L. monocytogenes ΔaroA/B trpS (WL-150), pTRPS was used as host strain in which pTRPS was replaced by the plasmids described in (Fig. 1). E. coli DH10b was used as host for all DNA cloning experiments.

DNA-delivery plasmids were derived from pSP118 22. For the construction of pSP118-P_4xTETP, P_4xTETP was PCR amplified with primers 4xTETP-PstI-for (5'-ctatcgataggtaccctgcagagttcctgccag-3') and 4xTETP-NotI-rev (5'-cacaaggcggccgcgaactggctaattggagtcac-3') using pGL3 4xTETP (kindly provided by S. Hemmi) 15 as template. The 4xTETP-fragment and pSP118 were ligated after cutting with PstI and NotI resulting in pSP118-P_4xTETP. Egfp was subcloned from pSP118-P_CMV-EGFP 22 into pSP118-P_4xTETP by NotI restriction and subsequent ligation from which pSP118-P_4xTETP-EGFP was obtained. Fcu1 was PCR amplified using primers FCU1-NotI-Kozak-for (5'-aaaaaagcggccgctcgccaccatggtgacagggggaatggc-3') and FCU1-NotI-rev (5'-aaaaaagcggccgcttaaacacagtagtatctgtc-3') with pCIneo-FCU1 as template 27, thereby introducing NotI-sites and a Kozak sequence for efficient translation initiation at the 5'-end of the open reading frame. The resulting fragment was introduced into pSP118 and pSP118-P_4xTETP using NotI restriction resulting in pSP118-P_CMV-FCU1 and pSP118-P_4xTETP-FCU1. Plasmids pSP118-P_CMV-PNP and pSP118-P_4xTETP-PNP were obtained after PCR amplification of a translation initiation signal from pCMVβ (Clontech) using primers TLI-NotI-for (5'-gtacccgcggccgcaattcccggggatcg-3') and TLI-rev (5'-aatgtgtggggtagccatggtgacttcttttttgct-3') and the open reading frame of deoD from E. coli chromosomal DNA using primers deoD-for (5'-agcaaaaaagaagtcaccatggctaccccacacatt-3') and deoD-NotI-rev (5'-aaaaaagcggccgcttactctttatcgcccagc-3') followed by recombinant PCR of the two fragments using TLI-NotI-for and deoD-NotI-rev. The resulting fragment was cloned into pSP118 and pSP118-P_4xTETP using NotI.

For protein delivery deoD was amplified using the primers deoD-BamHI-for (5'-aaaaaaggatccatggctaccccacacattaat-3') and deoD-SacISalI-rev (5'-aaaaaacagctggtcgacttactctttatcgcccagcag-3') with E. coli chromosomal DNA as template. The resulting framgment was cut with BamHI and SacI as was pSP2 P_CMV 22(thereby removing ply118, trpS and P_CMV) and ligation resulted in pUNK P_actA-SP-PNP. The trpS gene was again introduced as described previously 22 using the introduced SalI site resulting in the protein delivery plasmid pSP0-P_actA-SP-PNP.

COS-1 (Cercopithecus aethiops kidney fibroblasts), 4T1 (murine mammary gland tumor; kindly provided by E. Lukanidin), and B16 (murine skin melanoma; kindly provided by J. Becker) cells were cultured in RPMI 1640 medium supplemented with 2 mM L-glutamine (Gibco, Germany) (RPMI) and 10% fetal calf serum (FCS, Biochrom, Germany), and were maintained at 37°C in a 5% CO₂ atmosphere.

For prodrug treatment, cells were seeded in 24-well plates 4 days prior prodrug addition. Bactofection was carried out as reported previously 22 and started 3 days before addition of prodrugs. In brief, cells were washed with RPMI, infected for 1 h with a multiplicity of infection (MOI) of 5 bacteria per cell, washed again with RPMI, incubated with gentamicin-containing medium (100 μg/ml) for 1 h which was replaced by medium containing 10 μg/ml gentamicin for subsequent culture.

For protein-delivery, cells were infected with MOI 200 five hours prior to prodrug addition for 1 h. Cells were washed with RPMI and cultivated with gentamicin-containing medium (100 μg/ml) which was replaced with medium containing 10 μg/ml gentamicin after another 1 h. All cells were trypsinized, diluted, and reseeded in 96 well plates (approximately 2 × 10³ cells per well resuspended in 100 μl RPMI containing 10% FCS) 1 h before prodrug addition. Prodrugs were then added in another 100 μl medium resulting in a final concentration of 50 μM for MePdR (Gibco, Germany), 88 μM for Fludarabine (kindly provided by Schering, Germany) and 1 mM for 5-FC (Sigma, Germany) and the prodrug containing medium was left on the cells till the end of the experiment.

