In vivo gene therapy for colon cancer using adenovirus ... - Nature

Gene Therapy (2001) 8, 1547–1554  2001 Nature Publishing Group All rights reserved 0969-7128/01 $15.00 www.nature.com/gt

RESEARCH ARTICLE

In vivo gene therapy for colon cancer using adenovirus-mediated, transfer of the fusion gene cytosine deaminase and uracil phosphoribosyltransferase GA Chung-Faye, MJ Chen, NK Green, A Burton, D Anderson, V Mautner, PF Searle and DJ Kerr CRC Institute for Cancer Studies, University of Birmingham, Birmingham, UK

Virus-directed enzyme prodrug therapy (VDEPT) utilising cytosine deaminase (CD) converts 5-fluorocytosine (5-FC) into the chemotherapy agent, 5-fluorouracil (5-FU), and has entered into a clinical trial for metastatic colon cancer. To improve this system, a replication-deficient adenovirus, containing a bifunctional fusion gene, CD:uracil phosphoribosyltransferase (UPRT), was constructed (AdCDUPRT). UPRT enhances the conversion of 5-FU into its active metabolites, which inhibit DNA and RNA synthesis. In vitro, AdCDUPRT infection of colon cancer cells resulted in a marked increase in sensitisation to 5-FU, compared with AdCD-infected or uninfected cells. The corollary is a 苲100-fold and 苲10 000fold increase in sensitisation to 5-FC in AdCDUPRT-infected cells, compared to AdCD-infected and uninfected cells,

respectively. There was a strong bystander effect in vitro, 70% of tumour cells were killed by 5-FC when only 10% of cells expressed CDUPRT. In vivo, athymic mice with colon cancer xenografts treated with intratumoral AdCDUPRT and intraperitoneal 5-FC, significantly reduced tumour growth rates compared with untreated controls (P = 0.02), whereas AdCD/5-FC treated mice did not. At higher AdCDUPRT virus doses, 5-FC and 5-FU were equally effective at delaying tumour growth compared with controls. In summary, VDEPT for colon cancer utilising AdCDUPRT is more effective than AdCD and the bifunctional CDUPRT gene enables the use of either 5-FC or 5-FU as prodrugs. Gene Therapy (2001) 8, 1547–1554.

Keywords: adenovirus; VDEPT; 5-fluorocytosine; 5-fluorouracil; uracil phosphoribosyl transferase; colon cancer

Introduction Colorectal cancer (CRC) is the second commonest cause of cancer mortality in western countries. The liver is the most frequent site for metastases and up to 80% of patients will die with hepatic metastases. Less than 5% of liver metastases are resectable and the median survival in patients with unresectable liver metastases is 10.6 months.1 In metastatic disease, even with the most effective chemotherapy regimen of infusional 5-fluorouracil (5-FU), folinic acid in combination with irinotecan, the median survival is only 17.4 months.2 Hence, for most CRC patients, the prognosis remains poor and novel approaches are needed. Virus-directed enzyme prodrug therapy (VDEPT) has been proposed as an alternative treatment modality for advanced colorectal cancer. This involves the gene transfer of a non-mammalian enzyme into tumour cells, which converts an inactive prodrug into a highly toxic metabolite in the tumour milieu. Enzyme-prodrug systems aim to reduce the dose-limiting toxicities associated with systemic chemotherapy, such as myelosuppression and gastro-intestinal toxicity, by pro-

Correspondence: GA Chung Faye, CRC Institute for Cancer Studies, University of Birmingham, Edgbaston, Birmingham B15 2TA, UK Received 27 April 2001; accepted 20 July 2001

duction of the active cytotoxic species directly within the tumour and its surroundings. A commonly utilised enzyme prodrug combination is the E. coli enzyme, cytosine deaminase (CD), and 5-fluorocytosine (5-FC).3,4 CD converts 5-FC, a nontoxic antifungal agent, into 5-FU, one of the most effective anti-neoplastic agents in CRC. VDEPT strategies utilising CD/5-FC have shown antitumour effects in vitro and in vivo5 and have entered a phase I clinical trial. Patients with hepatic metastatic CRC undergo intratumoral injection of adenovirus encoding CD, under the control of a cytomegalovirus promoter (CMV), in combination with oral 5-FC.6 The cytotoxicity of 5-FU is determined by its conversion into 5-fluoro deoxyuridine monophosphate (5FdUMP) and 5-fluorouridine triphosphate (5-FUTP) as depicted in Figure 1. 5-FdUMP inhibits the enzyme, thymidylate synthase, thus depleting replicating cells of thymidine nucleotide precursors during DNA synthesis, whereas 5-FUTP inhibits RNA synthesis. The rate-limiting step in the generation of these active species is the formation, via a series of enzymatically catalysed reactions, of an intermediary metabolite, 5-fluorouridine monophosphate (5-FUMP) as shown in Figure 1. Also, small quantities of 5-FU may be converted directly into 5-FUMP, by uridine-5⬘-monophosphate synthase, the mammalian homologue of orotate phosphoribosyltransferase. However, it has been reported that the E. coli

