Endothelial cell transfection with cationic liposomes ...

2 downloads 0 Views 363KB Size Report
MB and KF are supported by a grant from the Crusaid. Star Foundation. We would also like to thank the BBSRC,. MRC, the Royal Society and Genzyme for ...
Gene Therapy (1998) 5, 614–620  1998 Stockton Press All rights reserved 0969-7128/98 $12.00 http://www.stockton-press.co.uk/gt

Endothelial cell transfection with cationic liposomes and herpes simplex-thymidine kinase mediated killing K Fife1, M Bower1, RG Cooper2, L Stewart2, CJ Etheridge2, RC Coombes1, L Buluwela3 and AD Miller2 Departments of 1Medical Oncology and 3Biochemistry, Charing Cross and Westminster Medical School; and 2Department of Chemistry, Imperial College of Science, Technology and Medicine, London, UK

Endothelial cells are a promising target for cancer gene therapy because neoangiogenesis is vital for the supply of oxygen and nutrients to solid tumours. However, endothelial cells have been reported to be difficult to transfect. We demonstrate high rates of transfection with the reporter gene pSV40␤gal using DC-Chol/DOPE cationic liposomes and lower rates with the novel polyamine cationic lipo-

somes ACHx/DC-Chol/DOPE and ACO/DC-Chol/DOPE. Endothelial cells transfected with HSV-thymidine kinase using DC-Chol/DOPE demonstrated 3 log10 increased cytotoxicity compared with controls when exposed to the prodrug ganciclovir, thereby demonstrating significant biological effect.

Keywords: endothelial cell; cationic liposome; gene therapy; HSV-thymidine kinase

Introduction The growth of a tumour beyond 1 mm3 is dependent upon the formation of a collateral vascular supply. Some tumour cells switch to an ‘angiogenic phenotype’, provoked by changes in the microenvironmental equilibrium between positive and negative regulators of angiogenesis produced by the tumour cells. New blood vessel formation then occurs and further tumour growth is provoked by twin effects of the endothelium on tumour. The ‘perfusion’ effect allows delivery of oxygen and nutrients and the ‘paracrine’ effect involves the release of several growth factors which act on the tumour cells.1 The actual process of angiogenesis requires proteolysis of the basement membrane and extracellular matrix, disruption of intercellular junctions, increases in permeability, migration, chemotaxis and proliferation of endothelial cells to form tubular structures. Damage to a single vessel can cause the death of thousands of dependent tumour cells, and only a small number of endothelial cells need to be affected to disrupt a vessel.2 Tumour-related endothelial cells, although morphologically abnormal and rapidly proliferating, are not malignant and do not mutate into therapy-resistant forms. In health, endothelial cells are quiescent and have a low turnover time except during wound healing, the menstrual cycle and placental formation. The importance of angiogenesis in the growth of tumours has been recognised and tumour vasculature is a potential target for anticancer therapy.3,4 In this work we aimed to develop a drug sensitivity gene therapy strategy targeting endothelial cells. This Correspondence: K Fife, Department of Medical Oncology, Charing Cross and Westminster Medical School, St Dunstan’s Road, London W6 8RP, UK Received 31 July 1997; accepted 18 December 1997

