Substitution of the Adenovirus Serotype 5 Knob with a Serotype 3 Knob Enhances Multiple Steps in Virus Replication¹

CRAd³ vectors have emerged as novel therapeutic agents for a variety of neoplastic diseases, which has led to their rapid translation into human clinical trials (1, 2, 3). From a conceptual standpoint, a CRAd agent’s antitumor effect is measured by the ability to productively infect and selectively replicate in target tumor cells and ultimately induce oncolysis. This process involves (a) efficient target cell infection, (b) tumor-specific replication, and (c) effective dispersion of the viral progeny. The degree that each of these steps is optimized determines the efficacy of the virotherapy. For example, it is generally considered that the binding of the virus to the target cell defines the efficiency of the infection (4, 5, 6, 7, 8). Also, a greater degree of tumor-specific replication correlates to a higher therapeutic index. Likewise, the capacity of the vector to efficiently disperse enhances the pervasiveness of the infection (9, 10, 11, 12). Deficiencies at these key steps likely explain the inefficacy of CRAds in human clinical applications to date. Furthermore, correction of these deficiencies will need to be considered in the design and utilization of new CRAds.

Although tumor-specific replication, efficiency of infection, and dispersion all represent critical determinants of therapeutic effect, CRAd development has primarily focused on improving tumor- specific replication. Many tumor cell types have been reported to express relatively low levels of the primary Ad5 receptor CAR, which essentially renders the tumor cell resistant to Ad5 infection (13, 14, 15, 16, 17). Clearly, in the context of CRAd agents, this CAR deficiency would adversely impact both the initial infection and the dispersion steps. The recognition that CAR deficiency is a limiting factor in Ad use has led to the development of strategies to alter Ad5 tropism to achieve CAR-independent gene delivery. In this regard, CAR-independent Ad5 vectors have been developed that enter cells through a variety of nonnative pathways via the epidermal growth factor, fibroblast growth factor, and folate receptors and integrins (14, 18, 19, 20, 21). These modifications in Ad5 tropism have also resulted in infection efficiency gains that directly translate into improved anticancer gene therapy effects.

In this study, we investigated a CAR-independent gene transfer approach to improve the potency of an Ad5-based CRAd. To accomplish this, we used a different Ad serotype because other Ad serotype receptors have been proposed as targets for tumor cell infection (22, 23, 24, 25, 26, 27, 28, 29, 30). For example, the Ad3 receptor, which is distinct from CAR, has been reported to be a better target for infection of ovarian cancer, SCCHN, and B cell lymphomas (8, 22, 24). Therefore, we evaluated the use of a chimeric Ad5 vector that has Ad3 tropism (Ad5/3) that was generated by substitution of the Ad3 knob domain into the Ad5 fiber protein. We hypothesized that the chimeric Ad5/3 vector would efficiently infect, replicate, and lyse SCCHN tumor cells that are poorly infected by Ad5. To test this, key steps in the replication cycle were compared between the chimeric Ad5/3 vector and its unmodified Ad5 counterpart. At each step of the virus life cycle, from entry to oncolysis, the chimeric Ad5/3 vector was progressively more efficient. Importantly, the findings of this study suggest that modifications the Ad fiber protein can affect multiple steps in Ad replication. As well, we additionally demonstrate that alternative Ad receptors are useful targets for cancer gene therapy and that the efficacy of CRAd gene therapy may be improved by developing CAR-independent approaches.

The human SCCHN cell lines, SCC-25, FaDu, and SCC-4, were purchased from the American Type Culture Collection (Manassas, VA). FaDu and SCC-25 cells were cultured in Eagle’s MEM with Earle’s Balanced Salt Solution containing 1% l-glutamine, 1 mm nonessential amino acids, and 10% FCS (Life Technologies, Inc., Gaithersburg, MD). SCC-4 cells are cultured in DMEM:Ham’s F12 containing 1% l-glutamine, 1 mm sodium pyruvate, 10 mm glucose, 0.2 mg/ml hydrocortisone, and 10% FCS. The HEK293 cells were purchased from Microbix Biosystems, Inc. (Toronto, Ontario, Canada), and cultured in MEM with Earle’s Balanced Salt Solution containing 1% l-glutamine and 10% FCS. All cell lines were maintained at 37°C in 5% CO₂ atmosphere.

