Improvement of Antitumor Activity by Gene Amplification with a Replicating but Nondisseminating Adenovirus

Gene therapy is a promising approach for cancer treatment; however, efficacy of current vectors remains insufficient. To improve the success of suicide gene therapy, we constructed a replication-competent adenoviral vector that has its protease gene deleted and expresses bacterial cytosine deaminase fused with bacterial uracil phosphoribosyltransferase (CU). The prodrug, 5-fluorocytosine, is transformed into the highly toxic and tissue-diffusible 5-fluorouracil by CU in infected cells. This vector is incapable of producing infectious particles but is able to undergo a single round of replication, thereby increasing transgene copy number and expression. In the presence of 5-FC, compared with the first-generation vector (AdCU), the replication-competent vector, Ad(dPS)CU-IRES-E1A, was significantly more efficacious for in vitro tumor cell killing and in bystander assays, whereas 25-fold fewer viral particles were required in a three-dimensional spheroid model. For in vivo experiments, in which virus was injected into preestablished intracranial glioma xenografts, followed by 5-FC treatment, mice receiving Ad(dPS)CU-IRES-E1A had significantly smaller tumors at 35 days postinjection as well as significantly longer median survival than mice treated with the replication-deficient, protease-deleted vector [Ad(dPS)CU]. In an immunocompetent syngeneic model, Ad(dPS)CU + 5-FC-treated mice had a median survival of only 23 days, whereas Ad(dPS)CU-IRES-E1A + 5-FC-treated animals had a survival of 57.1% at 365 days. In conclusion, Ad(dPS)CU-IRES-E1A in the presence of 5-FC produces more potent tumoricidal effects than its replication-deficient counterparts.


Introduction
Gene therapy directed at cancer aims to eradicate tumor cells by restoring tumor-suppressor genes, down-regulating oncogenes, forcing tumor cells into apoptosis, or by directly killing tumor cells with suicide genes (1).However, the efficacy of current approaches remains insufficient (2), limited in part by poor transduction as viral vectors reach only a fraction of the tumor cells.For example, in spheroid models in vitro, we and others have shown that only the superficial cell layers are transduced with adenovirus vectors (3,4).Similarly, a clinical study with an adenovirus vector expressing p53 revealed that only cells in the vicinity of the injection site were transduced (5).Hence, cancer gene therapy strategies are highly dependent on potent bystander effects to target all tumor cells (1).Yet, even with the strong bystander effect provided by the prototypical suicide genes based on Herpes simplex virus thymidine kinase or the Escherichia coli cytosine deaminase, poor in vivo transduction results in high dilution of the toxic product formed by conversion of the prodrugs, explaining in part the poor efficacy obtained in clinical trials (6).To compensate for the small number of transduced cells, a higher number of infectious viral particles are required [in terms of higher multiplicity of infection (MOI)] to improve therapeutic outcome.However, increasing the particle number may not be practical due to increased virus shedding and concomitant systemic toxicity (7)(8)(9).Thus, recent attention has focused on oncolytic vectors to overcome this limitation.In the past few years, several adenoviruses bearing mutations that render their dissemination tumor specific have been described in the literature (10,11).To date, these strategies have not resulted in vectors with sufficient antitumor activity to sustain complete tumor regression, although a few positive instances were observed (10).Poor transduction and poor dissemination in vivo seem to be responsible for this outcome.
In this study, we have explored another approach whereby instead of relying on the dissemination of viral particles as a means of amplification of the therapeutic efficacy (as in oncolytic viruses), we have used genome amplification to increase the number of therapeutic gene copies, thus potentially increasing antitumor activity.The replication-competent platform we chose is a mutant adenovirus vector (12) that is deleted in the protease gene (PS), which is a late protein responsible for capsid protein maturation and viral assembly (13).In the absence of PS, the adenovirus [Ad(dPS)] replicates its DNA normally, but fails to form infectious particles and is incapable of dissemination.
The recombinant adenovirus vectors we generated also express as a fusion protein the suicide gene cytosine deaminase:: Duracil phosphoribosyl transferase (CD:: DUPRT; ref. 14), which converts 5fluorocytosine (5-FC), a prodrug, into 5-fluorouracil (5-FU), a commonly used chemotherapeutic agent that is toxic to dividing cells in its phosphorylated form.The UPRT activity directly phosphorylates 5-FU to form 5-FU-monophosphate (5-FUMP), thereby bypassing the relatively slow phosphorylation of 5-FU in mammalian cells (15).We hypothesized that viral DNA replication would increase the transgene copy number and consequently transgene expression, thereby transforming transduced cells into megafactories for the production of 5-FU and 5-FUMP.
