Gene Therapy (2002) 9, 946–954  2002 Nature Publishing Group All rights reserved 0969-7128/02 $25.00 www.nature.com/gt RESEARCH ARTICLE Oxygen-sensitive - gene therapy for the eradication of radiation-resistant solid tumours

AV Patterson1,7, KJ Williams1, RL Cowen1, M Jaffar1, BA Telfer1, M Saunders2, R Airley3, D Honess4, AJ Van der Kogel5, CR Wolf6 and IJ Stratford1 1School of Pharmacy and Pharmaceutical Sciences, University of Manchester, Manchester, UK; 2Paterson Institute for Cancer Research, Christie Hospital, Manchester, UK; 3School of Pharmacy and , Liverpool John Moores University, Liverpool, UK; 4The Gray Laboratory Cancer Research Trust, Mount Vernon Hospital, Northwood, UK; 5Institute of Radiotherapy, University of Nijmegen, Nijmegen, The Netherlands; 6Cancer Research UK Molecular Pharmacology Unit, University of Dundee, UK; and 7Auckland Cancer Society Research Centre, The University of Auckland, Auckland, New Zealand

Overwhelming clinical and experimental data demonstrate HRE-P450R and empty vector transfectants had compara- that tumour hypoxia is associated with aggressive disease ble hypoxic fractions and were refractive to single dose and poor treatment outcome as hypoxic cells are refractive radiotherapy of up to 15 Gy. However, combining a prodrug to radiotherapy and some forms of . However, of RSU1069 with a reduced radiotherapy dose of 10 Gy rep- hypoxia is rare in physiologically normal tissues representing resents a curative regimen (50% tumour-free survival; day a tumour-specific condition. To selectively target this thera- 100) in the HRE-P450R xenografts. In complete contrast, peutically refractive cell population, we have combined 100% mortality was apparent by day 44 in the empty vector bioreductive chemotherapy with hypoxia-directed gene ther- control xenografts treated in the same way. Thus, an oxy- apy. We have transfected the human fibrosarcoma cell line, gen-sensitive gene-directed enzyme prodrug therapy HT1080, with a hypoxia-regulated expression vector enco- approach may have utility when incorporated into conven- ding the human flavoprotein cytochrome c P450 reductase tional radiotherapy and/or chemotherapy protocols for (HRE-P450R). This conferred hypoxia-dependent sensitivity loco-regional disease in any tissue where hypoxia is a to the alkylating nitroimidazole prodrug RSU1069 in vitro, contra-indication to treatment success. with a greater than 30-fold increase in oxic/hypoxic cytotox- Gene Therapy (2002) 9, 946–954. doi:10.1038/sj.gt.3301702 icity ratio compared with controls. Xenografts of both the

Keywords: hypoxia; bioreductive ; enzyme-prodrug therapy; cytochrome c P450 reductase

