-protease sequence of foot and mouth disease virus (FMDV) (Figure 1A). The 2A-encoding sequence was included to induce the cleavage of the vdIFN-γ protein from the 34 amino acid-long peptide that was translated from the Enh sequence. It has been shown that by incorporating an enhancer element within SFV vector, the expression of the heterologous protein can be increased by up to 10-fold due to enhanced translation (7,66). In contrast to the SFV-Ifng construct, we did not incorporate the enhancer element into the SFV-Tnfa vector aiming to retain low levels of translated TNF-α if necessary since lower levels of TNF-α might be important for normalization of tumor vasculature (46). The previously constructed vector pSFV-DsRed, which encode a fluorescent protein (Figure 1B), was used as a control to detect cell infection. All vectors comprised the prokaryotic SP6 RNA polymerase promoter for in vitro transcription of recombinant SFV replicons, as well as the vectors encoded SFV replicase complex, which consists of non-structural proteins (nsP) 1-4 ( Figures 1A-C). The SFV structural proteins C, p62, 6 K, and E1, which are necessary for the formation of infectious virus particles, were encoded by the SFV-Helper1 vector (Figures 1C,D). The heterologous genes (HGs) and the genes encoding the structural SFV proteins were inserted downstream of the SFV 26S promoter (Figures 1A-C).
In this study, replication-deficient rSFV particles were produced by BHK-21 cells that were co-transfected with RNA transcribed from the SFV-Helper1 vector and RNA transcribed from a vector encoding a HG (Tnfa, Ifng, or DsRed) ( Figure 1D).

Mouse and human lung carcinoma cells are susceptible to rsFV infection, While Macrophages are resistant
We wanted to use rSFV to deliver cytokine genes into cancer cells for subsequent production of the tumor-suppressive factors. Cell susceptibility to viral infection is a prerequisite for successful gene delivery into cells using a viral vector. We investigated the ability of rSFV particles to infect both cancer cells and macrophages using the reporter vector SFV-DsRed. We infected cells with SFV-DsRed at a concentration 15 virus particles per cell (i.e., MOI = 15). We observed the expression of the DsRed fluorescent protein using fluorescence microscopy and quantified the DsRed-expressing cells using flow cytometry at 24-h post-infection (Figures 2A-E). In all the tested cell types, non-infected cells were used as the negative control for DsRed expression. The BHK-21 cell line was used as a positive control due to being highly susceptible to SFV infection, as demonstrated by flow cytometry analysis that showed that 88.2% of the BKH-21 cells were infected under the experimental conditions used in this study (Figure 2A). We verified that both the murine (LLC) and the human (A549) lung carcinoma cell lines were infected by the rSFV particles because 41% of both cell types were DsRedpositive (Figures 2B,C). However, we also found that neither the mouse BMDMs nor the murine macrophage cell line J774A.1 was susceptible to SFV because less than 1.5% of the cells were DsRed-positive (Figures 2D,E). Macrophages remained resistant to SFV infection also when we increased the concentration of SFV particles during infection (data not shown).
To confirm these findings in human macrophages, we differentiated macrophages from blood-derived monocytes and infected the cells with SFV-DsRed virus particles at a MOI = 15 (calculated according to the titer that was determined in BHK-21 cells). The SeV is known to infect human macrophages (67), and we therefore included a SeV that encoded GFP (SeV-Gfp) at a MOI = 15 as a positive control. In parallel, we infected the human lung carcinoma cell line A549 with both types of recombinant viruses at a MOI = 15, and used the infected A549 cells as a positive control for SFV infection. To calculate the percentages of infected cells, we counted the total number of cells as well as the number of fluorescent protein-expressing cells within the same viewfields (multiple viewfields per sample) using phase-contrast and fluorescence microscopy at 24 h post-infection. In these experiments, 30% of the positive control sample (the human cancer cell line A549) was infected with SFV-DsRed, whereas the HMDMs remained uninfected (Figure 3, middle panel). Contrary to our findings for SFV, SeV-Gfp infected both HMDMs and A549 very efficiently, with 82 and 95% of the HMDMs and A549 cells, respectively, being positive for GFP (Figure 3, lower panel). Taken together, these data show that lung carcinoma cells are efficiently infected by rSFV, whereas primary mouse and human macrophages are resistant to SFV infection, suggesting that rSFV may be used to deliver genes for cancer therapy without killing tumor-associated macrophages.

