A Genetically Engineered Influenza A Virus with ras-Dependent Oncolytic Properties

INTRODUCTION

The type I IFN-induced cellular antiviral response is in the first line of defense against a viral infection by the host. One major antiviral effector induced by IFN is PKR 3 (1 , 2) . PKR is a serine/threonine protein kinase that dimerizes and autophosphorylates upon activation (3 , 4) . The activated form of PKR is capable of blocking protein synthesis through its ability to phosphorylate the α subunit of eukaryotic translation factor 2. This mechanism inhibits viral replication. To counteract the antiviral effects of IFN induction and PKR activation, many eukaryotic viruses have developed strategies to block the activity of PKR (5) . In the case of influenza A virus, PKR-blockage during influenza virus infection involves the viral nonstructural protein NS1 (6) . Previously, we generated an influenza A/PR/8/34 virus that lacks the NS1 gene (delNS1 virus; Ref. 7 ). Cells not permissive for delNS1 virus replication produced infectious particles when the infected cells were incubated with 2-AP, a chemical inhibitor of PKR (8) . To analyze the relevance of this observation on the organismal level, we determined the replication properties of the delNS1 virus in gene-knockout mice devoid of PKR. Although the delNS1 virus failed to replicate in the lungs of WT mice, it grew almost as efficiently as the PR8-wt virus in PKR-knockout mice. Moreover, we showed that delNS1 replicates to comparable levels in gene-knockout mice lacking STAT1 (7 , 8) . The latter protein is essential in the IFN-signaling pathway leading to transcriptional induction of PKR. These results suggested that counteracting the PKR-mediated antiviral response is the main function of influenza virus NS1 protein.

Previously, it was shown that activated (oncogenic) ras induces a cellular inhibitor of PKR (9) . Activated ras is found in ∼30% of all malignant human tumors (10) . Thus, we reasoned that an activated ras-signaling pathway might allow replication of the delNS1 virus. We demonstrate here that the delNS1 virus replicates in otherwise nonpermissive cells upon transformation with activated N-ras. As a result, infection with delNS1 virus leads to lysis of these tumor cells. The lack of PKR activation in cells expressing activated ras seems to be the basis for the oncolytic properties of the delNS1 virus. Moreover, we show that this selective replication property of the delNS1 virus in malignant N-ras-expressing cells can be exploited for a tumor-ablative therapy in a SCID mouse model.

MATERIALS AND METHODS

Cells and Viruses.

518-neo cells (neomycin control transfectants) and 518-L1 (oncogenic N-ras transfectants; Ref. 11 ) were grown in DMEM (Bio-Whittaker, Verviers, Belgium) containing 10% FCS. HMEC-1, an immortalized human dermal microvascular endothelial cell line (12) was kindly provided by Francisco J. Candal (Centers for Disease Control and Prevention, Atlanta, GA) and was cultured in DMEM supplemented with 10% FCS, 2 mm glutamine, 1 μg/ml hydrocortisone (Sigma Chemical Co., St. Louis, MO), 10 ng/ml human epidermal growth factor (Sigma Chemical Co.), and antibiotics. The primary melanocytic cell line NHEM (Szabo Scandic, Vienna, Austria) was grown in melanocyte growth medium (Clonetics Cambrex, East Rutherford, NJ). Primary keratinocytes were cultivated in keratinocyte growth medium (Clonetics Cambrex). U937 (American Type Culture Collection) cells were cultivated in DMEM containing 10% FCS. Tu26 4 were propagated on RPMI 1640 medium containing 10% FCS, established from human hepatoma, and underwent more than 50 passages. VL-2 and VL-4 (13) were propagated on RPMI 1640 medium containing 10% FCS. For tumor xenotransplantation, 518-L1 cells were harvested, washed twice with PBS, and resuspended in physiological NaCl. Vero cells adapted to growth on serum-free media were obtained from American Type Culture Collection and grown in serum-free AIMV media (Life Technologies, Inc. Gaithersburg). For virus growth on the tumor cell lines Tu 26, VL-2 VL-4, and U937, we infected cells with a m.o.i. of 0.1 in the presence of 3 μg/ml trypsin in the media. For viral growth on 518-derived cells, no trypsin was added. Mutations in the hot-spot codons (12, 13, and 61) of K-ras, N-ras, and H-ras were determined in tumor cell lines as described (14) .

