Enhanced antitumor activity of P450 prodrug-based gene therapy using the low K_m cyclophosphamide 4-hydroxylase P450 2B11

Gene-directed strategies that increase the chemosensitivity of tumor cells by stimulating tumor cell-catalyzed activation of anticancer prodrugs offer a unique opportunity to improve the efficacy of cancer chemotherapeutic agents (1–3). These gene-directed enzyme prodrug therapies (GDEPT) use a variety of anticancer prodrugs, including agents activated by bacterial, viral, and mammalian enzymes. Tumor-selective delivery of a prodrug activation gene can be achieved in a variety of ways, including the use of viral vectors that incorporate tumor-specific promoters (4) or selectively replicate in tumor cells that present specific genetic defects, such as p53 deficiency (5), or are characterized by pathophysiologic conditions, such as hypoxia (6). One GDEPT strategy uses genes that encode mammalian cytochrome P450 enzymes (7, 8), which metabolize anticancer prodrugs, including the oxazaphosphorines cyclophosphamide and ifosfamide, to active, cytotoxic metabolites (9, 10). Prodrug-activating P450 enzymes are expressed at high levels in liver but are generally deficient or present at very low levels in human tumors (11–14). Consequently, substantial increases in intratumoral production of active drug metabolites and enhanced anticancer activity may be realized by delivery of prodrug activation P450 genes to tumor cells (9, 15, 16).

Cyclophosphamide is commonly given in combination with other chemotherapeutic drugs in both adjuvant and high-dose chemotherapy settings and shows considerable clinical activity against a broad range of tumors, including breast cancer, endometrial cancer, and various leukemias and lymphomas (17). Preclinical P450 GDEPT studies using cyclophosphamide have shown a substantial improvement in antitumor activity associated with a decrease in systemic toxicity compared with conventional chemotherapy. An important feature of this P450-based gene therapy is its use of established, clinically effective anticancer prodrugs, such as cyclophosphamide, whose active 4-hydroxy-metabolite exerts bystander cytotoxicity that is not limited to immediately adjacent tumor cells (8, 18). Further enhancement of P450 GDEPT can be obtained by coexpression of P450 with the flavoenzyme NADPH P450 reductase (P450R; refs. 18, 19), by combination of cyclophosphamide with other prodrugs, including P450/P450R-activated bioreductive drugs (20, 21), and by administration of cyclophosphamide using a metronomic schedule (22), which is associated with antiangiogenic activity (23). P450 gene transfer can be achieved both in cell culture and in human xenograft models by using tumor cell replication-conditional herpesvirus (24, 25) or adenovirus vectors (26) to effect efficient, tumor-selective P450 gene delivery. Recent phase I/II clinical trials support the clinical development of P450 GDEPT based on studies using cellular vectors to deliver P450 2B1 to advanced pancreatic cancer patients treated with ifosfamide (27, 28) and retroviral vectors to deliver P450 2B6 to advanced breast cancer and melanoma patients given oral cyclophosphamide (29).

P450 enzymes, such as P450 2B1 and P450 2B6, metabolize many drug substrates, including cyclophosphamide and ifosfamide, with a high K_m, in the millimolar range (30, 31). High drug concentrations are therefore required for efficient prodrug activation and bystander tumor cell killing, limiting the utility of P450 GDEPT. Enhanced enzyme activity may potentially be achieved by using a low K_m P450 enzyme, which could lead to increased activity at the sub-K_m plasma levels of cyclophosphamide and ifosfamide typically seen in cancer patients (32, 33). We have recently identified a canine enzyme, P450 2B11, as a low K_m catalyst of cyclophosphamide and ifosfamide activation whose kinetic properties in an in vitro reconstituted enzyme system (K_m, 80–160 μmol/L; ref. 34) are far superior to those of P450 2B6 (K_m, ∼1,900 μmol/L; ref. 31). It is not clear, however, whether the low K_m of P450 2B11 will be manifest in the environment of an intact cell or whether a low K_m will translate into improved GDEPT activity in an animal tumor model in vivo.

Presently, we investigate whether efficient activation of cyclophosphamide and ifosfamide and effective tumor cell killing can be achieved at micromolar prodrug concentrations by introduction of P450 2B11 in combination with P450R and whether further increases in antitumor activity can be obtained by combining a replication-defective adenoviral vector encoding P450 2B11 and P450R with Onyx-017 (26), a replication-conditional oncolytic adenovirus. We also evaluate the effect of retroviral expression of P450 2B11 in a scid mouse tumor xenograft model and, in the context of an intact liver, on the therapeutic responses to cyclophosphamide compared with those seen with P450 2B6. Our findings establish that P450 2B11, with its low K_m and high catalytic efficiency, has the potential to substantially improve the utility of P450 GDEPT applications for cancer treatment.

Cyclophosphamide, chloroquine, and puromycin were purchased from Sigma Chemical Co. (St. Louis, MO). 4-OH-cyclophosphamide and 4-OH-ifosfamide, chemically activated derivatives of cyclophosphamide and ifosfamide, respectively, were obtained from Dr. Ulf Niemeyer (Baxter Oncology GmbH, Frankfurt, Germany). Fetal bovine serum (FBS) and DMEM were purchased from Life Technologies (Grand Island, NY). The conditional-replicating adenovirus Onyx-017 (26) was obtained from Onyx Pharmaceuticals, Inc. (Richmond, CA). Human tumor cell lines U251 (brain), PC-3 (prostate), and MCF-7 (breast) were obtained from Dr. Dominic Scudiero (National Cancer Institute, Bethesda, MD; ref. 35). MCF-7/4HC cells, which display ∼6-fold resistance to activated cyclophosphamide (4-OH-cyclophosphamide), are described elsewhere (36). Human tumor cells were grown at 37°C in a humidified, 5% CO₂ atmosphere in RPMI 1640 containing 5% FBS, 100 units/mL penicillin, and 100 μg/mL streptomycin. Rat 9L gliosarcoma cells were grown in DMEM containing 10% FBS and antibiotics.