Introduction to hypoxia as a potent physiological stimulus that pro- motes tumour progression, driving genetic change and Poor tumour oxygenation, due to an inadequate selecting for aggressive neoplastic phenotypes.3,4,7 supply, is a well-documented feature of most solid Consequently, an important goal for cancer research tumours leading to regions of acute and/or chronic has been to develop treatment strategies that selectively hypoxia. Conclusive experimental and clinical research target this resistant tumour cell population. One demonstrates the impact of hypoxic tumour cells upon approach has been the use of chemotherapeutic agents the therapeutic outcome of cancer radiotherapy and some that are specifically cytotoxic to hypoxic cells.8 These 1–5 chemotherapy. Hypoxic cells are approximately three bioreductive ‘’ can be reduced by cellular times more radio-resistant than normoxic cells which is reductases to species that cause DNA damage under thought to result primarily from the ability of oxygen to hypoxic conditions, but in the presence of molecular oxy- react with, and chemically modify the initial radiation- gen they are efficiently back-oxidized to the non-toxic induced DNA radicals. Although less clinical evidence is parent compound, such that little or no DNA damage available regarding the impact of hypoxia on the failure occurs. Of importance is the relative toxicity of bioreduc- of chemotherapy there is direct evidence that hypoxic tive under hypoxic versus aerobic conditions. Pref- cells within tumour xenografts are refractive to clinically erably, the drug should possess a high hypoxic cell cyto- relevant chemotherapeutic agents and it is reasonable to toxicity ratio (HCR; the ratio of the dose required to kill infer that the quiescent nature of hypoxic cells may ren- a proportion of aerobic cells to that required to kill an der them insensitive to agents that target rapidly divid- equal proportion of the same cells exposed to the drug 6 ing cells. Experimental and clinical evidence also points under hypoxic conditions), leading to maximum hypoxic tumour cell killing with minimal toxicity to aerobic tissues. Correspondence: KJ Williams, School of Pharmacy and Pharmaceutical Alternative strategies to exploit hypoxia have been Sciences, University of Manchester, Oxford Road, Manchester, UK developed in the field of gene therapy. Hypoxia not only Received 26 November 2001; accepted 14 February 2002 provides a powerful physiological stimulus, but also Oxygen-sensitive enzyme-prodrug therapy AV Patterson et al 947 results in a complex transcriptional cascade in order to hypoxia, we can simultaneously increase the therapeutic facilitate cell survival under hypoxic stress. This is achi- index of the therapy without increasing systemic toxicity. eved by the up-regulation of genes supporting anaerobic metabolism and new blood vessel development, such as Results phosphoglycerate kinase-1 (PGK-1), lactate dehydrogen- ase A (LDH-A) glucose transporter-1 (Glut-1) and vascu- Hypoxia-induced human cytochrome c P450 reductase lar endothelial growth factor (VEGF), amongst many expression in a human tumour cell line others. This coordinated hypoxic gene activation is A bicistronic vector, pCIneo PGK SV40mp P450R IRES mediated through the transcription factor HIF-1 EGFP (pHRE-P450R/EGFP), encoding for the co- (hypoxia-inducible factor-1) that binds to a common expression of human P450 reductase (P450R) and hypoxia-responsive element (HRE) found within the enhanced green fluorescent protein (EGFP) was con- enhancer region of all HIF-1 responsive genes.9,10 Fur- structed (Figure 1). A trimer of the hypoxia-responsive thermore, by introducing an HRE sequence into an element (HRE) of the mouse phosphoglycerate kinase appropriate expression cassette, hypoxic regulation of an gene-1 (PGK-1) was used as the hypoxic enhancer and exogenous therapeutic gene can be achieved. The oxygen cloned in the positive orientation upstream of the SV40 ‘sensing’ pathway has been demonstrated to be univer- minimal promoter.13 A control vector, lacking only the sally functional in mammalian cells11 and unlike the lim- P450R cDNA, was also constructed (pHRE-EGFP) and ited occurrence of other tumour-specific transcription fac- each vector was independently transfected into the tors, HIF-1 is reported to be expressed in 70% of solid HT1080 cell line to allow the selection of EGFP matched tumours,12 making HRE-regulated gene therapy an P450 reductase-expressing and empty vector control attractive strategy that can be applied to a broad range clones. Clones were isolated from each transfected popu- of cancers. lation and a detailed comparison of growth kinetics (both Hypoxia-selective gene therapy was first exemplified in vitro and when grown as xenografts in vivo) and using HRE-regulated expression of the suicide gene E. hypoxic gene induction was made in order to avoid clon- coli cytosine deaminase, demonstrating hypoxia-specific ing artifacts. This led to the selection of transcriptionally sensitisation of tumour cells to the prodrug 5-fluorocyto- equivalent GFP/R9 (P450R expressing) and the GFP5 sine.13 Subsequent studies have similarly demonstrated (empty vector) clones. the efficacy of HRE-regulated Herpes simplex virus 1- thymidine kinase activity to mediate ganciclovir sensitis- In vitro parameters of the human cytochrome c P450 ation.14 Although HRE-mediated prodrug activating reductase-expressing and empty vector clones enzyme expression offers transcriptional selectivity over The growth characteristics of both of these cell lines were non-targeted enzyme prodrug therapies, we report a identical in vitro (population doubling time 15.5 h (±1 h)) novel strategy that further enhances this approach by and expression of EGFP in air and following 18 h making hypoxia a prerequisite for both gene expression exposure to hypoxia was also matched. Analysis of EGFP and drug activation. This is achieved by using an oxygen- presentation by flow cytometry (refer to Materials and sensitive bioreductive agent as a prodrug in an O2-sensi- methods for details) revealed a median peak fluorescence tive gene directed enzyme prodrug therapy (GDEPT) of 48 ± 3 and 42 ± 1 for the GFP/R9 and GFP-5 clones, approach. For this we have used the nitroimidazole respectively, in air. This was induced approximately two- RSU1069 (1-[3-aziridinyl-2-hydroypropyl]-2-nitroimid- fold in both cell lines in response to hypoxia and 3 h re- azole) that is reduced by a series of single electron trans- oxygenation (median peak fluorescence of 79.2 ± 8.7 for fers to generate highly reactive toxic species.15,16 This GFP/R9, and 72.7 ± 10 for GFP-5 cells), rising to 3.5-fold results in a large oxic/hypoxic selectivity being recorded following 24 h re-oxygenation. The ability of re-oxygen- for RSU1069 being 30–200-fold more potent in the ation to enhance output from HRE-reporter constructs absence of oxygen. This compares with a value of two has been documented previously.13 P450R activity levels reported for mitomycin C.17 Several endogenous were also assessed in air and hypoxia. Aerobic P450R may catalyse the reduction of RSU1069, how- enzyme activity in the GFP/R9 cells was 137 ± 39 nmol ever we have harnessed the reductive capacity of a domi- cytochrome c reduced minϪ1 mgϪ1 protein. Eighteen nant nitroreductase, NADPH cytochrome P450 (c) hours of hypoxic exposure induced the activity by 2.2- reductase (P450R). This flavoenzyme will readily donate fold and re-oxygenation led to further increases in single electrons to any prodrug with an appropriate activity to a maximum at 6 h re-oxygenation of 540 ± 63 redox potential and we have previously established that nmol cytochrome c reduced min-1 mg-1 protein (Figure 2). constitutive over-expression of P450R confers sensitivity P450R activity remained above basal aerobic levels for to most classes of bioreductive agents, including nitro- approximately 4 days after hypoxic exposure (Figure 2). compounds such as RSU1069.18 Basal P450R expression in the GFP-5 cells was 12 nmol In this paper we restrict over-expression of P450R to cytochrome c reduced minϪ1 mgϪ1 protein which was hypoxic tumour cells by including the HRE from PGK-1 further suppressed upon hypoxic exposure to 7 nmol within our therapeutic gene cassette. We have previously cytochrome c reduced minϪ1 mgϪ1 protein. demonstrated that PGK-1 HRE-mediated gene expression The relatively low EGFP expression in response to is activated in tumour cells shown to be radio-resistant hypoxic exposure, compared with this strong induction in vivo.13 By combining HRE-driven P450R expression in P450R expression, probably reflects the fact that the and the bioreductive drug RSU1069 with radiation, we IRES used to generate the constructs expressed by the aim to kill the remaining radiation-resistant hypoxic two cell lines was attenuated to give preferentially higher tumour subpopulation. In addition, by employing this out-put from the primary coding sequence, and that refined GDEPT approach in which both enzyme EGFP protein folding and consequently functionality is expression and drug toxicity are dependent upon influenced by oxygen availability.19

