Conditionally Replicative Adenovirus with Tropism Expanded towards Integrins Inhibits Osteosarcoma Tumor Growth in Vitro and in Vivo

Osteosarcoma (OS) is one of the most common primary tumors of bone. It is highly malignant and typically affects children and young adults (1). Despite improvements in the treatment of OS, there are still too many patients who cannot benefit from current treatment modalities (2, 3). The overall survival with an aggressive chemotherapy regimen before and after surgery now varies between 50% and 65% (4, 5, 6). In addition, current chemotherapeutic agents have a broad spectrum of side effects (7). Therefore, new therapeutic options, such as gene therapy, are warranted. In the field of adenoviral vector (Adv)-based gene therapies for OS, several approaches have been explored, such as suicide gene therapy (8), tumor suppressor gene therapy (9), and cytokine-based gene therapy (10). Results of these studies have not yet been translated into clinical trials. Important limitations in this regard are the failure of nonreplicating Adv to achieve sufficient tumor-cell transduction and effective solid-tumor penetration.

The development of conditionally replicative adenoviruses (CRAds) has addressed these limitations. CRAds represent promising agents for the treatment of solid tumors, because CRAds have intrinsic replication capacity that allows in situ amplification and extensive tumor infection through lateral spread (11). CRAds have been designed following two different strategies, i.e., by controlling expression of essential early adenovirus genes with tumor-specific promoters (e.g., see Refs. 12, 13) and by deleting viral genes encoding proteins that are necessary to complete the viral lytic cycle in normal cells, but not in cancer cells (14, 15, 16).

A critical step determining CRAd efficacy is the infection efficiency dependent on expression of the primary adenovirus receptor, the Coxsackie adenovirus receptor (CAR; Ref. 17). Low CAR expression on primary-tumor cells has been described to limit the efficacy of adenovirus-based therapy for several types of cancers, including brain, bladder, and pancreatic cancer (18, 19, 20, 21). Recently, we demonstrated that Adv-infection efficiency is also compromised in OS, because of low or absent CAR expression (22). Therefore, other, more effective routes of infection for OS should be explored. Previously, it has been shown that incorporation of a cyclic Arg-Gly-Asp (RGD-4C) sequence that interacts with α_ν integrins into the adenovirus fiber-knob circumvents low CAR expression by providing an alternative viral entry pathway (23). In addition, recent studies have shown that a CRAd encoding retinoblastoma protein binding-deficient E1A proteins and the RGD-4C-fiber, Ad5-Δ24RGD (24), exhibited oncolytic potency on CAR deficient cancer cells (25, 21). Therefore, we hypothesized that Ad5-Δ24RGD could also be effective in OS tumors.

In this study, we first explored the expression of integrins on OS cell lines and primary OS cells. Next, we tested the effect of the RGD-4C insertion in the fiber knob of a nonreplicative Adv on OS transduction. Finally, we assessed the efficacy of Ad5-Δ24RGD on a broad panel of primary OS cell cultures in vitro and on primary OS s.c. xenografts in vivo.

MG-63 (Ref. 26; courtesy of Dr. C. Löwik, Leiden University Medical Center, The Netherlands), MNNG-HOS (American Type Culture Collection, Manassas, VA; Ref. 27), SaOs-2 (Ref. 28; courtesy of Dr. F. van Valen, Westfalische Wilhelms-Universitat Münster, Germany), and CAL-72 (Ref. 29; courtesy of Dr. J. Gioanni, laboratoire de Cancerologie, Faculte de Medicine, Nice, France) OS cell lines were maintained in DMEM supplemented with 10% FCS, 50 IU/ml penicillin, 50 μg/ml streptomycin, and 2 mm l-glutamine (all from Life technologies B.V., Breda, The Netherlands) at 37°C in a 5% CO₂ humidified atmosphere.

Fresh tumor material from patients suspected of having a classic high-grade OS was processed directly after open biopsy surgery and kept under sterile conditions at 4°C. The biopsies were washed in sterile PBS, cut into small pieces and digested with Liver Digestion Medium (Life technologies B.V.) during four consecutive 30-min incubations in a water bath at 37°C. After each round of digestion, the cells were collected and washed. Finally, all cells were pooled and washed. Cells were brought into culture and used for experiments at passage 0–5. Confirmation of OS tumor cell morphology of all short-term cultures was performed by histopathology. Patient characteristics have been described previously (22).