The amount of viable cells after prodrug addition was measured using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, Sigma), following the manufacturer's instructions. At 3, 5, or 7 days after prodrugs were added, medium was replaced by 50 μl MTT solution (2.5 mg/ml MTT dissolved in RPMI without phenol red) and incubated for 4 h at 37°C in a 5% CO₂ atmosphere. After removal of the MTT solution 1 N HCl diluted in isopropanol was added and the probes were measured at a wavelength of 570 nm. Non-prodrug treated cells were used as reference and were considered as 100% viable.

HPLC assay was performed as described earlier 14. In brief, 100 μl cell supernatant was boiled for 15 min and cell debris was removed by centrifuging at 14000 rpm for 15 min at room temperature. Thirty microliters of the supernatant were injected onto a Waters 625 LC-system chromatogroph equipped with a Waters 486 tunable absorbance detector (Waters GmbH, Germany). A Nucleosil 120-5 C₁₈ (4 × 250 mm) column (Machery Nagel, Germany) was used with a mobile phase of 50 mM NH₄H₂PO₄ and 90% acetonitrile (Roth, Germany) (flow rate of 1.0 ml/min). MePdR (and MeP) were detected at a UV wavelength of 254 nm, and were identified and quantified by comparing their retention times and absorption spectra with authentic samples.

Three days after infection, bactofected cells were trypsinized and resuspended with PBS. Cell viability was determined by staining cells with propidium iodide (PI, 1.0 μg/ml). Since nonviable cells tend to fluoresce at a similar wavelength as EGFP, PI-positive cells were gated out from the measurement. A minimum of 5 × 10⁴ cells were then measured using an Epics XL flow cytometer (Beckman Coulter) and data analysis was performed using WinMDI 2.8 (J. Trotter 1993–1998). Cells infected with corresponding strains not encoding EGFP served as negative control. Cells infected for 4 h with WL-150 pSP0-P_actA-GFP were analyzed with the same method.

PNP was overexpressed in E. coli M15 using the pQE30 QIAexpress-system (Qiagen, Germany). Therefore, deoD was PCR amplified using 6His-BamHI-deoD-for (5'-gataaaggatccgctaccccacacatt-3') and deoD-HindIII-rev (5'-caattaaagcttatcgcccagcagaac-3') with E. coli chromosomal DNA as template and cloned into pQE30 using BamHI and HindIII restriction sites. 6His-PNP expression was induced by addition of 1 mM IPTG and prepared under native conditions on a Ni-NTA column as described by the manufacturer. Lysis buffer contained 50 mM NaH₂PO₄, 300 mM NaCl, and 10 mM imidazole; wash buffer contained 50 mM NaH₂PO₄, 300 mM NaCl, and 20 mM imidazole; and elution buffer contained 50 mM NaH₂PO₄, 300 mM NaCl, and 250 mM imidazole.

Fifty microgram purified 6His-PNP were suspended in Montanide adjuvans (Merck, Germany) and injected intraperitoneally into CD-1 mice. After 4 weeks, this procedure was repeated and serum was harvested 10 days later.

For Western blot analysis 50 ml of exponentially growing Listeria-cultures were centrifuged and the pellet was resuspended in 100 μl/OD₆₀₀ 4× Laemmli buffer. Proteins in the supernatant were precipitated on ice for one hour with 10% Trichloroacetic acid. Pelleted proteins were washed with acetone and dissolved in 500 μl 4× Laemmli buffer.

Proteins from mammalian cells were isolated from cells either bactofected for 3 d or infected with the protein-delivery strain for 4 h. Isolation was performed from a 10 cm culture dish using 500 μl RIPA lysis buffer (0.15 M NaCl, 50 mM Tris-HCl, 1% NP-40, 0.5% Deoxycholic acid, 0.1% SDS) with subsequent removal of DNA using sepharose. 15 μl of the supernatant were mixed with 15 μl 4× Laemmli buffer.

After incubation of the protein solution at 98°C for 5 min 30 μl were loaded on a 15% SDS-polyacrylamid gel and electrophoresis was performed by the method of Laemmli 28. Immunoblotting was performed by a semidry method, with Hybond-ECL nitrocellulose membranes (Amersham Biosciences, Germany). Equal protein load was confirmed by Ponceau S-staining of the nitrocellulose membrane. After incubation with horseraddish peroxidase-conjugated secondary antibodies (Dianova, Germany) and a chemiluminescens-based immunoblot assay (ECL, Amersham Biosciences, Germany) was performed according to the provided instructions.