The retroviral plasmids pBabe-2B6-P450R-puro (26), pBabe-2B1-P450R-puro, and pBabe-2B11-P450R-puro are based on the pBabe series (37), which contains a puromycin resistance gene transcribed from an internal SV40 promoter. These retroviruses generate a bicistronic mRNA transcribed from the 3′-long terminal repeat of the viral vector, which encodes the indicated P450 2B cDNA linked to human P450R cDNA by an internal ribosome entry signal (IRES). pBabe-2B1-P450R-puro was constructed by excising the rat P450 2B1 coding sequence from pBabe-2B1-puro (38) using EcoRI and ApoI followed by ligation of the fragment to the EcoRI site of pBabe-IRES-P450R-puro (26). pBabe-2B11-P450R-puro was constructed by excising the P450 2B11 coding sequence, including 147 bp of 3′-untranslated region sequence, from plasmid pCD39 (Dr. James R. Halpert, University of Texas Medical Branch, Galveston, TX; ref. 39) using EcoRI and ApoI followed by subcloning into the EcoRI site of pBabe-IRES-P450R-puro.

Transfection of the ecotropic packaging cell line Bosc 23 (40) with the above-described pBabe-2B-P450R-puro plasmids, harvesting of retroviral supernatants, and infection of rat 9L gliosarcoma cells were carried out as detailed elsewhere (18, 41), with the following modifications. Bosc 23 cells were plated at 2.5 × 10⁶ in a 60-mm dish. Fresh medium (4 mL) containing chloroquine (25 μmol/L) was added to the cells 16 hours later. The cells were cotransfected 1 hour later with 40 μg pBabe-derived plasmid DNA and 7.5 μg pKat (42) using a calcium phosphate precipitation method. Pools of cells resistant to puromycin (2.5 μg/mL) were selected beginning 48 hours later over a 2- to 3-day period (18) with an overall survival rate of >80%. The resultant drug-resistant pools of 9L/P450 cells were propagated, lysed by sonication, and then assayed for P450R-catalyzed, NADPH-dependent cytochrome c reduction (ΔA₅₅₀) at 30°C in 0.3 mol/L potassium phosphate buffer (pH 7.7) containing 0.525 mg/mL cytochrome c. 9L cells infected with empty pBabe retrovirus (9L/pBabe cells, also called 9L control cells) served as P450-deficient controls. P450 protein expression was verified in each 9L/P450 cell pool by Western blot analysis of 50 μg sonicated cell lysate protein using rabbit anti-P450 2B1 polyclonal antibody. Lymphoblast-expressed P450 2B1 and P450 2B6 (BD-Gentest, Inc., Woburn, MA) and bacterial-expressed P450 2B11 (34) were used as standards.

The chemosensitivity of 9L/P450 cells was determined using a 4-day growth inhibition assay. Cells were plated overnight in triplicate at 4,000 per well of a 48-well plate. The cells were then treated for 4 days with cyclophosphamide at concentrations specified in each experiment. Cells remaining at the end of day 4 were stained and quantified using a crystal violet/alcohol extraction assay (18). Data are expressed as cell number (A₅₉₅) relative to drug-free controls, mean ± SD for triplicate samples, unless indicated otherwise. Effective drug concentrations for 50% response (EC₅₀) were determined based on incubations with cyclophosphamide or ifosfamide (0–120 μmol/L for 9L/2B11 cells and 0–1,000 μmol/L for 9L/2B1 and 9L/2B6 cells). Sigmoidal dose-response analysis using a variable slope was done using the nonlinear regression curve fit program of GraphPad Prism (GraphPad Software, San Diego, CA). Data were plotted as percentage of cells killed versus log₁₀ of the molar concentration of cyclophosphamide and ifosfamide.

Apparent K_m and V_max values for 9L/P450 cell-catalyzed cyclophosphamide and ifosfamide 4-hydroxylation were determined as follows. Cells (1 × 10⁵) were seeded in 12-well plates and grown for 48 hours. Cells were treated for 4 hours with 0 to 6 mmol/L cyclophosphamide or ifosfamide in 2 mL culture medium containing 5 mmol/L semicarbazide to trap the 4-hydroxy-metabolite.

Aliquots of medium (500 μL) were removed and treated with ZnSO₄ and Ba(OH)₂ to precipitate protein and then derivatized to 7-hydroxyquinoline by treatment for 25 minutes at 90°C with 100 μL freshly prepared fluorescent reagent (60 mg 3-aminophenol and 60 mg hydroxylamine-HCl in 10 mL of 1 N HCl). Samples were analyzed by high-performance liquid chromatography (HPLC) and quantified by comparison with standard curves of metabolite prepared using 4-OOH-cyclophosphamide dissolved in culture medium (43, 44). In some cases, the derivatized samples were stored at −20°C in the dark before HPLC analysis.

Mixtures of wild-type 9L cells and 9L/P450 cells containing increasing percentages of 9L/2B6, 9L/2B1, or 9L/2B11 cells (0%, 5%, 10%, 20%, and 100% of total cells) were plated overnight in 1 mL culture medium to give a total of 10⁵ cells per well of a 12-well plate. The cells were then treated with cyclophosphamide for a total of 96 hours, with the cyclophosphamide-containing culture medium changed after the first 48 hours. Cells were then stained and quantified using the crystal violet assay described above.