OS cells were incubated for 1 h on ice with anti-integrin α_vβ3 and α_vβ5 mouse monoclonal antibodies (α_vβ3 MAB 1962 and α_vβ5 MAB 1976; CHEMICON International Inc., Temecula, CA), 1:100 diluted in PBS containing 0.1% BSA, followed by FITC-conjugated rabbit-antimouse IgG antibody 1:100 (DAKO, Glostrüp, Denmark) for 1 h on ice. The samples were fixed in PBS/1% formaldehyde and analyzed for integrin expression on a FACscan (Becton Dickinson) using Cell Quest software within 24 h after fixation. Integrin expression was considered positive when the relative median fluorescence intensity of integrin stained cells over control cells was >2.

A recombinant E1-deleted Adv expressing the luciferase reporter gene under the control of the cytomegalovirus promoter, AdCMVLuc, was provided by Dr. R. D. Gerard (University of Texas Southwestern Medical Center, Dallas, TX). A similar Adv, containing an integrin targeting RGD-4C peptide within the fiber HI-loop, AdLucRGD, has been described previously (23). The CRAd AdΔ24 (30) was made by homologous recombination in “E1-complementing 293” cells between the pXC1 (Microbix Biosystems, Toronto, Canada) derivative pXC1-Δ24, carrying a 24-bp deletion in the pRb-binding CR2 domain in E1A (15), and pBHG11 (Microbix Biosystems). Ad5-Δ24RGD (31) carries the same Δ24 mutation, plus the RGD-4C fiber modification and adenovirus E3 region from pVK503 (23). Replication-deficient Adv were propagated on 293 cells; conditionally replicative adenoviruses were propagated on A549 cells (American Type Culture Collection). Viruses were purified using cesium chloride gradient banding and titrated in parallel by end-point-limiting dilution on 293 cells to determine plaque-forming units (pfu).

OS cell lines were plated in a 24 wells plate at 2.5 × 10⁴ cells/well. Primary OS cells were plated similarly or in a 96-well plate at 5 × 10³ or 10⁴ cells/well, depending on the amount of primary OS cells available. To allow adherence, cells were incubated overnight in DMEM/10% FCS medium. Adenoviruses were diluted in DMEM/2.5% FCS, and cells were infected in triplicate at the indicated multiplicity of infection (MOI). One hour after infection, the adenovirus-containing medium was removed and fresh DMEM/10% FCS medium was added. Cells were cultured at 37°C until analysis of gene transfer by luciferase-activity measurement or of cell viability by WST-1 conversion assay (see Colorimetric WST-1 Cell Viability Assay).

Twenty-four hours postinfection, cells were assayed for luciferase expression using the Luciferase Chemiluminescent Assay System (Promega, Madison, WI). Medium was removed from the cells, and luciferase reporter lysis buffer was added to the wells followed by a freeze thaw step. Luminescence was measured during 10 s immediately after initiation of the light reaction in a Luminat LB 9507 luminometer (EG&G Berthold, Bad Wildbad, Germany). Values were protein normalized (Bio-Rad Protein Assay, Bio-Rad laboratories, Veenendaal, The Netherlands). Results were expressed as relative light unit(s) (RLU) per microgram of protein.

Cell viability was quantified using the colorimetric WST-1 conversion assay, based on the cleavage of the tetrazolium salt WST-1 by mitochondrial dehydrogenases in viable cells (Roche Diagnostics, Manheim, Germany), 7 or 11 days after infection. The culture medium was removed and replaced by 100 μl of 10% WST-1 in culture medium. Culturing was continued until the formation of formazan dye was optimal. The A₄₅₀ was measured on a Bio-Rad Model 550 (Hercules, CA) microplate reader. WST-1 conversion was expressed as a percentage of the conversion by uninfected control cells.