Different DNA-delivery plasmids were constructed and introduced into an attenuated L. monocytogenes (Lm) strain carrying deletions in aroA, aroB 29, and trpS 22 (WL-150). All DNA-delivery plasmids constructed in this study were derived from the recently described vectors pSP118 and pSP0. These plasmids were stabilized by inserting the trpS gene (essential for cell viability) into the plasmid and simultaneous deletion of the chromosomal trpS copy from the genome. The plasmid pSP118 carried in addition the recently designed autolysis cassette consisting of a phage lysin gene (ply118) under the control of the listerial actA promoter P_actA 21. The introduced genes encoding the prodrug/drug converting enzymes used in this approach were placed under the control of the promoter P_CMV or that of the melanoma-specific P_4xTETP promoter (Fig. 1a).

For the expression analysis we used COS-1 cells (transformed green African monkey kidney cells) that can be transformed with the Listeria as DNA carrier at high rates 22, B16 melanoma cells deriving from a fast growing murine melanoma, and 4T1 cells deriving from a stage 4 breast cancer. These cancer cell lines were transformed at different efficiencies by WL-150 pSP118-P_CMV-EGFP as determined by flow cytometry of the EGFP-expressing cells (Fig. 2). Transformation with WL-150 pSP118-P_4xTETP-EGFP carrying egfp cDNA under the control of the melanoma-specific P_4xTETP promoter led to EGFP-expressing B16 melanoma cells at a frequency of 3.6% while COS-1 cells yielded less than 0.3% and 4T1 cells only about 0.04% EGFP-expressing cells (Fig. 2), demonstrating the specificity of the P_4xTETP promoter for the B16 melanoma cells in the bactofection approach.

For this study we used two different genes encoding prodrug-drug converting enzymes. Gene fcu1 encodes a fusion protein of yeast cytosine deaminase and uracil phosphoribosyl transferase which is highly active in converting of 5-Fluorocytosine (5-FC) into the toxic 5-Fluorouracil and further to 5-Fluoruridine (5-FU) 27. The gene deoD (E. coli) encodes a purine-nucleoside phosphorylase (PNP) which converts the prodrug 6-Methylpurine deoxyriboside as well as 9H-Purin-6-amine, 2-fluoro-9-(5-O-phosphono-β-D-arabinofuranosyl) (Fludarabine) to the cell toxic compounds 6-Methylpurine (MeP) and 2-Fluoroadenine (F-Ade) respectively 4. The expression of the prodrug/drug converting gene constructs (PNP (for MePdR and Fludarabine) and FCU1 (for 5-FC)) were placed under the control of P_CMV or P_4xTETP. The efficiency of prodrug-drug conversion was monitored by elimination of cells in the three different cancer cell lines (COS-1, B16 and 4T1) 3 days after DNA delivery. To determine the cell growth the cells were diluted and reseeded in 96 well plates prior to addition of the indicated prodrugs which were added at a final concentration of 50 μM for MePdR, 88 μM for Fludarabine, and 1 mM for 5-FC. The results are shown in Fig. 3(a–i) as the % of viable cells after prodrug treatment compared to untreated cells. Cell viability was measured by the MTT-test.

The data indicate that all three enzymes were functional in COS-1 cells and lead to rapid cell death (Fig. 3a,d,f) when WL-150 was used to introduce the expression cassettes under control of the P_CMV promoter. COS-1 cells without infection or transformed with Lm strains encoding enzymes not able to convert the applied prodrug, consequently were not affected by prodrug addition (Fig. 3a,d,f). In contrast, 4T1 cells were inhibited efficiently and highly specifically by both 5-FC upon infection with WL-150 pSP118-P_CMV-FCU1 and MePdR when infected with WL-150 pSP118-P_CMV-PNP. Fludarabine proved to be rather toxic for 4T1 cells and a concentration of 88 μM resulted already in the inhibition of non-infected cells (Fig. 3e). Unspecific obstruction was also observed when using lower concentrations (44 and 30 μM). Less than 15 μM Fludarabine allowed complete survival of the cells but not enough prodrug was converted to the toxic component to affect the 4T1 cells upon infection with WL-150 pSP118-P_CMV-PNP (data not shown).

In B16 melanoma cells cell-specific expression of PNP and FCU1 was obtained using the melanoma-specific P_4xTETP-promoter (Fig. 3c,i). However, cell inhibition after treatment of the transformed cells with the prodrugs 5-FC, and MePdR, respectively, was not as efficient as with the same drugs after enzyme production using P_CMV. Eradication of the entire cell population was obtained with the P_CMV constructs only (Fig. 3c,f,i).