initially involved optimising in vitro endothelial transfection efficiency using clinically relevant transfection methods with cationic liposomes carrying a reporter gene (lacZ; ␤-galactosidase). Liposomal transfection, although less efficient, has a better safety profile than viral transduction, avoiding the problems of insertional mutagenesis with retroviruses, immunogenicity with adenoviruses and the acquisition of replication competence causing infection. In addition, one of the liposomes tested, DC-Chol/DOPE, formulated from 3␤-[N(N⬘,N⬘-dimethylaminoethane)carbamoyl]cholesterol (DCChol) and dioleoyl L-␣-phosphatidylethanolamine (DOPE) (Figure 1), has had approval for human gene therapy trials.5,6 Two polyamine cationic liposomes ACHx/DC-Chol/DOPE and ACO/DC-Chol/DOPE, formulated from cationic polyamine lipids 3-aza-N1– cholesteryloxycarbonylhexane-1,6-diamine (ACHx or CJE52) (Figure 1) and 4-aza-N1–cholesteryloxycarbonyloctane1,8-diamine (ACO or B130) respectively (Figure 1),7 were also tested for endothelial transfection efficiency. These latter two liposomes were known to form smaller liposome–DNA complexes than DC-Chol/DOPE and were therefore expected to transfer across the endothelial cell membrane more efficiently than other cationic liposome systems tested, in the light of previous experience.7 The most efficient liposome was used to transfect endothelial cells with herpes simplex thymidine kinase (HSVtk), a viral enzyme to which the prodrug ganciclovir (GCV; Cymvene, Roche Products Ltd, Welwyn Garden City, UK) binds avidly and is phosphorylated. Ganciclovir triphosphate acts as a false base and is incorporated into elongating DNA causing chain termination and apoptotic death in dividing cells. There are two advantages of this system in the tumour angiogenesis situation: liposomes cause transient duration of transfection and GCV only kills dividing cells. Therefore normal quiescent

Liposomal endothelial cell transfection and killing K Fife et al

615

Figure 1 Structures of lipids used to formulate cationic liposomes.

vasculature should not be affected to any significant extent. In addition, the HSVtk/GCV system exhibits the ‘bystander effect’.8 It has been demonstrated that as few as 10–20% of targeted cells need to be HSVtk positive;9 neighbouring cells die secondarily to the bystander effect. This is thought to be because of transfer of the nucleotide toxin between cells in culture, but may also be immunologically mediated in vivo.10

Results Transfection The absolute amounts of pSV40␤gal DNA used per transfection were 1.25, 2.5 and 5 ␮g. The ratio of liposome:DNA was varied for each absolute quantity of DNA. For DC-Chol/DOPE liposomes (Figure 1), liposome:DNA ratios between 1.6:1, and 12:1 (wt:wt) were used. With ACHx/DC-Chol/DOPE liposomes (Figure 1), liposome:DNA ratios between 3:1 and 8:1 (wt:wt) were used while ratios between 1:1 and 6:1 (wt:wt) were tested with ACO/DC-Chol/DOPE liposomes (Figure 1). In general, lower liposome:DNA ratios were tested for the polyamine cationic liposomes because toxicity had been demonstrated at high ratios of DC-Chol/DOPE:DNA.

HUV-EC-C cells are non-immortalised and were obtained at low passage number; these are therefore most representative of in vivo conditions. The percentages of transfected HUV-EC-C cells using DC-Chol/DOPE liposomes are shown in Figure 2. The number of cells in the transfected wells was compared with that in the control (nontransfected) wells so that toxicity of the transfection reagent could be assessed. The percentage of cells surviving the transfection procedure is indicated in the graphs as numbers in bold type at the base of the columns. Percentage transfection was calculated as the number of blue staining (␤-galactosidase expressing) cells as a percentage of the average number of control cells; this calculation therefore takes into account the toxicity of the procedure by using the number of cells that would have been present in the absence of transfection as the denominator. Error bars signify one standard error, calculated as the standard deviation of transfection efficiencies divided by the square root of the number of readings. Transfection with 1.25 ␮g of DNA was nontoxic and produced the highest efficiency, with a rate of 5.8% (standard error 0.7) at a DC-Chol/DOPE:DNA ratio of 6.4:1. Toxicity increased and efficiency decreased with increasing amounts of DNA used.

Liposomal endothelial cell transfection and killing K Fife et al

616

Figure 2 Transfection of HUV-EC-C cells with pSV40␤gal using the cationic liposome DC-Chol/DOPE at different DC-Chol/DOPE:DNA ratios. Amount of pSV40␤gal: (`) 1.25 ␮g; (첸) 2.5 ␮g; (왎) 5 ␮g. Numbers in bold refer to the percentage of cells surviving transfection as compared to control, as an indication of toxicity. Error bars: one standard error.