Macrophages remain Viable after challenge with sFV Particles
We wanted to test macrophage viability after challenge with rSFV particles. Both J774A.1 macrophages and LLC cancer cells were subjected to rSFV-DsRed particles at MOI = 10 during 80-min infection protocol. Cell infection rate was determined 48 h post-infection by quantifying DsRed-expressing cells using flow cytometry. Cells were also stained with annexin V-FITC and DAPI followed by flow cytometry analysis to determine cell viability.
As expected, we observed that 98% of non-infected macrophages stayed viable after 48 h as determined using flow cytometry and retained round morphology, which is characteristic to J774A.1 cells in normal culture conditions (Figure 4A). Also 48 h after challenge with rSFV-DsRed particles 90% of macrophages remained viable, whereas 6% of macrophages were single-positive for Annexin V (Figure 4B). Interestingly, light microscopy revealed that macrophages changed morphology 48 h after challenge with rSFV by gaining spindle-like or polygonal shape (Figure 4B). LLC cells were used in parallel as a positive control for cell infection. As expected, the noninfected LLC cells showed 92% viability and were characterized   Figure 4C). However, the infected LLC cells underwent cell death 48 h after infection with rSFV, and only 35% of LLC cells were viable ( Figure 4D). We observed that 33% of rSFV-challenged LLC cells were double-positive for Annexin V-FITC and DAPI suggesting apoptotic/necrotic cell death. A LLC population of 30% was single-positive for DAPI 48 h after infection suggesting affected cell membrane permeability ( Figure 4D). Light microscopy revealed that LLC cells possess apoptotic morphology characterized by round swollen cells with translucent cytoplasm after infection with rSFV ( Figure 4D). We observed microscopically that both DsRed-positive (infected) and DsRed-negative (non-infected) cells undergo apoptosis after challenge with rSFV particles (Figure 4D, red and white arrows), whereas part of infected DsRed-expressing cells remain viable 48 h post-infection ( Figure 4D, yellow arrows). Flow cytometry analysis revealed that 69% of all DsRed-expressing cells undergo cell death characterized mostly by being double-positive for Annexin V-FITC and DAPI ( Figure 4E). Our data show that macrophages are not only resistant to SFV infection but also remain viable after challenge with rSFV particles.

sFV infection inhibits growth of Murine lung carcinoma cells
It has been shown that SFV infection induces cellular death via p53-independent apoptosis (21). To determine whether rSFV has a direct effect on cancer cell growth, we used in vitro assay to evaluate proliferation of infected LLC cells. The cells were cultivated until approximately 70-80% confluency, and they were then infected with SFV-Ifng or SFV-Tnfa viral particles at MOI = 15. To determine cell growth, a radioactive thymidine, which incorporates into new DNA strands during cell division, was added to the cell cultures at 47 h post-infection. The experimental setup is illustrated in Figure 5A. BHK-21 cells were used in parallel as a positive control for efficient cell infection by SFV and showed nearly complete growth inhibition ( Figure 5B). Growth inhibition was similar independent on whether SFV carried a gene encoding TNF-α or IFN-γ ( Figure 5B). It was observed using light microscopy that BHK-21 cells lost confluency and gained apoptotic cell morphology characterized by translucent cytoplasm, granularity, or round shape 47 h post-infection ( Figure 5C). LLC cell growth was inhibited by half 47 h after infection with SFV-Ifng and by 80% after infection with SFV-Tnfa ( Figure 5D). Light microscopy revealed that LLC cells acquired apoptotic cell morphology characterized by round shape, granularity, and translucent cytoplasm 47 h after infection with either SFV-Ifng or SFV-Tnfa viral particles ( Figure 5E). The data show that rSFV particles encoding murine IFN-γ or TNF-α efficiently inhibited LLC cell growth at 47 h post-infection.