Ad5Luc and Ad5/3Luc were generated at the Division of Human Gene Therapy (UAB). Ad5Luc.RGD, which contains an integrin binding motif (RGD) inserted in the HI loop of the fiber knob region, has been described previously (15, 21). These three viruses are E1⁻ replication-defective Ad vectors with cytomegalovirus promoter-driven luciferase reporter genes inserted in the E1 region. Ad5Luc3 and Ad5/3Luc3, previously described by Mittal et al. (31) and Krasnykh et al. (32), respectively, are E1⁺ replication-competent vectors with the luciferase reporter gene inserted into the E3 region under the control of the simian virus 40 promoter. Both Ad5/3Luc and Ad5/3Luc3 have modified fiber genes that contain an Ad5 shaft region and Ad3 knob region (nts 647-1208 of accession no. X01998 M12411). Viruses were propagated and titered (33) to determine the number of pfu on 293 cells and purified by double centrifugation on cesium chloride gradients as described elsewhere (34). The physical titers or total vp were determined spectrophotometrically by measuring the absorbance at 260 nm where 1 absorbance unit is equivalent to 1.1 × 10¹² vp (35). The vp:pfu ratios for the Ad vectors used in the study were the following: Ad5Luc, 20; Ad5/3Luc, 41; Ad5Luc.RGD, 48; Ad5Luc3, 21; and Ad5/3Luc3, 22.

To metabolically label Ad particles with a radioactive tag, 5× T175 flasks of HEK293 cells (∼90% confluent) were infected with either Ad5Luc3 or Ad5/3Luc3 (20 vp/cell). Twenty-four h later, 0.1 mCi of deoxy-[1′,2′,2,8-³H]-adenosine 5′-triphosphate (9.25 MBq; Amersham Pharmacia Biotech) was added to each flask. The infection was continued until extensive cytopathic effect was observed (∼42 h later). Infected cells were then harvested and virus was purified as described above. The radioactivity was determined by scintillation counting and physical titer was determined spectrophometerically. The specific activity of Ad5Luc3 (6.4 × 10¹¹ vp/ml) and Ad5/3Luc3 (4.2 × 10¹¹ vp/ml) was 6.5 × 10⁻⁵ cpm/vp and 7.9 × 10⁻⁵ cpm/vp, respectively, using a 60% counting efficiency estimation for tritium.

Female athymic nu/nu mice were purchased from The Jackson Laboratory at 4–6 weeks of age. They were housed in the UAB Animal Facility in cages fitted with microisolator tops and allowed sterile food and water ad libitum. Animals were treated humanely using approved procedures in accordance with the guidelines of the UAB Institutional Animal Care and Use Committee polices and federal guidelines mandated by United States Department of Agriculture Animal Welfare Regulations.

The [³H]-labeled viruses (above) were used for the binding assays. SCCHN cells were incubated with equivalent amounts of virus (6.5–7.9 × 10⁴ cpm/10⁵ cells) for 2 h at 4°C. Unbound virus was removed by washing the cell monolayers twice with cold media. The cells and [³H]-labeled viruses were then harvested by trypsonization. Radioactivity was measured by scintillation counting to determine the percentage of bound virus (bound ³H counts divided by total input ³H counts × 100) as described above.