The current work compared the replicating/nondisseminating Ad(dPS)CU-IRES-E1A that expresses CD:: DUPRT to a nonreplicating adenovirus vector also expressing CD:: DUPRT that we have previously characterized in glioblastoma cells (3).Additionally, we tested in vitro a replicating and disseminating adenovirus vector, Ad(PS+)CU-IRES-E1A, in comparison with its nondisseminating counterpart.Furthermore, we have benchmarked in the spheroid model the efficacy of these adenovirus vectors versus oncolysis using the wild-type adenovirus serotype 5 (Ad5wt).Finally, in vivo, we compared the replicating and the nonreplicating viruses in intracerebral glioma models using both human xenografts in athymic mice as well as a syngeneic glioma mouse model.We examined the ability of the viruses to cause tumor regression and to improve overall survival.

Materials and Methods
Cell lines and cell culture.The human glioblastoma cell lines U87 and U251 were obtained from American Type Culture Collection (Manassas, VA).The ovarian cancer cell lines Tov21G and Ov90 were derived from primary human tumors (16).The 293-PS-CymR cell line is a clone derived from the 293 cell line that has been described elsewhere (17).Human cell lines were grown in DMEM supplemented with 5% fetal bovine serum (FBS) and 2 mmol/L L-glutamine.The murine glioma cell line GL261 was obtained from National Cancer Institute-Frederick (Frederick, MD) and grown in RPMI supplemented with 10% FBS.
Construction of adenoviral recombinants.All of the viruses described here were constructed with the AdenoVator system (Q-Biogene, Carlsbad, CA).The suicide gene, composed of the fusion of the CD and the UPRT genes (called CD:: DUPRT or CU), was obtained from InvivoGen (San Diego, CA).The transgene was first subcloned in the shuttle vector pAdenoVator-CMV5(CuO)-IRES-GFP, in which the expression can be silenced in the presence of the cumate repressor (CymR) that binds to the cumate operator (CuO) downstream of the start site. 8Regulation of the transgene expression was necessary because our previous studies had shown that viruses expressing the CD:: DUPRT gene from the strong CMV5 promoter were unstable.The CD:: DUPRT gene was amplified from pGT60-codaupp (InvivoGen), using primers that introduced BglII restriction sites at each end of the fragment.The transgene was subcloned into the BglII site of the shuttle vector and was verified by sequencing.Then, the E1A gene, including its viral poly(A) signal, was inserted in place of the green fluorescent protein (GFP) downstream of the internal ribosome entry site (IRES) to coexpress CD:: DUPRT and E1A.Using purified DNA from Ad5 as template, the E1A gene was amplified with the primers that included NheI and MluI restriction sites at the 5 ¶ and the 3 ¶ ends, respectively: 5 ¶-CCTAGCTAGCATGAGACATAT-TATCTG-3 ¶ and 5 ¶-CGACGCGTCGGAGGTCAGATGTAACCA-3 ¶.The junction between the IRES and E1A was designed so that the ATG of E1A would correspond to the 11th ATG of the IRES as it was shown to be the most efficient ATG (18).This was achieved by replacing the fragment spanning the PmlI to NheI with the PCR product generated with the following primers: 5 ¶-CTAGCTAGCCATGGTTGTGGCAAGCTTATC-3 ¶ and 5 ¶-GCCACGTGTA-TAAGATAC-3 ¶.The final construct was sequenced at the junction of the IRES and E1A to confirm the configuration of the second transgene.The recombinant adenovirus (AdDPS-CuO-CU-IRES-E1A) was generated using the AdEasy system (Q-Biogene) with the AdDPS backbone and following the manufacturer's procedures.
Western blot.Cells were plated in six-well plates at a density of 2.5 Â 10 5 , transduced the next day with the appropriate adenovirus vector at a MOI of 50, and incubated for 72 h.Protein lysates were quantified with the detergent-compatible protein assay (Bio-Rad, Mississauga, Canada) and diluted in Laemmli buffer.Proteins were separated on 10% polyacrylamide gel (Invitrogen, Burlington, Canada) and transferred onto nitrocellulose membranes.Blots were incubated with either the anti-codA (kindly provided by InvivoGen) or an anti-adeno (3), washed, and incubated with secondary antibody coupled to horseradish peroxidase (Amersham, Baie d'Urfe, Canada).Signal detection was done by enhanced chemiluminescence (Amersham).
Cell viability assays.Adenovirus vectors were assessed using a doubling dilution assay.Cells were plated at a density of 5 Â 10 3 per well in 96-well plates.The next day, cells were transduced with doubling dilution of adenovirus vectors starting at a MOI of 10 in 50 AL for 5 h, and then 5-FC was added with 50 AL of medium.Each condition had eight replicates, and each experiment was done twice.Cells were incubated for 6 days and analyzed by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assays.Cells were incubated with MTT (Sigma, Oakville, Canada) at a final concentration of 1 mg/mL for 4 h.Then, medium was removed and DMSO was added.Absorbance was measured at 490 nm.For the experiments with spheroids, the incubation time was extended to 8 h followed by an overnight incubation with DMSO in the dark.
