E1A, E1B Double-Restricted Adenovirus with RGD-Fiber Modification Exhibits Enhanced Oncolysis for CAR–Deficient Biliary Cancers

Purpose: Cancers of biliary system represent highly malignant diseases of dismal prognosis. We have previously introduced AxdAdB3, an E1A, E1B double-restricted oncolytic adenovirus, which showed excellent oncolytic efficacy for approximately half of the biliary cancer lines with an enhanced safety to normal cells. The purpose of this study was to evaluate whether RGD-fiber modification (AxdAdB3-F/RGD), which enables integrin-dependent infection, can improve the infectivity and efficacy of AxdAdB3 for biliary cancers.

Experimental Design: Expressions of adenoviral receptors, coxsackievirus adenovirus receptor (CAR) and integrins (α_vβ₃ and α_vβ₅), were compared with the level of infectivity of LacZ-expressing replication-defective adenoviruses with wild-type fibers or RGD-modified fibers in a panel of biliary cancer cell lines in vitro. Viral replication and cytotoxicity in vitro of AxdAdB3-F/RGD, a novel E1A, E1B double-restricted replication-selective adenovirus with RGD-modified fibers, were compared with those of its parent virus, AxdAdB3, in various biliary cancer cells and in normal cells. In vivo antitumor effects of these oncolytic viruses were compared in a xenograft tumor model.

Results: Expression of CAR significantly correlated with the adenovirus infectivity, whereas integrin α_vβ₅ was abundantly expressed in almost all biliary cancer cells. Whereas AxdAdB3 effectively replicated and lysed only the biliary cancer cells with a preserved expression of CAR, AxdAdB3-F/RGD exhibited efficient replication and potent oncolysis in both CAR-positive and CAR-negative biliary cancer cells. AxdAdB3-F/RGD showed attenuated replication and little cytopathy in human normal cells (i.e., hepatocytes, WI-38 cells) as well as AxdAdB3. Furthermore, in nude mice with s.c. xenografts of CAR-deficient human biliary cancer, i.t. AxdAdB3-F/RGD therapy caused a marked inhibition of tumor growth.

Conclusions: The RGD-fiber modification strategy enhanced the infectivity, replication, and oncolytic effects of the E1A, E1B double-restricted oncolytic adenovirus for CAR-deficient biliary cancers. In addition, it preserved the merit of excellent safety of the double-restricted virus for normal cells. These results suggest a potential use of this agent for the treatment of biliary cancers.

Cancers of the biliary systems (gall bladder and bile duct) represent highly malignant diseases of dismal prognosis in humans because of the late diagnosis, high incidence of postsurgical local-regional recurrence, and frequent distant metastasis (1–3). The diseases are curable at their early stages, but only 10% to 30% of the patients can be considered for curative surgery and the 2-year survival rate is only ∼4% to 8%, with median survival of <6 months. Therefore, development of new approaches with better antitumor activity, such as gene therapy, merits a high priority, but few studies are available for biliary cancers. Previous studies on cancer gene therapy, which mostly used nonreplicating viruses as vectors, have yielded disappointing results, primarily because of the insufficient gene transduction to uninfected neighboring cells that limit their antitumor effects (4, 5).

Conditionally replicating adenoviruses (CRAd), which are capable of cancer-selective replication and oncolysis, have recently received widespread attention as potentially ideal tools for innovative cancer therapy (5–7). For instance, ONYX-015 (dl1520), a mutant adenovirus carrying a deletion in the p53-binding M_r 55,000 protein encoded by E1B, enables selective replication in tumor cells with dysfunctional p53 signaling pathway (8). Clinical trails of ONYX-015 in combination with chemotherapy have yielded remarkably good efficacy and safety in patients with head and neck cancers (9, 10). Another CRAds with a mutation in the pRb-binding domain of E1A have been shown to replicate in tumor cells with disrupted Rb signaling pathway (11–13). Although these single-restricted CRAds exhibited potential for cancer therapy, these viruses do replicate and cause some cytopathic effects in normal cells in vitro (11, 14, 15). Moreover, the clinical trial of intralesional ONYX-015 in patients with hepatobiliary cancers showed sufficient safety but limited therapeutic effects (partial response, 6.3%; prolonged disease stabilization, 6.3%; refs. 16, 17). These studies suggest that further efforts are necessary for biliary cancers to develop CRAds that exert more selective replication and effective oncolysis than ONYX-015.