To quantify Ad genome translocation to the host cell nucleus, 5 × 10⁶ SCCHN cells were mock-infected or infected with the [³H]-labeled viruses (100 vp/cell) in 5 ml/flask of growth media containing 2% FBS for 2 h at 4°C and then transferred to a 37°C incubator. Six h later, the cells were harvested by trypsonization. Nuclear extracts were then prepared by resuspending the cell pellet in 1 ml of hypotonic buffer containing 10 mm Tris-Cl (pH 7.9), 10 mm KCl, 0.1 mm EDTA, and 0.1 mm EGTA on ice for 15 min to allow the cells to swell. The cells were then lysed by 20 strokes in a Dounce homogenizer. The cell lysate was centrifuged for 8 min at 3000 × g at 4°C to collect the nuclear pellet. The nuclear pellets were solubilized in scintillation fluid (UniverSol ES; ICN, Costa Mesa, CA) and radioactivity was measured as described above.

To quantify Ad transgene expression, 10⁵ SCCHN cells were plated into 4 replicate wells of 12-well plates in the presence of 2 ml of culture media and were allowed to adhere overnight. The cells were then infected with either Ad5Luc, Ad5/3Luc, or Ad5Luc.RGD (50 vp/cell) in 300 μl/well of growth media containing 2% FBS for 2 h at room temperature. The cells were washed twice with growth media and then incubated with fresh growth media at 37°C in 5% CO₂. Thirty-six h after infection, the cells were rinsed with PBS and assayed for luciferase expression. For all luciferase enzyme assays, the cells were lysed in 200 μl of Promega (Madison, WI) lysis buffer. Ten μl of each sample were subsequently mixed with 50 μl of Promega luciferase assay reagent according to the manufacturer’s instructions and duplicate determinations of replicate samples were assayed in a Berthold luminometer.

SCCHN cells were infected with Ad5Luc3 or Ad5/3Luc3 (10 vp/cell) as described above. At 0, 3, 6, 12, and 24 h postinfection, total RNA was isolated using the RNeasy Mini kit (Qiagen). Before RT-PCR, all RNA samples were treated with DNase I for 15 min, 37°C. DNase I was then inactivated by heating at 72°C for 15. All subsequent RT-PCR assays were performed using the OneStep RT-PCR kit (Qiagen). For the nonquantitative analyses, E1A primers (5′-ACGGTTGCAGGTCTTGTCATTATCA-3′ and 5′-AAGCAAGTCCTCGATACATTCCA-3′) and GAPDH gene primers (5′-TCCCATCACCATCTTCCA-3′ and 5′-ACCTTCTACCACTACCCT-3′) were used. Because this E1A primer pair flanks an intron it can be used to distinguish between PCR products generated from E1A mRNA templates (351 bp) and E1A DNA templates (478 bp). In all cases, DNase I pretreatment of the total RNA was optimized to eliminate contaminating E1A DNA. The RT-PCR thermal cycles were as follows: 50°C (30 min) and 94°C (15 min) for 1 cycle; 94°C (1 min), 55°C (0.5 min), and 72°C (1 min) for 25 cycles; and 72°C (10 min) for 1 cycle.