Mouse lung carcinoma cells Produce and secrete sFV-encoded TnF-α and iFn-γ
In the next set of experiments, we sought to determine whether LLC cells secrete cytokines after infection with SFV-Tnfa and SFV-Ifng virus particles. In this experiment, BHK-21 cells were used as a positive control. The LLC and BHK-21 cells were   cultivated until 50 and 100% confluency, respectively, before they were infected with SFV-Tnfa or SFV-Ifng at a MOI = 1, MOI = 10, or MOI = 40 (calculated according to the titration in BHK-21 cells). After 24 h, the levels of secreted vdIFN-γ and vdTNF-α in the cell cultivation medium were analyzed using Luminex bead-based assay. LLC cells infected with SFV-Tnfa secreted vdTNF-α (range, 10 3 -10 4 pg/mL for all tested MOIs) ( Figure 6A). There was no clear correlation between the MOI of the virus particles and the level of vdTNF-α ( Figure 6A). The vdTNF-α was not detected in the medium of non-infected cells or in the medium of cells infected with SFV-Ifng, confirming that the secreted vdTNF-α was derived from the Tnfa gene that was delivered using SFV ( Figure 6A). In LLC cells, infection with SFV-Ifng resulted in the secretion of vdIFN-γ into the growth medium at a concentration ranging from 2.9 to 9.9 × 10 2 ng/mL at 24 h after infection ( Figure 6B). The concentration of vdIFN-γ appeared to correlate with the concentration of viral particles, and the highest tested MOI resulted in the highest level of vdIFN-γ ( Figure 6B). The vdIFN-γ was not detected in media obtained from non-infected cells or from cells infected with SFV-Tnfa, indicating that the vdIFN-γ originated from the SFV-delivered cytokine gene (Figure 6B). In conclusion, these results show that both BHK-21 and LLC cells produce and secrete vdTNF-α and vdIFN-γ when infected with the rSFV vectors described in this study.

Vector-Derived TnF-α is Functional in inducing cell Death in l929 Murine Fibrosarcoma cells
To determine whether the vdTNF-α is biologically active, we utilized a known property of TNF-α which is to induce necrotic cell death in the murine fibrosarcoma cell line L929 by binding to TNF receptor-1 (68). In these experiments, L929 cells were cultivated until 90% confluency; we then added rTNF-α or vdTNF-α at concentrations varying from 2.2 to 60 ng/mL to the cells. The culture medium of BHK-21 cells infected with SFV-Tnfa particles was used as the source of vdTNF-α. At 24 h after treatment, the morphology of the L929 cells was evaluated using bright-field microscopy. Then cells were stained with annexin V-FITC to determine phosphatidylserine exposure, whereas staining with PI was used to determine loss of membrane integrity in order to quantify the numbers of viable (Annexin V−PI−), apoptotic (Annexin V+PI−), and necrotic (Annexin V+PI+) cells using flow cytometry (Figures 7A-J). Untreated L929 cells were used as a control, and viable cells were defined as those that were annexin V-negative/PI-negative ( Figure 7A). Treatment with staurosporine was used as the control for apoptotic cell death (65), where annexin V-positive/PI-negative cells were regarded as early apoptotic, whereas annexin V-positive/PI-positive cells were regarded as late apoptotic undergoing secondary necrosis ( Figure 7B).  In the absence of any treatment, the L929 cells reached 100% confluency after 24 h and had a normal fibroblast-like, slightly polygonal cell morphology (Figure 7A). The untreated L929 cells were also analyzed using flow cytometry and used to set the gates for the viable (Annexin V−PI−) cell population, which accounted for 99% of all events (Figure 7A). After cells were treated with 1 µM staurosporine, cell confluency was lost, and an apoptotic cell morphology characterized by cell shrinkage, a round shape and granularity was observed (Figure 7B). Only 11% of the staurosporine-treated cells were viable. The remaining cells were undergoing apoptosis and secondary necrosis (Figure 7B). In cells that were incubate with increasing concentrations of rTNF-α or vdTNF-α (2.2, 6.7, 20, or 60 ng/mL), we observed a concentration-dependent increase in the proportion of apoptotic (Annexin V+PI−) and necrotic (Annexin V+PI+) cells (Figures 7C-J) in a similar manner between rTNF-α and vdTNF-α. Treatment with vdTNF-α and rTNF-α at the lowest concentration (2.2 ng/ mL) induced apoptosis in 9-13% and necrosis in 19-20% of the cells (Figures 7C,D), whereas the highest concentrations of rTNF-α and vdTNF-α (60 ng/mL) induced apoptosis in 18-23% and necrosis in 46-62% of the cells (Figures 7I,J). Moreover, after treatment with rTNF-α or vdTNF-α, we observed necrotic cell morphology characterized by round, swollen cells with translucent cytoplasm. These data verify the functionality of the vdTNF-α protein because it was able to induce apoptosis and necrosis in L929 cells in a concentration-dependent manner that was similar to that observed for rTNF-α.