More recently, our group has introduced a CRAd that has mutations in both E1A and E1B (18). We showed that the double-mutant CRAd, AxdAdB3, had potent oncolytic activities to several biliary cancer lines with an enhanced safety profile to human normal cells than the ONYX-015-type E1B single-restricted CRAd (18). A similar dual-mutant CRAd has also been shown by others to exhibit efficacy for glioma cells with a highly attenuated replication in normal astrocytes (19). A recent view that dual abnormalities in the pRb and p53 signaling pathways should be common features in various cancers (20) suggests the efficacy of the E1A, E1B double-mutant CRAds for various cancers. In addition, the enhanced safety profile of the CRAds would allow more effective treatment of cancers by increasing the dose of the CRAds or by combining with other therapeutic modalities (21, 22).

During these studies (18, 22), however, we found that the efficacy of AxdAdB3 (18) was insufficient in some biliary cancer lines. This might be due to the low expressions of coxsackievirus adenovirus receptor (CAR; ref. 23), the primary cellular receptor for adenovirus, in these cells. The CAR deficiency causing low adenovirus infectivity has emerged as a limiting factor for the adenovirus-mediated gene therapy as well as the CRAd-mediated oncolytic virotherapy (24). Thus, efforts have been made to develop adenoviruses (25, 26) and CRAds (27) with improved infectivity to cancer cells. Incorporation of an Arg-Gly-Asp (RGD) peptide sequence, which interacts with α_vβ integrins (28), into the adenovirus fiber knob circumvents low CAR expression in several cancers by providing an alternative viral entry pathway (25, 26). Recent studies have shown that an E1A single-restricted CRAd with RGD-modified fibers exhibits improved oncolytic potency on several CAR-deficient malignancies (29–31). However, little has been studied on the effects of the CRAds with RGD-modified fibers on normal cells.

Here we report on a novel E1A, E1B double-mutated CRAd in which the RGD peptides are inserted at the HI loop of the fiber knob domain. Our data showed that CAR deficiency underlies the primary obstacle for the CRAd therapy with wild-type fiber for biliary cancers and that the incorporation of RGD peptide to the fiber knob greatly improved the infectivity and the oncolytic efficacy for CAR-deficient biliary cancer cells of the double-mutated CRAd without sacrificing the excellent safety profile to normal cells. These results suggest that the infectivity-enhanced E1A, E1B double-restricted CRAd is a potentially effective tool for the treatment of biliary cancers.

Cell lines and cultures. Four human gall bladder cancer cell lines, TGBC-1, TGBC-2, TGBC-14, and TGBC-44, were established by Dr. T. Todoroki (Division of Surgery, University of Tsukuba, Tsukuba, Ibaraki, Japan; ref. 32). Two other gall bladder cancer cell lines, Mz-ChA-1 and Mz-ChA-2, and a bile duct cancer cell line, Sk-ChA-1, were obtained from Dr. A. Knuth (Johannes-Gutenberg University, Mainz, Germany; ref. 33). Another bile duct cancer cell line, KMBC, as well as two cholangiocellular carcinoma cell lines, KMC-1 and KMCH-1, were obtained from Dr. M. Kojiro (Kurume University School of Medicine, Kurume, Japan). Of these biliary cancer lines, it is known that TGBC-1 and TGBC-2 lines have deleted p53 (stop at codon 181; ref. 34) and deleted pRb.⁴