The PR8-wt virus was obtained by the plasmid-based transfection method described (15) . Briefly, plasmids coding for the NP and polymerase proteins and a plasmid coding for the PR8 NS gene segment were transfected into Vero cells. Transfected cells were then infected with the temperature-sensitive 25-A-1 virus (16) and incubated for 24 h at 37°C. Subsequently, transfectant viruses were selected by plaque purification of the supernatant at 39.5°C. The delNS1 virus was obtained in a similar manner, but with the open reading frame for the NS1 gene being deleted as described (7) . For analyses of replication properties, viruses were diluted in AIMV media (Life Technologies, Inc.), and the inoculum was applied at a m.o.i. of 0.1 in a volume of 100 μl to ∼10⁶ cells in 35 mm dishes. After 30 min, unadsorbed virus was removed by washing the cells twice with PBS, and AIMV media was added. Infected cells were incubated for 2 days at 37°C. For propagation of the viruses, Vero cells were infected at a m.o.i. of 0.1 and incubated with AIMV media containing 5 μg/ml trypsin for 3 days. Virus concentrations were determined using plaque assay on Vero cells (16) .

Immunofluorescence Analysis of Influenza Virus Infection.

Cells grown on chamber slides were infected with PR8-wt or delNS1 virus, at a m.o.i. of 3. Ten h postinfection, cells were fixed with aceton at 4°C. Thereafter cells were treated with PBS containing 1% FCS, incubated with a 1:500 dilution of the monoclonal antibody HT103, which binds to the NP of influenza virus 5 for 16 h at 4°C and washed with PBS. Cells were then incubated with a 1:350 dilution of an FITC-labeled goat antimouse antibody for 1 h at room temperature and washed with PBS. Cells were photographed with an immunofluorescence microscope.

PKR-Phosphorylation.

Transfectant melanoma cell lines expressing wild-type (518-neo) or oncogenic ras (518-L1) were either mock-treated or infected with delNS1 or WT PR8 virus at a m.o.i. of 3. To investigate the effect of 2-AP on PKR expression, the drug was added to the medium at a concentration of 5 mm. After 5 h, cells were washed with a phosphate-free buffer [118 mm NaCl, 4.75 mm KCl, 1.2 mm MgCl₂, 0.26 mm CaCl₂, 25 mm NaHCO₃, and 20 mm HEPES (pH 7.5)], incubated in DMEM lacking both phosphate and pyruvate (Sigma Chemical Co.) containing 500 uCI/ml ³²P[P_i] (Amersham, Aylesbury, United Kingdom) for 2 h. After labeling, the cells were washed twice with ice-cold PBS and 10 mm EDTA (without Ca²⁺ and Mg²⁺), and lysed for 10 min on ice in lysis buffer. The extracts were diluted to 300 μl with immunoprecipitation buffer and pretreated with protein G-agarose (100 μl of a 50% v/v suspension in wash buffer) for 1 h on ice. The agarose was removed and PKR antibody B10 (Santa Cruz Biotechnology, Santa Cruz, CA) was added. After 1 h on ice, protein G-agarose (100 μl) was added, and the suspension was agitated for 30 min at 4°C. The beads were washed with wash buffer, heated for 2 min at 95°C in 2× SDS-PAGE sample buffer, analyzed by 10% SDS-PAGE and autoradiographed for 7 days to determine PKR phosphorylation. Quantitations were done using densitometric analysis of the autoradiographs.

Western-Blot Analysis.

Whole-cell protein extracts of ∼10⁶ cells were prepared. The amount of soluble proteins was then quantified by means of a modified Bradford analysis (Bio-Rad, Richmond, CA). Total protein (20 μg) was applied to each lane and separated by SDS-PAGE. Proteins were transferred to polyvinylidene difluoride membranes (Millipore, Bedford, MA) by Western blotting. Blots were incubated with the N-ras-specific monoclonal antibody F155 (Santa Cruz Biotechnology). The membranes were washed twice with blocking solution and incubated with goat antimouse alkaline-phosphatase-conjugated second-step antibody (Tropix, Bedford, MA). Detection was performed using CSPD chemiluminescence substrate (Tropix), and membranes were exposed to Hyperfilm enhanced chemiluminescence (Amersham).

Animal Studies.

All procedures involving experimental animals were performed according to protocols approved by the National Committee on Animal Welfare. Pathogen-free male SCID-beige mice, 6–8 weeks of age, tested for leakiness, were obtained from the Institute of Virology at the University of Veterinary Sciences in Vienna, Austria. The animals were housed in laminar flow racks and microisolator cages under pathogen-free conditions and received autoclaved food and water. The mice were assigned to control or treatment groups (n = 9 mice/group). Subcutaneous tumors were generated by implantation of 1.5 × 10⁶ N-ras transfectant human melanoma cells 518-L1 in the left lower flank of anesthetized mice. For treatment, animals were given intratumoral injections of 2 × 10⁶ pfu of influenza delNS1 virus in 100 μl physiological NaCl solution. Virus was injected at 48-h intervals 3 or 5 times. The control group was treated with physiological NaCl solution alone. Flank tumor length (L), width (W), and depth (D) were determined every second day. Tumor volume was determined by the formula L × W × D. SCID mice were evaluated for tumor growth. Animals were killed at day 23 after tumor implantation or when the tumor size increased to 12,000 mm³.