Female athymic nu/nu mice, weighing 25–35 g, obtained from Harlan-CPB (Austerlitz, The Netherlands) were used. Animals were kept under pathogen-free conditions and fed a standard laboratory diet (Hope Farms, Woerden, The Netherlands) ad libitum. The experimental protocols adhered to the rules outlined in the Dutch Animal Experimentation Act (1977) and the published Guidelines on the Protection of Experimental Animals by the council of the European Commission (1986). OS-1a (21) tumor pieces (3 × 3x3 mm), derived directly from a patient with OS and propagated s.c. in nude mice, were xenografted under the skin of nude mice. OS-1a xenografts maintained gross histological resemblance to the primary human tumor. In the patient, this tumor had been treated according to the COSS96 protocol and did not show any response to treatment. When the nodules reached 150–300 mm³, mice received intratumor injections on 5 consecutive days with 20 μl of control medium (n = 7) or 1 × 10⁷ pfu Ad5-Δ24RGD (n = 6). Injections into the tumor were performed from a different angle every day. Tumor size was monitored twice a week using digital calipers. Volume was calculated from the average of tumor length and width according to the formula 4/3πr³. The mice were sacrificed when the tumors reached a size of >1000 mm³. The two-tailed, unequal variance Student’s t test from SPSS (SPSS Inc., Chicago, IL) was used to compare tumor growth rate (time required to reach 5 times the initial tumor volume) between the different treatment groups.

When the tumor reached a volume of >1000 mm³, mice were sacrificed and tumors were dissected. Slides of 4% formalin-fixed and paraffin-embedded tumor were routinely stained with hematoxilin-eosin and immunohistochemically for adenovirus hexon with 1:1600-diluted goat anti-adenovirus AB1056 (CHEMICON International Inc.), followed by antigoat-biotin-streptavidin horseradish peroxidase (K377, DAKO). An antigen retrieval step with 10 mm Tris/1 mm EDTA was included. Positive anti-adenovirus stained areas were dissected from paraffin blocks. The dissected samples were worked up for electron microscopy by a one-step dehydration/postfixation technique (toluene-OsO4), followed by embedding in epoxy resin (32). Samples were examined for adenovirus particles by electron microscopy at ×24,000 and ×108,000 magnification.

Previously, we examined CAR expression on OS cell lines and primary OS cell cultures (22). All OS cell lines, except SaOs-2, and all primary cells showed low or even absent CAR expression. In search of alternative target molecules on OS that would allow more efficient CRAd infection, we examined α_vβ3 and α_vβ5 integrin expression by fluorescence-activated cell sorter analysis (Table 1). All OS cell lines and primary OS cell cultures expressed at least one of these integrins. To study the utility of integrins as targets for adenovirus infection of OS, we compared the gene-transfer efficiency of an adenovirus vector with native tropism (AdCMVLuc) to that of a derivative vector with RGD-4C fibers (AdLucRGD). As shown in Fig. 1, targeting toward integrins strongly enhanced gene-transfer efficiency on OS cell lines and primary OS cell cultures. In particular, the on-average two orders of magnitude augmented transduction of primary cells by AdLucRGD held promise for the utility of integrin-targeted CRAds for the treatment of OS.

To evaluate the efficacy of the CRAd AdΔ24 on OS cells, we infected four OS cell lines at high MOI (Fig. 2,A). AdΔ24 effectively killed all OS cell lines, suggesting that a CRAd with the Δ24-mutation could be effective for the treatment of OS. However, when we infected the same panel of OS cell lines or OS primary cell cultures with AdΔ24 at low-MOI, only CAR-positive SaOs-2 cells were killed (Fig. 2,B). In contrast, integrin-targeted Ad5-Δ24RGD was highly active in killing all OS cell lines and primary cell cultures at low MOI (Fig. 2 B). Thus, only Ad5-Δ24RGD was capable of effectively killing CAR-deficient primary OS cells.

Finally, we investigated if Ad5-Δ24RGD was also effective against CAR-deficient primary human OS tumors in vivo. Nude mice carrying established OS-1a xenografts derived from a chemotherapy-resistant primary tumor were treated by intratumor injections with a total virus dose of 5 × 10⁷ pfu Ad5-Δ24RGD. The growth curves of tumors injected with virus or control medium are shown in Fig. 3. The median tumor growth rate was calculated for both treatment groups. Control tumors had a median tumor growth rate of 14 ± 4 days, whereas tumors injected with Ad5-Δ24RGD exhibited a significant growth delay (median tumor growth rate of 25 ± 4 days; P < 0.05).

Histological analysis of tumors exceeding 1000 mm³ showed classical OS features with osteoid calcified areas. In tumors retrieved from Ad5-Δ24RGD-treated animals, classic OS features could still be observed, together with posttherapeutic changes of necrosis within viable tumor fields surrounded by fibrosis (Fig. 4,A). Anti-adenovirus staining revealed viral replication throughout treated tumors, mainly in tumor cells at the border of areas with necrosis (Fig. 4,B). Hexon staining was absent in nontreated tumors (data not shown). Areas with positive anti-adenovirus staining were sampled for electron microscopy, which showed 65-nm dense, rounded adenovirus particle inclusions in tumor cells (Fig. 4, C and D). Electron microscopy of nontreated tumors did not show such inclusions (data not shown). Altogether, these findings showed a presence and replication of Ad5-Δ24RGD in treated tumors for up to 25 days postinjection.

