Systemic Administration of a Recombinant Vaccinia Virus Expressing the Cytosine Deaminase Gene and Subsequent Treatment with 5-Fluorocytosine Leads to Tumor-specific Gene Expression and Prolongation of Survival in Mice¹

Metastatic colon cancer is the second leading cause of death from malignancy in the United States and Europe, with 80% of these patients developing metastases in the liver. Surgical resection is the only curative option in this situation, but fewer than 10% of patients are candidates for surgery because of multifocality and/or the extent of the disease (1). The results of other treatments, in particular radiotherapy, immunotherapy, and systemic chemotherapy, are still disappointing and have failed to provide curative potential (2). Thus, novel therapeutic strategies including gene therapy are worthy of investigation.

Current gene therapy approaches include the transfer of genes to induce antitumor immune responses, regulate growth, or sensitize tumor cells to specific treatments. The latter concept (suicide gene therapy) comprises the introduction of a nonmammalian gene encoding for a prodrug-converting enzyme into the tumor or tumor vicinity (3). The theoretical background for this approach is comprised of the different sensitivities of tissues toward cytotoxic agents subsequent to the artificial insertion of a foreign gene, also called “mosaicism” (4, 5). This approach hypothetically would open a therapeutic window that cannot be achieved by conventional treatment approaches. The efficacy of suicide gene therapy is, therefore, dependent on the magnitude of sensitivity differences between tumors and normal tissues, which itself is directly correlated with the specificity and efficiency of the vector system used. The specificity of this system relies on tumor-specific gene expression that in most experiments is achieved by local injection of the vector. Ultimately, the ideal tumor-directed gene therapy system should demonstrate tumor-specific expression of the foreign gene after simple systemic vector application. An optimal vector system would yield gene expression sufficient in magnitude and duration to allow for tumor-specific prodrug conversion eventually leading to the elimination of all of the tumor cells (6). Multiple viral and nonviral vector systems have been explored (7), with tumor-specific transduction efficiency remaining the main problem for successful in vivo application of the concept (8, 9, 10).

Equally important for the successful in vivo application of suicide gene therapy as selective transduction efficacy is the efficiency of the enzyme/prodrug system used (11). Several systems have been described (12, 13, 14, 15, 16), including herpes simplex virus TK/ganciclovir, varicella zoster virus TK/6-methoxypurine arabinucleoside, xanthine-guanine phosphoribosyl-transferase/6-thioxanthine, 6-methylpurine deoxyriboside/purine nucleoside phosphorylase, and 5-FC/CD.³

On the basis of the extensive clinical use of 5-FU in colon cancer, the 5-FC/CD system is the most compelling for experimental treatment of metastasized colon cancer (17). Furthermore, for the 5-FC/CD system, a better “bystander effect” was described in direct comparison with other systems (18, 19, 20) because of the better diffusibility of the drug, which increases the in vivo potential of the system. The principal effectiveness of the system in the treatment of experimental colon cancer has been shown in vitro and in animal tumor models after intratumoral vector-mediated gene administration (21).

We have demonstrated previously (22) that a TK− recombinant vaccinia virus may be a promising candidate for efficient tumor-specific gene delivery. We here describe in detail the pattern of gene expression after i.v. vector administration in a murine model of liver metastases in both immunocompetent and athymic mice. We then demonstrate the in vivo application of this vector system in a suicide gene therapy setting using systemically delivered vaccinia-expressing CD followed by 5-FC to treat liver metastases, leading to complete tumor regression and cure in a significant proportion of the animals.

The recombinant vaccinia viruses used are based on a vaccinia shuttle plasmid, pCB023II, which is described in detail elsewhere (22). Briefly, the multiple cloning site of pBluescript KS II (Stratagene, La Jolla, CA) was inserted in between the TK flanking region of pSC65 (kindly provided by B. Moss, Laboratory of Viral Diseases, National Institute of Allergy and Infectious Diseases, Bethesda, MD;23). A gpt gene under the control of the p7.5 vaccinia early/late promoter was inserted for selection purposes (24). To prevent antisense transcription from native vaccinia promoters, an early termination signal (TTTTTNT) was designed for all of the three possible open reading frames and inserted just downstream of the left vaccinia virus TK flanking region (25). Any gene within the multiple cloning site is driven by a synthetic early/late promoter (23, 26, 27).