Similarly, ECV304 cells (Figure 3) demonstrated highest transfection efficiencies and lowest toxicity when 1.25 ␮g of DNA was used, with a maximum efficiency of 9.5% (standard error 1.2) at a DC-Chol/DOPE:DNA ratio of 1.75:1. In contrast to HUV-EC-C cells, the efficiency appeared to decline with increasing DC-Chol/ DOPE:DNA ratios. The toxicity increased considerably with 2.5 ␮g and 5 ␮g DNA. By comparison, transfection with ACHx/DC-Chol/DOPE liposomes was observed with optimal transfection efficiencies of only 2.1% for HUV-EC-C cells and 3.2% for ECV304 cells. In addition, some liposome-induced cytotoxicity was also observed at the higher liposome:DNA ratios, particularly with HUVEC-C cells. Transfection with ACO/DC-Chol/DOPE liposomes was markedly less efficient. The optimal transfection efficiency obtained with HUV-EC-C cells was barely 0.04% while that for ECV304 cells was 0.68%. ACO/DC-Chol/DOPE liposomes also appeared to be moderately toxic to ECV304 cells but not to HUV-EC-C cells. By comparison, best transfection efficiencies obtained with the commercially available cationic liposome Transfectam (Promega) were 1.74% for HUV-EC-C cells and 0.6% for ECV304 cells using the same conditions (data not shown). In summary, the best transfection efficiency of the most relevant endothelial cell line, HUVEC-C, was achieved with DC-Chol/DOPE which has the advantage of having already been granted approval for the use in human gene therapy trials.5 Toxicity was negligible under the conditions giving the best transfection efficiencies.

Cytotoxicity Owing to their high transfection efficiency, DCChol/DOPE liposomes were used to cotransfect cells with the HSVtk and puromycin resistance plasmids. Clones were isolated by puromycin selection. The cytotoxic effect of GCV on nontransfected and transfected HUV-EC-C cells is shown in Figure 4. Cytotoxicity was measured by counting transfected and nontransfected cells with a Coulter counter after 6 days of exposure to GCV, and the percentage of surviving cells was compared. One of two HSVtk-transfected HUV-EC-C clones demonstrated an IC50 (drug concentration which is cytotoxic to 50% of the cells) for GCV which was 3 log10 lower than nontransfected controls; the results of two experiments are shown (Figure 4). This shows a clear biological effect of GCV on HSVtk-transfected cells. One of four ECV304 clones had an IC50 for GCV almost 2 log10 lower than nontransfected ECV304 controls (Figure 5), the greatest differential being demonstrated at a GCV concentration of 1 ␮g/ml. The other clones did not show increased sensitivity to GCV possibly because they had been transfected with the puromycin resistance but not HSVtk plasmid, or the HSVtk plasmid, which is acquired episomally with liposome-mediated transfection, may have been lost during passaging of the clones. Bystander effect The results of this experiment are shown in Figure 6. In the absence of GCV, mixed cultures of the HSVtktransfected HUV-EC-C clone and nontransfected HUV-

Liposomal endothelial cell transfection and killing K Fife et al

617

Figure 3 Transfection of ECV304 cells with pSV40␤gal using the cationic liposome DC-Chol/DOPE at different DC-Chol/DOPE:DNA ratios. Amount of pSV40␤gal: (`) 1.25 ␮g; (첸) 2.5 ␮g; (왎) 5 ␮g. Numbers in columns refer to the percentage of cells surviving transfection as compared with control, as an indication of toxicity. Error bars: one standard error.

Figure 4 Toxicity of GCV to control and HSVtk-transfected HUV-EC-C cells: results of two experiments. Control (nontransfected) HUV-EC-C cells: (—쮿— ···왏···); HSVtk-transfected cells: (—왖— ···䉭···). Error bars: one standard error.

Figure 5 Toxicity of GCV to control and HSVtk-transfected ECV304 cells. Control (nontransfected) HUV-EC-C cells: (—쮿—); HSVtk-transfected cells: (···왖···). Error bars: one standard error.

