A potentially promising treatment of metastatic cancer is the systemic delivery of oncolytic adenoviruses. This requires engineering viruses which selectively replicate in tumors. We have constructed such an oncolytic adenovirus, OAS403, in which two early region genes are under the control of tumor-selective promoters that play a role in two key pathways involved in tumorigenesis. The early region E1A is controlled by the promoter for the E2F-1 gene, a transcription factor that primarily upregulates genes for cell growth. The E4 region is under control of the promoter for human telomerase reverse transcriptase, a gene upregulated in most cancer cells. OAS403 was evaluated in vitro on a panel of human cells and found to elicit tumor-selective cell killing. Also, OAS403 was less toxic in human hepatocyte cultures, as well as in vivo when compared to an oncolytic virus that lacked selective E4 control. A single intravenous injection of 3 × 10 12 vp/kg in a Hep3B xenograft mouse tumor model led to significant antitumor efficacy. Additionally, systemic administration in a pre-established LNCaP prostate tumor model resulted in over 80% complete tumor regressions at a tolerable dose. Vector genome copy number was measured in tumors and livers at various times following tail vein injection and showed a selective time-dependent increase in tumors but not livers over 29 days. Furthermore, efficacy was significantly improved when OAS403 treatment was combined with doxorubicin. This virus holds promise for the treatment of a broad range of human cancers including metastatic disease.

The use of oncolytic adenoviruses as a cancer therapeutic is dependent on the lytic properties of the viral life cycle, and the molecular differences between tumor cells and nontumor cells. One strategy for achieving safe and efficacious adenoviral therapies is to control expression of viral early gene(s) required for replication with tumor-selective promoter(s), particularly those active in a broad range of cancer cells. The retinoblastoma tumor suppressor protein (Rb) pathway is dysregulated in a majority of human cancers. The human E2F-1 promoter has been shown to be selectively activated/derepressed in tumor cells with a defect in the Rb pathway. Ar6pAE2fE3F and Ar6pAE2fF are oncolytic adenoviral vectors (with and without the viral E3 region, respectively) that use the tumor-selective E2F-1 promoter to limit expression of the viral E1A transcription unit, and, thus, replication, to tumor cells. We demonstrate that the antitumor activity of Ar6pAE2fF in vitro and in vivo is dependent on the E2F-1 promoter driving E1A expression in Rb pathway-defective cells, and furthermore, that its oncolytic activity is enhanced by viral replication. Selective oncolysis by Ar6pAE2fF was dependent on the presence of functional E2F binding sites in the E2F-1 promoter, thus linking antitumor viral activity to the Rb pathway. Potent antitumor efficacy was demonstrated with Ar6pAE2fF and Ar6pAE2fE3F in a xenograft model following intratumoral administration. Ar6pAE2fF and Ar6pAE2fE3F were compared with Addl1520, which is reported to be molecularly identical to an E1B-55K deleted vector currently in clinical trials. These vectors were compared in in vitro cytotoxicity and virus production assays, after systemic delivery in an in vivo E1A-related hepatotoxicity model, and in a mouse xenograft tumor model after intratumoral administration. Our results support the use of oncolytic adenoviruses using tumor-selective promoter(s) that are activated or derepressed in tumor cells by virtue of a particular defective pathway, such as the Rb pathway.

Adenovirus serotype 5 (Ad5)-based vectors can bind at least three separate cell surface receptors for efficient cell entry: the coxsackie-adenovirus receptor (CAR), alpha nu integrins, and heparan sulfate glycosaminoglycans (HSG). To address the role of each receptor involved in adenoviral cell entry, we mutated critical amino acids in fiber or penton to inhibit receptor interaction. A series of five adenoviral vectors was prepared and the biodistribution of each was previously characterized in mice. To evaluate possible species differences in Ad vector tropism, we characterized the effects of each detargeting mutation in non-human primates after systemic delivery to confirm our conclusions made in mice. In non-human primates, CAR was found to have minimal effects on vector delivery to all organs examined including liver and spleen. Cell-surface alpha nu integrins played a significant role in delivery of vector to the spleen, lung and kidney. The fiber shaft mutation S*, which presumably inhibits HSG binding, was found to significantly decrease delivery to all organs examined. The ability to detarget the liver corresponded with decreased elevations in liver serum enzymes (aspartate transferase [AST] and alanine transferase [ALT]) 24 hr after vector administration and also in serum interleukin (IL)-6 levels 6 hr after vector administration. The biodistribution data generated in cynomolgus monkeys correspond with those data derived from mice, demonstrating that CAR binding is not the major determinant of viral tropism in vivo. Vectors containing the fiber shaft modification may provide for a detargeted adenoviral vector on which to introduce new tropisms for the development of targeted, systemically deliverable adenoviral vectors for human clinical application.

