Targeting Adenovirus to the Serotype 3 Receptor Increases Gene Transfer Efficiency to Ovarian Cancer Cells¹

Epithelial ovarian cancer is the leading cause of gynecological cancer mortality in the United States with an estimated 23,400 new cases and 13,900 deaths in 2001. Most cases of ovarian cancer are diagnosed at an advanced stage, resulting in a 5-year overall survival of <30%, despite aggressive surgical debulking and chemotherapy (1). Thus, novel strategies for treatment of advanced stage ovarian cancer resistant to traditional therapeutic modalities are needed.

Gene therapy represents a promising treatment alternative that has recently displayed some clinical utility (2). In the context of ovarian cancer, Ads³ have shown promise in vitro and in animal models (3). In most cases Ad vectors have been based on serotype 5, because of its capability of mediating high levels of transgene expression, its ability to transduce dividing and nondividing cells, and its broad tissue tropism. However, the efficiency of Ad5 gene transfer may closely correlate with the cell surface density of its primary receptor, CAR (4, 5, 6). Unfortunately, the expression of CAR is highly variable, and often low, on ovarian and other primary cancer cells, resulting in relative resistance to Ad5 infection (7, 8, 9, 10, 11). On the basis of this concept, strategies to modify Ad tropism to circumvent CAR deficiency have used heterologous retargeting complexes or genetic capsid modifications (12). A relatively unexplored variation of the latter approach is the substitution of the knob domain of Ad5 with knobs from alternate Ad serotypes (13, 14, 15, 16). Specifically, Ad3 has a distinct, but unidentified receptor and, therefore, different tissue tropism (17, 18).

In this study, we have constructed an Ad containing a chimeric fiber with the knob domain of Ad3 in the Ad5 capsid (Ad5/3luc1), which redirects binding of the vector to the Ad3 receptor. We show that this genetically modified Ad vector significantly enhances the infectivity of human ovarian adenocarcinoma cell lines and primary human ovarian cancer cells. Also, our results suggest correlation between transgene transfer and receptor expression as measured by a novel knob binding assay. Genetic retargeting with the Ad3 knob may allow more efficient tumor cell transduction in the context of in vivo gene delivery and, thus, may offer the potential to improve Ad-based ovarian cancer gene therapy approaches.

Human ovarian adenocarcinoma cell lines OV-4, SKOV3.ip1, and Hey were obtained from Dr. Timothy J. Eberlein (Harvard Medical School, Boston, MA), and Dr. Janet Price and Dr. Judy Wolf (both from M. D. Anderson Cancer Center, Houston, TX), respectively. The human ovarian teratocarcinoma cell line PA-1 was obtained from the American Type Culture Collection (Manassas, VA). The 293 human transformed embryonal kidney cell line was purchased from Microbix (Toronto, Ontario, Canada). All cell lines were cultured at 37°C in media recommended by suppliers in a humidified atmosphere of 5% CO₂.

Fresh malignant ascites fluid samples from patients with pathologically confirmed ovarian adenocarcinoma were obtained from the University of Alabama at Birmingham Hospital. Cancer cells were purified by a previously described immunomagnetic-based method (19). Briefly, ovarian cancer cells were initially bound with a murine anti-TAG-72-antibody (CC-49) and then collected with magnetic beads coated with antimouse-IgG.

Two replication-incompetent Ad vectors containing a firefly luciferase transgene cassette in place of the deleted E1 region were used. Ad5luc1 was generated in our laboratory and described previously (20). The genome of Ad5/3luc1 (Ad containing chimeric fibers with the tail and shaft domains of Ad serotype 5 and the knob domain of serotype 3) was constructed by homologous DNA recombination in Escherichia coli using the previously described plasmids pNEB.PK.F5/3 (13) in a scheme described by Dmitriev et al. (21). The vector of interest was rescued by transfecting 293 cells with the resultant Ad genome. The viruses were propagated on 293 cells and purified on cesium chloride gradients. The VP concentration was determined at 260 nm, and a standard plaque assay on 293 cells was performed to determine infectious particles. The ratio of VP:infectious particles was 5.24 and 45.7 for Ad5luc1 and Ad5/3luc1, respectively.

Recombinant Ad5 and Ad3 fiber knob proteins with N-terminal 6xHis tags were expressed in E. coli using the pQE30 expression vector (Qiagen, Valencia, CA) and purified on nickel-nitrilotriacetic acid agarose columns (Qiagen) as recommended by the manufacturer and described elsewhere (13). The concentration of the purified proteins was determined by Bio-Rad DC protein assay (Bio-Rad, Hercules, CA). The ability of each knob protein to form a homotrimer was verified by Western blot of unboiled samples. The primary antibody used in detection was Penta-His antibody (Qiagen) and the secondary antibody was peroxidase-conjugated goat antimouse IgG (Sigma Chemical Co., St. Louis, MO).