Mz-ChA-1, Mz-ChA-2, and Sk-ChA-1 had mutated p53 (single base pair mutation; ref. 34) and wild-type pRb but deleted p16 (35). KMBC cells had wild-type p53 (34) and pRb but deleted p16 (35). HEK293 cells were purchased from American Type Culture Collection (Manassas, VA). A human fibroblast cell line, WI-38, and a lung cancer cell line, A549, were purchased from RIKEN Cell Bank (Ibaraki, Japan). Primary cultures of human hepatocytes and human intestinal epithelial cells were purchased from Applied Cell Biology Research Institute (Kirkland, WA). All biliary cancer cells (including six gall bladder cancer cells, two bile duct cancer cells, and two cholangiocellular carcinoma cells), HEK293 cells, WI-38 cells, and A549 cells were maintained in DMEM supplemented with 10% fetal bovine serum, 0.2% sodium bicarbonate, 2 mmol/L glutamine, penicillin (100 units/mL), and streptomycin (100 μg/mL) in a humidified atmosphere with 5% carbon dioxide at 37°C. The primary cultures of human hepatocytes were maintained in a CS-C complete medium kit (Cell Systems, Kirkland, WA). For passage of these primary cells, culture plates coated with type I collagen were used. The WI-38 cells and the primary human hepatocytes were starved by serum deprivation for 24 h before each experiment.

Construction of adenoviral vectors. Generation of AxdAdB3, an E1A, E1B double-restricted adenovirus with a mutated E1A and deleted E1B-55kD, was previously described (18). Recombinant adenoviruses with RGD-fiber mutation were constructed essentially according to the procedure previously described (36, 37). Briefly, we constructed a cosmid pWEAxKM-F/RGD, which contains the RGD-4C amino acid in the HI loop of the fiber knob domain between amino acid residues 546 and 547 with an E3 deletion. The amino acid sequence of the RGD-mutation is as follows: T₅₄₆CDCRGDCFCP₅₄₇. AxdAdB3-F/RGD was generated by cotransfecting 293 cells with pWEAxKM-F/RGD cosmid DNA, which had been digested with ClaI and PacI, together with the EcoRI- and AseI-digested DNA-TPC (terminal protein complex) of AxdAdB3. A wild-type adenovirus5 (Ad5-wt) was purchased from American Type Culture Collection. AxCAZ3-F/RGD (RGD 2194; ref. 36) and AxCAlacZ (RDB 2726; ref. 37), adenovirus vectors expressing Escherichia coli lacZ gene under the control of CAG promoter with or without an RGD peptide in the HI loop of the fiber knob domain, respectively, as well as Ax1w1 (RDB 1746; refs. 38, 39), a mock vector, were obtained from the RIKEN DNA Bank (Ibaraki, Japan). All adenoviral vectors were propagated in HEK293 cells and then purified by CsCl gradient centrifugation and stored at −80°C. The titers of adenoviruses were determined by the standard plaque-forming assay using HEK293 cells and the numbers of adenoviral particles were calculated as absorbance at 260 nm (A₂₆₀; ref. 40). All CRAds were successfully made, and any contamination with wild-type E1A, E1B or wild-type fiber was not detected by PCR and sequencing (data not shown).

Expressions of CAR and integrins on biliary cancer cell lines. For flow cytometry analysis, cells grown in 10-cm dishes were washed twice with PBS containing 0.02% EDTA and harvested. Cells were washed twice with PBS and resuspended with PBS containing 2% fetal bovine serum. The cells were then incubated for 30 min on ice with one of the following primary antibodies: 2 μg/mL RmcB (a mouse anti-CAR monoclonal antibody; Upstate, Charlottesville, VA), 7 μg/mL LM609 (a mouse anti–α_vβ₃ integrin monoclonal antibody; Chemicon International, Temecula, CA), or P1F6 (a mouse anti–α_vβ₅ integrin monoclonal antibody; Covance, Princeton, NJ), or normal murine immunoglobulin G [negative control; Sigma (St. Louis, MO) and Santa Cruz Biotechnology (Santa Cruz, CA)]. Subsequently, the cells were washed twice with 0.2% fetal bovine serum-PBS and incubated for additional 30 min with 10 μg/mL FITC-conjugated goat anti-mouse immunoglobulin G (Sigma and Santa Cruz Biotechnology). After washing twice with 2% fetal bovine serum-PBS, the cells were analyzed by FACScan using CellQuest software according to the manufacturer's instructions (Becton Dickinson, San Jose, CA). For Western blot analysis of CAR expression, total cell lysates were prepared by incubating cells for 2 h at 4°C in radioimmunoprecipitation assay buffer [150 μmol/L NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 10 mmol/L EDTA, and 50 mmol/L Tris (pH 7.4)]. Proteins (15 μg) from each sample were then subjected to 10% SDS-Tris-glycine gel electrophoresis. The membrane was incubated with a solution containing 1:1,000 dilution of primary antibody, RmcB. The secondary antibody was a horseradish peroxidase–conjugated sheep anti-mouse immunoglobulin G polyclonal antibody. Signals were visualized with the enhanced chemiluminescence-plus system (Amersham Biosciences Corp., Piscataway, NJ).