The CD gene was amplified from genomic DNA of Escherichia coli via PCR using pfu polymerase (Stratagene, La Jolla, CA). The CD gene was then ligated into the pCRII cloning vector (Stratagene), and the correct gene sequence was confirmed by automated sequencing (ABI Prism, Perkin-Elmer Applied Biosystems, Norwalk, CT). Finally, the CD gene was ligated into pCB023II, resulting in pMP-CD (Fig. 1).

Recombination of vaccinia viruses (strain Western Reserve) was performed using CaCl precipitation (28). Recombinant vaccinia viruses were subsequently isolated by mycophenolic acid selection, and the final viral construct was designated vvCD.

Another recombinant vaccinia virus, vvLuc, expressing the firefly luciferase gene, was used as a reporter vector and is described elsewhere (22).

Recombinant viruses were amplified in HeLa S3 cells using spinner flasks as described previously (28) and titered in standard plaque-forming assays using CV-1 cells.

MC38 is a murine colon adenocarcinoma derived from C57/BL mice (29). Human ovarian cancer cells HeLa S3 and monkey kidney cells CV-1 were obtained from the American Type Culture Collection (Manassas, VA) and used for generation, amplification, and titration of recombinant vaccinia viruses. All of the cell lines were cultured in DMEM supplemented with 10% heat-inactivated fetal bovine serum, 2 mm l-glutamine, 1% penicillin/streptomycin, 1% gentamicin, and 0.2% fungizone (Biofluids, Rockville, MD) at 37°C in a 5%-CO₂ incubator.

vvCD (10⁸ pfu) was incubated in a 6-well plate (Corning Costar, Cambridge, MA) with 1 μg of psoralen (4′ -aminomethyl-trioxalen, Calbiochem, La Jolla, CA) in 1 ml of HBSS and 0.1% BSA for 10 min. The plate was then radiated with 365-nm-long wave UV light using a Stratalinker 1800 (Stratagene) for 4 min (30). Virus samples were assayed for remaining replication competent virus (28).

MC38 cells were plated into flat-bottomed 96-well plates (Corning) at a concentration of 5000 cells/well and grown overnight. The media was then replaced with 100 μl of HBSS/0.1% BSA and inactivated vvCD at a multiplicity of infection of 1. After 2 h, 200 μl of complete DMEM was added to each well, and the plates were incubated overnight. After 24 h, the media were changed to complete DMEM containing the experimental concentrations of 5-FC (Sigma, St. Louis, MO). In daily intervals thereafter, cytotoxicity assays were performed to determine cell viability. MTT (50 μl, Sigma) was added to each well in fresh media and incubated at 37°C for 4 h. The fluid was then aspirated, and 120 μl of DMSO (Fluka, Switzerland) were added. The plates were wrapped in aluminum foil, placed on an orbital shaker, and shaken at 1500 rpm for 30 min at room temperature. The absorbance at 570 nm was measured on a plate reader (Multiskan MDD/340, Titertek, Huntsville, AL). All of the experiments were performed at least in triplicate; means and SDs were recorded. Cell viability is reported as the ratio (percent) viable cells normalized to untreated control samples.

Female C57/BL6 and athymic NCR-nude mice, 8–12 weeks old, were obtained from the NIH small animal facility and housed in barrier care rooms. Animals had unlimited access to food and water and were housed at a maximum of five mice per cage. All of the animal protocols were approved by the NIH Animal Care and Use Committee and conducted in strict compliance with the guidelines established by the NIH Animal Research Advisory Committee.