For quantitative RT-PCR, total RNA was isolated and treated with DNase I as described above. The forward primer (5′-AACCAGTTGCCGTGAGAGTTG-3′), reverse primer (5′-CTCGTTAAGCAAGTCCTCGATACAT-3′) and 6-carboxyfluorescein labeled probe (5′-CACAGCCTGGCGACGCCCA-TAM RA) were used to amplify the E1A mRNA and the forward primer (5′-GGTTTACATGTTCCAATATGATTCCA-3′), reverse primer (5′-ATGGGATTTCCATTGATGACAAG-3′), and 6-carboxyfluorescein-labeled probe (5′-CGTTCTCAGCCTTGACGGTGCCA-3′) was used to amplify GAPDH. All primers and probes were designed using the Primer Express 1.0 software (Perkin-Elmer, Foster City, CA) following the recommendations of the manufacturer. RT-PCR was then performed in a mixture with a final volume of 9 μl/reaction containing 1× TaqMan EZ buffer, 3 mm of manganese acetate, 300 μm dATP, dCTP, dGTP, 600 μm of dUTP, 100 nm of forward and reverse primer, 100 nm probe, 0.1 units/μl of rTth DNA polymerase, 0.01 units/μl of AmpErase UNG, 0.025% BSA, and RNase-free water. For each experiment, a known amount of E1A template RNA (10⁸, 10⁶, 10⁴, and 10² copies/μl) was used as a standard cure to quantify the E1A copy numbers of the experimental samples. A known amount of total RNA (25, 5, 1, and 0.1 ng/μl) was used to generate a standard curve to quantify total RNA based on GAPDH copies. To generate a standard curve for determining E1A copy numbers, known amounts of E1A cRNA were used after in vitro transcription (Ambion MAXIscript In Vitro Transcription Kit, Ambion, Austin, TX). A sample of the total RNA (1.0 μl) from each experimental sample or the standard curve samples (total RNA and E1A cRNA) was added to the 9-μl RT-PCR mixture in each reaction capillary tube. Three no-template controls received 10 μl of reaction mixture with 1 μl of water. All capillaries were then sealed, mixed, and subjected to TaqMan Real-Time PCR in a LightCycler System (Roche Molecular Biochemicals, Indianapolis, IN). Thermal cycling conditions were as follows: 2 min at 50°C; 30 min at 60°C; 5 min at 95°C; and 40 cycles of 20 s at 94°C and 1 min at 62°C. Data were analyzed with LightCycler software.

SCCHN cells were infected with either Ad5Luc3 or Ad5/3Luc3 at 10 vp/cell. After the 2-h adsorption period at 4°C, the cell monolayers were washed twice with fresh growth media and then cells were incubated in 1 ml of fresh growth media at 37°C in a CO₂ incubator. Forty-eight h after infection, the cells were harvested by scraping into the growth media. The cells were subjected to four freeze-thaw cycles and then centrifuged at 10,000 × g for 15 min at room temperature. The amounts of virus present in the resulting supernatants were then analyzed by a plaque assays on 293 cells as described previously (33).

SCCHN cells were either mock infected or infected with different amounts of Ad5Luc3 or Ad5/3Luc3. Twelve h later, the cells were trypsinized, counted, and plated (5000 cells/well) into 12-well culture plates in four replicates. Five days after infection, the number of viable cells was determined using the CellTiter 96 AQueous NonRadioactive Cell Proliferation Assay kit (Promega). This colorimetric assay measures the conversion of a tetrazolium compound to formazan by viable cells. On the assay day, media from each well were aspirated, replaced with 400 μl of fresh growth media containing 80 μl of the cell proliferation assay reagent, and incubated at 37°C for ∼1 h. Four h before the start of the assay, known numbers of uninfected cells were plated in triplicate wells to generate a standard curve for determining the number of viable cells in the experimental samples. These cells received the cell proliferation assay reagent at the same time as the experimental cells. A 100-μl aliquot of the media from each well was transferred to a 96-well microtiter plate. and the absorbance at 490 nm was measured in a plate reader (Molecular Devices, Menlo Park, CA). Data collected by the plate reader was analyzed by the SOFTmax software package (Emax Molecular Devices, Menlo, CA), and the number of viable cells was determined. In other experiments, the cells were infected as described (above) and then stained with crystal violet 3 days after infection.

FaDu cells were mock infected or infected (100 vp/cell) with either Ad5Luc3 or Ad5/3Luc3 as described above. Next, a mixture of uninfected (97.5%) and either mock- or virus-infected (2.5%) FaDu cells (5 × 10⁶ total cells) were then injected s.c. into the flank regions of athymic nu/nu mice. Beginning 10 days later, tumor volumes (width × length × height) were recorded every day for the remainder of the experimental period.

