D2001 Nature Publishing Group 0929-1903/01/$17.00/+0 www.nature.com/cgt

Highly efficient transduction of endothelial cells by targeted artificial virus-like particles Kristina MuÈller,1 Thomas Nahde,2 Alfred Fahr,1 Rolf MuÈller,2 and Sabine BruÈsselbach 2

1Institute of Pharmaceutical Technology and Biopharmaceutics, Philipps University, Marburg 35032, Germany; and 2Institute of Molecular Biology and Tumor Research (IMT), Philipps University, Marburg 35033, Germany.

Targeting the tumor vasculature by gene therapy is a potentially powerful approach, but suitable vectors have not yet been described. We have designed a new type of liposomal vector, based on the composition of anionic retroviral envelopes, that is serum-resistant and nontoxic. These artificial virus-like envelopes (AVEs) were endowed with a cyclic RGD-containing peptide as a targeting device for the avû3 -integrin on tumor endothelial cells (ECs). The packaging of plasmid DNA complexed with low-molecular- weight, nonlinear polyethyleneimine into these AVEs yielded artificial virus-like particles (AVPs) that transduced ECs with efficiencies of up to 99%. In contrast, transduction of a variety of other cell types by these RGD±AVPs was comparably inefficient under the same experimental conditions. This EC selectivity was mediated, in part, but not exclusively, by the RGD ligand, as suggested by the reduced, but still relatively high, transduction efficiency seen with AVPs lacking RGD. The interaction of anionic lipids of the AVPs with ECs may therefore contribute to the observed selective and highly efficient transduction of this cell type. These findings suggest that the targeted AVEtechnology is a useful approach to create highly efficient nonviral vectors. Cancer Gene Therapy (2001) 8, 107±117

Key words: Nonviral vectors; artificial virus-like envelope (AVE); artificial virus-like particle (AVP); anionic liposome; polyethyleneimine; endothelial cells; integrin targeting; cyclic RGD.

mong the new experimental approaches to cancer Examples are vascular endothelial growth factor receptor II Atreatment, gene therapy has gained particular attention. (FLK-1, KDR),5 transforming growth factor binding 6,7 8 Clinical studies showed that cancer gene therapy can protein CD105/endoglin, vû3 integrin, and CD13/ potentially be made to work, but the lack of adequate aminopeptidase N.9 That these surface molecules can be technologies is evident. This applies, in particular, to the used for the successful targeting of drugs to the tumor blood efficiency and specificity of tumor cell transduction, vessels has been demonstrated in several studies. Thus, indicating that the design of more efficacious and site- antiangiogenic strategies have made use of antibodies selective vectors is urgently required. A second serious neutralizing KDR/FLK-1;10 ± 12 antibodies directed to problem is the poor accessibility of many tumor cells to any endoglin have been employed to direct a ricin A conjugate kind of drug, in particular vectors carrying nucleic acids, as a to the tumor vasculature;13 and a cyclic RGD peptide has been consequence of the defective tumor vasculature and the high used to target doxorubicin to the tumor site.14 Cyclic interstitial pressure.1 The concept of endothelial tumor cell RGD peptides have also been used successfully for the targeting2 is attractive because the tumor blood vessels are delivery of both viral and nonviral vectors to ECs,15 ± 19 more readily accessible than the actual tumor cell compart- and therefore, seem to be suitable for tumor EC targeting. ment. In addition, endothelial cells (ECs) are unknown to The vasculature is a highly attractive target also for gene acquire resistance to treatment,3 and the endothelium therapy because it is comprised of a large number of represents a target that is largely independent of tumor potential target cells and is easily accessible via the blood type.4 Finally, as tumor ECs are essential for the nutrition stream. Ideally, an EC-directed vector should be safe, and growth of the tumor cells, any therapy targeting tumor nontoxic, nonimmunogenic, and cell type±specific; it should ECs should have a dramatic ``bystander effect''. transduce EC cells with high efficiency; and Ð with respect Several marker proteins up-regulated in tumor endothe- to clinical applicability Ð it should offer the possibility of lium have been identified. Among these membrane- large-scale production, stability of the product, and ease of associated proteins, some are of particular interest because handling. To date, a vector combining all these features has they can be exploited in targeting the tumor vasculature. not been described. Viral vectors are often highly efficient, but safety and Received June 23, 2000; accepted November 7, 2000. immunogenicity are issues of potential concern, and the Address correspondence and reprint requests to Dr. Rolf MuÈller, Institute limited transgene size often poses a serious obstacle. of Molecular Biology and Tumor Research (IMT), Philipps University, Nonviral vectors, on the other hand, frequently face the Marburg 35033, Germany. E-mail address: [email protected] problem of low transduction efficiency. Different concepts