To investigate the ability of recombinant Ad5 and Ad3 knobs to block infection by the Ad of the corresponding serotype, infection with Ad5luc1 and Ad5/3luc1 was performed in the presence of the purified knob proteins. Monolayers of SKOV3.ip1 cells in 24-well plates were preincubated with increasing concentrations of Ad5 or Ad3 knob in 100 μl of DMEM:F12 with 2% FBS for 10 min at room temperature. Ad5luc1 or Ad5/3luc1 was added at 5000 or 200 VP/cell, respectively, diluted in 100 μl of DMEM:F12 with 2% FBS, followed by a 30-min incubation at room temperature. The cells were then washed once with DMEM:F12 containing 2% FBS, and complete medium was added. After 24 h of incubation at 37°C, the cells were lysed and a luciferase assay was performed with the Luciferase Assay System (Promega, Madison, WI). The protein concentration of the cell lysates was determined as above to allow normalization of the gene expression data for the number of cells. Background luciferase activities were subtracted from the readings.

Cells grown in T75 flasks were washed with PBS, harvested by incubating with 0.53 mm EDTA in PBS, and resuspended in PBS containing 1% BSA. Cells (2 × 10⁵) were incubated with 20 ng of either Ad5 or Ad3 recombinant knob protein in 200 μl of PBS-BSA, or with buffer only, for 1 h at 4°C. Cells were washed twice with 4 ml of PBS-BSA and incubated with 300 μl of a 1:125 dilution of Tetra-His antibody (Qiagen) for 1 h at 4°C. The cells were washed once with 4 ml of PBS-BSA and incubated with 300 μl of a 1:100 dilution of the secondary FITC-labeled goat antimouse IgG (Sigma Chemical Co.) for 1 h at 4°C. After the cells were washed as described above, 2.6 μg/ml propidium iodide (Sigma Chemical Co.) was added to sort out dead cells from the sample; then, 2 × 10⁴ cells (OV-4) or 10⁴ cells (other cell lines) were analyzed immediately by flow cytometry at the University of Alabama at Birmingham FACS Core Facility. The FITC-positive (live) cell population for each cell line was determined by gating cells incubated with buffer only (negative control) at 1%.

Cells in 24-well plates were infected for 30 min at room temperature at 50, 200, and 1000 VP/cell by adding Ad5luc1 or Ad5/3luc1 diluted in 200 μl of DMEM:F12 with 2% FBS. Cells were washed once with DMEM:F12 containing 2% FBS, and complete medium was added. The luciferase assay was performed 24 h postinfection as described above.

Due to variable expression of CAR on human primary cancer cells, the utility of Ad5 as a cancer gene therapy vector could be compromised. Fortunately, native Ad5 tropism can be modified to circumvent CAR deficiency and to enhance infectivity. One approach is retargeting Ads by creating chimeric fibers possessing knob domains of alternate serotypes. To this end, we constructed nonreplicating Ads with either the native Ad5 fiber protein (Ad5luc1) or a chimeric fiber with the knob from Ad3 fiber (Ad5/3luc1). To evaluate the receptor binding properties of Ad5/3luc1 in comparison with Ad5luc1, infections of SKOV3.ip1 cells were performed in the presence of purified, trimeric recombinant Ad5 and Ad3 knob proteins (Fig. 1). Transgene expression mediated by Ad5luc1 decreased with increasing concentrations of Ad5 knob in a dose-dependent manner. In contrast, the purified Ad3 knob exhibited only limited ability to block the infection with Ad5luc1 (Fig. 1,A). Similarly, Ad3 knob blocked Ad5/3luc1 infection, whereas Ad5 knob had only minimal effect (Fig. 1 B). The data obtained in this competition experiment demonstrate that the relevant recombinant knob protein is able to block the infection of SKOV3.ip1 cells with either Ad5luc1 or Ad5/3luc1 in a dose-dependent manner, whereas gene transfer is only minimally inhibited by high concentrations of the irrelevant knob. Therefore, our results support the existence of a distinct receptor for Ad3, as suggested previously (13, 14, 17, 18). More importantly, these results confirm that Ad5/3luc1 is retargeted to the Ad3 receptor.