combined Treatment of TnF-α and iFn-γ induces cell Death in Mouse lung carcinoma cells
It has been previously shown that LLC cells are resistant to cell death induced by TNF-α (69). Based on previous studies, where IFN-γ was shown to enhance TNF-α receptor expression in cancer cells (51,52), we wanted to test whether IFN-γ may sensitize LLC cancer cells to TNF-α treatment. Therefore, we tested induction of cell death in LLC cells after combined treatment of recombinant cytokines TNF-α and IFN-γ. LLC cells were single-treated with either TNF-α or IFN-γ for 48 h. LLC cells were treated with TNFα in combination with IFN-γ for 24 h or 48 h. LLC cells were also pretreated with IFN-γ for 24 h followed by 24 h treatment with TNF-α. L929 fibroblasts, which were treated with TNF-α, were used as a positive control. Untreated cells were used as a negative control. Cell morphology after treatments was evaluated using bright-field microscopy. Cells were stained with annexin V-FITC to determine phosphatidylserine exposure, whereas staining with PI was used to determine loss of membrane integrity in order to quantify the numbers of viable (Annexin V−PI−), apoptotic (Annexin V+PI−), and necrotic (Annexin V+PI+) cells using flow cytometry (Figures 8A-H).
In the absence of any treatment for 48 h, 99% of L929 control cells and 94% of LLC cancer cells remained viable and showed normal cell morphology (Figures 8A,B). The positive control of L929 cells, which were treated with TNF-α for 48 h, lost their confluency and showed apoptotic morphology characterized by round swollen cells with translucent cytoplasm (Figure 8C). Only 26% of the TNF-α-treated L929 cells were viable (Figure 8C). By contrast, LLC cells were resistant to treatment with TNF-α for 48 h, showed 86% viability and retained normal cell morphology ( Figure 8D). Treatment with IFN-γ for 48 h was not cytotoxic to LLC cells, and 90% of the treated LLC cells remained viable ( Figure 8E). Interestingly, combined treatment of LLC cells with both cytokines for 24 h also did not induce cytotoxicity in LLC cells ( Figure 8F). However, a prolonged incubation of LLC cells with both cytokines for 48 h induced cytotoxicity in LLC cells resulting in 38% cells being double-positive for Annexin and   (a,B) left untreated for 48 h and used as negative controls or (c) L929 cells were treated with 50 ng/mL TNF-α for 48 h and used as a positive control. LLC cell viability was retained after treatment with (D) 50 ng/mL TNF-α or (e) 100 ng/mL IFN-γ and after (F) simultaneous treatment with both 50 ng/mL TNF-α and 100 ng/mL IFN-γ for 24 h. Cell death was induced in LLC cells after (g) simultaneous treatment with 100 ng/mL of IFN-γ in combination with 50 ng/mL of TNF-α for 48 h and after (h) LLC pretreatment with 100 ng/mL of IFN-γ for 24 h followed by treatment with TNF-α for 24 h. The resulting cell monolayers were analyzed using bright-field microscopy (scale bar, 50 µM). Cell death was determined using flow cytometry analysis after cell staining with annexin V-FITC (depicted on the x-axis) and propidium iodide (PI, depicted on the y-axis). Annexin V-positive/PI-negative cells were regarded as apoptotic, whereas annexin V-positive/PI-positive cells were regarded as necrotic. Experiment was performed in duplicates and repeated two times with standard error not exceeding 10% between independent experiments. One representative experiment is shown, where the percentages of the four distinct cell populations are averages of duplicates with SEM <5%. Figure 8G). Also pretreatment of LLC cells with IFN-γ for 24 h followed by treatment with TNF-α for 24 h was cytotoxic to LLC cells and induced apoptosis in 10% and necrosis in 41% of the cells (Figure 8H). Cell morphology of LLC cells changed after the combined treatments of LLC with both cytokines simultaneously for 48 h or with IFN-γ pretreatment followed by TNF-α (Figures 8G,H). After combined treatments with both cytokines, we observed cells with apoptotic cell morphology as well as live cells that had spread in an elongated shapes possibly due to access of free space resulting from cells dying in proximity SFV Encoding TNFα and IFNγ Frontiers in Immunology | www.frontiersin.org

PI (
November 2017 | Volume 8 | Article 1667 (Figures 8G,H). All in all, combined treatment with IFN-γ and TNF-α was effective in inducing cell death in LLC cancer cells, which were resistant to single-treatments with these cytokines.

