Materials and methods

Virus construction and preparation

Adenovirus shuttle vectors for Ad.BG and Ad.IR-BG were based on pE1sp1a (Microbix, Toronto, Canada) for E1-deleted virus construction. Ad.IR-BG vector containing the -galactosidase gene under the control of RSV promoter was constructed as described previously.¹³ Ad.Luc vector with CMV-Luciferase gene was originally constructed through a homologous recombination in bacteria using the vectors pAdEasy-1 and pAdTrack-CMV. Vectors were propagated by the Vector Core Facility of the University of Michigan (Ann Arbor, MI) according to the protocol described previously.¹² Virus was titered determined by plaque assay and checked for contamination with wild-type virus by PCR analysis, which consistently yielded less than one wild-type viral genome in 1 10⁹ PFU.¹¹ Ad.BG control vector containing RSV-driven -gal gene was provided by the Vector Core Facility (University of Michigan, Ann Arbor, MI).

Cell culture, viral infection, and -gal assays

Human hepatocellular carcinoma cells (SMMC7721 and MHCC97) were a kind gift from Dr Tang (Liver Cancer Institute, Fudan University, Shanghai, P.R. China) and were maintained in DMEM media containing 10% FBS and antibiotics at 37°C with 5% CO₂. Human colon (LoVo and HT29) cancer cell lines were obtained from American Type Culture Collection (Manassas, VA). The rat liver epithelial cell line (WB) (provided by Dr James Trosko at Michigan State University, MI) was cultured in RPMI media containing 10% FBS and antibiotics. The normal human colon epithelial cell line (NCM460) was obtained from INCELL Corp. (San Antonio, TX) and was maintained in recommended medium (INCELL M310A). Normal human hepatocytes (WRL68, ATCC) were maintained in DMEM as recommended. Stable cell lines expressing luciferase (LoVo-Luc and SMMC7721-Luc) for tumor implantation in mice were established by infecting parental LoVo and SMMC7721 with a lentivirus vector containing CMV-driven luciferase gene.

Cells at 2 10⁴/well in 24-well plates were infected by Ad.BG or Ad.IR-BG (2 10⁷ PFU) for x-gal staining or by a mixture of Ad.BG or Ad.IR-BG with Ad.Luc vector (2 10⁶:2 10⁶ PFU) for -gal activity measurement in the presence or absence of hydroxyurea (5 or 10 mM) for 72 hours. For -gal staining, cells were fixed in 1% paraformaldehyde prior to x-gal staining as previously described.⁵ Activity assays for -gal and luciferase were performed on cell lysates harvested after viral infection using commercially available -gal or luciferase assay systems (Promega, Madison, WI) according to the manufacturer's recommended protocols. The -gal activity was expressed as units corrected by luciferase activity for each cell line.

PCR analysis for viral DNA recombination and replication

LoVo, SMMC7721, WB, and NCM460 cells (1 10⁵/well in six-well plates) were infected by Ad.BG. or Ad.IR-BG for 4 hours at multiplicity of infection (MOI) to produce approximately equal infection frequency followed by extensive washing in PBS to remove free virus particles from the media. Cells were then lysed immediately or at 4 days after continued culture for cellular DNA purification using Genomic DNA purification Kit (Promega, Madison, WI). The recombination of viral DNA in infected cells was detected by PCR analysis using primers designed to detect recombination (Rec) (Fig 2a): Rec 5' AACGCCATTTGACCATTC and Rec 3' CGGGCCTCTTCGCTATTAC. Products from standard PCR reaction (94°C 1', 55°C 1' and 72°C 1' for 25 cycles) were then separated on 1% agarose gel. The viral DNA content, an indicator for viral DNA replication during infection, was quantified by real-time PCR analysis using an Opticon System (MJ Research, San Francisco, CA). The PCR reaction was performed in a 20 l volume consisting of 1 l of sample, 10 l of QuantiTect SYBR Green PCR Master Mix (Qiagen, Valencia, CA), and 1 l of each primers (0.5 M) according to the manufacturer's suggested conditions. These primers were designed for the -gal gene internal sequence (Int) (Fig 2a) that included Int 5' CACGGCAGATACACTTGCTG and Int 3' ATCGCCATTTGACCACTACC. The number of viral DNA copies was calculated from a standard curve of Ad.BG or of Ad.IR-BG versus PCR cycle number, and is presented as log expression of DNA molecules per l of sample.