Virus-directed cancer gene therapy GA Chung-Faye et al

1548

Figure 1 The metabolism of uracil/5-fluorouracil (5-FU) during DNA synthesis and the conversion of 5-FU into its active metabolites. UPRT converts 5-FU directly to 5-FUMP, which is then converted to its active metabolites 5-FdUMP and 5-FUTP, thus reducing degradation of 5-FU by the competitive enzyme, DPD. 5-FC, 5-fluorocytosine; 5-F-uracil, 5fluorouracil; 5-F-***, 5-fluoro-***; CD, cytosine deaminase; UPRT, uracil phosphoribosyltransferase; UMP, uridine monophosphate; dUMP, deoxy uridine monophosphate; dTMP, deoxy thymidine monophosphate; 2⬘d-5Furidine, 2⬘ deoxy 5-fluorouridine; 5-FUTP, 5-fluorouridine triphosphate; TS, thymidylate synthase; UDP, uridine phosphorylase; UDK, uridine kinase; TK, thymidine kinase; DPD, dihydropyrimidine dehydrogenase.

enzyme, uracil phosphoribosyltransferase (UPRT), can markedly potentiate the antitumour effect of 5-FU by converting it directly into its intermediary 5-FUMP, thereby leading to more efficient generation of its active metabolites, 5-FdUMP and 5-FUTP.7 In vivo, adenovirusmediated UPRT gene transfer to colon cancer xenografts has been shown to enhance the antitumour effect of 5FU.8 Thus, the rate-limiting step in the conversion of 5-FU into its active metabolites appears to be the production of 5-FUMP, which may be circumvented by the use of the UPRT gene. In the present study we investigated adenovirusmediated, CDUPRT fusion gene transfer in a VDEPT approach for treating colon cancer in vitro and in vivo. The rationale for using the CDUPRT fusion gene was that the CD gene would convert 5-FC to 5-FU, which would then be converted by UPRT to its active metabolites, thus, potentiating the antitumour activity. Also as UPRT has been shown to enhance 5-FU cytotoxicity, we compared the relative antitumour effects of the following enzyme/prodrugs combinations, CDUPRT/5-FC, CDUPRT/5-FU, as well as the conventional CD/5-FC. The vectors were E1-, E3-deleted, replication-deficient adenoviruses (Ad), with the transgene under the control of the cytomegalovirus (CMV) immediate–early promoter. Another aim of the study was to determine whether there was a bystander effect with CDUPRT/5FC as it was not known whether the bystander effect would be altered due to the different kinetics of 5-FU metabolism. The bystander effect with the CD/5-FC combination is due to diffusion of the 5-FU generated to surrounding cells,9 therefore it is conceivable that more efficient conversion of 5-FU into its active metabolites may reduce the pool of diffusible 5-FU to the surrounding cells, thereby reducing the bystander effect. Gene Therapy

Figure 2 AdCDUPRT sensitises colon cancer cells to 5-FU. SW480 (a) and WiDr colon cancer cells (b) were infected with virus, then exposed to varying concentrations of 5-FU for 4 days. Cell viability assays were performed and show that infection of the cells with the AdCDUPRT virus increased their sensitivity to 5-FU, in a virus dose-dependent manner, compared with AdCD-infected and uninfected cells. The error bars represent the standard error of the mean.