Liposomal endothelial cell transfection and killing K Fife et al

618

Figure 6 Effect of exposure to 1 mg/ml GCV for 6 days on mixed cultures of HSVtk-transfected and nontransfected HUV-EC-C cells. (왎) Without GCV; (`) GCV 1 ␮g/ml. Error bars: one standard error.

EC-C cells demonstrated decreasing survival at 6 days as the proportion of HSVtk-transfected cells increased, falling to 42% when 100% of cells were HSVtk transfected. This may have been because the transfected cells were less robust owing to the lengthy procedure of puromycin selection. On exposure to 1 ␮g/ml GCV, 10% of the pure HSVtk-transfected culture survived at 6 days, ie GCV caused a four-fold reduction in the number of HSVtk transfected cells surviving. In the presence of GCV, the mixed cultures demonstrated higher than expected toxicity when 80% of the culture was composed of HSVtktransfected HUV-EC-C cells; based on the figures above (ie 100% survival of the 20% of nontransfected cells and 25% survival of the 80% HSVtk-transfected cells), 40% survival would be expected if there was no excess killing of nontransfected cells whereas the actual cell survival was 26%. The cultures composed of 60% or less of HSVtktransfected HUV-EC-C cells did not show significantly higher than expected cell death. These results suggest that in mixed cultures containing a high proportion of HSVtk-transfected cells, the bystander effect may be occurring and leading to the death of nontransfected cells.

Discussion Endothelial cells, particularly of human origin, are recognised to be difficult to transfect. Plasmid DNA alone led to a transfection efficiency in primary cultured HUVEC (human umbilical vein endothelial cells) of less than 0.001% which increased to just 1% with a retroviral vector.11 However, 88% transduction of HUVECs with an adenoviral vector has been reported.12 Efficient retroviral transduction (although no figures were given for efficiency rates) has been demonstrated in sheep and rat endothelial cells.13,14 Rates of retroviral transduction of rabbit and bovine aortic endothelial cells of 0.5–1% and 9–17%, respectively, have been obtained by one group; the efficiency in bovine cells was increased to 50–90% by including DEAE dextran in the transduction process but

this resulted in 20–30% cytotoxicity.15 There are few reports of liposomal transfection of endothelial cells. Lipofectamine transfected porcine endothelial cells with 2–15% efficiency, whereas the adenovirus transduction rate was 100%.16 Primary cultured HUVEC cells could not be transfected with Transfectam (Promega, Madison, WI, USA),17 but better results were reported with Lipofectin (Gibco Life Technologies, Paisley, UK) with 10–20% of cells staining positive for the reporter gene CAT (chloramphenicol acetyltransferase).18 Our results are comparable with the best liposomal transfection rates reported so far,18 and have the advantage of using a cationic liposome system noted for low toxicity in vivo.5,6 There is a wide variation in transfection efficiency between different cell types. Polyamine cationic liposomes might be expected to transfect cells more efficiently since these liposomes are smaller and possess higher zeta potentials (see Materials and methods: liposomes) suggesting that they should in principle bind DNA more efficiently and correspondingly transfect cells more effectively. Certainly, this has been our experience with respiratory epithelial cells.7 Moreover, Takeuchi et al19 have also noted a direct correlation between cationic liposome zeta potentials and transfection efficiency in their studies on NIH3T3, HeLa and Cos-7 cells. However, this does not appear to be the case here with endothelial cells showing uniformly higher rates of transfection with DC-Chol/DOPE (Figure 1) than with the polyamine cationic liposomes tested. This study demonstrates efficient endothelial cell kill with the HSVtk/ganciclovir method; the IC50 for HSVtktransfected HUV-EC-C cells was 3 log10 lower than that for nontransfected cells. This drug sensitivity gene therapy strategy has been shown to be effective in many other cell types. For example, lipofectamine transfected a melanoma B16 cell line with 10% efficiency and the IC50 for unselected cultures was 15 ␮g/ml GCV.20 These results were confirmed in vivo in a mouse model.21 Mesothelial cells transduced with HSVtk showed a 3–4 log10 reduction of IC50 for GCV compared with nontransduced cells.22 Similar positive results have also been demonstrated in vitro and in vivo for several other cell types including glioma23,24 and ovarian carcinoma.9 Several clinical trials are now in progress using the HSVtk drug sensitivity gene paradigm, and tumour responses have been observed.25 Other drug sensitivity gene systems such as cytosine deaminase/5-fluorocytosine26 are also in development and initial results in a human breast cancer trial are encouraging (K Sikora, personal communication). In combination with the cationic liposome DCChol/DOPE, the HSVtk/ganciclovir strategy is a promising basis for gene therapy of endothelial cells, and would be expected to have low toxicity compared with viral methods. Adenoviral vectors have demonstrated toxicity in clinical trials causing mucosal inflammation,27,28 and retroviruses have the potential for insertional mutagenesis. To reduce possible systemic toxicity in vivo, liposomal uptake by the reticuloendothelial system may be reduced by using polyethylene glycol (PEG) modified cationic liposomes. Tissue specificity may also be increased by injecting the complexes locally or intraarterially. We are currently developing ligand–liposome constructs to target endothelial cells with a high degree of specificity which would enhance the value of this