Immunohistochemical Analysis.

To detect viral protein in tumor tissue, immunohistochemical analysis of s.c. established 518-L1 tumors was performed. Tissue samples of s.c. 518-L1 tumors, which had been injected with delNS1 virus 48 h before, were excised form the killed animal, rinsed with sterile PBS, and embedded into paraffin. Histological sections of the tumors were stained with H&E to confirm malignant tissue. Immunofluorescent analysis was carried out using a rabbit anti-influenza antiserum against PR8 virus, which detects total influenza virus protein. As a second antibody, we used an antirabbit horseradish peroxidase-labeled antibody.

Statistical Analysis.

To analyze the growth of flank tumors, an ANOVA was used to compare the area under the curve for tumor volume. Ps <0.05 were considered significant. For statistical data analysis, the SAS 8.0 (SAS Institute, Inc., North Carolina) statistical package was used. All data are presented as mean + SD of mean.

RESULTS

Growth of delNS1 Virus Is Dependent on Expression of Oncogenic ras in Vitro.

We first investigated whether expression of activated N-ras allows a productive infection of the delNS1 virus in cells otherwise not permissive for the virus. For this purpose, we took advantage of the melanoma transfectant cell line 518-L1, expressing high levels of oncogenic N-ras, and the corresponding cell line 518-neo, expressing wild-type ras only. The difference in ras expression of the two cell lines was confirmed by Western blot analysis (Fig. 1A) ⇓ . Although a viral titer of 7 × 10⁵ pfu/ml was determined in the supernatant of PR8-wt virus-infected cells expressing wild-type ras (518-neo), infection of these cells with the delNS1 virus did not result in the detectable release of infectious particles into the cell supernatant (Fig. 1B) ⇓ . In contrast, the delNS1 replicated to titers as high as 8.3 × 10⁴ pfu/ml in cells expressing N-ras (518-L1). In the latter cell line, the PR8-wt virus grew to titers of 1.4 × 10⁶ pfu/ml (Fig. 1B) ⇓ .

To determine whether abortive replication of the delNS1 virus was attributable to a block in viral entry, we analyzed the susceptibility of the cells to infection by immunofluorescence microscopy using an antibody against the NP, a protein expressed early in the replication cycle of influenza virus. Infection of the cell lines with delNS1 and PR8-wt virus as a positive control showed comparable levels of NP (Fig. 2) ⇓ . This finding suggests that the expression of N-ras in the melanoma cells did not alter viral entry or affect early protein synthesis of the NP.

We then determined the CPE induced by viral infection in the 518-derived cell lines. PR8-wt virus infection induced a complete CPE in both 518-neo and 518-L1 cells at a m.o.i. of 0.03 72 h postinfection. At this low m.o.i., the delNS1 virus induced complete CPE in the N-ras-transformed cell line only (Fig. 3) ⇓ . However, at an m.o.i. of 3, complete CPE could be induced by both viruses in 518-neo and in 518-L1 cells, respectively. Because infectious virus was not observed in the supernatants of 518-neo cells after delNS1 virus infection, we conclude that the delNS1 virus is not able to complete its replication cycle in these cells. As a consequence, the virus fails to spread from cell to cell and does not produce a visible CPE at a low m.o.i. However, delNS1 induces complete CPE after infection at a high m.o.i., suggesting that virus-infected cells are killed despite the abortive nature of the infection.

Growth of delNS1 Virus in 518-L1 Cells Is Associated with Inhibition of PKR during Virus Infection.