Adeno-2B6, a replication-defective, E1/E3 region-deleted adenovirus encoding P450 2B6 linked to P450R via an IRES sequence, was described previously (26). A corresponding adenovirus expressing P450 2B11 and P450R (Adeno-2B11) was constructed using the Adeno-X expression system (Clontech Labs, Inc., Palo Alto, CA). Retroviral plasmid pBabe-2B11-P450R-puro was linearized using EcoRI, blunt ended, and the P450 2B11-IRES-P450R expression cassette was then excised with DraI and ligated into the DraI site of pShuttle (Clontech Labs). The resulting plasmid, pShuttle-2B11-P450R, was used to construct the recombinant adenovirus, Adeno-2B11, as described by the manufacturer. The cytomegalovirus immediate-early promoter of pShuttle drives Adeno-2B11-directed expression of P450 2B11 and P450R and ensures robust expression in a wide range of cell lines and tissues. The functionality of Adeno-2B11 was initially determined by assaying P450R activity (cytochrome c reduction) in cell lysate prepared from Adeno-2B11-infected human kidney HEK 293 cells. Typically, a 4-fold increase in P450R activity was observed 48 hours postinfection when compared with cells infected with a control replication-defective adenovirus, adeno-PmeI (constructed with the empty pShuttle vector of Adeno-X system).

Adenoviral stocks were propagated in HEK 293 cells grown in 100-mm dishes at 37°C in a humidified 5% CO₂ incubator in DMEM containing 25 mmol/L glucose, 100 units/mL penicillin, 100 μg/mL streptomycin, and 10% heat-inactivated FBS (26). The purified virus was desalted by dialysis at 4°C against 10 mmol/L Tris (pH 8.0), 1 mmol/L MgCl₂, and 10% glycerol. The effectiveness of desalting was verified by conductivity measurements. Viral titers were determined as described in the Adeno-X Rapid Titer kit manual (Clontech Labs) and typically ranged from 10¹⁰ to 10¹¹ plaque-forming units/mL for both adeno-P450 viruses. Purified virus was stored in aliquots at −80°C.

U251 cells were seeded overnight in six-well plates at 10⁵ per well and then infected for 3 to 4 hours with Adeno-2B6 or Adeno-2B11 in 1 mL RPMI 1640 containing 5% FBS at multiplicities of infection (MOI) of 25, 50, or 100 plaque-forming units/cell. Each adeno-P450 virus was used at MOI 12 in experiments involving coinfection with the tumor cell-replicating adenovirus Onyx-017 (26). Fresh culture medium (2 mL) was then added without changing the medium, and the incubation with virus was continued for ∼20 hours. The cells were then changed to virus-free culture medium followed by an additional 24 hours incubation. The cells were then either (a) washed with PBS and lysed by sonication in 50 mmol/L potassium phosphate buffer (pH 7.4) containing 1 mmol/L EDTA and 20% glycerol and the lysates were assayed for P450R activity as described above for 9L cells or (b) the cells were treated for 4 hours with 100 or 250 μmol/L cyclophosphamide in 3 mL fresh RPMI 1640 containing 5% FBS and 5 mmol/L semicarbazide. Aliquots of culture medium (500 μL) were then removed and assayed for 4-OH-cyclophosphamide by HPLC as described above.

Cell lines used for cytotoxicity experiments (U251, PC-3, MCF-7, and MCF-7/4HC) were seeded in 24-well plates at 14,000 cells per well and allowed to establish growth overnight. The cells were then infected for 3 to 4 hours with Adeno-2B6 or Adeno-2B11 at MOI 0, 10, 25, or 50 in 0.2 mL RPMI 1640 containing 5% FBS. Fresh culture medium (0.8 mL) was then added followed by incubation with the virus for an additional 20 hours. The culture medium was then removed and the cells were treated with various concentrations of cyclophosphamide in 1 mL fresh RPMI 1640 containing 5% FBS. After 2 days of cyclophosphamide treatment, the medium was replaced with 1 mL fresh cyclophosphamide-containing culture medium for an additional 4 days. Cells remaining at the end of the 6-day cyclophosphamide treatment period were stained with crystal violet and quantified (A₅₉₅) as percent cell survival relative to drug-free controls. In experiments where the cells were infected by adeno-P450 in combination with Onyx-017, the levels of Adeno-2B6 and Adeno-2B11 were decreased to MOI 6 (for MCF-7 cells), MOI 8 (for U251 cells), or MOI 32 (for MCF-7/4HC and PC-3 cells).

The bystander activity of Adeno-2B11 was evaluated by infection of U251 cells, plated at 3 × 10⁴ per well in 24-well plates 24 hours before infection with Adeno-2B11 (MOI 0 or 25) for a total of 20 to 24 hours as described above. Cells were then placed in fresh medium with or without 20 μmol/L cyclophosphamide for 2 days, at which time fresh medium ± cyclophosphamide was reapplied for an additional 5 days. Cells cultured for 2 days after removal of the adenovirus were fixed in 4% paraformaldehyde for 30 minutes and then washed twice for 5 minutes with PBS for immunostaining with anti-P450 2B11 antibody.

Samples were blocked with 3% FBS in PBS for 20 minutes. Anti-P450 2B11 antibody (kindly provided by Dr. James R. Halpert; 1:1,000 dilution in PBS containing 3% FBS) was applied to the cells for 1 hour at 37°C in a volume of 300 μL. The cells were then washed with PBS twice for 5 minutes. Secondary anti-rabbit antibody (1:500 dilution in PBS containing 3% FBS) was applied to the cells for 1 hour at 37°C. Cells were washed twice with PBS and P450 2B11-positive staining was visualized with the TMB substrate kit for peroxidase (Vector Laboratories, Burlingame, CA). P450 2B11-stained cells were then counted as a percentage of the overall U251 population. Cells were also stained with crystal violet to determine cell survival after 7 days of cyclophosphamide exposure.