Results AdCDUPRT sensitises colon cancer cells to 5fluorouracil Expression of the CD-UPRT fusion protein in cells was predicted to increase their sensitivity to 5-FU. To test this hypothesis, SW480 colon cancer cells were infected with AdCDUPRT at multiplicities of infection (MOI) 5 and 25 p.f.u. per cell, before exposure to 5-FU at a range of concentrations. Uninfected cells, and cells infected with 25 p.f.u. per cell AdCD, served as controls, and were similarly treated with 5-FU. As shown in Figure 2a and Table 1, the uninfected SW480 cells had an IC50 of 5-FU (ie the drug concentration which gives a 50% reduction in cell number) of 6 ␮g/ml. AdCD-infected cells had a similar IC50 of 5-FU of 10.6 ␮g/ml, which was not statistically different from uninfected cells. In marked contrast, infection of the cells with the AdCDUPRT virus increased the sensitivity of the cells to 5-FU, in a virus dose-dependent manner. The IC50 of 5-FU was reduced by 43-fold at MOI 25, and by seven-fold at MOI 5, compared with uninfected cells. These results are therefore consistent with the hypothesis that UPRT expressed by cells after infection by AdCDUPRT improved conversion of 5-FU to its active metabolites as AdCD-infected cells were not sensitised to 5-FU, compared with uninfected cells. Similar results were obtained in another colon cancer cell line, WiDr (Figure 2b, Table 1). In preliminary experiments, these cells were found to be more susceptible to adenoviral infection than SW480 cells, so this experiment was performed using virus MOI of just 1 and 5 p.f.u. per cell to avoid excessive cell kill due to nonspecific virus toxicity. There was no significant difference in sensitivity


1549

Table 1 IC50 of 5-FU in SW480 and WiDr cells from Figure 2a and b Cell type

SW480

WiDr

Virus and MOI

IC50 of 5-FU (␮g/ml)

95% confidence interval (␮g/ml)

Decrease in IC50 from uninfected cells

None AdCD (25 MOI) AdCDUPRT (5 MOI) AdCDUPRT (25 MOI)

6.02 10.57 0.91 0.14

3.13–11.58 7.96–14.04 0.57–1.46 0.12–0.17

0.6 7 43


10.25 15.37 1.99 0.21

1.24–84.83 7.29–32.43 0.99–4.01 0.10–0.42

0.7 5 49

to 5-FU between the uninfected (IC50 10.3, CI 1.2–84.8) and AdCD-infected cells uninfected (IC50 15.4, CI 7.3– 32.4), whereas infection with AdCDUPRT decreased the IC50 by five-fold at 1 p.f.u. per cell (IC50 2.0, CI 1.0–4.0) and 苲50-fold at 5 p.f.u. per cell (IC50 0.2, CI 0.1–0.4) compared with uninfected cells. Enhanced sensitisation of colon cancer cells to 5-FC by AdCDUPRT Cytosine deaminase can convert 5-FC to 5-FU, and hence infection with AdCD (25 p.f.u. per cell) sensitised the cells to 5-FC, reducing the IC50 57-fold compared with uninfected cells (Figure 3a and Table 2). However, infec-

Figure 3 AdCDUPRT also sensitises colon cancer cells to 5-FC. SW480 (a) and WiDr colon cancer cells (b) were infected with virus and exposed to varying concentrations of 5-FC for 4 days. Cell viability assays were then performed. Infection with AdCD sensitised the cells to 5-FC, reducing the IC50 by between 35- and 57-fold compared with uninfected cells (Table 2). Infecting cells with AdCDUPRT at the same MOI resulted in a far greater increase in sensitivity to 5-FC, reducing the IC50 by between 苲2800 and 9700-fold compared with uninfected cells (Table 2) and by between 苲80 and 170 compared with AdCD-infected cells. Infecting cells with AdCDUPRT at a lower MOI still resulted in an increased sensitisation to 5-FC compared with uninfected cells and gave comparable levels as AdCD-infected cells at higher MOI. The error bars represent the standard error of the mean.

tion of the cells with AdCDUPRT at the same MOI resulted in a far greater increase in sensitivity to 5-FC, with an almost 10 000-fold reduction in IC50 from 4729 ␮g/ml in uninfected cells to 0.49 ␮g/ml in AdCDUPRT-infected cells. Infecting cells with AdCDUPRT at a lower multiplicity of 5 p.f.u. per cell still resulted in greater than 100-fold sensitisation compared with uninfected cells. A similar increase in sensitivity to 5-FC was seen in WiDr cells infected with AdCDUPRT (IC50 1.5, CI 0.8– 2.8), compared with AdCD-infected (IC50 120, CI 63–226) and uninfected cells (IC50 4178, CI 3174–5499) (see Figure 3b and Table 2). Previous experiments where cells were infected with a similar replication-deficient adenovirus with the green fluorescent protein as the transgene (AdGFP) as a control, did not show a statistically significant increase in sensitivity to 5-FC compared with uninfected cells (data not shown). Measurement of 5-FU released into culture medium Improved intracellular metabolism of 5-FU may diminish the amount released to the surrounding fluid, thus reducing the bystander effect. To investigate this, in the SW480 experiment described previously, the concentration of 5FU in the culture medium was measured at the end of 4 days incubation with a range of 5-FC concentrations. The culture medium from uninfected cells contained no detectable 5-FU after incubation with 5-FC. Infection with either AdCD or AdCDUPRT at MOI 25 p.f.u. per cell showed approximately 50% conversion of the input 5-FC (15, 150 or 1000 ␮g/ml) to 5-FU (Figure 4). As previously shown (Figure 2a, b), these concentrations of 5-FU are sufficient to result in significant cytotoxicity to uninfected colon cancer cells, thus implying a significant bystander effect may be obtained by 5-FU diffusion. The measured 5-FU levels were approximately 25% lower in the AdCDUPRT-infected cultures than in cultures infected with Ad-CD, however, these differences did not reach statistical significance. These data confirm that the 150fold greater sensitivity to 5-FC of the AdCDUPRTinfected cells relative to AdCD-infected cells (Table 2) cannot be explained by increased production of 5-FU. Moreover, after infection by AdCDUPRT at the lower MOI of 5 p.f.u. per cell, markedly less 5-FU was released into the culture medium (Figure 4, P = 0.0001, two-way analysis of variance). However, this still resulted in an Gene Therapy