Figure 2.

Replication-dependent viral DNA recombination and gene expression. (a) Cells in 24-well plates were infected by Ad.BG or Ad.IR-BG (MOI 100) in the presence or absence of hydroxyurea (5 mM) for 72 hours. Cells were fixed in paraformaldehyde and were stained by x-gal for -gal expression. (b) Illustrations of viral structure for Ad.BG, Ad.IR-BG and recombination form of Ad.IR-BG (Ad.IR-BG). Ad-ITR: adenovirus inverted terminal repeat; : viral packaging signal: IR; inverted repeat; PA: poly A sequence; Ad.E2,3,4: adenovirus DNA containing E2, 3, and 4 genes; Orientations for RSV promoter and -gal gene (5'–3') are indicated. Expected PCR products using primers for viral DNA recombination (Rec: 5'-AACGCCATTTGACCATTC, 3'-CGGGCCTCTTCGCTATTAC) and -gal internal sequence (Int: 5'-CACGGCAGATACACTTGCTG, 3'-ATCGCCATTTGACCACTACC). (c) LoVo, SMMC7721, WB, and NCM cells were infected with Ad.BG or Ad.IR-BG (MOI 10, 10, and 500, respectively) for 4 hours. After continued incubation for 72 hours in the presence or absence of hydroxyurea, cellular DNA was purified as described in Materials and methods for PCR analysis using Rec primers.

Full figure and legend (145K)

Tumor implantation, virus administration, and gene expression analysis

In all, 2 10⁶ LoVo-Luc and SMMC7721-Luc cells were injected subcutaneously in 7–8-week-old nude mice (Nu/Nu CD-1, female, Charles River Laboratories, Wilmington, MA) for 2 weeks. To ensure a single intrahepatic tumor nodule for in vivo imaging analysis, these subcutaneous tumors were dissected and minced into pieces (1 1 1 mm). Viable tumor tissues were then embedded directly into the livers of mice by using nylon suture (5-0, Ethicon, Inc., Cornelia, GA) through laparotomy. Tumor size was verified 3 weeks after tumor implantation by bioluminescence, as described previously.⁶ Mice bearing LoVo or SMMC7721 xenografts were administrated 4 10⁹ PFU of Ad.BG or Ad.IR-BG through the tail vein. Tumors and normal tissues including liver, lung, spleen, kidney, colon, and bone marrow were collected for homogenization or DNA purification at 72 hours after viral infusion for -gal activity measurement or PCR detection of viral DNA recombination, respectively. All animal-related procedures were performed according to the established procedures of the University of Michigan Laboratory Animal Maintenance Manual.

Data analysis and statistics

All values are expressed as meanSE and were compared by ANOVA analysis. Data were considered significantly different when P <0.05.

Results

Tumor-specific gene expression by Ad.IR-BG vector

We began by infecting human colon carcinoma (LoVo and HT29) and hepatocellular carcinoma (SMMC7721 and MHCC97) cells, normal human colon epithelial (NCM460) cell, and normal rat hepatocytes (WB) with either Ad.BG or Ad.IR-BG. We found that Ad.BG infection resulted in similar -gal expression in all cell lines (4 10^-2 U) (Fig 1). In contrast, Ad.IR-BG infection achieved significantly greater -gal expression in tumor cells (33.8 10^-2, 24.1 10^-2, 28.9 10^-2, and 14.0 10^-2 U for LoVo, HT29, SMMC7721, and MHCC97, respectively), whereas much lower -gal activities were observed in normal cells (0.23 10^-2 and 0.25 10^-2 U for WB and NCM460, respectively) (Fig 1) at 72 hours. Notably, however, -gal expression in tumor cells 24 hours after infection was similar regardless of whether Ad.BG or Ad.IR-BG was used (data not shown). This suggested that enhanced gene expression produced by Ad.IR-BG infection in tumor cells involves mechanisms in addition to transient gene expression utilizing the existing viral DNA templates.

Figure 1.

Ad.IR-BG-mediated tumor-selective gene expression. Human colon carcinoma cells (LoVo and HT29), hepatocellular carcinoma cells (SMMC7721 and MHCC97), rat hepatocytes (WB), and human colon epithelial cells (NCM) were infected by a mixture of Ad.BG or Ad.IR-BG and Ad.Luc vector (1:1 at MOI 100 each) for 72 hours. Activities for -gal and luciferase were measured from cell lysates by assay kits as described in Materials and methods. The -gal expression is expressed as units of activity normalized by luciferase activity for each cell line (U 10^-2). n=4, ^*P < 0.05.