Bystander assays.The bystander assays were done as previously detailed (3), except that each condition had eight replicates and each experiment was done twice.
Spheroids.U87 and U251 spheroids were generated as previously described (3).For Tov21G and Ov90 spheroids, confluent cells were trypsinized and 15 AL of medium containing 4 Â 10 3 cells were carefully dropped on a 150-mm Petri dish cover and inverted back on the Petri dish containing 20 mL of medium.After 1 week, the spheroids were transferred, one per well, in 24-well plates covered with 2% Seaplaque agarose (Cambrex, Rockland, MA).Transduction was done as previously described.The prodrug 5-FC was added the day after transduction at a concentration of 500 Amol/L, unless otherwise specified.
In vivo experimentation.To prepare tumor cells for intracerebral stereotactic injection into the caudate, subconfluent cells were detached in 0.5 mmol/L EDTA, centrifuged, and then resuspended in an appropriate volume of HBSS.U87LacZ human glioma cells (1 Â 10 5 in 3 AL HBSS) were implanted intracerebrally into athymic nude nice, and GL261 murine glioma cells (2 Â 10 5 in 3 AL HBSS) were implanted similarly into wild-type C57B/l6 mice as outlined previously (19).Ten days later, mice were injected intratumorally with virus (3 AL), either Ad(dPS)CU [5 Â 10 10 plaque-forming units (pfu)/mL] or Ad(dPS)CU-IRES-E1A (6 Â 10 9 pfu/mL), by reentering the burr hole created by the initial injection.Some groups of mice received a second injection of virus 21 days after the last virus injection.5-FC was administered i.p. for 5 days at 1 or 2 days after virus injection at a dose of 500 mg/kg twice daily.During long-term survival studies, animals were monitored daily regarding nutrition, hydration, agility, etc. Animals were euthanized if they showed any signs of lethargy, significant neurologic morbidity, or severe weight loss.All experiments were carried out according to the guidelines of the institutional Animal Care Committee.Survival was assessed with Kaplan-Meier analysis.For tumor volume assessment, mice were sacrificed 35 days posttumor implantation.The entire brain was removed and frozen in 2-methylbutane cooled with liquid nitrogen.The brain was cut into 20-Am-thick coronal sections to reconstruct the hgalactosidase-expressing tumors (19) to calculate tumor volumes.The sections were stained histochemically with X-gal, followed by H&E staining.Tumor volumes were calculated using the following formula: where V is the volume, a is calculated as (section number at the end of the tumor À section number at the start of the tumor) Â section thickness, and b is the length of the greatest longitudinal of the tumor.
Analysis of immune cells by flow cytometry.Tumor-bearing mice were anesthetized with 10 mL/kg of somnitol and perfused transcardially with PBS containing 6,000 units of heparin per liter.The forebrain was dissected and placed in HBSS.Samples were mechanically dissociated, passed though a 70-Am filter, then centrifuged.The cell pellets were resuspended and blocked at room temperature for 20 min with 24G2 hybridoma supernatant (anti-Fc receptor) containing 100 Ag/mL rat immunoglobulin, 2% FBS, and 0.01% sodium azide.Fluorescence-activated cell sorting buffer (HBSS, 2% FBS and 0.01% sodium azide) was added to the samples after incubation and then centrifuged.Cell pellets were resuspended and incubated with the following anti-mouse monoclonal antibodies for 1 h at 4jC: FITC-conjugated anti-NK-1.1;R-phycoerythrin-conjugated anti-CD45 (30-F11); peridin chlorophyll protein-Cy5.5-conjugatedanti-CD11b; and allophycocyanin-conjugated anti-TCR-h chain (H57-597).All antibodies were obtained from BD PharMingen (San Jose, CA).Following incubation, samples were washed and cell pellets were resuspended with PBS and analyzed by flow cytometry using a FACSCalibur (Becton Dickinson, San Jose, CA).