Cancer Gene Therapy, Vol 8, No 2, 2001: pp 107±117 107 108 MUÈ LLER, NAHDE, FAHR, ET AL: TRANSDUCTION OF ENDOTHELIAL CELLS BY TARGETED ANIONIC LIPOSOMES have been developed for the generation of nonviral vectors, lamellar liposome suspension was placed in a 3-mL glass e.g., cationic lipid±DNA complexes,20 polycation±DNA vessel cooled by an ice/water mixture and sonicated for 15 complexes,21,22 and liposome-entrapped polycation-con- seconds in an MSE-Soniprep 150 (Zivy, Oberwil, Switzer- densed DNA (LPD).23 However, these complexes are often land) equipped with a titanium tip. Sonication was repeated physico-chemically or biologically unstable, serum-sensi- 10 times; each sonication period was followed by a 30- tive, or toxic as a consequence of unspecific interactions with second pause, allowing the suspension to cool. The the biological environment.23,24 Promising results have been resulting liposome suspension was extruded through obtained with Sendai virus±fused liposomes (HVJ lipo- polycarbonate membrane filters with a pore size of 50 nm somes),25 but a potential problem may be the presence of using a standard device (LiposoFast2 ) 46 purchased from viral proteins in these vectors and their large size (400 nm). Avestin (Ottawa, Canada). The liposomes were used up 2 An attractive alternative is the use of artificial virus-like months after preparation, during which time no significant envelopes (AVEs) for the encapsulation of condensed increase in liposome size could be detected. plasmid DNA.26 AVEs mimic the lipid composition of retroviruses. These natural lipids are anionic and, in contrast Targeting motif to their artificial cationic counterparts, interact only weakly Using solid-phase synthesis (Applied Biosystems Peptide with their biological environment and therefore are nontoxic. Synthesizer, Foster City, CA), a cyclic peptide with the We have taken this approach further by combining AVEs amino acid sequence CDCRGDCFC and having an with the advantages of the cationic polymer, polyethyle- 22 additional arginine at the N-terminus was synthesized. neimine (PEI), which not only condenses the DNA but Cyclic condensation of the peptide was accomplished by is also believed to act as an endosomolytic agent and to stirring an aqueous solution of the synthesized peptide under protect the DNA from cytoplasmic nucleases. In addition, access of ambient air. Completion of cyclic condensation we have equipped these artificial viral particles (AVPs) was verified by high-performance liquid chromatography. with a targeting device for activated ECs, i.e., a cyclic 27 After high-performance liquid chromatography purification, RGD peptide thought to interact with v 3 integrin. We the peptide was lyophilized and stored at 48C. refer to these targeted nonviral vectors as RGD±AVPs. Covalent attachment of RGD peptide to liposomal surface (RGD±AVE) MATERIALS AND METHODS Activation of the N-glutaryl-DPPE carboxyl group at the AVEs liposomal surface 47 was achieved by adding 3.5 mg of 1- Throughout the preparation of AVEs,synthetic phospholipids ethyl-3- (3-dimethylaminopropyl) carbodiimide to 400 L were used without further purification. 