Infectivity of adenoviruses. To estimate the infectivity of Ad-F/wt, biliary cancer cell lines and A549 cells were seeded in 24-well plates at a density of 10⁵ per well 1 day before the infection. Cells were infected with AxCAlacZ at several multiplicities of infection [MOI; 0.1-500 plaque-forming units (pfu)/cell] for 1 h on ice, washed with PBS, and supplemented with fresh culture media, and were incubated at 37°C overnight. Cells were then stained with 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside 24 h after the infection and 100 cells were counted under a microscope. The 50% infectivity, which means the MOI of the adenovirus supposed to infect half of the cells, was calculated from the regression analysis. To compare the infectivity of Ad-F/RGD with Ad-F/wt, different biliary cell lines were infected with AxCAlacZ or AxCAZ3-F/RGD in the same manner. Twenty-four hours after the infection, cells were washed twice with PBS and harvested by scraping. β-Galactosidase assay was done using the β-Gal Assay Kit (Invitrogen, Carlsbad, CA) following the manufacturer's instruction. The amount of β-galactosidase was calculated from absorbance at 420 nm (A₄₂₀).

Viral replication. Cells were seeded in 24-well plates in the appropriate medium at a density of 5 × 10⁴ per well and were infected with various adenoviruses at a MOI of 1. Fresh medium was replaced 6 h after the infection. The cells and the supernatants were collected on the day infected and 1 to 5 days after the infection. Then, the cells were sonicated and titrated by the standard plaque-forming assay using HEK293 cells (40).

Cytopathic effects. Biliary cells were seeded in 96-well plates at a density of 2 × 10³ per well. WI-38 cells and the primary cultures of human normal cells were seeded in type I collagen–coated 96-well plates at a density of 5 × 10³ per well in appropriate medium. All cells were mock infected or infected with either AxdAdB3, AxdAdB3-F/RGD, or Ad5-wt at a MOI of 1 and 10. Viable cell numbers were evaluated by a colorimetric WST-1 assay (Takara Bio, Inc., Shiga, Japan) as previously described (18).

Animal studies. All mice received human care in compliance with the guidelines for the care and use of laboratory animals in research. Four-week-old female BALB/c nu/nu athymic mice were purchased from CLEA Japan (Tokyo, Japan) and were quarantined for 1 week. Xenografts were established by s.c. injection of 2 × 10⁷ TGBC-2 cells in 100 μL of Matrigel (BD Biosciences, San Jose, CA) into the right flank of each mouse. Tumors were measured with calipers and the volume of each tumor was calculated as 0.4 × longest diameter × width². When the tumor volume reached ∼150 mm³, animals were randomly assigned to one of the following groups and received i.t. injection of 100 μL of PBS or various adenoviruses (2 × 10⁸ plaque-forming units/d on days 1-3) followed with or without i.p. 5-fluorouracil injection (10 mg/kg/d) on days 7 to 12: (a) PBS group, (b) 5-fluorouracil group (i.p. 5-fluorouracil on days 7-12), (c) Ax1w1 group (i.t. Ax1w1 alone), (d) AxdAdB3 group (i.t. AxdAdB3), and (e) AxdAdB3-F/RGD group (i.t. AxdAdB3-F/RGD).