The clinical course of OS demands the development of new therapeutic options. CRAds represent promising agents for the treatment of solid tumors. However, their efficacy in OS tumors may be hampered by inefficient cell entry because of a lack of CAR expression (22). We, thus, hypothesized that the efficacy of CRAds against OS could be improved by expanding their tropism toward αν integrins. This was supported by our findings described herein that αv integrins are expressed on primary OS cells and that a replication-deficient Adv, which was genetically modified to expose the RGD-4C peptide in the fiber knob, exhibited dramatically enhanced gene transfer into OS cell lines and, more importantly, into primary OS cells.

We chose to investigate the effect of CRAd integrin targeting using AdΔ24 as a backbone. AdΔ24 encodes E1A proteins with a deletion of eight amino acids in the pRb-binding CR2 domain that disables their capacity to release E2F from preexisting pRb/E2F complexes, thereby conferring selective replication in cells with already dysfunctional pRb control (13). It has already been shown that infection with AdΔ24-type CRAds effectively kills cancer cells in vitro and inhibits tumor growth in vivo, and replication is reduced in nonproliferating normal cells and in cancer cells with restored pRb function (13, 14). Because genetic alterations in the Rb pathway are frequent in OS (24, 33), AdΔ24-type CRAds appear to be useful agents for treatment of this disease. Notably, however, the concept of integrin targeting is relevant for any kind of CRAd, irrespective of the genome modification conferring selective replication.

Enhancing the infection efficiency of AdΔ24 by integrin targeting translated into increased oncolytic effects. On a broad panel of primary OS cells infected at low MOI, Ad5-Δ24RGD showed complete oncolysis, whereas AdΔ24 induced no detectable cell kill at all. Because AdΔ24, in contrast to Ad5-Δ24RGD, lacks the adenovirus E3 region, differences in oncolytic effect may not only be explained by the expanded tropism of Ad5-Δ24RGD. However, AdΔ24 did effectively kill the CAR-positive OS cell line SaOS-2 at low MOI, suggesting that the infection enhancement of Ad5-Δ24RGD contributed to its efficacy on CAR-negative OS cells.

Primary cells derived directly from patient tumor samples represent an in vitro model closer to the patient compared with cell lines. Therefore, the observation that Ad5-Δ24RGD showed complete oncolysis of primary OS cells at an MOI of only 0.1 pfu/cell, whereas AdΔ24 required 100 pfu/cell for this effect, is of significant importance. The in vitro efficacy of Ad5-Δ24RGD was confirmed in a relevant in vivo OS tumor model. Ad5-Δ24RGD caused a significant tumor-growth delay of human primary, chemotherapy-resistant, CAR-deficient OS xenografts in nude mice. Histological analysis showed that Ad5-Δ24RGD replicated in OS tumor cells for at least 25 days. However, the CRAd did not completely eradicate the s.c. xenografts. This can best be explained by inefficient dispersion of the CRAd through the tumor mass after intratumoral injection. Barriers within the established tumor, such as connective tissue and tumor matrix, may limit the spread of virus (34). In OS, osteoid calcification could be such a hurdle for viral therapy that needs to be overcome.

Altogether, our data demonstrate that targeting a CRAd toward integrins enhances its cytotoxicity on OS dramatically and warrants further exploration of Ad5-Δ24RGD for its utility in OS treatment. The main challenge in the treatment of OS, however, is the eradication of metastases. Patients with OS do not die from the primary tumor; they die from metastatic disease, typically in the lungs. Notably, we have found that α_vβ3 and α_vβ5 integrins are expressed on OS lung metastases (data not shown). Clearly, intratumoral injection as evaluated in this study is not a feasible option for treatment of lung metastases. However, systemic delivery is now being explored with several viro-therapy agents in clinical trials (35), and, thus, the use of Ad5-Δ24RGD could perhaps be considered for the treatment of OS.

We thank D.C.J. Mastenbroek and M.v.d. Bergh Weerman for their expert technical assistance, as well as K. Suzuki for providing Ad5-Δ24RGD.