Liposomal endothelial cell transfection and killing K Fife et al

system in vivo. Transcription regulators such as endothelial-specific promoters may also be incorporated into the transfected gene construct. These mechanisms could reduce the dose of GCV required to achieve a biological effect, in the absence of improved transfection efficacy. The concept of targeting tumour vasculature as therapy for cancer and other diseases involving neovascularisation appears to be valid; several antiangiogenic agents have been shown to promote tumour regression3,4 and some are now in clinical trial. In view of the cascade effect of microvessel damage on tumour viability,2 transfection of only a minority of endothelial cells would be necessary for the gene therapy approach to antiangiogenesis, and the low level of bystander effect demonstrated in this study would not be a disadvantage. However, in vivo, the bystander effect is partly immunologically mediated and further investigation of this is warranted. In conclusion, we have shown that endothelial cells can be transfected with cationic liposomes with acceptable efficiency and high levels of cell kill can be achieved with HSVtk/ganciclovir gene therapy. These in vitro optimisation experiments provide a basis for developing the strategy in vivo together with endothelial targeting mechanisms.

Materials and methods Cells HUV-EC-C, a near diploid cell line derived from normal human umbilical cord vein, was obtained from the American Type Culture Collection (ATCC, Rockville, MD, USA). The cells were obtained at passage 13 and have a similar immunophenotype to primary endothelial cells, strongly retaining the endothelial markers Ulex europaeus agglutinin I (UEA I) and von Willebrand factor (vWF). They were maintained in Ham’s F12 medium (Gibco) with 10% heat inactivated fetal calf serum (Gibco), 30 ␮g/ml Endothelial Cell Growth Supplement (ECGS, Sigma), 100 ␮g/ml heparin, 100 ␮g/ml penicillin, 100 ␮g/ml streptomycin and 2 mm glutamine (Gibco). ECV304 is an immortalised endothelial cell line of human umbilical vein origin.29 These cells were acquired at passage 137 and maintained in M199 medium (Gibco) with 10% heat inactivated fetal calf serum, 100 ␮g/ml penicillin, 100 ␮g/ml streptomycin and 2 mm glutamine. They are also UEA I and vWF positive. Liposomes DC-Chol was synthesised as described previously30 and formulated into cationic liposomes with DOPE as follows. DC-Chol (3 mg, 6 ␮mol) and DOPE (0.3 ml, 3 mg, 4 ␮mol) (supplied at 10 mg/ml in CHCl3 by Sigma, Poole, UK) were combined in freshly distilled CH2Cl2 (5 ml) under nitrogen. Following this, 20 mm N-(2-hydroxyethyl)piperazine-N⬘-(2-ethanesulfonic acid) (Hepes) (Sigma) buffer, pH 7.8 (5 ml), was added and the mixture sonicated for 3 min. Organic solvents were removed under reduced pressure and the resulting liposome suspension was then sonicated for a further 3 min. Liposome preparations were stored at 4°C before use. ACHx or CJE52 and ACO or B130 were synthesised as described previously.7 ACO has been reported elsewhere as lipid 48, although no synthetic procedures were described, or characterisations reported.31 ACHx/DC-