Statistical analysis. Relationship between the expressions of each cellular receptor and the infectivity of adenoviruses was assessed by correlation analysis. The significance of differences between groups was analyzed by Student's unpaired two-tailed t test using the Statview program (SAS Institute, Inc., Cary, NC) and was considered statistically significant when adjusted P < 0.05.

Transduction efficiency of adenovirus with wild-type fiber and its relationship with the expression of CAR and integrins in human biliary cancer cells. To clarify whether the insufficient efficacy of AxdAdB3 in some biliary cancer cells was attributable to its low adenovirus infectivity, we first determined the transduction efficiency of adenovirus in several human biliary cancer lines. A lung cancer cell line, A549 (Fig. 1A, bottom), which is known to express high levels of CAR and is susceptible to adenovirus (41), was used as a positive control. The 50% adenovirus transduction efficiency in A549 cells was 3.5 pfu/cell, indicating that about half of the cancer cells expressed β-galactosidase through the infection and gene transduction by AxCAlacZ at a MOI of 3.5. In biliary cancer cells, the 50% transduction efficiencies of Mz-ChA-2, Sk-ChA-1, TGBC-44, and KMCH-1 cells were 3.3, 3.5, 7.4, and 14.9 pfu/cell, respectively. These cells were found to be relatively sensitive to adenovirus infection. Other six biliary cancer cells were found less susceptible to adenovirus infection because the 50% transduction efficiency exceeded 100 pfu/cell. Among these, TGBC-1 cells showed the lowest adenovirus infectivity (the 50% transduction efficiency was 366.6 pfu/cell), and β-galactosidase was transduced in only 16% of TGBC-1 cells following AxCAlacZ infection even at a high MOI of 100.

We next examined the expression levels of CAR and α_vβ₃- or α_vβ₅-type integrins in human biliary cancer cells and A549 cells in relation to the transduction efficiency of adenovirus [Fig. 1A (top) and B]. A549 cells highly expressed CAR on both flow cytometry and Western blots, consistent with a previous report (41). By contrast, 6 of 10 biliary cancer cells expressed CAR at low levels (1-23%). In particular, both flow cytometry and Western blotting analysis showed completely negative expression of CAR in TGBC-1 and TGBC-2 cells. A significant negative correlation was observed between the expression levels of CAR and the 50% transduction efficiency (correlation coefficient −0.623, P = 0.0269). However, some exceptions were recognized. TGBC-44 cells expressed CAR at low levels but showed efficient transduction of adenovirus. Contrariwise, Mz-ChA-1 cells were refractory to adenovirus-mediated gene transfer despite high level of CAR expression. α_vβ₅-type integrin was expressed at high levels in almost all biliary cancer lines (64-99%) and A549 cells (96%). High level of α_vβ₃-type integrin was observed only in TGBC-2 cells. There was no significant correlation between the expression level of α_vβ₃- or α_vβ₅-type integrins and the 50% transduction efficiency of adenovirus with wild-type fiber.

Transduction efficiency of adenovirus with or without RGD-fiber modification in biliary cancer cells. Based on the uniformly high expression of α_vβ₅-type integrin, in contrast to variable expression levels of CAR, in most biliary cancer cells, we next examined the transduction efficiency of the adenovirus with RGD-fiber modification (AxCAZ3-F/RGD), in comparison with that of the adenovirus with wild fiber (AxCAlacZ) in various biliary cancer cell lines by chemiluminescence assay (Fig. 2). In the biliary cancer cells expressing high levels of CAR, such as Sk-ChA-1 and Mz-ChA-2 cells, both AxCAlacZ and AxCAZ3-F/RGD showed similar efficient transduction of β-galactosidase in a dose-dependent manner, and no significant difference in the transduction efficiency was observed between the two adenoviruses at any MOIs tested. By contrast, AxCAZ3-F/RGD provided the two CAR-deficient biliary cancer lines, TGBC-1 (Fig. 2C) and TGBC-2 cells (Fig. 2D), with significantly increased β-galactosidase activities compared with AxCAlacZ (P < 0.05). AxCAZ3-F/RGD required ∼10-fold lower MOI than AxCAlacZ to transduce and express β-galactosidase activity of similar level in TGBC-1 and TGBC-2 cells.