Full figure and legend (46K)

Ad.IR-BG-mediated. gene expression requires viral DNA replication and recombination

To determine whether viral DNA replication and recombination were involved in Ad.IR-BG-mediated selective gene expression in tumor cells, hydroxyurea was used in cell infections as a ribonucleotide reductase inhibitor to block DNA synthesis. Hydroxyurea had no effect on the transgene expression of either normal or tumor cells by Ad.BG infection (Fig 2a), suggesting that Ad.BG-mediated gene expression does not require DNA replication. In contrast, Ad.IR-BG-mediated gene expression in tumor cells was significantly suppressed by incubation with hydroxyurea (Fig 2a), suggesting that Ad.IR-BG-mediated gene expression in tumor cells requires DNA replication.

To further determine whether -gal expression mediated by Ad.IR-BG infection was the result of DNA recombination between viral DNAs, specific primers were used in PCR amplification for viral DNA of the recombination form (Rec) (Fig 2b) from lysates of infected cells. LoVo, SMMC7721, NCM460, and WB cells were chosen to represent colon carcinoma metastases, primary liver cancer, and normal cell controls, respectively. As expected, the native form (0.45 kb fragment) but not the recombination form (1.10 kb fragment) was detected in all Ad.BG-infected cells regardless of the presence of hydroxyurea (Fig 2c). In contrast, recombination was dramatically inhibited in the presence of 5 mM hydroxyurea and was completely blocked by 10 mM hydroxyurea in all Ad.IR-BG infections (Fig 2c), indicating that recombination between viral DNAs requires DNA replication. These results demonstrated that Ad.IR-BG selectively transduced tumor cells as a result of viral DNA recombination that is activated by DNA replication in cells.

Viral DNA replication is prerequisite for the selectivity of gene expression in Ad.IR-BG-infected. tumor cells

To quantify viral DNA replication following viral infection, we developed a real-time PCR approach to measure viral DNA content in cell lysates. With the use of virus stocks as standards and Int (Fig 2b) as reaction primers, this method proved to be highly accurate and sensitive in quantifying adenovirus virions ranging from 10 to 1 10⁹ particles (data not shown). We found minimal replicated viral DNA in normal colon epithelial cells (NCM460) after a 4-day culture regardless of whether Ad.BG or Ad.IR-BG was used for infection (Fig 3a and b). As expected, viral DNA replication was observed in viral infected LoVo and SMCC7721 cells. However, the magnitude of viral DNA replication was similar in LoVo or SMCC7721 cells infected by either Ad.BG (1.5 0.2 to 2.2 0.3 and 1.3 0.1 to 2.0 0.2, respectively) (Fig 3a) or Ad.IR-BG (1.4 0.1 to 2.0 0.3 and 1.4 0.1 to 2.1 0.3, respectively) (Fig 3b). Although viral DNA replication was also noticeable in hepatocytes (WB) in contrast to normal colon epithelium (NCM460) followed by viral infection (Fig 3a and b), Ad.IR-BG infection resulted in minimal gene expression in both cells (Fig 1). This suggested that the selective -gal expression in Ad.IR-BG-infected tumor cells is not the direct outcome of increased viral DNA copies. The fact that hydroxyurea inhibited Ad.IR-BG-mediated gene expression (Fig 2a and c) suggested that viral DNA replication, although not a direct determinant, is required for Ad.IR-BG-mediated DNA recombination and selective gene expression in tumor cells.

Figure 3.

Real-time PCR analysis of viral DNA replication in cells. LoVo, SMMC7721, WB, and NCM cells were infected by Ad.BG or Ad.IR-BG (MOI 10, 10, 500, and 100, respectively) in six-well plates for 4 hours before washing with PBS to remove free virus in media. Incubations were continued in fresh media for 4 days. Cellular DNA were prepared from cell lysates at Day 0 and Day 4 for real-time PCR analysis. The PCR reaction was performed using 1 l of DNA sample and Int as primers by an Opticon System (MJ Research) described in Materials and methods. DNA molecules were calculated from standard curve and were expressed as log of DNA molecules/l. (a) Ad.BG infections, (b) Ad.IR-BG infections, n=5, ^*P < 0.05.