Gene Therapy Oxygen-sensitive enzyme-prodrug therapy AV Patterson et al 948

Figure 1 Structure of the hypoxia-regulated therapeutic cassette showing the mechanism of prodrug activation only under hypoxic conditions. The expression cassette was stably inserted into the genome of HT1080 fibrosarcoma cells to generate the GFP/R9 clone. Hypoxic regulation of human cytochrome P450 reductase (P450R) expression is afforded by a trimer of the HRE sequence from the PGK-1 gene. P450R expression is mediated via the transcription factor HIF-1 and yields the cytotoxic moiety from the bioreductive prodrug only under hypoxic conditions. The construct used to generate the control GFP-5 clone lacked the P450R cDNA.

Figure 3 Comparative growth rate of GFP/R9, GFP-5 and HT1080 wild- Figure 2 A time course profiling P450R activity in the GFP/R9 during type (wt) xenografts. Tumours were initiated from the intra-dermal injec- and post-hypoxic exposure. The level of P450R activity (represented as tion of 5 × 106 cells and measured daily once a palpable tumour formed. cytochrome c reduced min-1 mg-1 protein) in the GFP/R9 cells was determ- Data represent the mean ± standard error of the mean plotted until the ined during hypoxic exposure (open symbols) and upon subsequent re- first tumour within each group achieved the maximum designated volume oxygenation (closed symbols). 18 h hypoxic exposure induced P450R of 1000 mm3. All data sets were accrued from two or more independent activity by 2.2-fold and re-oxygenation led to further increases in activity experiments. to a maximum at 6 h re-oxygenation of 540 ± 63 nmol cytochrome c -1 -1 reduced min mg protein. P450R activity remained above basal aerobic 3 levels for approximately 4 days after hypoxic exposure. Data represent the 250 mm . This represented the designated treatment size mean of three independent experiments ± standard deviation. for use in subsequent studies, allowing a relative tumour volume four times that at the time of treatment (RTV4) to be used as the experimental endpoint, whilst staying In vivo parameters of the human cytochrome c P450 within the maximum tumour burden allowed for ethical reductase-expressing and empty vector clones reasons. Quantitative analysis of the hypoxic fraction As the purpose of these studies was to validate the poten- (HF) and vessel density (VD) following immunohistoch- tial of combining bioreductive chemotherapy and emical staining of pimonidazole adducts and mouse hypoxia-directed gene therapy both in vitro and in vivo, endothelial structures revealed no significant differences it was important that we did not select cell lines for in between the two tumour types (Figure 4). However, there vitro analysis that then differed markedly when grown was greater variability in both parameters in the GFP/R9 as xenografts in nude mice. Again, the inherent tumouri- tumours compared with GFP-5 tumours (GFP/R9 (n = 8), genecity and subsequent growth rate of both the HF 6.2% tumour area ± 5.0%, VD 89 ± 43 mmϪ2; GFP-5 GFP/R9 and GFP-5 cells was almost identical (Figure 3), (n = 5), HF 3.3% ± 1.4%, VD 108 ± 18 mmϪ2; P = 0.196 and no significant difference was seen in the time taken and P = 0.245 for HF and VD, respectively). Further to achieve two volume doublings between the two immunohistochemical analyses to simultaneously detect tumour types (8 ± 1.7 days for GFP/R9 (n = 12) versus P450R and pimonidazole binding revealed that P450R 9.3 ± 1.4 days for GFP-5 (n = 6), P = 0.213). In addition, was restricted to the hypoxic regions of the GFP/R9 the growth pattern mirrored that of un-transfected tumours (Figure 5a and b). In addition, expression of HT1080 cells (Figure 3). P450R in the GFP/R9 tumours co-localised with the Further studies were undertaken to address the vascu- endogenous HIF-1 regulated protein Glut-120,21 (data not larity and extent of hypoxia within tumours derived from shown). No P450R expression was detected in the each cell type. The hypoxic marker, pimonidazole, was hypoxic regions of GFP-5 tumours (Figure 5c and d). This administered to mice bearing tumours of approximately demonstrates that the transcriptional regulation afforded

Gene Therapy Oxygen-sensitive enzyme-prodrug therapy AV Patterson et al 949

Figure 4 Representative composite binary images of vascularity and hypoxic fraction in the GFP/R9 and GFP-5 tumour xenografts. Immunohistochemical studies determined the vascularity and extent of hypoxia within tumours derived from the GFP/R9 and GFP-5 cell types. Vascularity was determined by staining of mouse endothelial cells (red staining). The hypoxic marker, pimonidazole was administered before tumour extraction allowing for pimonida- zole adduct labelling (green staining). Quantitative analysis of the hypoxic fraction (HF) and vessel density (VD) following immunohistochemical staining of pimonidazole adducts and mouse endothelial structures revealed no significant differences between the two tumour types.