Replication of AxdAdB3 with or without RGD-fiber modification in human biliary cancer cells and normal cells. We examined the replication of the two CRAds and Ad5-wt in CAR-positive and CAR-negative human biliary cancer cells as well as in normal cells (Fig. 3). In the CAR-positive Mz-ChA-2 cells, both AxdAdB3 and AxdAdB3-F/RGD showed excellent replication: their titers increased 2,000- and 3,000-fold, respectively, of the initial titers by 5 days after the infection, compared with the 9,000-fold increase of the Ad5-wt. In the CAR-deficient TGBC-2 cells, AxdAdB3 (32-fold by 5 days) and Ad5-wt (99-fold by 5 days) replicated poorly, whereas AxdAdB3-F/RGD exhibited excellent replication: the titer of the latter adenovirus increased 5,900-fold of the initial titer, which is >180-fold of that of AxdAdB3, 5 days after the infection. By contrast, the replication of the two E1A, E1B-restricted CRAds was remarkably suppressed in the normal cells. In normal WI-38 fibroblasts, whereas the titer of Ad5-wt increased ∼63-fold by 5 days, both AxdAdB3-F/RGD and AxdAdB3 did not replicate at all. In addition, in primary hepatocytes, whereas Ad5-wt showed a robust replication (11,000-fold increase by 5 days after the infection), AxdAdB3 did not replicate at all and AxdAdB3-F/RGD showed limited replication (22-fold by 5 days), which was still ∼1/500 of that of Ad5-wt.

Cytopathic effects of AxdAdB3 with or without RGD-fiber modification in biliary cancer cells and normal hepatocytes. We examined whether the RGD-fiber modification enhances the oncolytic effects of AxdAdB3 for biliary cancers by comparing the cytopathic effects in vitro of AxdAdB3-F/RGD and AxdAdB3 (Fig. 4). In the two CAR-positive biliary cancer cells (Sk-ChA-1 and Mz-ChA-2 cells), both AxdAdB3 and AxdAdB3-F/RGD showed significant cytopathic effects at a low MOI of 1 (P < 0.01) and the effects increased in a time- and dose-dependent manner. There was no significant difference between the cytopathic effects of these two CRAds in both cancer cells. In the CAR-deficient TGBC-2 line, AxdAdB3 showed no cytopathic effect at a MOI of 10 and only mild cytopathic effect even at a high MOI of 100. By contrast, AxdAdB3-F/RGD showed significantly (P < 0.01) stronger cytopathic effects than AxdAdB3 in TGBC-2 cells, which increased in a time- and dose-dependent manner; TGBC-2 cells infected with AxdAdB3 survived and rather proliferated even on day 9 after the infection whereas cells infected with AxdAdB3-F/RGD (MOI of 100) entirely expired within 3 days. In another CAR-deficient TGBC-1 line, AxdAdB3 showed only mild cytopathic effect and total killing required a high dose (MOI of 30) and 9 days, whereas AxdAdB3-F/RGD exhibited remarkably faster and stronger cytopathic effect than AxdAdB3 and total killing was achieved within 3 days after the infection. Thus, RGD-fiber modification significantly enhanced the oncolytic efficacy of AxdAdB3 for CAR-deficient biliary cancers.

We next examined the cytopathic effects of AxdAdB3, with or without RGD-fiber modification, on normal primary hepatocytes (Fig. 5). Both AxdAdB3-F/RGD and AxdAdB3, at a MOI of 1, had almost no cytopathic effects on the primary hepatocytes, unlike the wild-type adenovirus, which caused total killing of the cells by day 9 (Fig. 5A). The two CRAd showed very limited cytopathic effects on these cells at a MOI of 10, and such effects were significantly (P < 0.01) weaker than those of Ad5-wt. These findings were also confirmed by light microscopic observation (Fig. 5B).