Total RNA was isolated using TRIzol reagent according to the manufacturer's instructions for monolayer cells. Reverse transcription to yield cDNA was carried out using GeneAmp RNA core kit (Applied Biosystems, Foster City, CA) using 1 μg total RNA in a volume of 20 μL according to the manufacturer's instruction. cDNA (4 μL; 1:50 dilution) in a total volume of 16 μL SYBR Green and PCR primers mixture was used for quantitative real-time PCR with the following primers: 5′-AAATCCTTCCTCAGGCTCCAA-3′ (P450 2B11 sense primer), 5′-GCCTCCCGTATGGCGTCTAT-3′ (P450 2B11 antisense primer), 5′-TTACAGTGCTCGCTTTGGTCTGT-3′ (adenovirus E3 sense primer), and 5′-TAAAGCTGCGTCTGCTTTTGTATTT-3′ (adenovirus E3 antisense primer). Samples were incubated at 95°C for 10 minutes followed by 40 cycles of 95°C for 15 seconds and 60°C for 1 minutes in an ABI PRISM 7900HT Sequence Detection System (Applied Biosystems). Results were analyzed using the comparative C_T (ΔΔC_T) method as described by the manufacturer. Data are presented as fold increase of each RNA compared with uninfected cell controls after normalization for the 18S RNA content of each sample.

Seven-week-old (25 to 29 g) male ICR/Fox Chase scid mice (Taconic Farms, Germantown, NY) were injected s.c. at each posterior flank with 4 × 10⁶ 9L/pBabe cells (control), 9L/2B6 cells, or 9L/2B11 cells in 0.5 mL serum-free DMEM using a 0.5-inch 29-gauge needle and a 1 mL insulin syringe. Tumor areas (length × width) were measured twice weekly using Vernier calipers (Manostat Corp., Herisau, Switzerland) and tumor volumes were calculated based on the following formula: volume = π / 6 (length × width)^3/2. When the tumors reached ∼1,000 to 1,200 mm³ in volume (23 days after tumor implantation), the mice were treated with cyclophosphamide using a metronomic schedule, 140 mg cyclophosphamide/kg body weight repeated every 6 days (22). Mice were divided into four groups (8–10 tumors per group): (a) 9L/pBabe tumors, untreated; (b) 9L/pBabe tumors, cyclophosphamide treated every 6 days; (c) 9L/2B6 tumors, cyclophosphamide treated every 6 days; and (d) 9L/2B11 tumors, cyclophosphamide treated every 6 days. Average tumor growth rates before cyclophosphamide treatment were similar in all four groups (average increases in tumor volume from days 14 to 23 after tumor implantation ranged from 3.8-fold to 5.6-fold). The effect of cyclophosphamide treatment on the growth rate of each tumor was calculated by comparison with the tumor volume on the day of the first cyclophosphamide injection (=100%) to facilitate comparisons of the effects of drug treatment on the growth of individual tumors that differ in size at the time of initial drug treatment.

9L cells were infected with retroviral particles encoding P450 2B11, transcribed as a bicistronic mRNA containing human P450R RNA linked via an IRES element. A pool of several thousand independent clones was obtained following puromycin selection, with each clone incorporating the P450-IRES-P450R cDNA at a random integration site and under the transcriptional control of the retroviral long terminal repeat promoter. 9L cell pools infected with a corresponding retrovirus coding for P450 2B1 or P450 2B6, IRES linked to P450R, were selected in parallel. P450 2B protein expression was readily detected by Western blot analysis of each 9L/P450 cell pool but not in 9L cells infected with the P450-deficient control retrovirus pBabe (Fig. 1A). Cellular P450 contents were 13, 27, and 16 pmol P450/mg total cell lysate for 9L/2B1, 9L/2B6, and 9L/2B11 cells, respectively. These P450 protein levels are similar to those obtained previously by retroviral expression of P450 2B6 (18). P450R activity measured in the 9L/P450 cell lysates was increased up to 2.7-fold compared with 9L control cells (Fig. 1B), confirming the coexpression of P450 and the IRES-linked P450R cDNA.

The P450-expressing 9L cells were sensitive to the cytotoxic effects of cyclophosphamide (Fig. 2A) and ifosfamide (Fig. 2B), whereas the P450-deficient 9L wild-type cells were insensitive to millimolar concentrations of both prodrugs. 9L/2B11 cells were the most susceptible to cyclophosphamide and ifosfamide cytotoxicity, with >90% growth inhibition observed at 50 μmol/L of each prodrug. The increased chemosensitivity of the 9L/2B11 cells was particularly striking at low concentrations of cyclophosphamide or ifosfamide (5–10 μmol/L), whereas the other 9L/P450 cells displayed little (9L/2B1 cells) or no growth inhibition (9L/2B6 cells), despite their similar levels of P450 and P450R expression. EC₅₀s determined for inhibition of 9L/2B11 cell growth by cyclophosphamide and ifosfamide (2.2–2.6 μmol/L) were 13- to 18-fold lower than for 9L/2B1 cells and 100- to 250-fold lower than for 9L/2B6 cells (Table 1).