Gene therapy has been hindered in part by the relative resistance of a number of different tumor types to Ad5 infection. This problem is especially important in the case of CRAd-based gene therapies, because the efficiency of virus replication directly correlates with the efficacy of the virotherapy. We have previously shown that Ad5 resistance can be overcome by rerouting entry of the vector to nonnative receptors such as epidermal growth factor receptor, fibroblast growth factor, folate, CD40, and integrins (7, 13, 14). In this study, we evaluated the use of Ad3 receptor targeting for increasing the efficiency of virus replication. Chimeric Ad5 vectors containing the knob region of Ad3 were compared with unmodified Ad5 vectors at several key steps of replication in a productive Ad infection, including binding, nuclear translocation, E1A transcription, translation, de novo virus production, and oncolysis.

The first step in Ad replication is the interaction of the virus with the host cell. In the case of Ad5, the knob region of the virus fiber protein binds to the Ad5 receptor CAR (36, 37). Other Ad serotypes such as Ad3 use different cell receptors for attachment and entry (38). This step is thought to largely determine the efficiency of the infection. We therefore used an in vitro assay to compare the binding activities of the Ad5 and chimeric Ad5/3 viruses to SCCHN cells. The genomes of each virus were metabolically labeled with tritiated dATP and then the viruses were purified by standard methods (34). Equivalent amounts of each virus were then incubated with SCCHN cells for 2 h at 4°C. Under these conditions, the virus binds to the cells but does not enter (39, 40). The amounts of bound and unbound virus were then measured by scintillation counting. For each cell line tested, the chimeric Ad5/3 virus had higher binding activity (Fig. 1 A). On average for the three cell lines tested, the chimeric Ad5/3 virus bound 4-fold better (from 2.9- to 5.4-fold) than the Ad5 virus. No difference in bound tritium counts was observed when the virus stocks were pretreated with DNase I, indicating that the tritium signal was attributable to intact virus particles rather than from contaminating-free viral DNA (data not shown). These data suggest that (a) the number of Ad3 receptors is higher on SCCHN cells compared with CAR and/or (b) that the Ad3 receptor-Ad3 knob interaction has a higher affinity compared with the CAR-Ad5 knob interaction.

After binding and internalization, disassembly of the capsid proteins allows the virus to escape from the endosome and enter the cytoplasm. The virus particles are then transported to the nuclear pore complex where the viral DNA is released into the nucleus to initiate viral gene expression (41). Therefore, we next compared the numbers of Ad5 and Ad5/3 viral genomes that reached the host nucleus shortly after infection (Fig. 1 B). In each SCCHN cell line tested, translocation to the nucleus was more efficient with the chimeric Ad5/3 virus. On average for the three cell lines tested, the chimeric Ad5/3 virus reached the host nucleus 16-fold more efficiently (from 7.9- to 27.6-fold) than the Ad5 virus. These results indicate that the chimeric Ad5/3 vector more rapidly reaches the nucleus of the infected host cell.

Once the virus reaches the nucleus, early gene transcription begins. The E1A transcript is produced first, and its gene product induces expression of the other early proteins encoded by the E1B, E2, E3, and E4 genes. Because the binding activities and rate of nuclear translocation of Ad5 and Ad5/3 differed, we reasoned that early gene expression would be different as well. Therefore, E1A transcription from the Ad5 and Ad5/3 vectors was compared at various time points after infection (0, 3, 6, 12, and 24 h). Total RNA was extracted and analyzed by RT-PCR using E1A-specific primer pairs (Fig. 2,A). E1A mRNA was detectable at the 12-h time point after infection with the Ad5 virus (Fig. 2,A, top panel) and was detectable at the 6-h time point after infection with the chimeric Ad5/3 virus (Fig. 2,A, bottom panel), indicating that infection by the chimeric Ad5/3 virus resulted in more E1A transcription products at earlier time points. Real-time TaqMan RT-PCR was also performed to quantitatively determine the amounts of E1A mRNA (Fig. 2,B). At each time point tested, higher copy numbers of E1A mRNA were detected after infection by the chimeric Ad5/3 vector. On average, the chimeric Ad5/3 virus produced 12-fold more E1A mRNA during the first 24 h of infection. However, the accumulation rate of E1A mRNA was higher for the Ad5 vector during the 3–6-h and 6–12-h intervals (Fig. 2 B, inset). During the 12–24-h interval, the accumulation rate of E1A mRNA was the same for both viruses. Overall, during the 3–24-h interval, E1A mRNA copy numbers increased by 2011- and 1372-fold after infection by the Ad5 and Ad5/3 viruses, respectively. These data demonstrate that infection by the chimeric Ad5/3 virus leads to higher total amounts of E1A transcripts; however, the rate of E1A mRNA transcription was somewhat higher for the Ad5 virus, suggesting slightly different kinetics of early gene transcription.