Previously we have shown that infection of normal cells with the delNS1 virus is associated with the activation of PKR, which prevents productive delNS1 virus infection. Cells which lack PKR allow replication of delNS1. We therefore hypothesized that PKR-activation is reduced in delNS1 virus-infected tumor cells, which express activated ras. To test this hypothesis, we determined the level of PKR-activation in infected cell lysates (Fig. 4) ⇓ . Baseline PKR activity was detected neither in mock-infected nor in PR8-wt virus-infected 518-neo or 518-L1 cells, respectively. In the delNS1-infected, replication-competent 518-L1 cells, the level of PKR activation was reduced ∼5-fold as compared with the level in nonpermissive infected 518-neo cells. The data suggest that virus-induced activation of PKR was inhibited because of the expression of oncogenic ras in the 518-L1 cells. To provide additional evidence that viral growth is dependent on the level of activated PKR, we determined PKR activation during viral infection in the presence of 2-AP. 2-AP is a chemical inhibitor of PKR activation. 2-AP treatment permits the delNS1 virus to replicate in this otherwise-nonpermissive cell line 518-neo (8) . As shown in Fig. 4B ⇓ , this 2-AP mediated rescue of delNS1 replication correlates with a reduction of activated PKR in delNS1 virus-infected cells. Thus, reduction of PKR activation by independent mechanisms lead to productive delNS1 virus replication.

Tumor-ablative Potential of the delNS1 Virus in Vivo.

To test the tumor-ablative potential of delNS1 virus-mediated oncolysis in vivo, N-ras transfectant 518-L1 cells were implanted into the left flank of SCID mice. In this tumor model, palpable tumors were established after ∼1 week. For tumor treatment, delNS1 virus was injected into the same site five times in 48 h intervals starting at day 5. Tumor growth was measured up to day 23 (Fig. 5) ⇓ . For statistical analysis, the area under the curve was determined. Tumor growth was substantially repressed in the treatment group (P = 0.0131). The experiment was repeated with similar results. The same dose of virus was also applied only three times at 2-days intervals starting at day 7. This treatment regimen still resulted in a significant reduction of tumor size (P = 0.014). However, the mean reduction of the tumor volume at the end of this experiment was ∼30% as compared with a reduction of 60% in the therapeutic regimen with five doses.

To determine whether the virus replicates in the tumor cells in vivo, we performed an immunofluorescence analysis, using an antiserum to influenza A virus, on the s.c. established tumors. Naive tumors of a size of 1000 mm³ were treated by injection of delNS1 virus or 0.9% NaCl. Staining was performed 48 h later. Cytoplasmic and membrane-bound staining was detected in the tumors, into which the virus was injected at the site of inoculation (Fig. 6B) ⇓ , but not in naive tumors (Fig. 6A) ⇓ . This indicates that viral replication in ras-transformed cells also occurred in tumor metastases in vivo. Adjacent muscle tissue was largely negative, although a faint staining above background level was still visible (Fig. 6C) ⇓ . This staining might be attributable to the abortive nature of the infection of the delNS1 virus in nonpermissive cells. It should also be noted that s.c. application of influenza delNS1 virus is completely apathogenic for SCID mice. No signs of disease, such as weight loss or ruffled fur, occurred after infection.

Growth of delNS1 Virus in Transformed Cell Lines of Different Histological Origins.

We determined whether productive replication of delNS1 also could be observed in other tumor cell lines. In the lung cancer cell lines VL-2 and VL-4, the delNS1 virus replicated to titers of 1 × 10⁴ and 2 × 10² pfu/ml, respectively. In contrast, no productive replication was detectable in the lymphoma cell line U937, and in the hepatoma cell line Tu26. In the cell lines which permit delNS1 virus replication, an oncogenic K-ras was detected. In contrast, no activating ras mutations were found to be present in the nonpermissive cell line Tu26. The U937 cell line was described to be negative for ras activation (17) . Although the absence of ras mutations do not rule out the lack of ras activation, the data indicate that the ras-dependent growth of the delNS1 virus is not restricted to the ras-transfectant cell line 518-L1.

Restricted Growth of the delNS1 Virus in Nonmalignant Cell Lines.

One requirement of a conditionally replicating virus for usage as a therapeutic agent is its restricted growth in nonmalignant cells. To investigate the selective replication properties of delNS1 virus in cells present in normal human skin, we infected primary keratinocytes, cultured primary melanocytes NHEM, and the microvascular endothelial cell line HMEC-1 with the delNS1 and the PR8-wt virus, respectively. Infections were done at a m.o.i. of 0.1 to allow multicycle replication. In neither of these cells, delNS1 virus infection led to the release of infectious particles into the supernatant. In contrast, these cell lines supported viral replication of the PR8-wt virus (Table 1) ⇓ .

DISCUSSION

We report that an influenza A virus lacking the NS1 open reading frame (termed delNS1 virus) may serve as a conditionally replicating oncolytic agent. The specificity of the delNS1 virus is based on the fact that replication of this virus is inhibited through activation of the ubiquitously expressed PKR. In tumor cells expressing oncogenic ras, activation of PKR is blocked. Correspondingly the delNS1 virus is able to grow in those tumor cells. Thus, a molecular marker which is highly associated with an oncogenic phenotype permits the lytic growth of an otherwise replication-defective virus.