Results
Description of the adenovirus genomes.The structures of the adenovirus vectors used in this study are presented in Fig. 1A.The two nonreplicating adenovirus vectors [AdCU and Ad(dPS)CU] consist of the same backbone and differ only in the presence of the PS gene [deleted in yPS (dPS)].These are first-generation adenovirus vectors, deleted in E1 and E3 regions and expressing the CD:: DUPRT fusion gene in a dicistronic cassette with GFP.The two replicating adenovirus vectors [Ad(dPS)CU-IRES-E1A and Ad(+PS)CU-IRES-E1A] express the same CD:: DUPRT fusion gene, but the GFP gene is replaced with the E1A gene, allowing DNA replication.The adenovirus vector containing the PS gene [Ad(+PS)CU-IRES-E1A] can package and disseminate after replication, whereas the dPS adenovirus vector can only replicate but cannot disseminate.Ad5wt was chosen to directly compare the suicide gene approach with oncolytic approaches, as Ad5wt should be as potent if not more so than most oncolytic adenovirus vectors described in the literature because it is not impaired by any mutations.The expression cassette is under the control of the CMV5(CuO) promoter.This promoter allows the repression of transcription during adenovirus vector production in 293 cells expressing the CymR repressor.This was needed for efficient adenovirus vector production. 8The expression cassette expresses the fusion gene CD:: DUPRT and the GFP gene, the latter being under the translational control of an IRES.Ad(dPS)CU is a variant of the AdCU that bears an additional deletion of its PS gene.Ad(dPS)CU-IRES-E1A has the same structure as the Ad(dPS)CU, but the GFP reporter gene was replaced with the E1A gene, rendering this adenovirus vector replicative.Ad(PS+)CU-IRES-E1A differs from the replicating/nondisseminating virus by having the PS gene, thus being both replicating and disseminating.Ad5wt was used in this study to benchmark the suicide gene approach to an oncolysis approach.B, suicide gene expression.C, viral protein expression in tumor cell lines infected with the adenovirus vectors.Tumor cells were plated in six-well plates at a density of 2.5 Â 10 5 , transduced the next day with adenovirus vectors at a MOI of 50, and incubated for 48 h.Blots were incubated with either anti-codA (B) or anti-Ad5 (C).In B, in the lane marked with *, cells were transduced with Ad(dPS)CU.
All the adenovirus vectors were validated by PCR-based assays for the structure of the CD:: DUPRT expression cassette, the deletion of the PS gene in Ad(dPS)CU and Ad(dPS)CU-IRES-E1A, and the presence of E1A genes in both replicating adenovirus vectors (Supplementary Fig. S1A).We also tested whether the presence of the prodrug 5-FC would affect the characteristics of the replicative adenovirus vectors.The addition of 5-FC reduced viral protein expression (Supplementary Fig. S1B) and yield of viral particles produced from Ad(+PS)CU-IRES-E1A by f4-fold (Supplementary Fig. S1C).This suggested that, in this backbone, oncolysis in combination with suicide gene expression may not be optimal.
Transgene expression and viral protein expression.The effect of the various adenovirus vector was evaluated in the human U87 and U251 and the murine GL261 glioma cell lines, as well as in the human Ov90 and Tov21G ovarian cancer cell lines.The human glioma cells were efficiently transduced at low MOIs, whereas the ovarian cancer cell lines and the murine glioma cells required higher MOIs, especially the Ov90 and GL261 cell lines that displayed low transduction even at MOI as high as 200 to 500 (data not shown and Supplementary Table S1).We evaluated by Western blot analysis the expression of the CD:: DUPRT fusion gene produced after transduction of these cell lines with the firstgeneration adenovirus vector and the two replicative adenovirus vectors (Fig. 1B).The AdCU-transduced cells expressed CD:: DUPRT as a single 78 kDa protein, whereas in cells transduced with the replicative adenovirus vectors, the expression of the CD:: DUPRT gene was increased and secondary degradation products appeared.When the expression of the viral proteins was analyzed, only trace amounts of proteins were detectable in cells transduced with the first-generation AdCU (Fig. 1C).In contrast, with both replicative adenovirus vector, high levels of viral proteins were detected, albeit to a lesser extent than with Ad5wt.Furthermore, the level of viral protein expressed was similar between cells transduced with either Ad(dPS)CU-IRES-E1A or Ad(PS+)CU-IRES-E1A, suggesting that the PS deletion did not affect the extent of adenovirus vector genome replication and transgene expression.
Comparison of cytotoxic activities in vitro.To evaluate whether Ad(dPS)CU-IRES-E1A can provide improved cytotoxic activity compared with AdCU, we used doubling dilutions of adenovirus vector to determine the MOI that would reduce cell viability to 50% at a constant concentration of 5-FC (Fig. 2).In the U87 glioblastoma cells, the replicating/nondisseminating Ad(dPS)CU-IRES-E1A was 45 times more potent than its firstgeneration counterpart, requiring infection at a MOI as low as 0.2 to inhibit 50% cell viability, whereas AdCU had to be used at a MOI of 9 to obtain the same result.In the absence of 5-FC, AdCU was not toxic at any of the MOIs tested, whereas in the absence of 5-FC, the replicating/nondisseminating adenovirus vector was as good as AdCU + 5-FC (compare Fig. 2A with B).The MOI required to reduce cell viability to 50% was compared for all the adenovirus vector in the various cell lines (Fig. 2C).The glioblastoma U87 cells and the ovarian cancer Tov21G cells were the most responsive, Figure 2. Cytotoxicity assays.Cells were plated at a density of 5 Â 10 3 per well in 96-well plates.The next day, cells were transduced with doubling dilutions of adenovirus vectors starting at a MOI of 10 for 5 h, and then 5-FC (25 Amol/L for U87 and 500 Amol/L for U251, Tov21G, and Ov90) was added.Cells were incubated for 6 d and analyzed by MTT assays.Curves obtained with the U87 cells (A, with 5-FC; B, without 5-FC).C, evaluation of the initial MOI required to produce 50% viability.Data were obtained from curves displayed as in (A ) and generated for each cell line.Each dilution had eight replicates and each experiment was done twice.