Chol/DOPE (4.5:1.5:4.0 mol:mol:mol ratio) and ACO/DC-Chol/DOPE (4.5:1.5:4.0 mol:mol:mol ratio) cationic liposomes were then prepared in a similar way to DC-Chol/DOPE liposomes, as described above. Liposome sizes were determined by photon correlation spectroscopy (Model N4 MD sub-micron particle analyser, Coulter Electronics, Luton, UK) and zeta potentials by laser doppler spectroscopy (Zetamaster 3000, Malvern Instruments, Malvern, UK). DC-Chol/DOPE liposomes were found to be approximately 500 nm in diameter with a zeta potential of around 50 ± 1 mV, ACHx/DCChol/DOPE liposomes were approximately 400 nm in diameter with a zeta potential of around 70 ± 1 mV, and ACO/DC-Chol/DOPE liposomes were approximately 300 nm in diameter with a zeta potential of around 75 ± 1 mV.

Plasmids pSV40␤gal (Promega) is a 6821 base pair plasmid comprising the SV40 promoter and enhancer driving the lacZ and ampicillin resistance genes. LacZ encodes the enzyme ␤-galactosidase which converts the substrate X-gal to a blue product. pAGO is a 6386 base pair plasmid comprising a 2025-bp PvuII fragment of herpes simplex virus type I DNA containing the HSVtk gene inserted at the PvuII site of pBR322.32 The plasmid pucPuro contains the SV40 early promoter driving the puromycin resistance gene contained in pBABEpuro33 cloned into puc18. Plasmid DNA was extracted and purified using the Qiagen 500 method (Qiagen, Crawley, UK). Transfection Cells were plated in six-well plates at a density of 1 × 105 cells per well. After 24 h, cells were washed once with PBS (Gibco) and once with optiMEM serum-free medium (Gibco). pSV40␤gal DNA and different ratios of DCChol/DOPE, ACHx/DC-Chol/DOPE or ACO/DCChol/DOPE cationic liposomes (Figure 1) were mixed (total lipid concentration was 1.2 mg/ml in each case). OptiMEM was then added to give a total volume of 1 ml/well and the solution was then incubated at room temperature for 10 min. The transfection mixture was laid over the cells, which were incubated at 37°C in 5% CO2 for 2 h. At the end of the incubation period, the transfection mixture was aspirated, cells washed twice in PBS and complete medium replaced. Two days later, the cells were fixed for 10 min on ice in 0.5% paraformaldehyde/PBS, washed in PBS MgCl2 then in 1 ml of detergent solution (2 mm MgCl2, 0.01% sodium deoxycholate, 0.02% Nonidet P40 and PBS pH 7.4) for 10 min on ice. Cells were then incubated at 37°C for 3 h in 1 ml of 2% X-gal. Cells successfully transfected with the ␤-galactosidase gene stained blue and their percentage was counted in multiple fields in each well using a 10 mm grid (Graticules, Tonbridge, UK). Transfection efficiency was calculated as the number of transfected cells as a percentage of control cells. Cytotoxicity Cells were cotransfected with two plasmid constructs, one containing the HSVtk gene (pAGO) and the other containing the puromycin resistance gene (pucPuro) to allow antibiotic selection of transfected clones. Plasmid DNA of pAGO and pucPuro were used in the weight ratio of 10:1 to maximise the chance of puromycin-