Next we evaluated the level of reporter transgene expression after infection by Ad5Luc and Ad5/3Luc vectors. In addition, we have previously shown that another genetically modified vector (Ad5Luc.RGD), which contains an RGD motif inserted into the HI loop region of the Ad5 knob, more efficiently infects SCCHN cells (13). Therefore, this vector was also used in the analysis of reporter transgene expression. We knew from previous experiments that replication-competent Ad systems produce higher levels of reporter gene products compared with replication-defective systems because of the amplification of genome numbers (42). In this experiment, it was important to uncouple early gene translation (i.e., luciferase) from genome replication, therefore this group of nonreplicative Ad vectors was used. SCCHN cells were infected with equal amounts of each of the Ad vectors, and transgene activity was measured 36 h later (Fig. 3). Consistent with our previous studies, reporter gene activity after infection by Ad5Luc.RGD virus was higher than Ad5 in each of the SCCHN cell lines tested. Strikingly, reporter gene activity after infection by the chimeric Ad5/3 virus was even greater than either Ad5Luc or Ad5Luc.RGD. On average for the three cell lines tested, the chimeric Ad5/3 virus demonstrated 53-, 131-, and 33-fold higher reporter gene levels than Ad5Luc at 50, 500, and 5000 vp/cell, respectively. These data suggest that the chimeric Ad5/3 virus is capable of higher protein expression that further augments its efficiency. In addition, the data demonstrates that at least two nonnative receptor pathways can be exploited to improve the efficiency of Ad infection in this cancer cell type.

Viral assembly begins in the cell nucleus at ∼24 h after infection (43) and continues for 2–3 days until the cells lyse and release the new virions. The differences in E1A transcription (Fig. 2) and transgene expression (Fig. 3) between the two viruses predicts more efficient production of progeny virus by the chimeric Ad5/3 vector. To test this possible outcome, SCCHN cells were infected with equal amounts (10 vp/cell) of either the Ad5 or chimeric Ad5/3 viruses. Forty-eight h after infection, the amount of de novo virus produced was determined by plaque assays (functional viral particles) on 293 cells and spectrophotometric analyses (total viral particles). The FaDu, SCC25, and SCC4 cells produced 403-, 281-, and 161-fold, respectively, more pfu of the chimeric Ad5/3 virus compared with the Ad5 virus (Fig. 4,A). One possible explanation for this result was that the HEK293 cells that were used to determine the virus titers have a disproportionate amount of the Ad3 receptor relative to the Ad5 receptor, which would have artificially skewed these results in favor of the chimeric Ad5/3 vector. To test this possibility, HEK293 cells were infected with equal numbers of vector particles (10 vp/cell), and the resulting numbers of pfu and vp were determined (Fig. 4,B). The HEK293 cells produced approximately equal amounts of functional and total particles of each virus, validating the results with the SCCHN cells (Fig. 4 A). The vp-to-pfu ratio, which is an index of the quality of the virus preparation, was 25 and 21 for Ad5Luc3 and Ad5/3Luc3, respectively, indicating that there was no difference in the capacity of the HEK293 cells to produce either virus. These data suggest that, in addition to binding and early gene transcription and translation, the chimeric Ad5/3 virus is more efficient at de novo virus production.