While 51 human adenoviral serotypes have been identified to date, the vast majority of adenoviral vectors designed for gene transfer have been generated in the adenovirus serotype 5 (Ad5) backbone. Viral infections caused by Ad5 are endemic in most human populations and the majority of humans carry preexisting humoral and/or cellular immunity to Ad5 which may severely limit the use of Ad5-based vectors for gene therapy applications. To circumvent this preexisting Ad5 immunity, we have identified Ad35 as an alternative adenoviral serotype to which the majority of humans do not have neutralizing antibodies. Importantly, Ad35 can be grown to high titers with a low particle-to-PFU ratio. As a prerequisite for the development of Ad35 for use as a gene transfer vector, a genome organization map was constructed using the available Ad35 sequence information, and E1a-deficient Ad35 vectors encoding marker genes were generated. Ad35 biodistribution in mice was assessed following intravenous administration and compared with that of Ad5. Extremely low levels of Ad35 were detected in all organs evaluated, including liver, lung, spleen, and bone marrow, while Ad5 displayed high transduction of these organs. Due to the lack of Ad35 liver tropism, minimal hepatotoxicity was observed in mice treated with Ad35. Furthermore, the half-life of Ad35 in mouse blood was found to be two to three times longer than that of Ad5. These data suggest that either mice do not express the Ad35 cell surface receptor or that Ad35 does not efficiently transduce mouse cells in vivo following systemic delivery. Therefore, to begin to elucidate the Ad35 cell entry mechanisms, in vitro competition studies were performed. These data demonstrated that Ad35 cell entry is CAR independent, and may involve protein(s) expressed on most human cells.

Systemic administration of adenoviral vectors leads to a widespread distribution of vector. Therefore, targeting of adenoviral vectors to specific tissues or cell types will require methods to ablate the normal tropism of the vector simultaneously with the introduction of new receptor specificities. To inhibit native receptor binding, we mutated residues in the AB loop of the adenovirus type 5 (Ad5) fiber. We genetically incorporated the S408E-P409A mutation, referred to as KO1, into the adenoviral genome alone or in combination with an RGD-targeting ligand in the HI loop of fiber. Transduction experiments confirmed that the KO1 mutation results in a significant reduction in fiber-dependent gene transfer on A549 and primary fibroblast cells that could be restored via the RGD-targeting ligand. Competition transduction experiments verified the receptor-binding properties of each vector on A549 and hepatocytes in vitro. Unexpectedly, in mice systemic delivery of the vector containing the KO1 mutation resulted in efficient liver transduction that was localized specifically to hepatocytes. We confirmed these results in three different mouse strains, indicating that hepatic adenoviral gene transfer may be independent of the coxsackievirus-adenovirus receptor and that in vivo retargeting will require further viral capsid modifications to generate a fully detargeted adenoviral vector upon which to introduce new tropisms.

Adenovirus binds to mammalian cells via interaction of fiber with the coxsackie-adenovirus receptor (CAR). Redirecting adenoviral vectors to enter target cells via new receptors has the advantage of increasing the efficiency of gene delivery and reducing nonspecific transduction of untargeted tissues. In an attempt to reach this goal, we have produced bifunctional molecules with soluble CAR (sCAR), which is the extracellular domain of CAR fused to peptide-targeting ligands. Two peptide-targeting ligands have been evaluated: a cyclic RGD peptide (cRGD) and the receptor-binding domain of apolipoprotein E (ApoE). Human diploid fibroblasts (HDF) are poorly transduced by adenovirus due to a lack of CAR on the surface. Addition of the sCAR-cRGD or sCAR-ApoE targeting protein to adenovirus redirected binding to the appropriate receptor on HDF. However, a large excess of the monomeric protein was needed for maximal transduction, indicating a suboptimal interaction. To improve interaction of sCAR with the fiber knob, an isoleucine GCN4 trimerization domain was introduced, and trimerization was verified by cross-linking analysis. Trimerized sCAR proteins were significantly better at interacting with fiber and inhibiting binding to HeLa cells. Trimeric sCAR proteins containing cRGD and ApoE were more efficient at transducing HDF in vitro than the monomeric proteins. In addition, the trimerized sCAR protein without targeting ligands efficiently blocked liver gene transfer in normal C57BL/6 mice. However, addition of either ligand failed to retarget the liver in vivo. One explanation may be the large complex size, which serves to decrease the bioavailability of the trimeric sCAR-adenovirus complexes. In summary, we have demonstrated that trimerization of sCAR proteins can significantly improve the potency of this targeting approach in altering vector tropism in vitro and allow the efficient blocking of liver gene transfer in vivo.