whereas U251 and Ov90 were less responsive due to moderate sensitivity to 5-FU in the case of the U251 cells and poor transduction in Ov90 (data not shown Supplementary Table S1).No differences were observed between Ad(dPS)CU and AdCU, confirming that the deletion of the PS gene had no effect on the cytotoxic activity of the vector.Moreover, when the replicating/ nondisseminating Ad(dPS)CU-IRES-E1A was compared with its PS+ counterpart, Ad(PS+)CU-IRES-E1A, the cytotoxic activity showed a modest improvement of 2-fold (Fig. 2C), demonstrating only poor oncolytic activity by Ad(PS+)CU-IRES-E1A.
Bystander activity of the replicating/nondisseminating adenovirus vector.To achieve clinical success, a strong bystander activity is of paramount importance.Therefore, we evaluated whether the benefits observed at low titers of virus with the Ad(dPS)CU-IRES-E1A would also translate into an improvement in a bystander assay.U87 glioblastoma cells transduced with either MOI of 10 or 100 were mixed at different with nontransduced (bystander) cells.Glioblastoma cells transduced with MOI of 10 of Ad(dPS)CU-IRES-E1A were more potent at reducing cell viability of the entire population than cells transduced with MOI of 100 of AdCU (Fig. 3A).When a MOI of 10 was used for both adenovirus vectors, the replicating/nondisseminating virus was 14 times better than AdCU (3% cells versus 42%) for killing 50% of the U87 cell population (Fig. 3B).Furthermore, similar results were obtained with the U251 (11% cells versus >100%) as well as the two ovarian cancer cell lines Tov21G (5% cells versus >100%) and Ov90 (2% cells versus 75%; Fig. 3B).Although at a MOI of 10, the first-generation AdCU did not display any bystander activity in U251, Tov21G, and Ov90 cells, with MOI of 100 bystander activity was observed at 11% in U251 and 12% in Tov21G (data not shown).
Comparison of cytotoxic activities in spheroid models.Monolayer cell cultures do not reproduce the three-dimensional structure of tumors.In cancer gene therapy, accessibility of cells to the viral vector is very important and specifically for adenovirus vector, transduction efficiency varies greatly between cells in monolayers and in spheroids (three-dimensional structures).We have previously shown that in spheroid cultures, only superficial cell layers are transduced with the first-generation AdCU, which coexpresses the GFP protein (3).Therefore, we tested the antitumoral activities of the various adenovirus vectors in U87 spheroids.Both the replicating/nondisseminating and the replicating/disseminating virus reduced viability close to or below 50%, respectively, at a lower viral dose (particle number) than the Figure 3. Analysis of bystander effect.Cells were plated at a density of 2 Â 10 5 in 12-well plates.The next day, they were transduced with adenovirus vectors at MOIs of 10 and 100.After 4 h, transduced cells were trypsinized and resuspended in 2 mL of medium.Doubling dilutions of cells were generated and plated into 96-well plates in 50 AL.Then, nontransduced cells (bystander cells) were trypsinized, diluted to 1 Â 10 5 /mL, and 50 AL were added to the wells containing the transduced cells.These dilutions were done with 5-FC (25 Amol/L for U87 and 500 Amol/L for U251, Tov21G, and Ov90).This resulted in the following ratios of transduced cells: 50%, 33%, 20%, 11%, 6%, 3%, 1.5%, 0.8%, 0.4%, and 0.2%.The cells were incubated for 6 d and analyzed by MTT assays.A, curves obtained with the U87 cells.B, evaluation of the IC 50 of transduced cells.Values for U87, U251, Tov21G, and Ov90 were obtained from curves using MOI of 10 as displayed in (A).Each condition had eight replicates and the experiment was done twice.first-generation AdCU (Fig. 4A).Furthermore, viability was still significantly reduced to levels around 50% even when the number of infectious particles of the replicating adenovirus vectors was decreased by 5-fold, suggesting that the number of viral particles of replicating adenovirus vectors can be reduced by 25-fold while still matching the performance of the first-generation AdCU (Fig. 4A).
None of the adenovirus vectors tested were successful at significantly reducing the viability of U251 spheroids (Fig. 4B), most probably due to the resistance of these p53 À cells to 5-FU (3).On the other hand, despite their poor transduction, Ov90 spheroids responded to treatment with 5 Â 10 6 pfu of the replicative adenovirus vectors Ad(dPS)CU-IRES-E1A and Ad(PS+)CU-IRES-E1A but not with the first-generation AdCU (Fig. 4B).This further suggests that the higher efficacy of the replicative platform can compensate for poor transduction.