Vector-Derived iFn-γ in combination with a Tlr2/1 agonist activates Macrophages toward a cancer-suppressive Phenotype
To verify that vdIFN-γ is a functional protein, we used an in vitro assay based on cancer cell growth inhibition mediated by activated macrophages. The macrophages were treated with mitomycin C to block their proliferation, and they were activated by applying either vdIFN-γ or rIFN-γ in combination with a second stimulus. Based on previous experiments performed in our lab, we chose to stimulate the BMDMs with IFN-γ and TLR2/1 agonist Pam3 because this combination leads to efficient inhibition of cancer cell growth (70). After the BMDMs were stimulated for 24 h, LLC cells were added, and the co-cultures were cultivated for an additional 20 h. Finally, to enable the detection of proliferating cancer cells, we added radiolabeled thymidine to the cultures at 24 h before they were harvested ( Figure 9A).
In the first experiment, we treated cells with different concentrations of rIFN-γ in combination with 100 ng/mL Pam3 to identify the lowest effective dose of IFN-γ that is necessary to activate BMDMs toward a cancer-suppressive phenotype ( Figure 9B). The results revealed that adding rIFN-γ within the range 100-1.56 ng/mL effectively inhibited cancer cell growth (cpm below 750 when the ratio of BMDMs:LLC cells was 20:1) in a concentration-dependent manner. Some growth inhibition was also observed at 0.098-0.39 ng/mL rIFN-γ, but minor or no effect was observed at concentrations ≤0.024 ng/mL ( Figure 9B).
All controls included in this assay produced the expected results (Figures 9B,C). Treating the BMDMs with mitomycin C resulted in counts lower than 650 cpm, confirming that the treatment hindered proliferation (Figures 9B,C; first group of bars from the left). Unstimulated macrophages did not inhibit growth in LLC cells since more than 15,000 cpm were measured in the wells (Figures 9B,C; second group of bars from the left). These results were comparable to the average cpm in LLC cells that were cultivated alone (Figures 9B,C; orange line). Treating cells with either Pam3, rIFN-γ, or vdIFN-γ alone resulted in growth rates similar to those observed in LLC cells cultivated alone (Figures 9B,C).
To analyze the functionality of vdIFN-γ, we treated BMDMs with the same concentrations that were used to test rIFN-γ. The vdIFN-γ was produced by infecting BHK-21 cells with SFV-Ifng viral particles and harvesting the cell culture medium to use as a source of vdIFN-γ. The results were similar to what was observed for rIFN-γ in that incubating macrophages with vdIFN-γ in combination with Pam3 resulted in efficient cancer cell growth inhibition in a concentration-dependent manner. The lowest effective concentration of vdIFN-γ was 6.25 ng/mL, which produced less than 1,050 cpm ( Figure 9C). In all the experiments, the most efficient inhibition of cancer cell growth was observed when the ratio of BMDMs to LLC cells was 20:1 or 10:1 (Figures 9B,C). Finally, the pattern of effective IFN-γ doses and the extent of growth inhibition they produced was similar between rIFN-γ and vdIFN-γ (Figures 9B,C). Based on these data, we conclude that vdIFN-γ has biological activity and can substitute for rIFN-γ.