Full figure and legend (47K)

Ad.IR-BG mediates selective gene expression in intrahepatic tumor xenografts

These promising in vitro data encouraged us to assess adenovirus-mediated gene expression in LoVo and SMMC7721 tumor xenografts. We found that intravenous administration of Ad.BG (4 10⁹ pfu) resulted in -gal expression in LoVo and SMMC7721 tumors (2.300.32 and 2.510.25 mU/mg, respectively) similar to the expression in liver of both groups (2.920.28 and 3.010.30 mU/mg, respectively) (Fig 4a). However, in animals receiving Ad.IR-BG, -gal expression was only detectable in LoVo and SMMC7721 xenografts (1.630.22 and 1.750.25 mU/mg, respectively); whereas background expression levels were observed in liver from both cell types (<0.1 mU/mg) (Fig 4a). Moreover, -gal expression in other normal tissues in Ad.IR-BG-treated mice was significantly lower in liver, lung, and spleen compared to Ad.BG-treated mice (Fig 4b). This result, in combination with the similar finding in cell culture, demonstrated that Ad.IR-BG vector could mediate tumor-specific gene expression.

Figure 4.

Selective -gal expression in tumor xenografts by Ad.IR-BG infusion. Nude mice bearing intrahepatic LoVo or SMMC7721 tumor xenografts were given Ad.BG or Ad.IR-BG vector at 4 10⁹ PFU/mice through the tail vein. Tissues were harvested 72 hours later and were homogenized for -gal expression. The -gal activity is expressed as milliunits (mU) per 1 mg of total proteins used in the assays. (a) -gal activities in liver and tumor xenografts, (b) -gal activities in other normal tissues, n=4, ^*P < 0.05.

Full figure and legend (68K)

We then determined whether this tumor-specific gene expression was the result of viral DNA recombination in tumor cells. By PCR analysis using primers (Rec) that recognize viral DNA of recombinant form, we found that only a small and nonrecombinant fragment (0.45 kb) was found in all tissues, liver, and tumor xenografts (LoVo or SMCC7721), in Ad.BG-treated mice (Fig 5a). In contrast, Ad.IR-BG injection resulted in viral DNA recombination detected by a long PCR fragment (1.10 kb) exclusively in LoVo or SMCC7721 tumor xenografts but not in the liver (Fig 5b). Consistent with the finding in cell culture, Ad.IR-BG produced a tumor-specific gene expression through a viral DNA recombination mediated by IR sequence.

Figure 5.

Ad.IR-BG-mediated viral DNA recombination in tumor xenografts. Nude mice bearing LoVo and SMMC7721 xenografts were infused with Ad.BG or Ad.IR-BG vector at 4 10⁹ PFU/mice through the tail vein. DNAs were prepared from tissues of the liver (L) and the tumors (T) at 72 hours after viral infusion. Viral DNA recombination was determined by PCR analysis using specific primers (Rec) as described in Figure 2b with the primers for -actin as controls. The amplified viral DNA of recombinant form (1.10 kb) was indicated representatively on agarose gel from three tumor xenografts for Ad.BG infusions (a) or Ad.IR-BG infusions (b).

Full figure and legend (132K)

Discussion

In this study, we found that a replication-activated adenovirus vector containing the IR sequence (Ad.IR-BG) selectively transduced human colon carcinoma and hepatocellular carcinoma cells as compared to normal colon epithelial cells and hepatocytes. The selective -gal expression mediated by Ad.IR-BG infection was the result of a replication-dependent recombination between viral DNAs specifically in tumor cells. Administration of Ad.IR-BG vector produced selective -gal expression in tumor xenografts for both colon carcinoma metastasis in liver (LoVo) and primary hepatocellular carcinoma (SMMC7721).

The major developments in achieving selective high-activity gene expression in tumor cells in adenovirus gene therapy involve the use of conditionally replicating adenovirus vectors. One such vector is ONYX-015 that contains the E1a gene and uses viral genomic deletion for E1b 55k to interrupt the virus life cycle in the presence of cellular p53 (in normal cells) but not mutant p53 (found in many tumor cells).^7,8,9,10 Although ONYX-015 has shown potential in various laboratory and clinical studies, a lack of correlation between p53 status and viral replication has been reported.¹⁵ Other vectors use tumor-specific promoters for E1 gene expression restricted to tumor cells to facilitate viral replication.^16,17,18 However, these vectors have limited application due to the relative selectivity of promoter activity in tumors. Recent studies suggest that DNA replication for E1-deleted adenovirus is supported by many human cancer cells, but is deficient in primary normal cells including fibroblast and liver.^13,19 Moreover, E1-deleted adenovirus replicates selectively in tumor cells though at low efficiency. The modified vector used in this study contains the IR sequence that mediates predictable viral DNA rearrangement that leads to selective transgene expression in tumors depending on DNA replication.¹³ Our study suggests that the E1-deleted adenovirus containing the IR sequence is superior to Ad.E1-deleted adenovirus in tumor-selective gene expression.