Figure 6 Hypoxia-regulated P450R expression hypersensitises GFP/R9 to Figure 5 Relative spatial distribution of P450R and pimonidazole binding treatment with the bioreductive drug RSU1069 in vitro. GFP-5 (ᮀ, ᭿) in GFP/R9 and GFP5 tumour xenografts. The distribution of P450R ᭺ b (panels a and c) and pimonidazole binding (panels b and d) was assessed and GFP/R9 ( , ) cells were exposed to RSU1069 for 18 h in air (closed by immunohistochemical staining on 3–4 ␮M sections taken from for- symbols) or under anoxic conditions (open symbols). Proliferation relative to untreated controls was ascertained 3 days after drug exposure by MTT malin-fixed, paraffin-embedded GFP/R9 (panels a and b) and GFP-5 ± tumour xenografts (panels c and d). P450R expression in the GFP/R9 assay. Data represent mean standard deviation values obtained from tumours (a) co-localises with hypoxic regions revealed by pimonidazole three independent experiments. staining (b). No P450R expression was detected in the GFP-5 tumours (c). Hypoxia-regulated human cytochrome c P450 reductase expression sensitises cells to bioreductive drug treatment in vitro by the HRE that we have used specifically restricts P450R The relative sensitivity of the GFP/R9 and GFP-5 cells to expression to the target tumour population. Reflecting bioreductive drug treatment was then analysed. Cells the variability in hypoxic fraction (pimonidazole binding) were exposed to the bioreductive drug RSU1069,15,16 in the GFP/R9 tumours, P450R activity measurements under aerobic or hypoxic conditions for a period of 18 h obtained for tumours excised at the designated treatment and proliferation relative to control cells ascertained 72 h size ranged from 50–225 nmol cytochrome c reduced later by MTT assay (Figure 6). Despite the difference in minϪ1 mgϪ1 protein. aerobic P450R activity in the two cell types, there was no

Gene Therapy Oxygen-sensitive enzyme-prodrug therapy AV Patterson et al 950 difference in the mean IC50 value obtained for RSU1069 selected clones that are transcriptionally equivalent using in air (47 ␮M for the GFP/R9, versus 43 ␮M for the GFP- a bicistronic vector encoding for HRE driven P450R and 5 cells; Figure 6). This is in part due to the stringency in EGFP reporter gene expression to generate the GFP/R9 the level of hypoxia required for RSU1069 activation, clone and an almost identical vector, lacking only P450R with maximal cytotoxicity being achieved below oxygen cDNA, to generate the control GFP-5 cells. We have care- concentrations of 0.1%.17 Exposure to RSU1069 under fully evaluated these clones to match EGFP expression in hypoxic conditions increased toxicity by 30-fold in the hypoxic and oxic conditions and cell growth rates in vitro ␮ GFP-5 cells, and a mean IC50 concentration of 1.4 M was and in vivo. This provided us with confidence that no achieved. However, in the GFP/R9 cells the enhanced observable artifacts, inherent to the process of cloning, cytotoxicity achieved under hypoxic conditions was have been introduced during the selection procedures. markedly increased to 1000-fold (mean IC50 hypoxia, To test the validity of our approach in vitro, we selected 0.046 ␮M; Figure 6). the bioreductive drug RSU1069 for its exquisite oxygen sensitivity. The oxygen cytotoxicity profile of this drug is Hypoxia-driven human cytochrome c P450 reductase a mirror image of that for radiotherapy in that cells at expression sensitises tumours to bioreductive drug oxygen concentrations rendering them resistant to radio- treatment in combination with radiotherapy in vivo therapy are sensitive to RSU1069. This is also reflected in We next sought to evaluate the efficacy of bioreductive the very high hypoxic cytotoxicity ratio of RSU1069 being chemotherapy and hypoxia-directed gene therapy in 30–200-fold more toxic to hypoxic cells dependent upon tumour xenografts derived from the GFP/R9 and GFP-5 the cell type used.17 In addition, we have preliminary cells. For these studies, the prodrug RB6145 (1-[3- data to demonstrate a limited bystander effect with (2-bromoethylamino)-2-hydroxypropyl]-2- RSU1069 in vitro,23 which is consistent with earlier nitroimidazole) was used, which when hydrolysed in vivo reports, although this may not have been expected from releases RSU1069.22 As highlighted above, RSU1069 an activated nitroimidazole. The results obtained demon- shows preferential toxicity towards hypoxic cells. Radio- strated that our gene therapy approach hypersensitised therapy was chosen to control the oxic cell population the GFP/R9 cells to RSU1069 in vitro specifically under within the tumours. From the pimonidazole studies, anoxic conditions. approximately 5% of the tumour cells in both xenograft GFP/R9 and GFP5 cells grown as xenografts were types were radiobiologically hypoxic. In spite of the well- allowed to reach exponential growth (Figure 3, tumour oxygenated nature of both the GFP/R9 and GFP-5 doubling time of approximately 4 days) before combined tumour types, a single dose of 10 or 15 Gy X-irradiation treatment with radiotherapy and RB6145 – a chemical was not curative. RTV4 was achieved within a mean time precursor of RSU1069 that exhibits reduced systemic tox- of 28 days after treatment with 10 Gy and 40 days after icity in vivo.22 This allowed us to evaluate the ability of treatment with 15 Gy for both tumour types. Adminis- our therapeutic approach to potentiate the radioresponse tering RB6145 immediately after 10Gy radiotherapy in of an established tumour with a clinically relevant the GFP-5 tumours increased tumour response compared hypoxic fraction.24 Consistent with our in vitro data, with radiotherapy alone, but this did not achieve statisti- hypoxia-regulated P450R expression proved a successful cal significance (P = 0.06). In striking contrast, 50% of the approach to enhance bioreductive hypoxic cell cytotoxic- GFP/R9 tumour-bearing mice treated with this regimen ity and hence improve the efficacy of radiotherapy. Our were tumour-free 100 days after therapy. Data are the combined therapy yielded a 50% cure rate. The inability mean of two independent experiments, у5 xenografts per to cure all tumours may reflect the number of cells within treatment group (Figure 7). In addition, the combination the mass exhibiting intermediate O2 tensions insufficient of RB6145 and 10 Gy radiotherapy was well tolerated. to invoke killing with radiotherapy, but too high to achi- Although one mouse was excluded, 14 out of the 15 GFP- eve adequate P450R expression and RSU1069 toxicity. 5 and GFP/R9 tumour-bearing mice treated exhibited no This may also reflect the five-fold range in P450R activity loss of condition or weight. Further, those mice whose levels measured in the tumours. In addition, this may GFP/R9 tumours were cured by the combination of suggest that the bystander effect of RSU1069 in vivo may radiotherapy and RB6145 showed no acute or late-normal be insufficient to kill these cells. Nevertheless, we have tissue damage in the field of radiotherapy. provided a proof of principle for an oxygen-sensitive GDEPT approach and envisage that with the rationale Discussion design of prodrugs specifically activated by P450R and exhibiting greater in vivo bystander properties than One of the major goals of cancer therapy is to achieve RSU1069 (which we are actively developing) a total tumour-selective toxicity. To this end hypoxia provides tumour response would be achievable. a powerful tumour-specific condition towards which the Several enzyme prodrug systems are available for therapy can be targeted. In addition, hypoxic cells within enhancing the selective toxicity of cancer chemotherapy. a tumour represent the most resistant cell population Of these, the most evaluated pairings are cytosine being refractive to radiotherapy and probably many deaminase/5-fluorocytosine (5-FC) and thymidine forms of chemotherapy. Here we have sought to specifi- kinase/ganciclovir (GCV), the latter having demon- cally hypersensitise this resistant population to bioreduc- strated no activity when included with standard therapy tive drug treatment using an oxygen-sensitive gene ther- (both surgical resection and radiotherapy) for treatment apy approach. This significantly increased the efficacy of of patients with glioblastoma multiforme.25 Indeed, in radiotherapy in vivo, achieving a 50% cure rate using this randomised multicentre phase III trial, standard ther- radiation doses that consistently failed to control the pri- apy plus adjuvant gene therapy neither increased time to mary tumour when administered alone. tumour progression nor improved overall survival when Throughout these experiments, we have stringently compared with radiotherapy and surgery alone. A poss-