Disseminated liver metastases were established using a modification of a previously described method (31). MC38 adenocarcinoma cells were harvested from cell culture flasks using trypsin-EDTA (Life Technologies, Inc., Grand Island, NY), washed three times in HBSS, and adjusted to a final concentration of 1.25 × 10⁶ cells/ml. Cell preparations were kept on ice until injection. Mice were anesthetized with ketamine (40 mg/kg, Fort Dodge, Fort Dodge, IA) and xylazine (12 mg/kg, Butler, Columbus, OH) given i.p. and were prepared for surgery under sterile conditions. Animals were positioned on cardboard in a right lateral position and washed with 70% ethanol. A left s.c. incision approximately 6 mm long was made and the peritoneum opened. The spleen was exposed and gently retracted, the gastrosplenic ligament and the short gastric vessels were identified and cut, leading to complete mobility of the spleen. The spleen was then exteriorized out of the abdominal cavity and positioned on a moisturized piece of sterile gauze (3 × 3 cm). The prepared cell suspension (200 μl, = 2.5 × 10⁵ MC38 cells) were then slowly injected into the spleen using a 30-g needle (Becton Dickinson, Franklin Lakes, NJ) introduced through the upper pole. After injection, the needle was slowly removed, and slight pressure was applied to the spleen to achieve hemostasis. At least 5 min was allowed for the tumor cells to flow into the splenic vein and thus into the portal venous circulation. The splenic pedicle was then tied with 4–0 silk (Ethicon, Somerville, NJ), and the spleen was removed. After splenectomy, the abdominal cavity was closed in one layer with 9-mm wound autoclips (Roboz Surgical, Rockville, MD).

This murine tumor model reliably yields disseminated metastases confined to the liver. We have induced metastases in more than 2500 animals with less than 1% perioperative mortality and more than 99.9% tumor “take.” After injection of 2.5 × 10⁵ cells as described, untreated animals develop microscopic liver metastases within 3 days and macroscopic disease around day 7 and die from tumorous hepatic replacement around 4 weeks after tumor inoculation. For the experiments described, virus injections were uniformly performed on day 12 after tumor inoculation.

For the reporter gene experiments, animals were injected i.v. with vvLuc 12 days after tumor inoculation. In immunocompetent C57/BL6 mice, 10⁸ pfu vvLuc were used, whereas in nude mice 10⁶ pfu of virus were administered. Cohorts of animals were killed at various time points. Tumors and normal tissues (liver, kidneys, pancreas, uterus, ovary, bowel, heart, lung, muscle, skin, brain, scar tissue) were harvested, immediately frozen, and stored at −70°C.

Luciferase activity was assayed using a commercially available assay system (Promega, Madison, WI). Frozen tissue samples were briefly thawed, homogenized and lysed in 750 μl of 1× reporter gene lysis buffer. Samples were centrifuged for 5 min at 13,000 rpm to pellet the cellular debris. Supernatant (10 μl) was added to 100 μl of luciferase assay reagent in an 8-mm × 50-mm disposable cuvette and immediately placed into a luminometer (TD-20/20, Turner, Sunnyvale, CA). After a 2-s delay, light emission was measured for 10 s. Each sample was measured in duplicate, and mean values were recorded. The concentration of total protein in each sample was then determined using a bicinchoninic acid protein assay kit (Pierce, Rockford, IL) using a BSA standard. All of the measurements were done in triplicate. The final firefly luciferase activity is expressed in relative light units per mg protein.

For the in vivo treatment experiments, groups of animals were injected with vvCD or control vector (10⁸ pfu for immunocompetent mice and 10⁶ pfu for nude mice) i.v. 12 days after tumor inoculation. Beginning 2 days after systemic virus administration, 5-FC was administered for a total of 7 days (daily dose: 1180 mg/kg body weight, administered i.p. in HBSS). Appropriate control groups were used in all of the experiments, and all of the experiments were repeated for confirmation. The end point for all of the animal experiments was survival. Animals surviving 120 days were killed and necropsied to confirm the absence of tumors.

Data are given as mean ± SD, and Student’s paired t test (2-sided) was used when appropriate. Survival analysis was done using the Kaplan-Meier method (32). Differences were assessed using the log-rank test (33).