In vivo antitumor effects of AxdAdB3 with or without RGD-fiber modification on the CAR-deficient biliary cancer xenografts. We next examined the antitumor efficacy in vivo of AxdAdB3 with or without RGD-fiber modification for s.c. xenografts of TGBC-2 cells in nude mice. The tumor growth curves of each treatment group are shown in Fig. 6A. The tumors treated with PBS (control) grew rapidly and reached ∼16-fold of the initial volumes on day 64. The tumors treated with Ax1w1 (mock) or AxdAdB3 also grew faster and reached 25- or 21-fold of the initial volumes, although there was no significant difference between these three groups. Mice that received i.t. AxdAdB3-F/RGD showed significantly inhibited tumor growth than the mice treated with PBS (control; P < 0.05) and the tumor volumes were just 1- to 2-fold of the initial volumes on day 64. Tumors were examined by electron microscopy on day 14 (Fig. 6C). In the tumors treated with Ax1w1 (mock), very few viral particles were detected. In the tumors treated with AxdAdB3, some viral particles could be detected in numerous viable tumor cells. In contrast, the tumors treated with AxdAdB3-F/RGD have undergone extensive necrosis, and numerous viral particles were detected, mostly in the scarcely remaining viable tumor cells around the necrotic area.

We showed that use of AxdAdB3-F/RGD, a novel E1A, E1B double-restricted CRAd that carries an RGB peptide in the HI loop of the fiber knob domain, can improve the infectivity and thus enhance the oncolytic effects in vitro and in vivo against CAR-deficient biliary cancer cells without significantly sacrificing the low toxicity of the double-restricted CRAd to human normal hepatocytes.

Our group has previously introduced AxdAdB3, the CRAd which has both mutant E1A and E1B (18). We showed that the double-mutant CRAd had potent oncolytic activities to several biliary cancer lines, with less toxicity to human normal cells than the ONYX-015-type E1B single-restricted CRAd. Recently, a similar dual-mutant CRAd has also been shown by others to exhibit efficacy for glioma cells with a highly attenuated replication in normal astrocytes (19). However, we found insufficient efficacy of AxdAdB3 for some biliary cancer cells. In addition, the clinical trial of intralesional administration of ONYX-015 for hepatobiliary cancers has shown some but not sufficient antitumor effects, which led us to the present study (11, 12).

We found that biliary cancer cells refractory to the treatment of CRAds with wild-type fibers (AxdAdB3 and AxE1AdB), such as TGBC-1 cells and TGBC-2 cells, express low levels of CAR, whereas the majority of biliary cancer cells highly express α_vβ₅-type integrins (Fig. 1). Adenovirus infectivity, evaluated by the transduction efficiency, significantly correlated with the expression levels of CAR and was a critical factor affecting the efficacy of the CRAd. However, some exceptions were also noted. For example, TGBC-44 cells, which allow efficient transduction of adenovirus, surprisingly expressed CAR at low levels. Although CAR represents the major host cell receptor for adenovirus binding and entry, recent studies have suggested that other molecules on host cells may also serve as receptors for adenovirus attachment. The α2 domain of MHC class I molecules has been reported to serve as a receptor for Ad5 particles based on competition experiments with phage display peptides (42). In addition, Dechecchi et al. (43) have recently suggested that heparin sulfate proteoglycans may also promote cell attachment of subgroup C viral particles. These molecules may play a significant role in the attachment of adenovirus to the adenovirus-susceptible cells with low expression of CAR, such as TGBC-44 cells. Contrariwise, Mz-ChA-1 cells were refractory to adenovirus-mediated gene transfer despite the high levels of CAR expression noted by flow cytometry and Western blotting. Possible explanation for this discrepancy, which needs to be tested in the future study, includes inappropriate distribution of CAR on the cell surface (44) or some mutation of CAR leading to its functional impairment.