Gutless adenoviral vectors are devoid of all viral coding regions and display reduced cytotoxicity, diminished immunogenicity, and an increased coding capacity compared with early generation vectors. Using hemophilia A, a deficiency in clotting factor VIII (FVIII), as a model disease, we generated and evaluated a gutless vector encoding human FVIII. The FVIII gutless vector grew to high titer and was reproducibly scaled-up from vector seed lots. Extensive viral DNA analyses revealed no rearrangements of the vector genome. A quantitative PCR assay demonstrated helper virus contamination levels of <2%, with the best preparation containing 0.3% helper virus. We compared the gutless vector with an E1/E2a/E3-deficient (Av3) early generation vector encoding an identical FVIII expression cassette following intravenous administration to hemophilia A mice. Gutless vector-treated mice displayed 10-fold higher FVIII expression levels that were sustained for at least 9 months. In contrast, mice treated with the Av3 vector displayed FVIII levels below the limit of sensitivity of the assay at 3 months. Assessment of hepatotoxicity by measuring the serum levels of liver enzymes demonstrated that the gutless vector was significantly less toxic than the Av3 vector at time points later than 7 days. At the highest dose used, both vectors caused a transient 10-fold increase in liver enzymes 1 day after vector administration, suggesting that this increase was caused by direct toxicity of the input capsid proteins. These data demonstrate that the gutless vector displayed increased duration and levels of FVIII expression, and was significantly less toxic than an analogous early generation vector.

ABSTRACT Retroviral vectors for gene therapy are designed to minimize the occurrence of replication-competent retrovirus (RCR); nonetheless, it is possible that a vector-derived RCR could establish an infection in a patient. Since the efficacy of antiretroviral agents can be impacted by interactions between virus, host cell, and drug, five commonly used antiretroviral drugs were evaluated for their abilities to inhibit the replication of a murine leukemia virus (MLV)-derived RCR in human cells. The results obtained indicate that the combination of nucleoside analogs zidovudine and dideoxyinosine with the protease inhibitor indinavir effectively inhibits MLV-derived RCR replication in three human cell lines. In addition, MLV-derived RCR was found to be inherently resistant to the nucleoside analogs lamivudine and stavudine, suggesting that mutations conferring resistance to nucleoside analogs in human immunodeficiency virus type 1 have the same effect even in an alternative viral backbone.

Hemophilia A is the most common severe hereditary coagulation disorder and is caused by a deficiency in blood clotting factor VIII (FVIII). Canine hemophilia A represents an excellent large animal model that closely mimicks the human disease. In previous studies, treatment of hemophiliac dogs with an adenoviral vector encoding human FVIII resulted in complete correction of the coagulation defect and high-level FVIII expression [Connelly et al. (1996). Blood 88, 3846]. However, FVIII expression was short term, limited by a strong antibody response directed against the human protein. Human FVIII is highly immunogenic in dogs, whereas the canine protein is significantly less immunogenic. Therefore, sustained phenotypic correction of canine hemophilia A may require the expression of the canine protein. In this work, we have isolated the canine FVIII cDNA and generated an adenoviral vector encoding canine FVIII. We demonstrate expression of canine FVIII in hemophiliac mice at levels 10-fold higher than those of the human protein expressed from an analogous vector. Canine FVIII expression was sustained above human therapeutic levels (50 mU/ml) for at least 1 year in hemophiliac mice.

While replication-defective retroviral vectors provide excellent vehicles for the long-term expression of therapeutic genes, they also harbor the potential to induce undesired genetic changes by random insertions into the host genome. The rate of insertional mutagenesis for retroviral vectors has been determined in several different assay systems; however, the rate at which such events induce cellular transformation has not been directly determined. Such measurements are critical to determining the actual risk of carcinogenesis resulting from retroviral gene therapy. In this study, the ability of a replication-defective retroviral vector, G1nBgSvNa, to induce cellular transformation in the BALB/c-3T3 in vitro transformation assay was assessed. The transformation frequency observed in vector-transduced BALB/c-3T3 cells, which contained one to six copies of integrated provirus, was not significantly different from that of untreated control cells. The finding that G1nBgSvNa was nontransforming in this assay indicates that the rate of transformation induced by retroviral insertions is less than the spontaneous rate of cellular transformation by BALB/c-3T3 cells, or less than 1.1 x 10(-5). These results are the first to define an upper limit for the rate of transformation induced by retroviral vectors.