The use of oncolytic Ad vectors of different serotypes for cancer gene therapy was first investigated nearly a half-century ago (44). These wild-type Ads were injected intralesionally, and responses were determined by the liquefaction and ulceration of the inoculated tumors. More recently, oncolytic CRAd vectors have been evaluated in human clinical trials (1, 2, 3). We therefore compared the Ad5 and Ad5/3 vectors as oncolytic agents for SCCHN gene therapy. SCCHN cells were infected with each virus, and ∼3 days later, the monolayers were stained with crystal violet to assess oncolysis (Fig. 5,A). For each cell line tested, more oncolysis was observed after infection by the chimeric Ad5/3 virus. To quantitatively evaluate oncolysis by the two viruses, SCCHN cells were infected with escalating amounts of each virus (0–1000 vp/cell), and then 5 days later, cell viability assays were performed (Fig. 5 B). The efficiencies of the two viruses were compared by estimating the amount of virus inoculum needed to achieve 40% oncolysis in each of the SCCHN cell lines. On average, 127-fold less of the chimeric Ad5/3 virus was needed to achieve an equivalent amount of oncolysis by the Ad5 virus. Collectively, these data demonstrate that the chimeric Ad5/3 virus is capable of highly efficient oncolytic replication in tumor cells that are relatively refractory to Ad5 infection.

We next determined if the in vitro results translated to an in vivo tumor rejection model. FaDu cells were mixed with mock-, Ad5/3Luc3-, or Ad5Luc3-infected FaDu cells at a ratio of 97.5-to-2.5, respectively, and then transplanted s.c. into athymic nu/nu mice (Fig. 6,A). In the mock-infected group, the tumor nodules grew to an average volume of 1800 mm³ by day 23 of the experiment, at which time, the mice had to be sacrificed because of tumor burden. Tumors in the Ad5/3Luc3-infected group reached an average maximum volume of 170 mm³ on days 4–6 and then became smaller for the remainder of the experimental duration, reaching an average volume of 78.4 mm³ on day 27. Five of 12 tumor nodules (42%) in this group completely regressed by day 27, and these animals remained tumor free for an additional 30 days. The pattern of tumor growth in the Ad5Luc3-infected group mirrored that of the Ad5/3Luc3-infected group for the first 16 days, after which the tumors resumed growth and obtained an average volume of 360 mm³ on day 27. Unlike the Ad5/3Luc3-infected group, none of the tumor nodules in this group completely regressed. Fig. 6 B shows representative mice in each experimental group on day 23. These data show that the higher replication efficiency of the chimeric Ad5/3 vector translates to an in vivo tumor model resulting in a substantially improved therapeutic effect.

The modification of vector tropism is an important issue for Ad development. Both untargeting of the vector from the native receptor and retargeting of the vector to heterologous receptors have been critical advancements for improving the use of Ad-based gene therapy. These developments were necessary to achieve targeting of specific cells and tissues, to increase specificity, and to increase the efficiency of infection. Immunological, chemical, and genetic modifications have each been applied to alter vector tropism. In addition, the native tropism has been altered by swapping the knob regions of the fiber proteins of different Ad serotypes. The resulting chimeric vectors have been shown to have different neutralization profiles (45) and efficiencies (8, 22, 24, 25, 26, 32, 46) for infecting various cell types. These studies typically use replication-incompetent Ad vectors with quantifiable reporter genes (e.g., luciferase, green fluorescent protein) for determining infection efficiency. To date, however, the use of chimeric Ad vectors as oncolytic agents has not been rigorously evaluated.