Comparison of cytotoxic activities in established xenograft models.As a first step to examining if Ad(dPS)CU-IRES-E1A is capable of causing tumor regression in vivo, we assessed tumor volume in athymic nude mice implanted intracerebrally with U87LacZ human glioma cells and injected intratumorally 10 days later with virus and treated with 5-FC (500 mg/kg twice daily given i.p.).At 35 days posttumor implantation, Ad(dPS)CU-IRES-E1Atreated mice (n = 6) had significantly smaller tumors (19.8 F 18.8 mm 3 ) than animals (n = 5) that received only 5-FC (159.6 F 52.30 mm 3 ; P = 0.0002, two-tailed t test).
To determine whether these results would translate into an extended life span for tumor-bearing mice, we did long-term survival studies.We first evaluated the best dosing regimen for 5-FC given the cytotoxic capacity of the replicating/nondisseminating virus alone (Fig. 2A).To allow the transduced cells to convert as much 5-FC as possible to accumulate the toxic 5-FU, before the virus itself eliminates the tumor cells, 5-FC was administered at either 1 or 2 days after virus injection.These studies revealed that survival of mice treated with 5-FC 1 day after virus injection with Ad(dPS)CU-IRES-E1A resulted in a significant increase in median survival compared with the cohort that was treated 2 days after virus injection (P = 0.0005).
In survival studies, we tested the effect of providing two injections of virus spaced 21 days apart, in which the first injection is administered 10 days after tumor implantation.5-FC was given the day after virus injection in both injection periods.Mice treated with Ad(dPS)CU-IRES-E1A and 5-FC had a much greater median survival of 117 days (n = 15) than any group tested (Fig. 5A).Ad(dPS)CU plus 5-FC-treated animals only had a median survival of 76 days, which is similar to the median survival obtained after a single virus injection (data not shown).Treatment with Ad(dPS)CU-IRES-E1A alone without 5-FC led to a small improvement in survival (median survival of 60.5 days) compared with untreated controls and mice treated with 5-FC only (median survivals of 47 and 53 days, respectively).Due to the capacity of Ad(dPS)CU-IRES-E1A to further prolong survival after a second injection compared with Ad(dPS)CU, the results support the idea
To assess if the immune system can be a contributor to the effect observed with Ad(dPS)CU-IRES-E1A, a syngeneic glioma model was used by implantation of GL261 murine glioma cells in wildtype C57Bl/6 mice.Adenovirus vector-mediated gene transfer is quite poor in these cells, with fewer than 1% of cells being transduced with an adenovirus vector lacZ recombinant virus at a MOI of 500 (data not shown; Supplementary Table S1).Nevertheless, infection at a MOI of 50 with Ad(dPS)CU-IRES-E1A decreased cell viability to 50% in the presence of 1.4 Amol/L of 5-FC, indicating that GL261 can be sensitized to killing with 5-FC and suggesting that the majority of the cell killing that occurs after addition of 5-FC is achieved through a bystander effect.
The same protocol as described above was used to evaluate the effect of Ad(PS)CU and Ad(dPS)CU-IRES-E1A (Fig. 5B).In this model, untreated mice have a median survival of 23 days, whereas mice treated with 5-FC alone have a median survival of 27 days.Ad(dPS)CU-IRES-E1A-injected animals treated with 5-FC had a dramatically higher survival rate.In fact, 365 days after tumor implantation, when the experiment was terminated, this treatment group had a survival rate of 57.1%.In contrast, animals treated with Ad(dPS)CU-IRES-E1A alone and mice treated with Ad(dPS)CU with 5-FC had median survivals of only 24 and 27 days, respectively, with survival curves very similar to control groups.We compared the extent of immune cell infiltration by flow cytometry, analyzing and quantitating T cells, macrophages, and natural killer (NK) cells in the brains of mice treated with the adenovirus vector and 5-FC at different times after adenovirus vector injection (3, 5, and 10 days).As shown in Fig. 6, for the 10-day time point, Ad(dPS)CU-IRES-E1A elicited greater macrophage and T-cell infiltration, which may be a major reason for its superior efficacy in this immunocompetent model.