The striking sensitivity of 9L/2B11 cells to low micromolar prodrug concentrations may reflect the low K_m for cyclophosphamide and ifosfamide that Escherichia coli–expressed P450 2B11 exhibits in vitro (34). To test this hypothesis, we investigated whether the cellular cyclophosphamide and ifosfamide 4-hydroxylase activity of each cell line is related to its cyclophosphamide and ifosfamide cytotoxicity profile. In particular, we wanted to determine if the low K_m that bacterial-expressed P450 2B11 exhibits for cyclophosphamide and ifosfamide in a reconstituted enzyme system (34) is also manifest when the P450 is expressed in intact tumor cells. Figure 3 presents the concentration-dependent profiles for 4-hydroxylation of cyclophosphamide (Fig. 3A) and ifosfamide (Fig. 3B) catalyzed by each 9L/P450 cell line. Cyclophosphamide and ifosfamide metabolism catalyzed by 9L/2B11 cells was saturated at a substantially lower prodrug concentration than 9L/2B1 or 9L/2B6 cells. Steady-state kinetic analysis revealed a K_m of 70 μmol/L for cyclophosphamide 4-hydroxylation by 9L/2B11 cells (i.e., 7- and ∼20-fold lower than the corresponding K_m exhibited by 9L/2B1 and 9L/2B6 cells, respectively; Table 1). 9L/2B11 cells also exhibited a higher V_max than 9L/2B1 and 9L/2B6 cells, resulting in a 13- to 29-fold higher overall catalytic efficiency (V_max/K_m). In the case of ifosfamide, the K_m of 9L/2B11 cells (35 μmol/L) was 23-fold lower than that of 9L/2B1 cells and 42-fold lower than 9L/2B6 cells. Overall, the increase in V_max/K_m for ifosfamide 4-hydroxylation by 9L/2B11 cells was 21- and 153-fold when compared with 9L/2B1 and 9L/2B6 cells, respectively (Table 1). These kinetic data are consistent with the cyclophosphamide and ifosfamide cytotoxicity profiles shown in Fig. 2.

Tumor cell expression of P450 2B enzymes, and the associated activation of cyclophosphamide and ifosfamide, leads to formation of cell-permeable metabolites that exert bystander cytotoxicity toward P450-deficient neighboring tumor cells (8, 18). The high prodrug metabolic activity of 9L/2B11 cells indicates that these cells produce high intracellular concentrations of activated cyclophosphamide and ifosfamide, which could, in turn, lead to rapid death of the P450 2B11-expressing cells, thereby limiting the overall bystander cytotoxic response. We therefore investigated the ability of 9L/2B11 cells to induce a bystander cytotoxic response. Figure 4 shows that cyclophosphamide treatment of 9L/2B11 cells led to substantial bystander killing of cocultured wild-type 9L cells even when the P450-expressing cells represented only 10% of the total tumor cell population (Fig. 4C). By contrast, 9L/2B6 cells exhibited little or no bystander killing at low concentrations of cyclophosphamide and exhibited only a modest bystander effect at high cyclophosphamide concentrations (Fig. 4A). For example, at 20% 9L/2B11 cells, 100 μmol/L cyclophosphamide killed 86% of the total cell population, corresponding to at least ∼82% bystander killing (see Fig. 4). The corresponding bystander killing by 9L/2B6 cells was only 15% or ∼5-fold lower. The bystander effect exerted by 9L/2B1 cells (Fig. 4B) was somewhat weaker than that of 9L/2B11 cells, killing 64% of the overall population when 9L/2B1 cells constituted 20% of the population. At low cyclophosphamide concentrations (10 μmol/L), 49% of bystander cells were killed when the population was seeded with 20% 9L/2B11 cells, whereas only 12% bystander killing was effected by 9L/2B1 cells under the same conditions. Further studies focused on the comparison of CYP2B11 with CYP2B6, which has recently been entered into clinical trials in patients with advanced breast cancer and melanoma (29).

A replication-defective adenovirus containing a P450 2B11-IRES-P450R expression cassette was engineered and designated Adeno-2B11. This adenovirus has the same design as Adeno-2B6 (26), which expresses P450 2B6 linked to P450R via an IRES sequence. Adeno-2B11 and Adeno-2B6 were compared with respect to their ability to induce P450R expression and cyclophosphamide activation in the human brain tumor cell line U251, assayed 48 hours after initiation of virus infection, at which time Adeno-2B6-dependent P450 expression is maximal (26). Both viruses induced dose-dependent increases in P450R activity, with a somewhat higher increase seen in cells infected with Adeno-2B6 compared with Adeno-2B11 (7-fold versus ∼5-fold increase at MOI 100; Fig. 5A). By contrast, there was substantially greater conversion of cyclophosphamide (100 μmol/L) to 4-OH-cyclophosphamide by cells infected with Adeno-2B11 compared with Adeno-2B6 (5.5 versus 0.2 μmol/L 4-OH-cyclophosphamide released into culture medium; Fig. 5B).

Adeno-2B6 and Adeno-2B11 both effected dose-dependent increases in cyclophosphamide toxicity toward U251 cells. However, the overall cytotoxic response was substantially greater with Adeno-2B11 and could be achieved at a much lower prodrug concentrations (Fig. 5D versus Fig. 5C), consistent with the metabolic data shown in Fig. 5B. For example, U251 cells infected with 50 MOI Adeno-2B6 showed moderate cyclophosphamide cytotoxicity (25% cell kill) at 50 μmol/L cyclophosphamide, whereas ∼85% cell killing was achieved with Adeno-2B11 under comparable conditions. The superiority of Adeno-2B11 was most apparent at 10 to 20 μmol/L cyclophosphamide, consistent with the low K_m (cyclophosphamide) of P450 2B11. The strong cytotoxicity conferred by Adeno-2B11 was associated with substantial bystander killing of the uninfected U251 cells in the population. Thus, although only 10 ± 2.5% of the U251 cells expressed P450 2B11 as shown by immunostaining of cells infected with MOI 25 Adeno-2B11 (Fig. 6B versus Fig. 6A), major cytotoxicity was still seen following treatment with 20 μmol/L cyclophosphamide, with the results most dramatic after a 7-day treatment period (Fig. 6C and D).