The clinical use of retroviral vector producer cells (VPCs) to deliver retroviral vectors efficiently to target cells has been investigated as a method to increase efficiency of gene delivery, presumably as a result of continued vector production in vivo. Studies were conducted in rats to evaluate the distribution of vector to distal organs and tissues as measured by transduction. Rats were treated with two doses of VPCs using two routes of administration: (1) subcutaneous injection, chosen to maximize both the dose and exposure of animals, thereby enabling identification of potential target organs under worst-case conditions; and (2) direct injection into brain parenchyma, chosen to mimic the intended clinical route of administration and provide an estimate of risk to patients receiving this therapy. Twelve organs or tissues were collected 7 days after administration of VPCs and analyzed by PCR for the presence of vector and vector producer cell sequences. Vector was detected most frequently at the site of injection by either route of administration. Less frequently, vector was detected in draining lymph nodes at the higher dose only using either route of injection. Single specimens of lung and contralateral skin were positive for vector following subcutaneous administration only. Vector was detected in gonadal tissue from a single low-dose male following subcutaneous administration, but this finding was not reproduced in any high-dose male or any males injected intracerebrally. In contrast, VPCs were detected only at the site of administration. The frequency of detection of VPCs 7 days after administration was higher when rats were injected by the intracerebral route. Based on these studies, gene transfer to distal organs or gonadal tissue following intracerebral administration of VPCs is not considered to be a risk to patients undergoing retroviral vector gene therapy for the treatment of brain cancer (glioblastoma multiforme; GBM).

A rapid, sensitive assay was developed that can detect the six species of Mycoplasmas that account for the vast majority of cell culture infections. This assay, a modification of the method published by Wong-Lee & Lovett [1], allows direct evaluation of culture medium by a single-step PCR method that utilizes primers complementary to conserved 16S rRNA sequences. Extensive testing of medium from uninfected cultures spiked with purified Mycoplasma DNAs showed that the method described in this report can detect the equivalent of one Mycoplasma in 15 μl culture medium; thus, evaluation of a single culture sample allows detection of Mycoplasmas in cultures infected with the equivalent of 10 or more Mycoplasmas per 15 μl (or ≥6.7×102 Mycoplasma equivalents/ml) with greater than 99.99% confidence. Comparison of results obtained with this PCR-based assay and a standard biological colony-forming assay revealed that the PCR assay is capable of detecting 0.0015-0.015 colony forming units, suggesting that the PCR assay may also be detecting nonviable Mycoplasmas. The high level of amplification achieved with this method allows direct detection of amplification products by ethidium bromide staining of agarose gels, and thus allows rapid screening of cell cultures.

Rapidly progressive T-cell lymphomas were observed in 3 of 10 rhesus monkeys several months after autologous transplantation of enriched bone marrow stem cells that had been transduced with a retroviral vector preparation containing replication-competent virus (R. E. Donahue, S. W. Kessler, D. Bodice, K. McDonagh, C. Dunbar, S. Goodman, B. Agricola, E. Byrne, M. Raffeld, R. Moen, J. Bacher, K. M. Zsebo, and A. W. Nienhuis, J. Exp. Med. 176:1124-1135, 1992). The animals with lymphoma appeared to be tolerant to retroviral antigens in that their sera lacked antibodies reactive with viral proteins and contained 10(4) to 10(5) infectious virus particles per ml. By molecular cloning and DNA sequencing, we have now demonstrated that the serum from one of the monkeys contained a replication-competent retrovirus that arose by recombination between vector and packaging encoding sequences (vector/helper [V/H] recombinant) in the producer clone used for transduction of bone marrow stem cells. Southern blot analysis demonstrated 14 or 25 copies of this genome per cell where present in two animals. The genome of a second replication-competent virus was also recovered by molecular cloning; it arose by recombination involving the genome of the V/H recombinant and endogenous murine retroviral genomes in the producer clone. Twelve copies of this amphotropic virus/mink cell focus-forming virus genome were present in tumor DNA of one animal, but it was not found in tumor DNA of the other two animals with lymphoma. Southern blot analysis of DNA from various tissues demonstrated common insertion site bands in several samples of tumor DNA from one animal, suggesting clonal origin of the lymphoma. Our data are most consistent with a pathogenic mechanism in which chronic productive retroviral infection allowed insertional mutagenesis of critical growth control genes, leading to cell transformation and clonal tumor evolution.