Using TLR ligands in combination with IFN-γ provides a classical set of stimuli that activates macrophages toward an M1-phenotype, which can produce NO through inducible NO synthase (70). We, therefore, determined NO production as a readout to verify that vdIFN-γ triggers macrophage activation potentially toward an M1-phenotype. We measured NO 2 −, which is the end product of the rapid oxidation of NO, in cell culture media using the Griess test. First, we verified that the BMDMs produced NO in a concentration-dependant manner at 24 h after stimulation with both rIFN-γ and Pam3 (Figure 10A). We then stimulated BMDMs with both vdIFN-γ and Pam3, and the results were similar to those observed for rIFN-γ in that the levels of NO 2 − varied according to the concentration of vdIFN-γ ( Figure 10B). The highest levels of NO 2 − were observed when rIFN-γ and vdIFN-γ were used in a range from 0.39 to 100 ng/mL. At IFN-γ concentrations ≤0.098 ng/mL, the levels of NO 2 − were lower (Figures 10A,B). Stimulation with rIFN-γ alone resulted in the production of only small amounts of NO 2 − in the cell culture medium (Figure 10A). The levels of NO 2 − were also low in cultures stimulated with vdIFN-γ alone ( Figure 10B) or Pam3 alone (Figures 10A,B). Thus, IFN-γ and Pam3 had a synergistic effect on the ability of macrophages to produce NO. The higher levels of NO 2 − that were observed following stimulation with higher doses of IFN-γ (Figures 10A,B) were well-correlated with the enhanced cancer cell growth inhibition that was observed in the growth inhibition experiments (Figures 9B,C). Taken together, data from the growth inhibition assay demonstrated that vdIFN-γ is a potent agent at very low concentrations that when combined with Pam3 induces an antitumor phenotype in macrophages that suppresses growth in LLC cells.

DiscUssiOn
In this study, we developed two new rSFV vectors encoding either murine TNF-α or IFN-γ and showed by in vitro studies that the two rSFV-encoded cytokines are functional. The rSFV efficiently infected mouse and human lung carcinoma cells in vitro, whereas murine and human macrophages were resistant to SFV infection. The rSFV inhibited LLC cell growth, induced cancer cell death and simultaneously exploited cancer cells for production of SFV-encoded TNF-α and IFN-γ at levels that are functional in vitro. The functionality of SFV-encoded TNF-α was shown via cell death induction in L929 cells. The rSFV-encoded IFN-γ activated macrophages toward a tumoricidal phenotype in vitro.
SFV vector, which provides high and transient expression of the transgene without further virus replication (20). In order to form viral particles, we used SFV-Helper1 system (20), which encodes SFV structural proteins. We produced infectious SFV-Tnfa and SFV-Ifng recombinant virus particles in BHK-21 cells at titers 10 7 -10 8 IFU/mL. Various mouse and human cancer cell lines are susceptible to SFV infection although the transduction rates may differ between FigUre 10 | The production of nitric oxide (NO) by macrophages at 24 h after activation was determined by analyzing NO2− levels in the cell culture medium. Mitomycin C-treated bone marrow-derived macrophages (BMDMs) were cultivated for 24 h before they were activated with the TLR2/1 agonist Pam3 (100 ng/mL) in combination with either (a) rIFN-γ at various concentrations ranging from 100 to 0.0015 ng/mL in 4-fold dilutions or (B) vdIFN-γ at concentratios ranging from 100 to 0.0015 ng/mL in 16-fold dilutions. Non-stimulated BMDMs and BMDMs that were stimulated with Pam3, rIFN-γ, or vdIFN-γ were used as the negative controls. Nitrite ( ) NO2− levels were measured in the macrophage culture medium at 24 h post-stimulation using the Griess test. Experiments were performed in triplicates and were repeated two times with similar results. The bars represent the mean values of triplicates ± SEM from one representative experiment.
cell types (71)(72)(73). The hamster kidney fibroblast cell line BHK-21, which is commonly used for production of SFV particles, is efficiently infected by SFV (71). In the current study, the high susceptibility of BHK-21 cells to SFV infection was confirmed. Two cell lines of particular interest for development of lung cancer immunotherapy are mouse and human lung carcinoma cell lines, LLC and A549, respectively, which proved to be susceptible to SFV infection with a comparable transduction level of 41% 24 h post-infection for both cell lines when SFV was used at MOI = 15. This is in agreement with a previous study showing that A549 cells are susceptible to SFV infection (72). Our results encouraged us using LLC cells as target murine cancer cells further because they were successfully infected by rSFV particles and were able to express rSFV-encoded proteins.