1550

Table 2 Cell type

SW480

WiDr

IC50 of 5FC in SW480 and WiDr cells from Figure 3a and b Virus and MOI

IC50 of 5-FC (␮g/ml)

95% confidence interval (␮g/ml)

Decrease in IC50 from uninfected cells


4729 82.96 40.78 0.49

337–6703 68.58–100.4 23.44–70.94 0.36–0.68

57 116 9651


4178 119.8 128.4 1.499

3174–5499 63.46–226.3 93.79–175.7 0.81–2.79

35 33 2787

Figure 4 The cell culture medium from Figure 3a was collected and 5FU levels were determined by HPLC for a number of starting concentrations of 5-FC. No 5-FU was detected in the culture medium from uninfected cells. Infecting SW480 cells with either AdCD or AdCDUPRT at MOI 25 p.f.u. per cell showed approximately 50% conversion of the input 5-FC (15, 150 or 1000 ␮g/ml) to 5-FU. Although 5-FU levels were 25% lower in the AdCDUPRT-infected cells compared with the CD-infected ones (25 p.f.u. per cell), these differences did not reach statistical significance. However, as expected, markedly less 5-FU was detected in the culture medium from cells infected with 5 p.f.u. per cell AdCDUPRT (P = 0.0001, two-way analysis of variance).

IC50 comparable to that achieved with 25 p.f.u. per cell AdCD (Table 2). A significant bystander effect is seen with AdCDUPRT and 5-FC To test the killing of uninfected cells in the presence of cells infected with Ad-CDUPRT, infected and uninfected SW480 cells were mixed in varying proportions. The mixed cultures were exposed to 50 ␮g/ml 5-FC, and their viability determined. This concentration of prodrug was chosen because it is nontoxic to uninfected cells (Figure 2a) and corresponds to the steady-state concentration pharmacokinetically achievable in patients. As expected, uninfected cells showed 100% viability; 70% of cells were killed when only 10% expressed CDUPRT, indicating a significant killing of uninfected, ‘bystander’ cells; and even when 100% of cells were infected with AdCD, only 60% cell killing was achieved at this prodrug concentration (Figure 5). The difference between these three groups was statistically significant (P = 0.0001, one-way analysis of variance). Increasing the proportion of CDUPRT-expressing cells to 50% achieved similar levels of cytotoxicity as when the total cell population expressed CDUPRT (苲80–85% cell killing). Gene Therapy

Figure 5 CDUPRT and 5-FC exhibit a strong bystander effect. CDUPRT-expressing SW480 cells (represented by solid bars) were mixed with varying proportions of wild-type cells and exposed to 5-FC at a dose of 50 ␮g/ml, which corresponds to the steady-state concentration pharmacokinetically achievable in patients. After 4 days the number of viable cells was then determined as described previously. Wild-type SW480 cells were resistant to 5-FC at that dose. However, 70% of cells were killed when only 10% expressed CDUPRT, whereas only 60% of cells were killed even when 100% of cells expressed CD (CD-100, represented by the dotted bar). The error bars represent the standard error of the mean.