To demonstrate that infection of MC38 cells with vvCD would lead to efficient prodrug conversion to 5-FU, in vitro cytotoxicity assays were performed. To avoid cytolytic vaccinia effects, vvCD was inactivated for these experiments. As shown in Fig. 2, the conversion of 5-FC to 5-FU leads to significant cell killing compared with the control groups. Neither the prodrug alone (no virus/5-FC), nor the vector alone (vvCD/no treatment), nor prodrug treatment in combination with an irrelevant vector (vvLuc/5-FC) decreased cell viability as compared with untreated controls, whereas the treatment group (vvCD/5-FC) showed 78% cytotoxicity after 7 days (P < 0.001).

After i.v. administration of vvLuc to tumor-bearing mice, reporter gene activity over time was determined. HBSS injections were used as negative controls. At various time points, groups of mice (n = 10 per time point) were killed, and tumor and tissue samples were harvested and assayed as described.

As shown in Table 1, significant reporter gene activity was measurable for about 1 week in immunocompetent animals. The gene expression in tumors was dramatically higher than in any other tissue. At the peak on day 4, tumor luciferase activity was increased 2600-fold compared with the ovary, the next highest organ (P < 0.00001). On day 6, luciferase activity in tumors was still higher than other organs at least 650-fold. In general, ovaries showed the highest gene expression among nontumor-bearing organs. Gene expression in other organs such as kidney or lung was usually about 4 logs lower than in tumor. Measurements from normal liver must be considered false-positive because of contamination with gene-expressing tumor cells.

In athymic mice, maximum luciferase activity in tumor occurred on day 8 (Table 2). Despite considerably lower virus dose (10⁶ as compared with 10⁸), luciferase activity in tumor was still 500-fold increased in comparison with ovary (P < 0.00001) and 50,000-fold as compared with lung. Significant gene expression in tumor was detectable for more than 5 weeks.

As shown in Fig. 3, a marked difference in luciferase activity of at least 100-fold (usually between 1,000- and 10,000-fold) was observed between tumor tissue and the next closest organ over the full duration of the experiments, both in immunocompetent and nude mice. The observed pattern in reporter gene activity supports the existence of a “therapeutic window” encouraging the usefulness of the approach in the treatment setting.

The murine liver metastases model was used for in vivo treatment experiments as described. Two days after i.v. administration of vvCD, prodrug was administered i.p. at 1180 mg/kg 5-FC daily from day 2 to 8.

As shown in Fig. 4, survival in the treatment group (vvCD/5-FC) was markedly improved as compared with controls (P < 0.001). Survival in all of the other groups did not differ irrespective of control group: (a) prodrug alone (no virus/5-FC); (b) vector alone (vvCD/no prodrug); (c) irrelevant vector (vvLuc/5-FC); or (d) untreated animals (no vector/no treatment). The median survival of treatment animals was increased by more than 50% (53 versus 34 days; P < 0.0001), and 4 (18.2%) of 22 animals in the treatment group survived indefinitely. No control animals survived.

Fig. 5 shows the liver of a representative treatment animal as compared with a control on day 18. Whereas the liver of the treatment animal (A) shows only a few small tumor nodules, the control mouse (B) has subtotal hepatic replacement with tumor.

A similar treatment experiment was performed in nude athymic mice. As shown in Fig. 6, the survival benefit for the treatment group was even more pronounced amid decreased immunological vector elimination. The median survival was increased by 103% (68 versus 33 days; P < 0.0001), and the cure rate in the treatment group was 30.4% (7 of 23 animals). All of the control animals died.

We observed moderate viral toxicity in nude mice indicated by clinical signs of weight loss, lethargy, and flu-like symptoms, usually occurring later than 6 weeks after vector administration. This did, however, not affect survival of these animals within the experimental period. In immunocompetent mice, no such symptoms were observed.

Additional control groups of animals treated without any vector and 5-FU alone at the maximal tolerated systemic dose of 30 mg/kg were included in all of the treatment experiments. Survival of these mice did not differ from that of other control groups (data not shown).