The RGD peptide in the penton base protein of adenovirus interacts with integrins and allows internalization of adenoviruses. Several reports indicated that incorporation of the RGD peptide to the COOH terminus (25) or the HI loop of the fiber knob domain (26, 29–31) enhances adenovirus-mediated gene transfer to CAR-deficient cells through the binding with integrins overexpressed in neoplastic tissues. As we compared the adenovirus infectivity, assessed by the lacZ transduction efficiency, with and without the RGD-fiber modification (AxCAZ3-F/RGD versus AxCAlacZ), we found that the RGD-fiber modification provided the adenovirus with a significantly increased infectivity for CAR-deficient biliary cancer cells (Fig. 2). Due to the increased infectivity, the CRAd with RGD-fiber modification showed an enhanced viral replication (Fig. 3) and enhanced oncolytic efficacies in vitro (Fig. 4) and in vivo (Fig. 6) for CAR-deficient biliary cancer cells. On the other hand, both adenoviruses with and without the RGD-fiber modification showed similar adenovirus infectivity and both CRAds exhibited similar levels of replication and oncolytic activity in the CAR-positive biliary cancers. These results support our hypothesis that RGD-fiber modification would enhance the oncolytic efficacy of the E1A, E1B double-restricted CRAd for biliary cancers.

Little has been studied about the effects of RGD-fiber modified CRAds on the normal cells including hepatocytes. Especially, liver toxicity due to adenovirus tropism to the liver (45, 46) has been one concern of the CRAd approach. We found that the replication of both CRAds with and without RGD-fiber modification (AxdAdB3 and AxdAdB3-F/RGD) was remarkably inhibited in normal WI-38 fibroblasts and primary hepatocytes (Fig. 3). In hepatocytes, whereas Ad5-wt showed a robust replication (11,000-fold increase by 5 days) and AxdAdB3 showed no replication, AxdAdB3-F/RGD showed a limited replication (22-fold by 5 days), which was still ∼1/500 of that of Ad5-wt. Furthermore, both CRAds had very limited cytopathic effects on the normal hepatocytes (Fig. 5). In addition, in the animal study, whereas extensive viral replication was noted in the tumor, no signs of virus-associated toxicities were observed in normal tissues. These data support that the therapy with AxdAdB3-F/RGD would be considerably safe because the replication of the E1A, E1B double-mutated CRAd is markedly inhibited in the normal cells that have intact functions of both pRb and p53 pathways. Having the two mutations in the CRAd to restrict its replication in normal cells would also decrease the risk of unwanted generation of wild-type adenovirus from the CRAd during its replication in vivo. It should be noted, however, that small amount of replication of AxdAdB3-F/RGD in normal cells (Fig. 3) may cause some unwanted in vivo effects and that nude mice are poorly permissive for human adenovirus. Thus, the true in vivo safety should be tested in the future study by a permissive immunocompetent animal model for adenovirus. Recently, Thomas et al. (47) have shown that Syrian hamster is a good animal model for such evaluation.

One might assume that the types of cancer responsive to the double-mutated CRAd are limited because it requires abnormalities in both pRb and p53 pathways. Indeed, we know the status of p53, pRb, and p16 in 6 of 10 biliary cancer lines used and five of the six have abnormality in p53 and six of the six lines have abnormality in either pRb or p16. However, it is recently assumed that such dual abnormalities in the pRb signaling pathway (pRb, p16, cyclin-dependent kinase 2, cyclin-dependent kinase 4/cyclin-dependent kinase 6, etc.) and in the p53 signaling pathway (p53, p14, murine double minute-2, ataxia telangiectasia mutated, etc.) are common features in most cancer cells (20). Therefore, we may reasonably expect the efficacy of the double-restricted CRAd for a variety of other malignancies as well. In addition, the enhanced safety profile of the CRAd may allow dose escalation of the CRAd therapy or its combination with potent cytotoxic genes or other therapeutic modalities (21). These issues should be examined in future studies.

We thank J. Sakamoto, N. Sugae, and K. Inatsuki for technical assistance.