We sought to establish that the level of Ad3 receptor would predict sensitivity of cells to the chimeric vector. Because the Ad3 receptor is not yet identified, we developed a novel knob binding assay to quantify the cell surface expression of CAR and the Ad3 receptor on human ovarian cancer cell lines (Fig. 2,A). For this analysis cells were incubated with recombinant, 6xHis-tagged Ad5 knob or Ad3 knob, followed by detection with anti-His antibody and FITC-conjugated antibody. PA-1 cells have been shown to express high levels of CAR (10). In contrast, OV-4 and SKOV3.ip1 display moderate or low levels of CAR (10, 22). The 293 cells were included as a CAR-positive control (21). This knob binding assay suggested that 293 cells express larger amount of CAR than Ad3 receptor (Fig. 2 B). Also, PA-1 displayed more CAR than the Ad3 receptor on the cell surface. In contrast, human ovarian adenocarcinoma cell lines OV-4 and SKOV3.ip1 expressed more Ad3 receptor than CAR. These results suggest higher expression of the Ad3 receptor relative to CAR on human ovarian adenocarcinoma lines (PA-1 is a teratocarcinoma).

We hypothesized that differential expression of CAR and Ad3 receptor could correlate with infectivity by serotype 5 or chimeric virus. Infection at 1000 VP/cell of Ad5/3luc1 resulted in 1.12-, 291-, 125-, 2.60-, and 116-fold higher luciferase expression in 293, OV-4, SKOV3.ip1, PA-1, and Hey cells, respectively, in comparison with Ad5luc1 (Fig. 3). The relative differences were very similar to the other amounts of virus. Non-adenocarcinoma cell lines expressing similar levels of CAR and Ad3 receptor in the knob binding assay (293 and PA-1) demonstrated comparable results in the gene transfer assay. In contrast, transgene expression in adenocarcinoma cell lines (OV-4, SKOV3.ip1, and Hey) was 2 orders of magnitude higher with Ad5/3luc1. Importantly, the receptor density, as estimated by the novel knob binding assay, correlated with reporter gene expression. The actual transgene expression in 293 cells was higher, perhaps resulting from E1 trans-complementation and viral replication. Thus, the amount of binding to primary receptor appears to be the major factor determining the efficiency of Ad-based gene delivery to target cells. Nevertheless, there are other receptors that could be sufficient to mediate the initial binding of Ad5 (23). Furthermore, expression of αv integrins may affect the infectivity of cells by Ad.

Human trials have suggested a discrepancy between cell line and clinical gene transfer efficiency (24, 25). To more closely model the clinical situation with the most stringent available substrate, gene transfer experiments were performed using unpassaged human primary ovarian adenocarcinoma cells, purified from malignant ascites fluid. Recent studies have shown that most primary ovarian cancer cells display moderate to low levels of CAR (10). In our experiments, ovarian cancer primary cells obtained from four patients demonstrated 5.18-, 16.5-, 5.63-, and 12.5-fold higher transgene expression when infected at 1000 VP/cell of Ad5/3luc1 in comparison with Ad5luc1 (Fig. 4 A–D, respectively). Infection at 50 and 200 VP/cell gave similar results. For some samples, the findings were also corroborated at 5000 VP/cell.⁴ Thus, an augmentation of luciferase activity was observed with the modified vector, but to a smaller extent than with ovarian adenocarcinoma cell lines. These results underline the necessity of analyzing primary tumor material in addition to established cell lines.

As with ovarian cancer, variable expression of CAR is documented in many other cancer types such as glioma, melanoma, bladder cancer, and rhabdomyosarcoma (7, 8, 9, 10, 11, 26). It is known that, for entry, viruses often exploit cellular receptors important in conserved pathways (27). Interestingly, previous studies suggest that CAR may act as a tumor suppressor, which could be linked to the frequent down-regulation seen in highly tumorigenic cells (28). Our results suggest that expression of CAR versus the Ad3 receptor is different on human ovarian cancer cells, and the density of Ad3 receptor is often higher. Although the receptor and its function are unknown, it is not impossible that the expression of the Ad3 receptor could be unrelated to the carcinogenic process and therefore unaffected by malignant progression. Thus, Ad3 receptor-mediated gene transfer could be advantageous in the context of advanced cancer.

In conclusion, we describe a chimeric Ad5/3luc1 incorporating the Ad3 knob in the Ad5 fiber. True genetic retargeting, as described here, could give an advantage in comparison with enhanced infectivity, i.e., viruses that continue to bind CAR despite tropism modification. Exploiting the different tropism of Ad3 led to enhanced infectivity of ovarian cancer cell lines and primary cells. Furthermore, we used a novel knob binding assay to investigate receptor concentration and found high expression of the Ad3 receptor on ovarian adenocarcinoma cells.

We thank Drs. Joanne T. Douglas, Dirk M. Nettelbeck, Natalya Belousova, and Yuji Heike for helpful discussions and technical advice.