In vivo, Ad CDUPRT and 5-FC or 5-FU have significant antitumour effects Athymic Balb C nu/nu mice with established subcutaneous human colon cancer xenografts (SW480) were treated with three consecutive daily intratumoral injections (108 p.f.u.) of either AdCD, AdCDUPRT or AdGFP. Mice in each group were subsequently randomised to received intraperitoneal (IP) administrations of either 5-FC, 5-FU or phosphate-buffered saline (PBS). Tumour volumes were measured weekly (Figure 6). The only group where there was a significant reduction of tumour growth rate over time, compared with the untreated controls, was the AdCDUPRT/5-FC-treated group (P = 0.02, random effects model, Table 3). Further analyses on the pooled groups of mice treated with either 5-FC, 5-FU or PBS (irrespective of virus) or AdCD, AdCDUPRT or AdGFP (irrespective of prodrug) showed no difference between the groups. In a smaller experiment to test a higher daily dose of administered virus (5 × 108 p.f.u.), mice with established tumours were treated with AdGFP/PBS (n = 6),


1551

Figure 7 In a similar experiment with a higher daily virus dose (5 × 108 p.f.u.), mice were treated with the following combination: AdGFP/PBS (n = 6), AdCDUPRT/5-FC (n = 7) and AdCDUPRT/5-FU (n = 7). There were significant reductions in tumour growth rates in both the AdCDUPRT/5-FC and AdCDUPRT/5-FU groups compared with the AdGFP/PBS control group (P = 0.02, random effects modelling). The error bars represent the standard error of the mean.

AdCDUPRT/5-FC (n = 7) and AdCDUPRT/5-FU (n = 7). There were significant reductions in tumour growth rates over time in both the AdCDUPRT/5-FC and AdCDUPRT/5-FU groups compared with the AdGFP/PBS control group (Figure 7, random effects model, P = 0.02). Thus, in vivo, AdCDUPRT in combination with either 5-FC or 5-FU appeared to have a similar antitumour effect at higher viral doses (5 × 108 p.f.u. per injection), but only AdCDUPRT/5-FC was effective at lower viral doses (108 p.f.u. per injection). Figure 6 AdCDUPRT and 5-FC have a significant antitumour effect in vivo. Balb C nu/nu mice with established subcutaneous SW480 colon cancer xenografts were treated with three consecutive daily intratumoral injections (108 p.f.u.) of either AdCD, AdCDUPRT or AdGFP. Mice in each group subsequently received IP administrations of either 5-FC, 5FU or PBS. For the purposes of clarity the experiment has been presented on three different graphs. The untreated control group is the same in all three graphs. The error bars represent the standard error of the mean. Compared with the untreated controls, there was a significant reduction of tumour growth rate only in the AdCDUPRT/5-FC-treated group (P = 0.02, random effects modelling, Table 3).

Table 3 Slope estimates from the random effects modelling using original tumour sizes Group

Untreated AdCDUPRT/5-FU AdCD/PBS AdCDUPRT/PBS AdGFP/5-FU AdGFP/PBS AdCD/5-FU AdGFP/5-FC AdCD/5-FC AdCDUPRT/5-FC

Slope estimate

P value from baseline untreated group

48.67 45.65 45.17 38.42 38.33 38.12 36.18 29.26 28.70 10.83

0.85 0.83 0.53 0.53 0.52 0.45 0.24 0.23 0.02

Discussion The combination of CD/5-FC in VDEPT has shown promise in preclinical studies and has entered a clinical trial using a replication-deficient adenovirus encoding CD similar to AdCD used in this study, in combination with oral 5FC.6 However, by utilising an adenovirus encoding the bifunctional fusion gene CDUPRT, we have shown increased sensitisation to 5-FC by 80- to 168-fold compared with AdCD, and by 2800- to 9600-fold relative to uninfected cells. The increased cytotoxicity of the AdCDUPRT/5-FC combination compared with AdCD/5-FC is attributable to enhanced 5-FU conversion to its active metabolites, mediated by the UPRT component of the fusion gene, as AdCDUPRT-infected cells had increased sensitivity to 5-FU, compared with Ad-infected and uninfected cells, whereas there was no increase in 5FU sensitisation in AdCD-infected cells compared with uninfected cells. Measurement of 5-FU levels in the culture medium of infected cells treated with 5-FC confirmed that the increased sensitisation achieved using AdCDUPRT compared with AdCD was not due to increased generation of 5-FU by AdCDUPRT-infected cells. Indeed, the measured levels of 5-FU generated from 5-FC in medium of AdCDUPRT-infected cells were about 25% lower than in medium of AdCD-infected cells, consistent with improved intracellular conversion of 5-FU to 5-FUMP due to the UPRT moiety; however, this reduction was not statistically significant. Even, in cells infected with lower doses of AdCDUPRT and consequently producing 100-fold less 5-FU in the culture Gene Therapy