Onyx-017 is an E1b-55-kDa gene-deleted adenovirus that harbors a wild-type E3 region, which contains a myriad of genes involved in cell death and other adenoviral processes (45). E1b-55-kDa functions include p53 degradation, RNA export, and host protein shutoff, and the loss of these E1b-55-kDa late functions is an important determinant of the tumor cell selectivity of the Onyx virus (46). Onyx-017 can also serve as a helper virus that facilitates intratumoral replication and cell-to-cell spread of replication-defective adenovirus, such as Adeno-2B6 (26). Given the extraordinary ability of P450 2B11 to activate cyclophosphamide at low prodrug concentrations, we investigated the effect of Adeno-2B11 + Onyx-017 on cyclophosphamide cytotoxicity. These experiments were carried out in U251 tumor cells, which are permissive to adenovirus infection, and PC-3 tumor cells, which are relatively resistant to infection (26).

Cells were infected with Adeno-2B11 or Adeno-2B6 at MOI 8 (U251 cells) or MOI 32 (PC-3 cells) either alone or in combination with Onyx-017. The higher adeno-P450 MOI was used for the PC-3 cells to compensate for the low adenovirus infectivity of this cell line. Substantial increases in cyclophosphamide cytotoxicity were seen in tumor cells infected with Adeno-2B11 + Onyx-017 compared with Adeno-2B6 + Onyx-017 (Fig. 7). For example, Adeno-2B11 combined with Onyx-017 (MOI 2) induced an 88% decrease in U251 cell survival at 50 μmol/L cyclophosphamide (Fig. 7B), whereas Adeno-2B6 + Onyx-017 effected <20% cell killing under comparable conditions (Fig. 6A). At higher MOIs, Onyx-017 further increased the cytotoxicity of low cyclophosphamide concentrations toward the Adeno-2B11-infected U251 cells. In the prostate carcinoma cell line PC-3, no cyclophosphamide-dependent cytotoxicity was seen with either adeno-P450 virus alone, owing to the very poor infectivity of these cells. Coinfection with Onyx-017 sensitized the cells to cyclophosphamide, with the effects being much more dramatic in the case of Adeno-2B11, particularly at low cyclophosphamide concentrations (Fig. 7D versus Fig. 7C).

By coamplifying the replication-defective Adeno-2B6 and spreading it to neighboring tumor cells, Onyx-017 can dramatically increase overall P450 metabolic activity (26). This spreading effect of Onyx-017 was also seen with Adeno-2B11 as shown by the ability of culture medium from Adeno-2B11 + Onyx-017 coinfected U251 cells, but not culture medium from cells infected with Adeno-2B11 alone, to deliver P450 2B11 to fresh U251 cells (data not shown). The extent to which Onyx-017 coamplifies Adeno-2B11 could be quantified by measuring P450 2B11 RNA levels in U251 cells infected either with Adeno-2B11 alone or with Adeno-2B11 + Onyx-017. Dramatic increases in P450 2B11 RNA as well as adenovirus E3 RNA were seen in the coinfected cells, showing the helper effect of Onyx-017 on Adeno-2B11 transgene expression (Table 2). Similar increases in P450 2B6 RNA were seen in Adeno-2B6 + Onyx-017 coinfected cells (data not shown). Furthermore, evaluation of P450 metabolic activity revealed substantially higher 4-OH-cyclophosphamide formation in cyclophosphamide-treated U251 cells infected with Adeno-2B11 + Onyx-017 compared with adeno-P450 2B6 + Onyx-017 (Fig. 7E) despite similar increases in P450R activity induced by Adeno-2B6 and Adeno-2B11 (data not shown). Thus, although the helper effect of Onyx-017 on the replication and spread of Adeno-2B11 and Adeno-2B6 is fundamentally the same, the superior metabolic activity of Adeno-2B11 results in much greater increases in tumor cell production of 4-OH-cyclophosphamide.

We next investigated the ability of Adeno-2B11 to kill MCF-7 breast cancer cells resistant to activated cyclophosphamide (MCF-7/4HC cells). A ∼5-fold higher MOI of adeno-P450 was used to infect the MCF-7/4HC cells in an effort to overcome the 6-fold drug resistance of this cell line (36). Figure 8A shows that the drug-sensitive parental MCF-7 cell line was sensitized to cyclophosphamide by Adeno-2B11 + Onyx-017 coinfection to a greater extent than by Adeno-2B6 + Onyx-017 coinfection. Onyx-017 (MOI 3) exerted little or no intrinsic cytotoxicity under these conditions, and Adeno-2B11 exhibited only moderate toxicity in the absence of Onyx-017. These treatments were much less toxic toward MCF-7/4HC cells as anticipated. Nevertheless, Adeno-2B11, but not Adeno-2B6, was able to induce cyclophosphamide-dependent killing of MCF-7/4HC cells when coinfected with Onyx-017 (Fig. 8B). Thus, the high metabolic activity of Adeno-2B11 can, in part, help overcome acquired resistance to activated cyclophosphamide.

9L tumor cells transduced with retrovirus encoding P450 2B11 or P450 2B6 and coexpressing P450R were grown as solid, s.c. tumors in scid mice. P450-deficient 9L tumors were studied in parallel as a control. Cyclophosphamide was given on an antiangiogenic (metronomic) schedule consisting of 140 mg cyclophosphamide/kg body weight i.p. repeated every 6 days (23) for up to 11 cyclophosphamide injections (22). This cyclophosphamide dose and schedule is accompanied by minimal host toxicity (22, 23) and contrasts to a maximally tolerated dose of 265 mg cyclophosphamide/kg body weight given as a single injection in the same scid mouse model (47).