In contrast to the tested cancer cell lines, macrophages were found to be resistant to SFV infection. Similar data were obtained with primary mouse and human macrophages, as well as with the mouse macrophage cell line J774A.1. Moreover, macrophages retained their viability after challenge with rSFV particles. The fact that macrophages are not killed by SFV but instead serve as effector cells responding to SFV-encoded factors may be an advantage for immunotherapy based on the activation of tumor-associated macrophages. Cancer cells, due to their susceptibility to SFV infection, may be exploited to produce such SFV-encoded factors. The mechanism by which macrophages remained resistant to SFV infection is unclear. The life cycle of replication-deficient rSFV particles is a several-step process comprising viral entry via endocytosis, a pH-dependent release of virus nucleocapsid from the endosome, capsid disassembly, viral RNA replication in the cytoplasm, and subsequent translation of the transgene (74). The suppression in either of the abovementioned steps may have resulted in lack of SFV-encoded transgene expression in macrophages as we observed. Furthermore, fresh isolated peritoneal mouse macrophages were also resistant to infection with the replication-deficient SFV-DsRed virus as was confirmed by absence of fluorescent signal in the infected cells (data not shown). In contrast to our results, a study from 1976 described that a replication-competent SFV was able to infect and multiply at very low yields within peritoneal mouse macrophage cultures in vitro (75). The authors tested the presence of SFV within cell growth medium of peritoneal macrophages post-infection by titrating plaque forming units in chick embryo cell cultures. Similarly, another study from 1988 reported low susceptibility of mouse spleen-derived Mac-1-positive cells to infection with replication-competent or first-generation avirulent SFV (76). The differences between the protocols and SFVs used in our study and studies by van der Groen et al. or Wu et al. make it difficult to compare the results. However, a possible explanation for the macrophage low susceptibility toward SFV in the studies mentioned (75,76) could be that the ex vivo isolated peritoneal or spleen cells, which the authors used as a source of macrophages, also contained other cell types that served as host cells for SFV multiplication. The main difference though is that the authors used replication-competent SFVs, including a wt SFV, which was isolated from mosquitos and was produced by replication in chick embryo cell cultures, or first-generation avirulent SFV. Our data based on use of pure macrophage cultures clearly show that mouse and human macrophages are resistant to infection with SFV particles that are based on the replication-deficient rSFV1 vector and that have been produced in BHK-21 cells.
Infection with SFV inhibited proliferation of BHK-21 and LLC cells in vitro independent on whether SFV-encoded TNF-α or IFN-γ. The inhibition of cancer cell growth is likely caused by the ability of SFV particles to induce p53-independent apoptosis in infected cells, which is linked to viral RNA replication and appearance of double-stranded RNA intermediates in cell cytoplasm (21). This property is of particular interest for lung cancer therapy since a mutated p53 is one of the key molecular markers in many solid cancers, including human lung cancers (77). In fact, the target LLC cancer cells used in this study were shown to undergo cell death after infection with replication-deficient rSFV particles. The infected cancer cells were shown to express rSFVencoding transgene while undergoing cell death. Thus, rSFV has therapeutic potential not only as a transgene delivery vehicle but may also elicit a direct oncolytic effect on cancer cells. Moreover, intratumoral injection of rSFV particles has been shown to inhibit tumor growth in preclinical mice models by induction of apoptosis (78).
Early after infection with rSFV particles the host cells shut down the translation of endogenous cellular proteins (79), whereas the heterologous protein encoded by rSFV is produced at high levels (80). In the present study, we embraced this property in order to produce tumor-suppressive cytokines within cancer cells. We showed that LLC cells can produce and release rSFVencoded TNF-α and IFN-γ in the cell culture medium at levels that are functional in vitro. Whereas vdIFN-γ was present at 100 ng levels in the supernatant, the amount of vdTNF-α was in several nanograms. The relatively low levels of vdTNF-α that we detected might be explained by the lack of translational enhancer in a vector, as well as by the fact that TNF-α is primarily produced as a transmembrane protein (81), and cleavage of the transmembrane form is needed before soluble TNF-α can be released into the cell culture medium (82).
Nevertheless, the levels of vdTNF-α detected were sufficient to elicit a functional response in vitro as shown by induction of cell death in mouse fibroblast cell line L929. TNF-α is known to induce either caspase-dependent apoptotic cell death or caspaseindependent programmed necrotic cell death (necroptosis) downstream of TNFR1 signalling depending on microenvironment and cell type (83,84). We induced mainly necrotic cell death in L929 cells in a dose-dependent manner after treatment with either vdTNF-α or rTNF-α, which is in agreement with previous reports showing that L929 cells undergo caspase-independent necroptosis in response to TNF-α treatment (68,85). In accordance with a previous study (65) we also observed an annexin V-positive/ PI-negative cell population after treatment with vdTNF-α, which may represent not only apoptotic but also primary necrotic cells according to studies showing phosphatidylserine exposure during early necrosis (oncosis) before the integrity of the membrane is lost (86,87). By inducing fibroblast cell death in vitro, we verified that SFV-encoded TNF-α is biologically active. However, in vivo effects of TNF-α differ and comprise increased vessel permeability used to improve drug penetration (43)(44)(45) and destruction of tumor vasculature (41,42) leading to hemorrhagic tumor necrosis (40,88). By using recombinant cytokines, we also showed that TNF-α is a cytokine that may induce cytotoxicity in mouse lung carcinoma cells but only when combined with IFN-γ. Such observations can be explained by a previously reported IFN-γ property to enhance TNF-α receptor expression (51,52), thus sensitizing cancer cells to TNF-α-induced cell death. Furthermore, in the context of cancer immunotherapy TNF-α may serve as an adjuvant in order to enhance T-cell infiltration (46,89).