Gene Therapy Oxygen-sensitive enzyme-prodrug therapy AV Patterson et al 951

Figure 7 Hypoxia-regulated P450R expression enhances the efficacy of combined bioreductive chemotherapy and radiotherapy in vivo. The graphs in panel (a) represent the time taken for GFP/R9 and GFP-5 tumours to reach RTV4. Three different treatment groups are shown: untreated (control), radiotherapy alone (10 Gy) and combined with bioreductive chemotherapy (10 Gy + RB6145). Total tumour eradication was achieved in 50% of the GFP/R9 tumours treated with 10 Gy + RB6145. This is illustrated in the growth profiles for each individual GFP/R9 and GFP-5 tumour given this treatment (panel b). RB6145 was administered at a dose of 250 mg/kg immediately after radiotherapy. ible contributing factor to the ineffectiveness of this gene the human P450R cDNA and enhanced green fluorescent therapy approach is the quiescent nature of hypoxic cells protein (EGFP), under the transcriptional control of the rendering them resistant to the cytotoxic metabolites of HRE from the PGK-1 gene, we constructed a bicistronic GCV. In addition, hypoxic cells have also been shown to expression cassette employing pCIneo as the vector back- be refractive to 5-fluorouracil, the of 5- bone (Promega, Southampton, UK). A synthetic oligonu- FC,6 suggesting that hypoxia may restrict the efficacy of cleotide containing three copies of the minimal 18 bp these approaches. PGK-1 HRE sequence TGTCACGTCCTGCACGAC and Taken together this paper provides a persuasive argu- BglII compatible 5’ and 3’ ends was cloned in the forward ment for exploiting hypoxia using an oxygen-sensitive orientation directly upstream of the SV40 minimal pro- GDEPT approach. In addition, the inadequate response moter (SV40mp) in the pGL3 vector (Promega). The PGK- of poorly oxygenated tumours to radiotherapy and HRE-SV40 promoter region was then isolated from pGL3 chemotherapy has set a precedent for evaluating the by Nhe1/HindIII digestion and cloned between SpeI and extent of tumour hypoxia before treatment. The growing HindIII restriction sites within pCIneo, to replace its CMV availability of technologies for measuring the level of enhancer/promoter sequence, creating pCIneo PGK hypoxia in human tumours using bioreductive agents, eg SV40mp. The P450R cDNA was excised from 26 27 pimonidazole; intrinsic markers, eg Glut-1; and the pBabe/puro by EcoRI/SalI restriction digestion and development of probes for use with non-invasive imag- inserted into pCIneo creating pCIneo PGK SV40mp P450R ing techniques, such as magnetic resonance spectroscopy (pHRE-P450R). To allow co-expression of the EGFP and positron emission tomography, is beginning to revol- reporter an internal ribosome entry site (IRES) and EGFP 28 utionise the individualisation of patient treatment. This fragment was created by combining pIRES and pEGFP- would identify a cohort of patients for which an oxygen- NI (Clontech, Palo Alto, CA, USA). Specifically EGFP was sensitive gene therapy approach would have great restricted from pEGFP-NI and directionally cloned into benefit. pIRES using SmaI/NotI. To facilitate cloning of IRES EGFP at the 3’ end of P450R the IRES EGFP fragment Materials and methods was excised by NheI/NotI digestion and inserted into pCIneo PGK SV40mp P450R (previously linearised with Vector construction XbaI/NotI) to create the final vector pCIneo PGK SV40mp The full-length cDNA for human P450 reductase (2.1 kb) P450R IRES EGFP (pHRE-P450R/EGFP). This vector was has previously been constitutively expressed within the used to generate the stable HT1080 cell line GFP/R9. A vector pBabe/puro.18 To achieve co-expression of both control vector lacking the therapeutic gene P450R was