Conditionally replicating oncolytic strains have been developed for several types of viruses including adenovirus (18 , 19) , herpes simplex virus (20 , 21) , poliovirus (22) , vesicular stomatitis virus (23) , and reovirus (24 , 25) . For some of these viruses, their oncolytic activity is already tested in the clinic (reviewed in Refs. 26 and 27 ). Moreover, bacteria such as Salmonella typhimurium also have been engineered to selectively target malignant tissue (reviewed in Ref. 28 ). This is the first description of a genetically engineered influenza virus (delNS1 virus) which has conditionally oncolytic properties. Because every microorganism family usually shows a certain extent of tissue tropism, we believe that tumor site and the molecular pattern expressed by the malignant cell might be relevant to which oncolytic microorganism will be beneficial for therapy. For this reason, we believe that the description of the oncolytic potential of a new virus family broadens the application of oncolytic viruses.

The delNS1 virus has a number of properties, which makes it suitable as an oncolytic strain for tumor therapy. (a) The delNS1 virus is apathogenic (8) . We therefore hypothesize that this agent would be associated with few side effects. The apathogenic phenotype is caused by a large deletion in its genome, which is unlikely to revert. It should be noted, that attenuated influenza A virus with multiple single point mutations have been shown to be safe when applied to humans (29) . (b) Because multiple serologically defined subtypes exist for influenza A viruses, different subtypes of the delNS1 virus can be constructed by exchanging the antigenic surface-glycoproteins of the virus. Therefore preexisting immunity of the host can be circumvented by choosing delNS1 variants with appropriate surface-glycoproteins. Also, the availability of such variants may allow repeated administration. (c) It is possible to construct influenza A viruses with shorter deletions in their NS1 protein. The length of the NS1 protein inversely correlates with the level of attenuation. This feature of the delNS1 virus allows the choice of the optimal length of the NS1 protein, which is associated with efficient tumor destruction but is still attenuated enough in the host to allow a safe application of the virus. Finally, we would mention also that the tumor ablative effect of the delNS1 virus is advantageous to the effect of cisplatin, a chemotherapeutic agent used in experimental melanoma therapy. In the 518-L1 SCID mouse model, cisplatin does not significantly reduce the tumor weight (11) .

Infection of a cell culture dish of the nonpermissible cell line 518-neo with the delNS1 virus at a m.o.i. >1 still leads to a CPE within 36 h. Obviously, infection of nonpermissive cells with the delNS1 virus leads to the death of the infected cell, despite that no detectable virus is produced. This delNS1 virus-induced cell death might be mediated by the induction and activation of PKR, because activated PKR has been shown to be associated with apoptotic cell death. Certainly alternative mechanisms of virus-mediated cell death can be envisioned. The induction of cell death by delNS1 virus infection of nonpermissible cells is a favorable effect, because it suggests that this virus has no potential for viral latency.

DelNS1 virus-mediated oncolysis might not be restricted to malignant cells with mutations in the ras proto-oncogene because PKR dysfunction is not restricted to the overexpression of oncogenic ras. For example, an aberrant function of upstream elements of the ras signaling pathway such as receptor tyrosine kinases could result in ras activation even in the absence of detectable ras mutations. Moreover, because PKR is an IFN-induced protein, low levels of activated PKR may also be caused by defects in the IFN pathway. Alterations in the IFN-dependent signal cascades, including changes in STAT1 and type I IFN receptor molecules, have been described to occur frequently in malignantly transformed cells. For example, several melanoma and lymphoma cell lines contain no or reduced levels of STAT1 (30 , 31) . In addition, leukemia cell lines were shown to be defective in IFN α and β genes (32) . In this regard, we have found efficient replication of the delNS1 virus in the lungs of STAT1 knockout mice (7 , 8) , supporting the assumption that this virus may replicate in tumor cells with defective IFN pathways. Experiments to investigate the potential of the delNS1 virus to serve as an oncolytic agent in malignant cells with various defects in the IFN pathway are currently being performed.

PKR antagonists expressed by viruses have been described for a number of virus families, including retroviruses (33) , herpesviruses (34) , vaccinia viruses (35) , and flaviviruses (36) . Deleting this very protein in a virus might be a reasonable strategy to design ras-dependent oncolytic viruses derived from these virus families. This article describes the oncolytic activity of genetically engineered influenza A virus and proposes a rational approach by which prototypes of oncolytic viruses can be designed.