1552

medium, there was still equivalent or improved cytotoxicity compared with AdCD-infected cells. These data confirm that the increased cytotoxicity of the AdCDUPRT/5FC combination compared with AdCD/5-FC is due to enhanced 5-FU conversion to its active metabolites, mediated by the UPRT component of the fusion gene. Thus, UPRT expression can overcome the rate-limiting conversion of 5-FU to 5-FUMP and hence increase its antitumour effects. The difficulty in achieving efficient gene delivery to all cells in a tumour mass in vivo is a major limitation for successful cancer gene therapy. Therefore, an important feature of VDEPT is the bystander effect, where surrounding, untransduced cells are also killed by the diffusion of toxic metabolites, which may partly compensate for limited gene delivery. The bystander effect described with CD/5-FC is due to diffusion of 5-FU to surrounding cells.9 It was not known whether the bystander effect would be altered in the CDUPRT/5-FC combination due to rapid conversion of 5-FU to its active metabolites, which may not diffuse readily into surrounding cells. However, the CDUPRT/5-FC combination demonstrated a strong bystander effect in vitro with significant cell killing when only 10% of cells express the transgene. This was better than when 100% of cells expressed CD, thus demonstrating the marked superiority of AdCDUPRT/5FC compared with AdCD/5-FC. Extrapolation of in vitro data is difficult but this suggests that if a small proportion of the tumour cell population expresses CDUPRT then a significant antitumour effect may be seen. 5-FC can be given orally, is well-tolerated and steady state levels of 50 ␮g/ml can be readily achieved in patients.10 At this concentration, 70% of tumour cells are killed in vitro when only 10% of cells express CDUPRT. Thus, the 5-FC levels achievable in humans are within the therapeutic range, provided there is sufficient CDUPRT expression. Other workers have described an immunologicallymediated ‘distant bystander effect’ whereby killing cancer cells in vivo with CD/5FC can cause immune rejection of unmodified tumour cells injected subsequently, or previously established.11 Although testing for immunologically mediated effects was beyond the scope of the present study, it is unlikely that the addition of the UPRT moiety would interfere with the mechanisms of the distant bystander effect. In vivo, a significant antitumour effect was seen with the AdCDUPRT/5-FC combination. At the daily virus dose of 108 p.f.u. per injection there was a significant reduction in the tumour growth rates in the AdCDUPRT/5-FC group compared with the untreated controls (P = 0.02, random effects model). Although there were reductions in the tumour growth rates in the AdCD/5-FC and AdGFP/5-FC groups compared with the untreated controls, these were not statistically significant (Table 3). It is possible that there may be a small antitumour effect with AdCD/5-FC in vivo which this study may not have been sufficiently powered to detect. In vitro, a weaker antitumour effect was seen with the AdCD/5-FC combination (Figures 3a and b) but there was no evidence that 5-FC alone or in combination with AdGFP (data not shown) exerted any antitumour effect. No difference was seen between the tumour growth rates in the AdCDUPRT/5-FU group (108 p.f.u. per injection), compared with the untreated controls. However, at a higher daily viral dose (5 × 108 p.f.u. per injection), simi-

Gene Therapy

lar reductions in the growth rates were seen in both the AdCDUPRT/5-FC and AdCDUPRT/5-FU groups compared with the AdGFP/PBS control group (P = 0.02, random effects modelling). This result may be explained by the fact that, in vivo, only a small proportion of cells express the transgene and relatively low levels of 5-FU can be administered due to systemic toxicity, thereby reducing the antitumour effect. In vitro, the therapeutic ratio of AdCDUPRT/5-FU, defined as: IC50 of 5-FU in uninfected cells/IC50 of 5-FU in AdCDUPRT-infected cells, varies between 42 and 50 (Table 1). By contrast, the therapeutic ratio of AdCDUPRT/5-FC, (IC50 of 5-FC in uninfected cells/IC50 of 5-FC in AdCDUPRT-infected cells) varies between 2800 and 9600 (Table 2) and 50-fold higher 5-FC doses were given to the mice. Thus, in vivo AdCDUPRT/5-FC, with its high therapeutic ratio, may be more effective than AdCDUPRT/5-FU at lower virus doses. However, provided a sufficiently high virus titre is administered, AdCDUPRT/5-FC and AdCDUPRT/5FU appear to be equally effective. In the clinical setting, infusional 5-FU regimens are already established as effective treatment for metastatic colorectal cancer and in the context of a VDEPT approach with AdCDUPRT, systemic 5-FU may provide an additional antitumour effect on disseminated CRC micrometastases. Another possibility is to administer both 5-FC and 5-FU in combination as this may provide additional antitumour benefits without significant increases in toxicity. Thus, VDEPT strategies using the bifunctional CDUPRT fusion gene in combination with either 5-FC or 5-FU as prodrugs, appear promising. The AdCDUPRT/5FC combination is more effective in vitro and in vivo than AdCD/5-FC, which has reached clinical trials. At sufficiently high doses in vivo, AdCDUPRT can also sensitise CRC cells to 5-FU, a treatment already shown to have survival benefits in metastatic CRC. Translation of this work into clinical trials may involve comparisons between 5-FC and 5-FU as prodrugs, or even both given in combination. Furthermore, a VDEPT approach using oncolytic adenoviruses encoding CDUPRT may improve transgene delivery and may also provide a synergistic effect of the virus with 5-FC or 5-FU, as has been reported in a phase II clinical study utilising an oncolytic adenovirus combined with 5-FU and cisplatin12 and this approach warrants further investigation.