Gene Therapy Oxygen-sensitive enzyme-prodrug therapy AV Patterson et al 952 also created, pCIneo PGK SV40mp IRES EGFP, by cloning (final concentration 200 ␮M) and the rate of reduction of the NheI/NotI IRES EGFP fragment into the multiple cytochrome c was monitored at 550 nm for 3 min against cloning site of pCIneo PGK SV40mp (pHRE-EGFP). This a blank without NADPH. Initial rates of reaction were vector allowed the generation of the transcriptionally based on the co-efficient of 21 mM/cm equivalent control HT1080 cell line, GFP-5. and expressed as nmol cytochrome c reduced min/mg/protein. Protein concentrations were determ- Cell culture ined by Bradford analysis using bovine serum albumen HT1080 (human fibrosarcoma) cells were maintained in as a protein standard. exponential growth phase in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% (v/v) foetal In vitro drug sensitivity calf serum (FCS), 2 mM glutamine and 0.2 mM non- Dose response curves were determined using the MTT essential amino acids in a 95% air:5% CO2 environment. proliferation assay, which is based on the ability of viable HT1080 clones GFP-5 and GFP/R9 were selected and cells to convert a soluble tetrazolium salt, MTT, into pur- maintained in DMEM supplemented with 10% FCS, 2 ple formazan crystals. For the hypoxic experiments, all mM glutamine, 0.2 mM non-essential amino acids and plastics and media were pre-incubated in catalyst- 0.2 mg/ml geneticin 418 sulphate (G418). All cell lines induced anoxia for 72 h before use to remove any and clones were free from mycoplasma residual oxygen. RSU1069 (10 mM stock in dimethyl (assayed using Mycotect; Gibco BRL). sulphoxide (DMSO)) was serially diluted using culture medium and aliquoted into 96-well plates (100 ␮l per Transfections and clonal selection of HT1080 cells well) at twice the required final concentration. GFP-5 and 5 × 106 cells in exponential growth were transfected with GFP/R9 cells (4 × 104 cells/ml) were seeded directly into 10 ␮g of linearised pCIneo PGK min SV40 P450R IRES the drug (100 ␮l per well) and cultured for 18 h under EGFP or pCIneo PGK min SV40 IRES EGFP control vector either hypoxic or aerobic conditions. To prevent growth DNA using standard electroporation techniques.29 Cells arrest of the cells under hypoxia, the medium was also were plated at low density and 48 h later were exposed supplemented with 300 ␮M 2-deoxycytidine. The cells to 0.2 mg/ml G418. Individual colonies were then iso- were then washed free of drug and allowed to grow in lated and EGFP expression was employed to monitor the fresh medium in air for 72 h. MTT was added (to a final stability of vector expression within clonal populations concentration of 0.5 mg/ml medium) and the cells incu- for at least 8 weeks by carrying out multiple passages bated for a further 3 h after which the culture medium either in the presence or absence of G418. The clones and unconverted MTT was removed. The formazan crys- were then analysed for PGK-1 HRE-mediated EGFP tals were dissolved in 0.2 ml of DMSO and the optical expression levels allowing the selection of trancription- density at 540 nm was measured on a multiwell spectro- ally equivalent control EGFP and P450R EGFP-expressing photometer. The optical density of the dissolved crystals