In this study, we showed that rSFV-encoded IFN-γ was functional in activating macrophages in vitro toward a tumoricidal phenotype, which is consistent with previous studies showing that IFN-γ renders macrophages tumoricidal in vitro (34)(35)(36). We observed that in order to activate macrophages, presence of both vdIFN-γ and the TLR2/1 agonist Pam3 was required. Also previous studies showed that two signals, IFN-γ in combination with a second signal are required to render macrophages tumoricidal (39) and able to induce NO production (90). It is known that classically activated or M1 macrophages increase the expression of inducible NO synthase and, thus, produce NO (91). We showed that vdIFN-γ synergized with TLR2/1 agonist Pam3 to induce NO production in macrophages in a concentrationdependent manner, similarly as observed by combining rIFN-γ with another TLR agonist, CpG-containing DNA, respectively, in a previous study (90). Our observation that high level of NO, which was produced by activated macrophages, correlated with increased cancer cell growth inhibition is in agreement with the early studies reporting that tumoricidial abilities of macrophages in vitro are based on NO being cytotoxic to cancer cells (92,93).
Macrophages are common immune cells in many tumors and often are considered to inhere a pro-tumoral phenotype associated with angiogenesis, metastasis, and suppression of T cell activation (94). By contrast, M1 macrophages suppress tumor growth and angiogenesis (95) and might be positively associated with patient survival (95)(96)(97). Several in vivo studies have shown the therapeutic benefit of redirecting tumor-associated macrophages to M1 phenotype by agents such as TLR9 ligand CpG in combination with antibody to IL-10R (88), TLR7/8 agonist 3M-052 (98), or liposomal nanoparticle-encapsulated hydrazinocurcumin (99). Previous studies suggest IFN-γ as a good candidate for tumorassociated macrophage re-education due to the ability of IFN-γ to re-polarize tumor-associated macrophages toward an M1 phenotype in vitro (100) and its ability to render macrophages tumor-suppressive in mouse tumor models in vivo (31,(37)(38)(39). The results from this study suggest that SFV-delivered vdIFN-γ is as efficient as rIFN-γ in rendering macrophages tumoricidal in vitro, which along with our finding that macrophages but not cancer cells are resistant to SFV infection encourages a possible application of SFV-Ifng vector for activation of tumor-associated macrophages toward tumor-suppressive M1 phenotype in further studies in vivo. Moreover, either IFN-γ or TNF-α delivery by rSFV vector may comprise not only a direct effect of the rSFVdelivered cytokines but may also benefit from the inherent ability of SFV vector to induce p53-independent apoptosis (21) and elicit endogenous type-I IFN responses, which may be required for the therapeutic effect of a vector-encoded cytokine (3).
In summary, the present study shows that rSFV can efficiently infect lung carcinoma cells in vitro and exploit them for production of functional rSFV-encoded TNF-α and IFN-γ. The rSFV-encoded IFN-γ was able to activate macrophages toward an antitumor phenotype in vitro. Our results set the basis for application of SFV-Tnfa and SFV-Ifng vectors to modulate the cytokine milieu in the tumor microenvironment in further studies using mouse models for lung cancer.

eThics sTaTeMenT
This study was carried out in accordance with the recommendations of The Norwegian Regulation on Animal Experimentation, Norwegian Food and Safety Authority.

aUThOr cOnTriBUTiOns
AZ and AC conceived the study. All authors contributed to the design of experiments, data analysis, and interpretation of data. BK, EM, PC, IJ, BS, and IØ performed experiments. BK wrote