The greater gene expression achievable by this strategy is only worthwhile if it is selectively activated in tumor cells. Notably, however, both vectors mediated viral DNA replication in rat hepatocyte WB cells, although viral DNA replication was absent in human fibroblasts, human mammary epithelium, normal human hepatocyte, and primary mouse hepatocytes (data not shown), which is consistent with previous studies.¹³ The WB cell is an immortalized rat liver hepatocyte cell line, a normal diploid epithelial cell with doubling time equal to that of the cancer cells used in this study. Although it lacks the transformed phenotypes represented in most cancer cells,²⁰ it is possible that dysregulated cell growth may provide extra factors to activate viral DNA replication. This may suggest that the insufficiency of viral DNA replication in normal cells is cell type-dependent. The cellular factors that complement E1 function in viral DNA replication need to be explored further. Hydroxyurea, an inhibitor of DNA replication, significantly reduced (at 5 mM) or completely blocked (at 10 mM) viral DNA replication, suggesting that DNA replication is required for viral DNA recombination. We also observed that although Ad.IR-BG infection mediated a low background level of viral DNA recombination in WB cells, the process is deficient in normal human hepatocyte and is more efficient in tumor cells. Thus, the mouse tumor xenograft model may provide us insight into adenovirus gene therapy for human disease. Importantly, however, the transgene expression mediated by Ad.IR-BG infection is highly selective in tumor cells (LoVo and SMMC7721) but defective in normal hepatocytes (WB), implying that other cellular mechanisms, in addition to selective viral DNA replication, could be involved in selective DNA recombination in tumor cells.

Although the replication-activated adenovirus has been shown to replicate in various human tumor cells,¹⁰ its oncolytic ability and the overall anticancer effect is probably not sufficient to be used alone. Our study indicated that systemic administration of Ad.IR-BG failed to delay tumor growth for both LoVo and SMMC7721 xenografts in nude mice (data not shown). Measurements of viral replication from tumors by real-time PCR analysis were similar to those from liver partly because of nonspecific viral targeting into liver at the time of injection or quick dissemination of progeny virus from tumor to liver (data not shown). Thus, efforts should be made to combine the use of the replication-activated adenovirus vector with other strategies in order to achieve a highly efficient and selective gene expression as well as anticancer effects. In particular, the combination of enzyme/prodrug with a replication-activated adenovirus vector would be anticipated to be a better approach for tumor treatment. Indeed, the delivery of a fusion gene of cytosine deaminase and herpes simplex thymidine kinase by the replication-competent adenovirus vector has already shown clinical promise in the treatment of recurrent prostate cancer.²¹