HT1080 cell lines. Briefly, cells were incubated in air or is proportional to the number of viable cells allowing IC50 hypoxia (catalyst-induced anoxia, Ͻ1 ppm oxygen; Bac- values (the concentration of drug required to reduce tron anaerobic chamber, Sheldon Manufacturing, Cor- optical density by 50% compared with untreated nelius, OR, USA) for 18 h followed by 3 h of re-oxygen- controls) to be calculated and used as a measure of ation. Cells were then detached using 2 mM EDTA and cellular sensitivity to the given treatment. resuspended in FACS flow (Becton Dickinson, Mountain View, CA, USA). EGFP expression was determined by Xenograft studies flow cytometry using a Becton Dickinson FACSort with Tumour xenografts of the GFP/R9 and GFP-5 clones a single laser excitation wavelength of 488 nm. EGFP were initiated from the interdermal injection, 1 cm from fluorescence (emission wavelength 508 nm) was meas- the tail base on the midline of the back of female nude ured in the FL-1 channel that detects emitted light of mice (cba nu/nu, aged 9–12 weeks), of 5 × 106 cells pre- wavelength 500–600 nm. EGFP fluorescence was plotted pared in a 0.1 ml volume of serum-free DMEM. Once versus number of cells and median peak values obtained palpable tumours were established, daily measurements for the air and hypoxic cell samples. of tumour dimensions were taken using calipers. Tumour bearing mice were randomly allocated into treatment Reagents groups (n = 5–10 per group) and all treatments were per- RSU1069 (1-[3-aziridinyl-2-hydroypropyl]-2-nitroimidazole) formed on tumours of approximately 250 mm3 in volume and its brominated precursor RB6145 (1-[3-(2-bromo- and experiments were terminated when tumours reached ethylamino)-2-hydroxypropyl]-2-nitroimidazole) were a volume four times that at the time of treatment (RTV4). prepared according to described methods.22 Mice were restrained whilst localised radiotherapy (X- rays) was delivered at a dose rate of 2 Gy/min. RB6145 NADPH:cytochrome P450 reductase activity was prepared in 0.9% w/v saline at pH 4 and adminis- determination tered immediately after radiotherapy at a dose of 250 NADPH:cytochrome P450 reductase activity was determ- mg/kg by intra-peritoneal injection. The bioreductive ined spectrophotometrically as the NADPH dependent hypoxic marker, pimonidazole (Hydroxyprobe-1; Natu- reduction of cytochrome c. The reaction comprised 400 ral Pharmacia International, USA)26 was administered at ␮l of cytochrome c (final concentration 50 ␮M), 100 ␮lof a dose of 80 mg/kg in 0.9% w/v saline to mice carrying 10 mM potassium cyanide (final concentration 1 mM), 250 mm3 tumours by intra-peritoneal injection, 2 h before 10–300 ␮g protein lysate (10–100 ␮l volume) and 100 mM death. All procedures were carried out in accordance phosphate buffer pH 7.6, to a total volume of 0.98 ml. with the Scientific Procedures Act 1986 and in line with The reaction was equilibrated to 37°C and was initiated the UKCCCR guidelines 1999, by approved protocols by addition of 20 ␮l of 10 mM NADPH to the test cuvette (Home Office Project Licence number 40-1770).