Discussion
Adenoviral vectors of serotype 5 are among the most widely used vectors for gene transfer in cancer clinical trials (20).These studies have identified factors that have a critical effect on the eventual therapeutic response.These include the efficiency of initial transduction of the tumor cells, which is dependent on the presence of the appropriate viral receptors; the penetration of the tumor mass by the viral vector, which is dependent on the dose (or titer) that is injected; and the targeting of metastases through systemic delivery, which is dependent on the transit time and resistance to inactivation in the circulatory system.Most primary tumors as well as tumor cell lines express very low levels of the Coxsackie and adenovirus receptor (CAR), the primary receptor for Ad5 (21)(22)(23)(24).The av-containing integrins, the low-affinity secondary receptors (25), which are up-regulated in many tumors (26), can somewhat overcome this limitation when higher titers are delivered to the tumor mass.Higher titers may also make up for the poor diffusion characteristics of the adenoviral particle.Hence, it was hypothesized that oncolytic adenoviral vectors would increase efficiency by providing a means to amplify virus in situ in a tumor-specific fashion (10).However, it has become apparent that barriers also exist for oncolytic adenoviruses.First, transduction is still limited by paucity of CAR on the neighboring, uninfected tumor cells, making them refractory to transduction or at least less susceptible, so that the need for higher doses may not be circumvented (27).Furthermore, if the oncolytic replication rate does not keep up with tumor cell proliferation rate, the MOI is effectively reduced.Also, there may be other barriers to dissemination of virus to neighboring cells (28,29).
A more recent approach has been the ''arming'' of oncolytic viruses with transgenes that have antitumoral activity (30), with the most commonly used ones being based on suicide genes (31)(32)(33)(34)(35).Although this has been somewhat successful in that increased efficacy has been obtained in some tumor models (36), but not in others (37), it has been reported that the converted metabolites of prodrugs are detrimental to oncolytic activity (38) as the viruses do not seem to replicate as well in the presence of agents that affect cell division.We made similar observations in that when the tumor cells were transduced with the replicative adenovirus vectors, the level of viral protein expressed was reduced in the presence of the prodrug (Supplementary Fig. S1B).Moreover, the obtained titers of the Ad(PS+)CU-IRES-E1A were reduced also by 4-fold in the presence of the prodrug (Supplementary Fig. S1C).This suggests that the combination of a suicide gene and oncolysis may not result in much improvement.
In this study, we evaluated and characterized the antitumor activity of the CU-expressing adenoviral vector Ad(dPS)CU-IRES-E1A, deleted in the PS gene, that is capable of one round of replication but is deficient in packaging.Our data indicate that this replicating/nondisseminating virus is more potent in tumor cell killing than its first-generation adenoviral counterpart as well as Ad5wt.The single round of replication increases the effective genome copy number in every infected cell, leading to the increased levels of the transgene (CU) as shown in all cell lines tested, including the murine glioma cell line, GL261 (Fig. 1B).
[Although mouse cells are only semipermissive to adenovirus infection due to several blocks in the late phase of the lytic cycle, early genes are expressed, including those of the E2 region encoding the proteins that are required for genome amplification, and, as a result, transgene expression is also increased in GL261 cells.]In practical terms, compared with the first-generation vector, much lower MOI was required to reduce cell viability by 50% for the same concentration of 5-FC (Fig. 2).The potent bystander effect of the replicating-nondisseminating vector is another important contributor to its cytotoxic capacity.The bystander effect is a significant consideration in gene therapy trials in that it is difficult to achieve 100% transduction within a tumor mass or in animal models with preestablished tumor (as in the models used here).Using Ad(dPS)CU-IRES-E1A, only 3% of cells transduced with a MOI of 10 can reduce viability of U87 to 50% in the presence of 25 Amol/L 5-FC (which compares favorably with the serum levels achieved in patients; ref. 39), indicating a considerable bystander effect when compared with AdCU, which required 42% (Fig. 3).The in vitro bystander effect is most evident in the three-dimensional spheroid assays, which most closely reproduce the structure of an actual tumor.As few as 2 Â 10 5 pfu of Ad(dPS)CU-IRES-E1A could elicit a reduction in viability of the U87 spheroid, a dose that was much lower than that required for the first-generation vector or for oncolysis effected by Ad5wt (Fig. 4).Cell death did not increase much with increasing dosage, presumably because only the superficial cell layers are transduced (3), and the 5-FC/5-FU diffuses poorly through the layers.The in vivo studies with the human glioma xenografts in the athymic mice validated these in vitro results, demonstrating that Ad(dPS)CU-IRES-E1A is capable of penetrating the three-dimensional mass of a preestablished tumor and of providing better suicide gene activity than the first-generation vector (Fig. 5).Importantly, in both in vivo models, E1A expression (in the absence of 5-FC) had modest or no antitumoral effects.
Because the immune system is a major contributor to the in vivo bystander effect (40)(41)(42), we evaluated the antitumoral activity of Ad(dPS)CU-IRES-E1A in a syngeneic glioma model.It is also possible that the immune system could hinder adenoviral infection by rapid clearance of the vector.In fact, Ad(dPS)CU-IRES-E1A + 5-FC was much more potent in the syngeneic model than in the athymic mouse model, with 57% cure rate (as defined by proportion of animals alive and healthy at 365 days after tumor implantation).This was all the more remarkable because GL261 cells are poorly transducible in vitro (<1% at MOI of 500 with AdCMVlacZ).Flow cytometry analyses of immune cell profiles at short periods after adenovirus administration showed that macrophages and T cells may contribute differentially to tumor regression in the Ad(dPS)CU-IRES-E1A group compared with the Ad(dPS)CU.Most importantly, these syngeneic mice did not show any overt side effects after vector administration: All mice that required euthanasia (f40% of treated animals) had large tumors and no indication of extensive inflammation (data not shown).