As alternative to the above described delivery of expression plasmid transfer for prodrug converting enzyme production into cancer cells, we constructed Lm strains capable to express and secrete the respective prodrug/drug converting enzyme directly into the host cell cytosol. Therefore, we chose PNP, since FCU1 is a yeast protein and may require posttranslational modification for its function.

To facilitate PNP secretion from Listeria predominantly in the cytosol of the Lm-infected cells the deoD gene (encoding PNP) was cloned in frame downstream of the promoter and the secretion signal sequence of the listerial actA gene. ActA is produced in the cytosol of Lm-infected mammalian cells where it is responsible for actin polymerization needed for intra- and intercellular movement of L. monocytogenes 30. The P_actA-SP-PNP cassette was inserted into the recently described Lm lethal balanced plasmid pSP0 31 resulting in pSP0-P_actA-SP-PNP which was then introduced into WL-150 (Fig. 1b). Western blot analysis of protein preparations from the supernatant and the cell pellet showed that PNP was efficiently secreted by this Lm strain (Fig. 4a). Supernatants of logarithmically grown WL-150 pSP0-P_actA-SP-PNP and (as control) WL-150 pSP0 cultured in BHI and supplemented with 1% (w/v) AmberliteTM XAD-4 to enhance PactA activity 32 were incubated with 100 μM MePdR for 6 h and conversion to the toxic MeP was determined by HPLC analysis (Fig. 4b). The results showed that 85% of MePdR were converted to MeP by the supernatant of a WL-150 pSP0-P_actA-SP-PNP culture while no conversion was obtained in the control supernatant. Similar results were obtained with lysed bacteria indicating that MePdR can not be converted by the L. monocytogenes derived purine deoxyribonucleoside phosphorylases (encoded by deoD and pnp) annotated in the Lm genome 33 (see also http://genolist.pasteur.fr/ListiList/).

After infection of 4T1, B16 and COS-1 cells with WL-150 pSP0-P_actA-SP-PNP at a MOI of 200 for 5 h the prodrug treatment was initiated. Prodrug-dependent inhibition of all cell types was observed when MePdR was added at a final concentration of 50 μM (Fig. 3a–c) to the cultures. However, the addition of 88 μM Fludarabine did not result in specific reduced proliferation of 4T1, B16 or COS-1 cells infected with this Lm secretion strain.

In a recent study Critchley et al. 2004 34 reported protein-delivery to be more efficient in comparison to DNA-delivery in prodrug therapy using engineered E. coli vehicles. The authors based their conclusion on the findings that a higher fraction of GFP expressing cells was found after protein-delivery than after DNA-delivery.

Our results shown in Fig. 3a–f of this paper suggest, however, a significantly higher conversion efficiency of the prodrug by the enzyme after DNA delivery than after the protein-delivery approach. To determine whether this difference was due to the use of L. monocytogenes as bacterial carrier WL150 carrying pSP0-P_actA-GFP encoding GFP under the control of the listerial actA promoter was employed as measure of GFP expression inside infected COS-1 cells using a flow cytometer. The results (Fig. 2) showed that the majority of the infected cells indeed did harbour GFP-expressing bacteria similar to the results obtained with the E. coli as carrier system. But, although the fraction of cells producing enzymes for prodrug conversion was higher with the protein-delivery approach, the amount of PNP produced in the entire cell populations is higher after DNA delivery (Fig. 5a). Furthermore, densitometrical scanning of the Western blot revealed a 4–6 lesser PNP expression in B16 cells when deoD was under the control of P_4xTETP promoter in contrast to the P_CMV construct.

In addition, a quantitative comparison of the PNP activity by the two approaches was performed by measuring the amount of the generated toxic compound (MeP). Culture media were taken 7 d after prodrug addition and the non cleaved MePdR concentration was measured by HPLC-analysis. The results shown in Fig. 5b confirmed findings described earlier in Fig. 3 and Fig. 5a.

The reduction of MePdR concentration correlated strictly with the amount of PNP generation. Strain WL-150 pSP118-P_CMV-PNP as vector resulted in close to 100% conversion of MePdR to MeP in COS-1 and B16 cells, and 80% conversion rate was achieved in 4T1 cells respectively. In contrast, strain WL-150 pSP118-P_4xTETP-PNP yielded about 40% prodrug conversion in B16 melanoma cells and no conversion was observed in 4T1 or COS-1 cells (Fig. 5b).

The conversion of MePdR after infection of the mammalian cells with PNP-enzyme secreting Lm reached only 25–55% in all the cell cultures.