Materials and methods All cell lines were obtained from the ATCC and grown in DME/Hepes, supplemented with 10% fetal calf serum, 2 mm glutamine, 100 IU/ml penicillin and 100 mg/ml streptomycin (Sigma, UK). Cell viability assays were performed as follows: cells were incubated with virus for 90 min, at the appropriate MOI, before replacement of medium; 24 h after incubation, 5000 cells were transferred to each of triplicate wells of 96-well plates; after a further 24 h to allow the cells to readhere, the medium was replaced with medium containing serial dilutions of prodrug and exposed to prodrug for 4 days before the number of remaining viable cells was estimated by the formazan-containing MTS assay, (CellTiter 96 AQueous one solution cell proliferation assay; Promega), according to the manufacturer’s instructions. Each experiment was performed independently two to three times; the results


were similar and the data from repeated experiments have therefore been combined in the figures shown. The level of 5-FU generated in the supernatant was measured by pooling the triplicate supernatant and analysis by HPLC. Bystander experiments were performed by mixing CDUPRT-expressing SW480 cells with varying proportions of uninfected cells. The CDUPRT-expressing SW480 cells had been previously exposed to high titres of virus for 24 h (50 p.f.u. per ml) and washed twice with PBS to remove any unattached virus. This process ensured that ⬎90% of cells expressed the transgene. The mixed cells were then exposed to prodrug for 4 days and cell viability was assessed as previously described. Adenovirus construction The CD gene was amplified from E. coli DNA with primers CCACCATGTCGAATAACGCTTTACAAACAATT and GGATCCTTATTAACGTTTGTAATCGATGGCTTC TG, and cloned downstream of the CMV promoter into the HpaI site of pxLNCX.13 The CMV-CD cassette was then excised with BamHI and cloned into the BamHI site of an E1-, E3-deleted left end adenoviral plasmid previously described.14 Recombinant viruses were harvested after homologous recombination, following transfection of the left end plasmid with its complementary right end plasmid into HER-911 cells, which provided E1 in trans.15 Monoclonal viruses were rescued after two rounds of plaque purification under agar and propagated in HER911 cells, followed by centrifugal purification in caesium chloride gradients. Virus titre was determined by plaques assays after serial dilutions of virus on HER-911 cells under agar, and expressed as p.f.u./ml. CDUPRT was obtained by PCR from the pGT65 plasmid (InvivoGen, USA), using the following primers CGAAGCTTCCACCATGGTGTCGAATAACG and CTT TGGTACGAAATAAGGATCCAAGCTTGC. The PCR fragment was then subcloned into a similar E1-, E3deleted, left hand adenoviral plasmid and recombinant virus was rescued as previously described. In vivo experiments All experimental procedures were performed with the authority of the Home Office, UK guidelines and in accordance with good animal practise. Balb C nu/nu mice, aged 6–8 weeks, were obtained from Harlan Laboratories. Tumours (SW480) grew in the mice after implanting 5 × 106 cells in 200 ␮l subcutaneously in the right flank. Established tumours were randomised into three groups and injected with three consecutive daily injections (50 ␮l each) of either AdCDUPRT (n = 21), AdCD (n = 21) or a similar adenovirus encoding green fluorescent protein (AdGFP, n = 21). Two days after the first viral injection, each group was subdivided into three further groups, each containing seven mice, and given 12 consecutive daily intraperitoneal injections of prodrug: 5FC in PBS (500 mg/kg), 5-FU in PBS (10 mg/kg) and PBS. Tumour volume estimations were made using the following formula, volume = length × width2/2. Tumours were measured weekly, using callipers, by a blinded observer. Mice were culled when tumours approached 10% of body weight. Statistical methods Cell viability graphs, IC50s and 95% confidence intervals were generated by GraphPad Prism (GraphPad