The present study evaluates the systemic application of a TK− recombinant vaccinia virus as a vector for tumor-directed gene therapy. The crucial issue for a suicide gene therapy approach to work effectively is tumor-specific gene delivery in vivo (9). Only if the gene encoding for the prodrug-converting enzyme is exclusively or at least predominantly expressed in tumors, can the hypothetical concept work. To investigate the properties and characteristics of the TK− vaccinia virus vector system, we established metastatic tumors of syngeneic colon carcinoma cells in the liver of C57BL/6 and athymic nude mice. We constructed recombinant vaccinia viruses expressing the firefly luciferase gene and the E. coli CD gene, respectively, and administered those viruses i.v. in our in vivo model of hepatic metastases. We show the feasibility of tumor-specific gene delivery using this vector. Functional CD expression was achieved and led to tumor cell killing in vitro and increased survival and cures in vivo.

The most important basis for the success of this approach is the unique ability of recombinant vaccinia viruses to specifically transduce tumors. Other viral vectors such as adenoviruses lack this tumor-tracking ability. Instead, adenoviruses have extensive hepatotrophic characteristics, leading to substantial hepatotoxicity (34). Because it is impossible to accurately determine hepatic gene expression in the present study because of the impossibility of exact separation of nontumorous hepatic tissue in a liver metastasis model (see Fig. 5), the assessment of vaccinia hepatotrophism has to rely on results from other tumor models. We have found no substantial hepatic gene expression after systemic administration of the TK− vaccinia virus in models of s.c. or i.p. tumors (22). We also observe some gene expression in other organs in our model, but the reported manifold increased expression of the genes transferred within tumors as compared with other tissues provides a window of opportunity for many treatment approaches, including the delivery of a prodrug converter-enzyme gene. Although some tumor specificity may be achievable in other vector systems using complicated modifications like tissue-specific or heat-inducible promoters (35, 36, 37), recombinant TK− vaccinia viruses are genuinely tumor-specific without such additional modifications that may decrease efficiency.

There are several possible explanations for the tumor selectivity of the TK− vector system, including the relative large size of the virus, specific receptors, and decreased immune clearance within tumors (38, 39). Most likely, however, the specificity is caused by the disruption of the vaccinia virus TK gene by the inserted transgene (40, 41). As a cytoplasmic virus, vaccinia relies on its own enzymes to create functional nucleotides for DNA synthesis and viral replication. By deleting the TK gene, the virus must then rely on host cell nucleotides for its replication. Senescent cells lack a large pool of functional nucleotides and, therefore, may not be permissive for viral replication. Cancer cells on the other hand have a higher percentage of cells synthesizing nucleotides for cell division, allowing for a pool of nucleotides for viral DNA synthesis and viral replication. There are, however, differences between cell lines in terms of permissiveness for recombinant vaccinia viruses (22), potentially limiting the tumor types in which the system is efficient.

Modified recombinant vaccinia viruses have also been used for other gene therapy approaches, including treatment concepts using cytokine genes (39, 42, 43), tumor antigens (44, 45), or costimulatory molecules (46) to enhance tumor immunogenicity (47). Those approaches rely on the high immunogenicity of the viruses used, which leads to long-lasting immunity against antigens encoded by the viral vectors (48). Although this most likely will preclude repeated application of the same vector (49), it also provides a safety feature for the systemic application of these viruses in the clinical setting (50), which is particularly important when the use of attenuated but replication-competent viral vectors is discussed. Also, the transgene expression demonstrated in nontumorous organs gives reason for concern, even when nonessential organs such as the ovaries are involved. Nevertheless, in clinical Phase I trials of recombinant vaccinia viruses in immunocompetent cancer patients, no significant vector-associated morbidity has been observed thus far (44, 50, 51).