619

Liposomal endothelial cell transfection and killing K Fife et al

620

resistant cells being cotransfected with HSVtk. DCChol/DOPE liposomes were used as the transfection agent at a liposome:DNA ratio of 5:1 (wt:wt). Cells (5 × 105) were plated in 10 cm plates and the transfection procedure described above was followed, using 5 ␮g total of DNA. Two days after transfection, puromycin was added to the cells at a concentration of 2 ␮g/ml medium and replaced every 3 days. Plating of clones was performed when isolated colonies had formed and most or all background cells had died. Cells were grown with puromycin to maintain selection; two HUV-EC-C clones and four ECV304 clones were successfully isolated. Transfected clones and control cells were plated in triplicate in 24-well plates at 5 × 104 cells per well and GCV added at concentrations of 0, 0.1, 1, 10, 50, 100, 500 and 1000 ␮g/ml. Medium/GCV was changed on the third day. After 6 days of exposure to GCV, cells were trypsinised and counted with a Coulter counter.

Bystander effect From the cytotoxicity experiments, it was established that at a GCV concentration of 1 ␮g/ml, approximately 99% of the HSVtk-transfected HUV-EC-C clone was killed compared with 10–20% of nontransfected cells. Different ratios of the HSVtk-transfected clone:nontransfected cells were mixed and plated in triplicate in a 24-well plate. The total number of cells per well was 3 × 104 and the following ratios of HSVtk-transfected clone:nontransfected cells were used: 0:100, 20:80, 40:60, 60:40, 80:20 and 100:0. The day after plating, GCV at 1 ␮g/ml was added. On the sixth day of exposure, cells were trypsinised and counted with a Coulter counter. A parallel control experiment was performed, using a 24-well plate with HSVtk-transfected clone:nontransfected cells in the same ratios but to which no GCV was added.

Acknowledgements MB and KF are supported by a grant from the Crusaid Star Foundation. We would also like to thank the BBSRC, MRC, the Royal Society and Genzyme for financial support.

References 1 Folkman J. Angiogenesis in cancer, vascular, rheumatoid and other disease. Nature Med 1995; 1: 27–31. 2 Denekamp J. Vascular attack as a therapeutic strategy for cancer. Cancer Metastasis Rev 1990; 9: 267–282. 3 Baillie C, Winslet M, Bradley N. Tumour vasculature – a potential therapeutic target. Br J Cancer 1995; 72: 257–267. 4 Folkman J. Clinical applications of research on angiogenesis. New Engl J Med 1995; 333: 1757–1763. 5 Caplen N et al. Liposome-mediated CFTR gene transfer to the nasal epithelium of patients with cystic fibrosis. Nature Med 1995; 1: 39–46. 6 Nabel G et al. Direct gene transfer with DNA–liposome complexes in melanoma: expression, biologic activity and lack of toxicity in humans. Proc Natl Acad Sci USA 1993; 90: 11307–11311. 7 Cooper R et al. Polyamine analogues of 3␤-[N-(N⬘,N⬘dimethylaminoethane)carbamoyl]cholesterol (DCChol) as agents for gene delivery. Chem Eur J 1998; 4: 137–151. 8 Freeman S et al. The ‘bystander effect’: tumour regression when a fraction of the tumour mass is genetically modified. Cancer Res 1993; 53: 5274–5283. 9 Rosenfield M et al. Adenovirus-mediated delivery of the herpes simplex virus thymidine kinase gene selectively sensitizes human ovarian carcinoma cells to ganciclovir. Clin Cancer Res 1995; 1: 1571–1580.