Software). A two-way analysis of variance was used to assess differences between the viruses in the amount of log (5-FU) generated with varying exposures to concentrations of 5-FC. A one-way analysis of variance was used to assess the bystander effect between viruses on the log transformed cell viability data. The Duncan’s multiple range test then identified where any differences lay. Ninety-five per cent confidence intervals for the IC50s were calculated and used to identify any differences between viruses. In vivo, the relationship between the actual tumour volumes and the combinations of virus and prodrug was represented as a random effects model, which allowed the intercepts and regression coefficients to vary among mice.16 The significance of the deviations in slope estimates from the baseline estimate for the untreated group was assessed within the model. The analysis was carried out using SAS statistical software (SAS Institute, SAS Circle, Cary, NC, USA).

1553

Acknowledgements This work was supported by the Cancer Research Campaign, UK and the Medical Research Council, UK.

References 1 Ballantyne GH, Quin J. Surgical treatment of liver metastases in patients with colorectal cancer. Cancer 1993; 71: 4252–4266. 2 Douillard JY et al. Irinotecan combined with fluorouracil compared with fluorouracil alone as first-line treatment for metastatic colorectal cancer: a multicentre randomised trial (published erratum appears in Lancet 2000 Apr 15;355(9212):1372). Lancet 2000; 355: 1041–1047. 3 Mullen CA, Kilstrup M, Blaese RM. Transfer of the bacterial gene for cytosine deaminase to mammalian cells confers lethal sensitivity to 5-fluorocytosine: a negative selection system. Proc Nat Acad Sci USA 1992; 89: 33–37. 4 Austin EA, Huber BE. A first step in the development of gene therapy for colorectal carcinoma: cloning, sequencing, and expression of Escherichia coli cytosine deaminase. Mol Pharm 1993; 43: 380–387. 5 Hirschowitz EA et al. In vivo adenovirus-mediated gene transfer of the Escherichia coli cytosine deaminase gene to human colon carcinoma-derived tumors induces chemosensitivity to 5-fluorocytosine. Hum Gene Ther 1995; 6: 1055–1063. 6 Crystal RG et al. Phase I study of direct administration of a replication deficient adenovirus vector containing the E. coli cytosine deaminase gene to metastatic colon carcinoma of the liver in association with the oral administration of the pro-drug 5-fluorocytosine. Hum Gene Ther 1997; 8: 985–1001. 7 Tiraby M et al. Concomitant expression of E. coli cytosine deaminase and uracil phosphoribosyltransferase improves the cytotoxicity of 5-fluorocytosine. FEMS Microbiol Lett 1998; 167: 41–49. 8 Kanai F et al. Adenovirus-mediated transduction of Escherichia coli uracil phosphoribosyltransferase gene sensitizes cancer cells to low concentrations of 5-fluorouracil. Cancer Res 1998; 58: 1946–1951. 9 Lan KH et al. Tumor-specific gene expression in carcinoembryonic antigen-producing gastric cancer cells using adenovirus vectors. Gastroenterology 1996; 111: 1241–1251. 10 Recommended 5-fluorocytosine dose in patients. British National Formulary 1999; 37: 276–277. 11 Pierrefite-Carle V et al. Cytosine deaminase/5-fluorocytosinebased vaccination against liver tumors: evidence of distant bystander effect. J Nat Cancer Inst 1999; 91: 2014–2019. 12 Khuri FR et al. A controlled trial of intratumoral ONYX-015, a selectively-replicating adenovirus, in combination with cisplatin and 5-fluorouracil in patients with recurrent head and neck cancer. Nat Med 2000; 6: 879–885. Gene Therapy


1554

13 Green NK et al. Sensitization of colorectal and pancreatic cancer cell lines to the prodrug 5-(aziridin-1-yl)-2,4-dinitrobenzamide (CB1954) by retroviral transduction and expression of the E. coli nitroreductase gene. Cancer Gene Ther 1997; 4: 229–238. 14 Latham JP, Searle PF, Mautner V, James ND. Prostate-specific antigen promoter/enhancer driven gene therapy for prostate cancer: construction and testing of a tissue-specific adenovirus

Gene Therapy

vector. Cancer Res 2000; 60: 334–341. 15 Fallaux FJ et al. Characterization of 911: a new helper cell line for the titration and propagation of early region 1-deleted adenoviral vectors. Hum Gene Ther 1996; 7: 215–222. 16 Laird NM, Ware JH. Random-effects models for longitudinal data. Biometrics 1982; 38: 963–974.