Untreated 9L tumors grew rapidly, necessitating that the animals be killed by day 49 post-tumor implantation (day 26 on the X axis in Fig. 9A). All of the 9L control and 9L/P450 tumors responded to 6-day repeated cyclophosphamide treatment as indicated by the substantial decreases in volume of each individual tumor (Fig. 9B–D). Tumor growth was arrested after the third cyclophosphamide injection (days 12–15 post-treatment) in the case of the 9L/2B6 and 9L/2B11 tumors. This was followed by major regression leading to an apparent elimination of all 10 of the P450 2B11-expressing tumors by the 7th cyclophosphamide injection (days 35–39 posttreatment; Fig. 9D). By contrast, only 3 of 10 tumors that expressed P450 2B6 were eliminated by day 43, and it was not until day 57 (after the 10th cyclophosphamide injection) that the 9L/2B6 tumors regressed to the point where they were no longer measurable. Control 9L tumors given the cyclophosphamide 6-day treatment regressed by an average of only 83% in contrast to the complete regression seen with 9L/2B6 and 9L/2B11 tumors (Table 3).

The much greater responsiveness of both 9L/P450 tumor types was also apparent from the period of continued tumor growth that immediately followed initiation of the 6-day cyclophosphamide treatment schedule (Table 3). Thus, the P450-deficient tumors continued to increase in volume by an average of 450% before they began to regress, whereas the P450 2B6 and P450 2B11 tumors increased by an average of only 250% to 280% and then began to regress.

Cyclophosphamide treatment was halted after the 11th cyclophosphamide injection (day 60) to ascertain whether any viable tumor cells remained. The P450-deficient 9L control tumors resumed aggressive growth beginning ∼18 days after cyclophosphamide treatment was discontinued (Fig. 9B). By contrast, the P450 2B6-transduced tumors exhibited an average tumor-free period of 30 days (range, 21–39 days), after which they resumed growth. The 9L/2B11 tumors, however, had a substantially longer tumor-free period (mean, 79 days; range, 60–100), and only 2 of the 10 tumors exhibited aggressive regrowth when the experiment was terminated on day 134 (Fig. 9D; Table 3).

P450 prodrug gene therapy has several unique characteristics that distinguish it from other cancer GDEPT strategies. One important feature is its provision of two distinct routes for prodrug activation, a hepatic route and an intratumoral route (primary tumor and its metastases), which results in the tumor being exposed to activated metabolites from both an extratumoral (systemic) compartment and a local, intratumoral compartment. Other GDEPT strategies use prodrugs that are not metabolized in the liver, limiting therapeutic activity to cells nearby those cells infected with the prodrug-activating gene therapy vector. Moreover, liver-activated P450 prodrugs, such as cyclophosphamide, exert both a systemic antitumor bystander effect and a powerful cytotoxic effect directed against tumor-associated endothelial cells (antiangiogenic activity). This latter effect is manifest when cyclophosphamide is given using a metronomic schedule (23) and greatly enhances the regression of tumors that coexpress P450 2B6 and P450R (22). However, a key limitation of P450-based gene therapy is the high K_m that characterizes most mammalian P450 enzymes with their drug substrates (e.g., K_m, 500–2,000 μmol/L for cyclophosphamide and ifosfamide with P450 2B1 and P450 2B6; refs. 30, 31). This leads to inefficient prodrug activation at pharmacologically relevant drug concentrations (e.g., ∼50–200 μmol/L cyclophosphamide and ifosfamide; refs. 32, 33). The goal of the present study, therefore, was to investigate whether P450 GDEPT can be enhanced using P450 2B11, a canine P450 enzyme that was recently identified as a low K_m catalyst of cyclophosphamide and ifosfamide activation (34) and displays kinetic properties far superior to those of P450 enzymes 2B1 and 2B6, used previously in preclinical and clinical GDEPT studies.

We first investigated the chemosensitization of 9L gliosarcoma cells to cyclophosphamide and ifosfamide following infection with retrovirus that delivered P450 2B1, 2B6, or 2B11 at similar expression levels. Each P450 was delivered in combination with P450R, which catalyzes the rate-limiting electron transfer step required for all microsomal P450 drug oxidations and can markedly improve P450 metabolic activity and antitumor responses when introduced into tumor cells in combination with P450 (18, 19). The resultant pool of 9L/2B11 cells was highly sensitive to low micromolar concentrations of cyclophosphamide and ifosfamide (EC₅₀, 2.2–2.6 μmol/L) in contrast to the corresponding 9L/2B1 and 9L/2B6 cell pools, which required up to 100- to 250-fold higher prodrug concentrations for effective cell killing. The enhanced chemosensitivity of 9L/2B11 cells was associated with correspondingly higher cyclophosphamide and ifosfamide 4-hydroxylation (prodrug activation) and K_m values for prodrug metabolism measured in intact cells (K_m, 35–70 μmol/L) that were 20- to 40-fold lower than 9L/2B6 cells (K_m, 1,380–1,460 μmol/L; Table 1). Although P450 2B11 exhibits 22% to 26% amino acid difference from P450s 2B1 and 2B6, its unique kinetic properties are most likely due to a limited number of amino acid substitutions at the active site. These include a valine residue at position 114, which when introduced into P450 2B1 decreases the K_m toward cyclophosphamide and ifosfamide up to 4-fold (34). The presence of this and other naturally occurring “mutations” in P450 2B11 gives rise to a 29- to 153-fold higher overall catalytic efficiency (V_max/K_m) for cyclophosphamide and ifosfamide 4-hydroxylation, with corresponding increases in cyclophosphamide cytotoxicity, in tumor cells that express P450 2B11 compared with P450 2B6.