Local or intratumoral delivery of prodrug converter genes or immune response-activating genes has yielded promising results (52, 53), but the ultimate goal of gene therapy is the clinically simple systemic delivery of a gene-encoding vector (54). An interesting approach has recently been described using the hepatotrophic properties of adenoviruses to “sensitize” the vicinity of liver metastases toward subsequent prodrug application (55). Although the authors remain optimistic about the toxicity of this concept, others have described severe hepatotoxic side effects of adenoviruses (7, 34, 56). In contrast, recombinant vaccinia viruses do not lead to significant hepatic toxicity (41, 44). Locoregional application strategies may, however, provide quantitative improvements in the efficacy of our gene delivery system and warrant further investigation.

We have previously suggested that a TK− vaccinia virus shows higher efficiency than other viral vectors (22), but an effective antitumor response still relies on the presence of a bystander effect in vivo. It has been estimated that even with the most efficient gene delivery systems, only a relatively small percentage of tumor cells will express the transgene in vivo (7, 9). As a replicating virus, the virus alone will kill a small percentage of cells within the tumor. We have not seen an antitumor effect of virus without prodrug at the doses used in these experiments. In other experiments at higher viral titers, we have seen an oncolytic effect of the virus.⁴ The addition of a suicide gene should allow for more diffuse killing of tumor cells based on a bystander effect before immunological clearance of the vector. The CD/5-FC system provides a potent bystander effect (3), especially compared with the herpes simplex virus TK/ganciclovir system, in which the existence of a notable bystander effect remains controversial (9, 11, 55, 57). After conversion from 5-FC, 5-FU readily diffuses across cell membranes and involves untransduced neighboring cells. Thus, delivery and expression of CD in a comparably small number of tumor cells will be sufficient to achieve a significant antitumor effect (58), which is also demonstrated by the results of our in vivo treatment experiments. A potential advantage to a suicide gene in a replicating vector is that the addition of the prodrug may control viral replication and toxicity. We may also find virus escape mutants that lose the CD gene after treatment with 5-FC. These issues will be addressed in future experiments.

The question whether only partial tumor regression and increased survival or complete tumor regression and cure can be observed in an individual animal, is most likely decided by timing and quantitative aspects of gene expression and local prodrug concentration. 5-FU induced cytotoxic death of the tumor cell also eliminates the vector, and the necessary optimal sequence of events within tumors may be influenced by many factors like local tumor architecture, vascular support, and local immune clearance of the vector (3). The increased local concentration of 5-FU, however, is the reason for the superiority of this approach over direct systemic 5-FU administration. In fact, it has also been reported by others that tolerable systemic 5-FU doses do not cause significant tumor regression (20, 21, 59), which was confirmed in our experiments.

Not surprisingly, intratumoral gene expression in our in vivo tumor model was sustained for a much longer period of time in immunocompromised animals. Without adequate T-cell response, vector elimination takes much longer in nude mice as demonstrated in our reporter gene studies. The similarly prolonged CD gene expression leads to a markedly increased survival benefit in the treatment situation as compared with immunocompetent animals, further indicating that quantitative factors play a major role in determining the success of achieving complete tumor eradication. One can envision that modifications in dosage and/or schedule of gene administration and prodrug treatment may further improve these quantitative balances, ultimately leading to increased cure rates in in vivo treatment experiments. In a clinical setting, temporary immunosuppression of patients may allow for a limited period of gene expression sufficient to achieve responses in otherwise untreatable palliative situations. Other similar approaches may help to overcome the presence of antivaccinia antibodies in the serum that exists in the majority of the current cancer patient population. It is also possible that a viral and suicide gene necrotic cell death may lead to a potent antitumor immune response. This will be addressed in future experiments.

In summary, we were able to demonstrate tumor regression and cures in an in vivo model of established liver metastases of colon adenocarcinoma based on tumor-specific delivery of a prodrug converter gene using the TK− vaccinia virus vector system. Optimization of gene delivery, vector and prodrug dosages, and administration modalities may further enhance the results of suicide gene therapy and contribute to the translation of these promising experimental systems into clinical practice.

We thank Dr. Steven Rosenberg for helpful comments and Barbara Owen for technical assistance with the manuscript.