Gene Therapy Oxygen-sensitive enzyme-prodrug therapy AV Patterson et al 953 Determination of tumour vascularity and hypoxic fraction 5 Brizel DM et al. Tumor hypoxia adversely affects the prognosis (pimonidazole binding) of carcinoma of the head and neck. Int J Radiat Oncol Biol Phys Staining for endothelial structures, pimonidazole binding 1997; 38: 285–289. and subsequent quantitative analysis of vessel density 6 Grau C, Overgaard JE. Effect of etoposide, carmustine, vincris- and hypoxic fraction have been extensively described tine, 5-fluorouracil, or methotrexate on radiobiologically oxic and hypoxic cells in a C3H mouse mammary carcinoma in situ. elsewhere.30 Briefly, 5-␮m sections were cut from snap Cancer Chemother Pharmacol 1992; 300: 277–280. frozen tumour specimens, fixed with acetone and incu- 7 Graeber TG et al. Hypoxia-mediated selection of cells with bated for 45 min with 9F1 antibody (rat monoclonal anti- diminished apoptotic potential in solid tumours. Nature 1996; body raised against mouse endothelium). Sections were 379:88–91. then sequentially treated for 30 min with tetramethyl rho- 8 Stratford IJ, Workman PB. Bioreductive drugs into the next mil- damine isothiocyanate (TRITC)-conjugated goat anti-rat lennium. Anticancer Drug Des 1998; 13: 519–528. and TRITC-conjugated donkey anti-goat antisera. Pimon- 9 Ratcliffe PJ, O’Rourke JF, Maxwell PH, Pugh CW. Oxygen sens- idazole adducts in the same sections were disclosed by ing, hypoxia-inducible factor-1 and the regulation of mam- overnight incubation with rabbit anti-pimonidazole anti- malian gene expression. J Exp Biol 1998; 201: 1153–1162. serum followed by 2 h incubation with FITC-conjugated 10 Semenza GL. Hypoxia-inducible factor 1: oxygen homeostasis donkey anti-rabbit antiserum. TRITC- and FITC-conju- and disease pathophysiology. Trends Mol Med 2001; 7: 345–350. gated antibodies were supplied by Jackson Immuno 11 Maxwell PH, Pugh CW, Ratcliffe PJ. Inducible operation of the Research Laboratories, West Grove, PA, USA. All quanti- erythropoietin 3’ enhancer in multiple cell lines: evidence for a widespread oxygen-sensing mechanism. Proc Natl Acad Sci USA tative assessment was made relative to the viable tumour 1993; 90: 2423–2427. area, ascertained from haemotoxylin and eosin-stained 12 Zhong H et al. Overexpression of hypoxia-inducible factor adjacent sections. 1alpha in common human cancers and their metastases. Cancer Res 1999; 59: 5830–5835. Immunohistochemical staining 13 Dachs GU et al. Targeting gene expression to hypoxic tumour Formalin-fixed, paraffin-embedded tumour specimens cells. Nature Med 1997; 3: 515–520. were sectioned at 3–4 ␮m thickness. P450R and pimonid- 14 Koshikawa N, Takenaga K, Tagawa M, Sakiyama S. Therapeutic azole adduct presentation was analysed in adjacent sec- efficacy of the suicide gene driven by the promoter of vascular tions. The rabbit polyclonal anti-human P450 reductase endothelial growth factor gene against hypoxic tumour cells. antibody used18 was applied at a 1/1000 dilution. Con- Cancer Res 2000; 60: 2936–2941. trols were treated with an identical dilution of rabbit 15 Adams GE, Ahmed I, Sheldon PW, Stratford IJ. RSU 1069, a 2- serum. The secondary antibody treatment (horseradish nitroimidazole containing an alkylating group: high efficiency as a radio- and chemosensitizer in vitro and in vivo. Int J Radiat peroxidase-conjugate) and visualisation of immunoreac- Oncol Biol Phys 1984; 10: 1653–1656. tivity (3,3-diaminobenzidine hydrochloride, DAB) was 16 Stratford IJ et al. RSU 1069, a nitroimidazole containing an aziri- achieved using reagents supplied in the Envision anti- dine group. Bioreduction greatly increases cytotoxicity under rabbit kit (DAKO, Glostrup, Denmark), according to the hypoxic conditions. Biochem Pharmacol 1986; 35: 105–109. manufacturer’s instructions. To visualize pimonidazole 17 Koch CJ. Unusual oxygen concentration dependence of toxicity adduct formation, sections were sequentially treated with of SR-4233, a hypoxic cell . Cancer Res 1993; 53: 3992–3997. 3% H2O2 in water (5 min, room temperature; RT), 0.05% 18 Patterson AV et al. Overexpression of human NADPH:cytoch- pronase in tris-buffered saline (TBS; pH 7.2; 5 min, RT) rome c (P450) reductase confers enhanced sensitivity to both tir- and 10% casein (10 min, RT), with TBS washes in apazamine (SR4233) and RSU1069. Br J Cancer 1997; 76: 1338– between. Hydroxyprobe-1MAb1 (mouse monoclonal; 1347. Natural Pharmacia) was applied at a 1/100 dilution 19 Coralli C et al. Limitations of the reporter green fluorescent pro- for 30 min at RT and staining visualised using the tein under simulated tumour conditions. Cancer Res 2001; 61: appropriate Envision kit (DAKO). 4784–4790. 20 Ebert BL, Firth JD, Ratcliffe PJ. Hypoxia and mitochondrial inhibitors regulate expression of glucose transporter-1 via dis- Acknowledgements tinct Cis-acting sequences. J Biol Chem 1995; 270: 29083–29089. 21 Wiesener MS et al. Constitutive activation of hypoxia-inducible The collaboration with Albert van der Kogel was initiated genes related to overexpression of hypoxia-inducible factor- via the Biomed II Concerted Action Programme 1alpha in clear cell renal carcinomas. Cancer Res 2001; 61: (Development of Methods for the Rapid Analysis of 5215–5222. Tumour Oxygenation to Allow Treatment Stratification; 22 Jenkins TC et al. Synthesis and evaluation of alpha-[[(2- contract BMH4983006). Many thanks are due to Hans haloethyl)amino]methyl]-2- nitro-1H-imidazole-1-ethanols as Peters. The Medical Research Council funded this work. prodrugs of alpha-[(1-aziridinyl)methyl]-2- nitro-1H-imidazole- 1-ethanol (RSU-1069) and its analogues which are radiosensitiz- ers and bioreductively activated cytotoxins. J Med Chem 1990; References 33: 2603–2610. 1 Gatenby RA et al. Oxygen distribution in squamous cell carci- 23 Cowen RL et al. Viral mediated gene delivery of hypoxia regu- noma metastases and its relationship to outcome of radiation lated reductive enzymes to activate bioreductive drugs and thus therapy. Int J Radiat Oncol Biol Phys 1988; 14: 831–838. improve the therapeutic index of these anti-tumour agents P295. 2 Okunieff P et al. Oxygen tension distributions are sufficient to Mol Ther 2001; 3: S103. explain the local response of human breast tumors treated with 24 Aquino-Parsons C, Luo C, Vikse CM, Olive PL. Comparison radiation alone. Int J Radiat Oncol Biol Phys 1993; 26: 631–636. between the comet assay and the oxygen microelectrode for 3 Teicher BA. Hypoxia and drug resistance. Cancer Metastasis Rev measurement of tumor hypoxia. Radiother Oncol 1999; 51: 179– 1994; 13:39–68. 185. 4 Hockel M et al. Association between tumor hypoxia and malig- 25 Rainov NG. A phase III clinical evaluation of herpes simplex nant progression in advanced cancer of the uterine cervix. Can- virus type 1 thymidine kinase and ganciclovir gene therapy as cer Res 1996; 56: 4509–4515. an adjuvant to surgical resection and radiation in adults with

Gene Therapy Oxygen-sensitive enzyme-prodrug therapy AV Patterson et al 954 previously untreated glioblastoma multiforme. Hum Gene Ther zole hypoxia probe N-(2-hydroxy-3trifluoropropyl)-2-(2-nitro-1- 2000; 11: 2389–2401. imidazolyl) acetamide (SR 4554, CRC 94/17): a non-invasive 26 Raleigh JA, Chou SC, Arteel GE, Horsman MR. Comparisons diagnostic probe for the measurement of tumor hypoxia by among pimonidazole binding, oxygen electrode measurements, magnetic resonance spectroscopy and imaging, and by positron and radiation response in C3H mouse tumors. Radiat Res 1999; emission tomography. Anticancer Drug Des 1998; 13: 703–730. 151: 580–589. 29 van den Hoff MJ, Moorman AF, Lamers WH. Electroporation 27 Airley R et al. Glucose transporter glut-1 expression correlates in ‘intracellular’ buffer increases cell survival. Nucleic Acids Res with tumor hypoxia and predicts metastasis-free survival in 1992; 20: 2902. advanced carcinoma of the cervix. Clin Cancer Res 2001; 7: 30 Bussink J et al. Optical sensor-based oxygen tension measure- 928–934. ments correspond with hypoxia marker binding in three human 28 Aboagye EO, Kelson AB, Tracy M, Workman P. Pre-clinical tumor xenograft lines. Radiat Res 2001; 154: 547–555. development and current status of the fluorinated 2-nitroimida-

Gene Therapy