Studies using Adeno-2B11, a replication-defective adenovirus delivering P450 2B11 in combination with P450R, further established the superior activity of this enzyme combination for P450 GDEPT using cyclophosphamide. The increased activity of P450 2B11 compared with P450 2B6 was manifest in human tumor cell lines readily infected by adenovirus (U251 cells) and in cells with a lower intrinsic susceptibility to adenovirus infection (PC-3 and MCF-7 cells). P450 expression and cyclophosphamide-dependent cell killing could be greatly enhanced by coinfection with the conditionally replicating, oncolytic adenovirus Onyx-017. Enhanced cyclophosphamide-dependent cell killing was also seen in an MCF-7 cell line resistant to activated cyclophosphamide (MCF-7/4HC cells), showing the potential of P450 GDEPT to extend the effectiveness of cyclophosphamide therapy to include drug-resistant tumor cells. Further increases in activity against tumor cells showing intrinsic or acquired drug resistance to P450 prodrugs may be achieved by directed P450 evolution (48) or through further optimization of the gene delivery vector to increase intratumoral P450 expression and P450 metabolic activity.

The use of a low K_m P450 enzyme, such as P450 2B11, could be particularly useful for GDEPT applications in certain patient populations, such as children with B-cell non-Hodgkin's lymphoma, where inadequate (hepatic) conversion of cyclophosphamide to active metabolites is associated with an increased risk of disease recurrence (49). Consequently, for patients with low rates of cyclophosphamide clearance and for individuals who produce significant quantities of inactive cyclophosphamide metabolites (49), P450 gene therapy may well prove to be clinically beneficial. This approach may be more effective than increasing the dose of cyclophosphamide, which can lead to saturation of the bioactivation pathway (50) and increased host toxicity. Other benefits of using P450 2B11, a canine enzyme, for P450 GDEPT include the potential for eliciting an antitumor immune response, leading to a systemic bystander effect that greatly increases the overall antitumor response. Indications of such an immunologic response were reported in a recent phase I trial of P450 GDEPT combined with low-dose cyclophosphamide in breast cancer patients (29). It should be noted, however, that such immune responses might negatively affect viral replication by increasing the elimination of vector-infected cells.

The bystander cytotoxic effect is critical to the success of any prodrug activation gene therapy strategy (51). One potential concern of using a highly active catalyst of cyclophosphamide activation, like P450 2B11, is that it may generate a high intracellular concentration of active drug, which in turn may accelerate death of the P450-expressing tumor cells, thereby limiting bystander cytotoxicity (52). Our studies showed, however, that 9L/2B11 cells mediate substantial bystander cytotoxicity, killing ∼70% to 75% of bystander tumor cells under conditions where cyclophosphamide-treated 9L/2B6 cells killed only up to 15% of bystander tumor cells. Similarly, in a population of U251 cells infected with Adeno-2B11 under conditions where only ∼10% of the cells express P450, >90% cell killing was observed at low-dose (20 μmol/L) cyclophosphamide. Further increases in bystander killing may be obtained by coexpression of the pan-caspase inhibitor p35 (53) or by introduction of other antiapoptotic factors that delay but do not abrogate cyclophosphamide-induced cell death (52).

The therapeutic potential of P450 2B11 was evaluated in vivo in scid mice in a tumor xenograft model. 9L tumors expressing a P450 2B11-IRES-P450R expression cassette fully regressed by day 39 of cyclophosphamide treatment compared with a 57-day cyclophosphamide treatment period required for comparable regression of 9L/2B6 tumors. Moreover, an extended tumor-free period was seen following cessation of cyclophosphamide treatment in the case of the 9L/2B11 tumors (up to 100 days) versus 9L/2B6 tumors (21–39 days). These findings highlight the advantages of using P450 2B11 to effect prodrug activation, particularly under conditions where intratumoral prodrug levels may be low (e.g., as a consequence of the reduced access that tumor cells may have to circulating cyclophosphamide as they regress in response to the antiangiogenic action of the 6-day repeating cyclophosphamide treatment schedule). In the case of 9L control tumors, cyclophosphamide effected delayed tumor regression as was seen previously (22). This can, in part, be attributed to apoptosis of the endothelial cells lining the tumor vasculature by liver-derived 4-OH-cyclophosphamide (23). Overall, the dramatic regression of both 9L/2B11 and 9L/2B6 tumors indicates a direct tumor cytotoxic action associated with intratumoral cyclophosphamide activation in addition to the endothelial cell-directed, antiangiogenic mechanism that is associated with liver cyclophosphamide activation, which may apply equally well to 9L control and 9L/P450 tumors.

The response of 9L/2B6 tumors to the metronomic, 6-day cyclophosphamide treatment schedule seen in the present study was less dramatic than that seen in a previous study, where cyclophosphamide had a more immediate inhibitory effect on tumor growth (22). Moreover, the initial response of 9L/2B11 tumors in the present study was not apparently different from that of 9L/2B6 tumors. One possible explanation could be that, in the present study, the 9L/P450 tumors were derived from pools of retrovirus-infected 9L cells, each composed of several thousand independent clones. Each clone may express the P450, and P450R, with a markedly different efficiency, depending on where in the genome the retroviral sequence is inserted, and as a consequence, each clone may respond to cyclophosphamide in an idiosyncratic manner. In contrast, the 9L/2B6 tumors studied earlier were derived from a unique high P450 activity clone selected for high P450 2B6 expression (22). Consequently, all of the 9L/2B6 tumor cells responded to cyclophosphamide rapidly and in a similar manner. This conclusion is supported by immunohistochemical analysis of the present 9L/2B6 and 9L/2B11 tumors, which revealed substantial heterogeneity with respect to P450 expression at the time of initial cyclophosphamide treatment, with many of the tumor cells expressing little or no detectable P450 2B protein.²

In conclusion, the present study shows the utility of the low K_m cyclophosphamide and ifosfamide 4-hydroxylase P450 2B11 for prodrug activation cancer gene therapy. The use of this enzyme in combination with P450R, coupled with a replication-conditional viral vector for gene delivery, may facilitate the development of more effective therapeutic approaches to cancer treatment.