10 Pope I, Poston G, Kinsella A. The role of the bystander effect in suicide gene therapy. Eur J Cancer 1997; 33: 1005–1016. 11 Sun B, Plumpton C, Sinclair J, Brown M. In vitro expression of calcitonin gene-related peptide in human endothelial cells transfected with plasmid and retroviral vectors. Neuropeptides 1994; 26: 167–173. 12 Lemarchand P et al. Adenovirus-mediated transfer of a recombinant human ␣1-antitrypsin cDNA to human endothelial cells. Proc Natl Acad Sci USA 1992; 89: 6482–6486. 13 Dichek D, Nussbaum O, Degen S, Anderson W. Enhancement of the fibrinolytic activity of sheep endothelial cells by retroviral vector-mediated gene transfer. Blood 1991; 77: 533–541. 14 Yao S et al. Expression of human factor IX in rat capillary endothelial cells: toward somatic gene therapy for hemophilia B. Proc Natl Acad Sci USA 1991; 88: 8101–8105. 15 Kahn M, Lee S, Dichek D. Optimization of retroviral vectormediated gene transfer into endothelial cells in vitro. Circ Res 1992; 71: 1508–1517. 16 Muller D et al. Catheter-mediated pulmonary vascular gene transfer and expression. Circ Res 1994; 75: 1039–1049. 17 Nathwani A et al. Efficient gene transfer into human umbilical endothelial cells allows functional analysis of the human tissue factor promoter. Br J Haematol 1994; 88: 122–128. 18 Powell J, van Zonneveld A, van Mourik J. Gene transfer into specific vascular cells. Eur J Vasc Surg 1992; 6: 130–134. 19 Takeuchi K et al. Effect of zeta potential of cationic liposomes containing cationic cholesterol derivatives on gene transfection. FEBS Lett 1996; 397: 207–209. 20 Calvez V et al. Virus-free transfer of the herpes simplex virus thymidine kinase gene followed by ganciclovir treatment induces tumour cell death. Clin Cancer Res 1996; 2: 47–51. 21 Soubrane C et al. Direct gene transfer of a plasmid carrying the herpes simplex virus-thymidine kinase gene (HSVtk) in transplanted murine melanoma: in vivo study. Eur J Cancer 1996; 32A: 691–695. 22 Smythe W et al. Use of recombinant adenovirus to transfer the herpes simplex virus thymidine kinase (HSVtk) gene to thoracic neoplasms: an effective in vitro drug sensitization system. Cancer Res 1994; 54: 2055–2059. 23 Takamiya Y et al. An experimental model of retrovirus gene therapy for malignant brain tumors. J Neurosurg 1993; 79: 104–110. 24 Barba D, Hardin J, Sadelain M, Gage F. Development of antitumour immunity following thymidine kinase-mediated killing of experimental brain tumours. Proc Natl Acad Sci USA 1994; 91: 4348–4352. 25 Roth J, Cristiano R. Gene therapy for cancer: what have we done and where are we going? J Natl Cancer Inst 1997; 88: 21–39. 26 Trinh Q et al. Enzyme/prodrug gene therapy: comparison of cytosine deaminase/5-fluorocytosine versus thymidine kinase/ ganciclovir enzyme/prodrug systems in a human colorectal carcinoma cell line. Cancer Res 1995; 55: 4808–4812. 27 Knowles M et al. A controlled study of adenoviral vectormediated gene transfer in the nasal epithelium of patients with cystic fibrosis. New Engl J Med 1995; 333: 823–831. 28 Crystal R et al. Administration of an adenovirus containing the human CFTR cDNA to the respiratory tract of individuals with cystic fibrosis. Nat Genet 1994; 8: 42–51. 29 Takahashi K et al. Spontaneous transformation and immortalization of human endothelial cells. In Vitro Cell Dev Biol 1990; 25: 265–274. 30 Alton E et al. Non-invasive liposome-mediated gene delivery can correct the ion transport defect in cystic fibrosis mutant mice. Nat Genet 1993; 5: 135–142. 31 Lee E et al. Detailed analysis of structures and formulations of cationic lipids for efficient gene transfer to the lung. Hum Gene Ther 1996; 7: 1701–1717. 32 Colbere-Garapin F et al. Cloning of the active thymidine kinase gene of herpes simplex virus type 1 in Escherichia coli K-12. Proc Natl Acad Sci USA 1979; 76: 3755–3759. 33 Morgenstern J, Land H. Advanced mammalian gene transfer: high titre retroviral vectors with multiple drug selection markers and a complementary helper-free packaging cell line. Nucleic Acids Res 1990; 18: 3587–3596.