Evidence that targeting Ad serotype receptors other than CAR may be of use was presented in previous studies that suggested the Ad3 receptor is expressed at high levels in EBV-infected B-lymphoid and head and neck squamous cell carcinoma cell lines (22, 24, 26, 38). More recently, using a modified flow cytometry method, Kanerva et al. (8) showed directly that the Ad3 receptor is expressed at higher levels than the Ad5 receptor in ovarian cancer cells. Importantly, these studies demonstrate that the Ad3 receptor can be exploited to overcome inefficient infection via the Ad5 receptor. On the basis of these observations, we hypothesized that chimeric Ad vectors can be used as more effective oncolytic viruses. To test this hypothesis, we used a model system of isogenically matched Ad vectors with either Ad5 or Ad3 tropism. Our study showed that each step of the virus replication cycle was enhanced by routing the vector through the Ad3 receptor. A modest enhancement in vector binding to the tumor cell via the Ad3 receptor was followed by a greatly amplified oncolytic outcome. By example, a 3–6-fold increase in binding (Fig. 1,A) translated into 2–2.5-log higher oncolytic activity (Fig. 5 B). As reported previously for ovarian cancer cells (8), it is likely that the SCCHN cells used in this study also express higher levels of the Ad3 receptor compared with the Ad5 receptor. It is difficult, however, to reconcile how such an apparently small difference in binding could have solely accounted for the logarithmic increases in virus production and oncolytic activity. One possible explanation is that number of vp/cell used in the binding assays (i.e., 10,000 vp/cell) was considerably higher than in the virus production and oncolysis experiments (i.e., 0.1–100 vp/cell). This was necessary to have a detectable radioactive signal (³H) by scintillation counting. The high number of vp/cell may have partially masked the differences in the binding activities of the two viruses, effectively under estimating the actual binding activities.

Although we initially expected that the enhancement in binding would correlate with subsequent downstream steps in virus replication, this study indicated a steady and progressive improvement in efficiency at each of the subsequent steps investigated (Fig. 7). Even if the conditions of the binding experiment underestimated the actual binding activities of the two viruses, the other experiments were performed under conditions that were similar to each other. It is therefore likely that additional mechanisms contributed to the differences between the two viruses. For example, using another subgroup B chimeric Ad5 virus (i.e., Ad5/7), Miyazawa et al. (23, 47) have shown that the intracellular trafficking is altered by the fiber protein from different Ad serotypes. Therefore, it is possible that mechanisms involved with routing of the viral DNA to the nucleus contributed to the higher efficiency of the chimeric Ad5/3 vector. More efficient nuclear translocation could contribute to higher levels of early transcription and translation products. Although we cannot rule out the possibility that the Ad5 and Ad5/3 vectors had different transcriptional kinetics, this seems unlikely because the rate of early gene transcription was similar for both vectors (Fig. 2 B, inset). In addition to its role in intracellular trafficking, the Ad fiber protein has been associated with intranuclear remodeling after viral infection (48). This Ad-induced remodeling is a requisite step for virus assembly and export. Furthermore, Legrand et al. (49) have shown that the fiber gene, although dispensable for particle formation, is involved with the production of fully mature and correctly assembled virions. Therefore, Ad5/3 may have an improved ability to amass in and escape from the infected cell. Whereas the replication of different Ad serotypes is likely to involve many conserved mechanisms, this study suggests that different Ad serotypes have evolved to achieve more efficient replication in their native host cells. Additional studies will, however, be needed to fully understand the underlying mechanisms that account for the replication efficiency of different Ad serotypes. It is also worth noting that the chimeric Ad5/3 vector does not efficiently infect cells that express low levels of the Ad3 receptor such the cervical cancer cell line C33A (data not shown). Therefore, it will be important to match different Ad serotype tropisms to appropriate tissue and cell targets in translational applications.

The development of oncolytic Ad vector systems is being rapidly pursued for gene therapy applications (1, 2, 3). These CRAds are based on Ad5 platforms that have been shown previously to have limited infection efficiency in numerous neoplastic tissues (7, 14, 17, 50, 51, 52). This study underscores the importance of tropism modification to improve the efficacy of CRAd-based gene therapy. Importantly, it shows the need of using alternate Ad serotype tropisms for tumors that are refractory to Ad5. Moreover, it reminds us of the need for the characterization of other Ad serotype receptors and the possible roles they may have in tumor development and therapies.

We thank Dr. Ming Wang for performing the quantitative RT-PCR analyses and Dr. Cynthia A. Derdeyn for critical review of the manuscript.