References

1. O'Connell MJ, Nagorney DM, Bernath AM, et al. Sequential intrahepatic fluorodeoxyuridine and systemic fluorouracil plus leucovorin for the treatment of metastatic colorectal cancer confined to the liver. J Clin Oncol. 1998;16:2528–2533. | PubMed | ISI | ChemPort |
2. Harmantas A, Rotstein LE, Langer B. Regional versus systemic chemotherapy in the treatment of colorectal carcinoma metastatic to the liver: is there a survival difference? Meta-analysis of the published literature. Cancer (Phila). 1996;78:1639. | Article | ChemPort |
3. Tang ZY. Hepatocellular carcinoma — cause, treatment and metastasis. World J Gastroenterol. 2001;7:445–454. | PubMed | ISI | ChemPort |
4. Grill J, Van Beusechem VW, Curiel DT, Gerritsen WR. Combined targeting of adenoviruses to integrins and epidermal growth factor receptors increases gene transfer into primary glioma cells and spheroids. Clin Cancer Res. 2001;7:641–650. | PubMed | ISI | ChemPort |
5. Zhang M, Li SP, Nyati MK, et al. Regional delivery and selective expression of a high-activity yeast cytosine deaminase in an intrahepatic colon cancer model. Cancer Res. 2003;63:658–663. | PubMed | ISI | ChemPort |
6. Nayti MK, Sreekumar SP, Li SP, et al. High and selective expression of yeast cytosine deaminase under a carcinoembryonic antigen promoter-enhancer. Cancer Res. 2002;62:2337–2342. | PubMed | ISI | ChemPort |
7. Bischoff JR, Kirn DH, Williams A, et al. An adenovirus mutant that replicates selectively in p53-deficient human tumor cells. Science. 1996;274:373–376. | Article | PubMed | ISI | ChemPort |
8. Johnson L, Shen A, Boyle L, et al. Selective replicating adenoviruses targeting deregulated E2F activity are potent, systemic antitumor agents. Cancer Cell. 2002;1:325–337. | Article | PubMed | ISI | ChemPort |
9. Khuri FR, Nemunaitis J, Ganly I, et al. A controlled trial of intratumoral ONYX-015, a selectively-replicating adenovirus, in combination with cisplatin and 5-fluorouracil in patients with recurrent head and neck cancer. Nat Med. 2000;6:879–885. | Article | PubMed | ISI | ChemPort |
10. Nemunaitis J, Cunningham C, Buchanan A, et al. Intravenous infusion of a replication-selective adenovirus (ONYX-015) in cancer patients: safety, feasibility and biological activity. Gene Therapy. 2001;8:746–759. | Article | PubMed | ISI | ChemPort |
11. Steinwaerder DS, Carlson CA, Lieber A. DNA replication of first-generation adenovirus vectors in tumor cells. Hum Gene Ther. 2000;11:1933–1948. | Article | PubMed | ISI | ChemPort |
12. Steinwaerder DS, Carlson CA, Lieber A. Generation of adenovirus vectors devoid of all viral genes by recombination between inverted repeats. J Virol. 1999;73:9303–9313. | PubMed | ISI | ChemPort |
13. Steinwaerder DS, Carlson CA, Otto DL, et al. Tumor-specific gene expression in hepatic metastases by a replication-activated adenovirus vector. Nat Med. 2001;7:240–243. | Article | PubMed | ISI | ChemPort |
14. Bernt KM, Steinwaerder DS, Ni SH, et al. Enzyme-activated prodrug therapy enhances tumor-specific replication of adenovirus vectors. Cancer Res. 2002;62:6089–6098. | PubMed | ISI | ChemPort |
15. Rothmann T, Hengstermann A, Whitaker NJ, et al. Replication of ONYX-015, a potential anticancer adenovirus, is independent of p53 status in tumor cells. J Virol. 1998;72:9470–9478. | PubMed | ISI | ChemPort |
16. Hallenbeck PL, Chang YN, Hay C, et al. A novel tumor-specific replication-restricted adenoviral vector for gene therapy for hepatocellular carcinoma. Hum Gene Ther. 1999;10:1721–1733. | Article | PubMed | ISI | ChemPort |
17. Li YH, Yu DC, Chen Y, et al. A hepatocellular carcinoma-specific adenovirus variant, CV890, eliminates distant human liver tumors in combination with doxorubicin. Cancer Res. 2001;61:6428–6436. | PubMed | ISI | ChemPort |
18. Matsubara S, Wada Y, Gardner TA, et al. A conditional replication-competent adenoviral vector, Ad-OC-E1a, to cotarget prostate cancer and bone stroma in an experimental model of androgen-independent prostate cancer bone metastasis. Cancer Res. 2001;61:6012–6019. | PubMed | ISI | ChemPort |
19. Nelson JE, Kay MA. Persistence of recombinant adenovirus in vivo is not dependent on vector DNA replication. J Virol. 1997;71:8902–8907. | PubMed | ISI | ChemPort |
20. Rae RS, Mehta PP, Chang CC, et al. Neoplastic phenotype of gap-junctional intercellular communication-deficient WB rat liver epithelial cells and its reversal by forced expression of connexin 32. Mol Carcinogen. 1998;22:120–127. | Article | ISI | ChemPort |
21. Freytag SO, Khil M, Stricker H, et al. Phase I study of replication-competent adenovirus-mediated double suicide gene therapy for the treatment of locally recurrent prostate cancer. Cancer Res. 2002;62:4968–4976. | PubMed | ISI | ChemPort |

Acknowledgements

We thank Mary Davis (Department of Radiation Oncology, University of Michigan) for reviewing this manuscript.