For gliomas, the intracerebral xenograft model, compared with a flank s.c.model, better approximates the in vivo situation in terms of gene expression induced by the implantation (43).Also, because clinical trials in glioblastoma multiforme have used mainly intratumoral injection of adenoviral vectors (5,44,45), the intracerebral tumor treatment also takes into consideration the limited amount of viral particles that can be delivered stereotactically.Hence, we opted to target intracerebral tumors in athymic and immunocompetent mice and observed the animals for a long period of follow-up (1 year).The response obtained in the athymic mice with the U87 xenografts compares very favorably with that reported for the conditionally replicative adenovirus vector Delta24, showing median survival of 117 days as opposed to 50 days in Delta24-injected mice (46) or 74 days in the ''armed'' Delta24-hyCD expressing the humanized form of the yeast CD gene (47).Remarkably, in the more invasive, syngeneic GL261 model (48), over 57% of the treated animals survived up to 1 year and showed no tumors when the experiment was terminated.These in vivo data underscore the potential this approach may have in a clinical setting.

Figure 1 .
Figure 1.Characterization of adenovirus vectors used in this study.A, schematic representation.AdCU vector is an adenovirus deleted in the E1 and E3 regions:The expression cassette is under the control of the CMV5(CuO) promoter.This promoter allows the repression of transcription during adenovirus vector production in 293 cells expressing the CymR repressor.This was needed for efficient adenovirus vector production.8The expression cassette expresses the fusion gene CD:: DUPRT and the GFP gene, the latter being under the translational control of an IRES.Ad(dPS)CU is a variant of the AdCU that bears an additional deletion of its PS gene.Ad(dPS)CU-IRES-E1A has the same structure as the Ad(dPS)CU, but the GFP reporter gene was replaced with the E1A gene, rendering this adenovirus vector replicative.Ad(PS+)CU-IRES-E1A differs from the replicating/nondisseminating virus by having the PS gene, thus being both replicating and disseminating.Ad5wt was used in this study to benchmark the suicide gene approach to an oncolysis approach.B, suicide gene expression.C, viral protein expression in tumor cell lines infected with the adenovirus vectors.Tumor cells were plated in six-well plates at a density of 2.5 Â 10 5 , transduced the next day with adenovirus vectors at a MOI of 50, and incubated for 48 h.Blots were incubated with either anti-codA (B) or anti-Ad5 (C).In B, in the lane marked with *, cells were transduced with Ad(dPS)CU.

Figure 4 .
Figure 4. Cytotoxic activity in spheroids.A, U87 spheroids were transduced with 1 Â 10 6 , 5 Â 10 6 , or 2 Â 10 5 pfu and treated with 500 Amol/L of 5-FC.At each dose, the shown viruses were used.The diameters of the U87 spheroids were f1 mm.Spheroids were incubated for 10 d after the treatment and analyzed by the MTT assay.NA, not assessed.B, U251 spheroids were treated with 5 Â 10 6 pfu and 500 Amol/L of 5-FC on days 0, 3, and 6.Viability was assessed on day 10.The diameters of these spheroids were f0.5 mm.Ov90 spheroids were treated with 5 Â 10 6 pfu and 500 Amol/L of 5-FC on day 0 only, and viability was assessed on day 10.Those spheroids had a diameter of f1 mm.

Figure 5 .
Figure 5. Kaplan-Meier survival curves.U87lacZ xenografts in athymic mice (A) and GL261 mouse glioma cells in C57/Bl6 mice (B).Animals were implanted with tumor cells and 10 d later were injected with the various viral recombinants, followed 24 h later by 5-FC treatment as described in Materials and Methods.Mice were monitored daily and were euthanized if they showed any signs of lethargy, significant neurologic morbidity, or severe weight loss.The experiment was terminated at 365 d after tumor implantation.

Figure 6 .
Figure 6.Quantitation of immune cells in the brain of tumor-bearing animals by flow cytometry analysis.C57Bl/6 mice were implanted with GL261 cells and 10 d later were injected with either Ad(dPS)CU or Ad(dPS)CU-IRES-E1A, followed 24 h later by 5-FC treatment.Ten days after viral recombinant injection, animals were euthanized and brain tissue was dissociated for immunostaining as described in Materials and Methods.Populations of immune cells (macrophages, NK cells, and T cells) were identified and quantitated by flow cytometry.Statistical significance was determined by t test (two-tailed).*, P < 0.05.