1,2-Dipalmitoleoyl- AVE and shaking the suspension for 5 hours in the dark. The sn-glycero-3-phosphoethanolamine (DPPE) and 1,2-dio- resulting active O-acyl-intermediate reacted with added leoyl-sn-glycero-3- [phospho- L -serine] (DOPS) were RGD peptide (250 g in 150 L buffer) overnight, yielding purchased from Avanti Polar Lipids (Alabaster, AL); 1,2- a covalent coupling of the peptide to the liposomal surface. dilauroyl-sn-glycero-3-phosphoethanolamine (DLPE) was The RGD±AVEs were separated from unbound peptide by obtained from Genzyme (Liestal, Switzerland); and choles- Sephadex G25 gel permeation chromatography in Tris buffer terol was from Calbiochem (San Diego, CA). All other 10 mM, pH 7.4. Coupling efficiency was monitored using a substances were of analytical grade. fluorescently labeled derivative of the RGD peptide. For this purpose, RGD was conjugated with 5- (4,6-dichlorotriazi- Synthesis of lipid anchor nyl)aminofluorescein (5-DTAF) 48 purchased from Mole- N-glutaryl-DPPE was prepared by dissolving DPPE in cular Probes (Eugene, OR). 5-DTAF was dissolved in anhydrous chloroform. Glutaric anhydride and water-free borate buffer (pH 9) and RGD peptide was added in a 1:4 pyridine were added and the solution stirred at 208C for 2 molar ratio (RGD/5-DTAF). The solution was stirred days. After this period, the products were dried by applying overnight and uncoupled 5-DTAF was separated from the vacuum to the solution and N-glutaryl-DPPE was purified labeled peptide by size exclusion chromatography using using preparative silica gel chromatography (Merck 60 F Sephadex G25. The purified fluorescently labeled RGD was 254) in chloroform/methanol/ammonia (65/35/3). Spots covalently linked to activated N-glutaryl-DPPE in AVEs as were dissolved in methanol, dried, and redissolved in described above. Fluorescence of the labeled AVE was read chloroform for further use. N-glutaryl-DPPE was identified using a spectrofluorimeter (model LS50B; Perkin Elmer, È by 1H-NMR as described previously.45 AVEs with the Uberlingen, Germany) and compared to an appropriately composition DOPS/DLPE/cholesterol/N-glutaryl-DPPE diluted RGD/5-DTAF solution. (3:3:3:1 mol/mol) were prepared as follows: a chloroform solution of the lipid mixture (total amount 10 mol) was Condensation of plasmid DNA dried into a thin film in the inner surface of a rotating 100- Low-molecular-weight, branched polyethyleneimine (PEI; mL glass vessel warmed by a water bath to 308C. Residual Lupasol G100, BASF, Ludwigshafen, Germany) was added chloroform was removed by vacuum desiccation for 15 to 15 g of plasmid DNA up to a ratio of polyethyleneimine minutes. The lipid film was hydrated either in 1 mL sterile 10 nitrogen:DNA phosphate of 20.7. Condensation of plasmid mM Tris buffer (pH 7.4) or in 1 mL sterile phosphate- DNA was monitored by dye exclusion technique using buffered saline (PBS). After 2hours, the resulting multi- Picogreen2 (Molecular Probes) as described previously.49

Cancer Gene Therapy, Vol 8, No 2, 2001 MUÈ LLER, NAHDE, FAHR, ET AL: TRANSDUCTION OF ENDOTHELIAL CELLS BY TARGETED ANIONIC LIPOSOMES 109

1.0

0.8

0.6

0.4 Intensity (%) 0.2

0

01 20 30 40 5 0 N/P ratio 0 Figure 1. Condensation of DNA by polyethylenimine determined by Figure 3. Cryogenic transmission electron microscopy (Cryo-TEM) dye exclusion. Six individual measurements with different DNA of RGD±AVPs complexed with PEI/plasmid DNA. The pictures show plasmid concentrations are shown to illustrate the high reproducibility encapsulation of condensed DNA in a lipid layer within discrete of the method. particles. The diameter of the larger liposomes shown is approxi- mately 140 nm.

In another set of experiments, protamine sulfate (USP quality; EliLilly and Co., Indianapolis, IN) was used as emission wavelength of 518 nm. Appropriate amounts of condensing agent. PEI were added to the wells to achieve the desired N/P ratios. Preparation of RGD±AVPs Size and zeta potential analysis of vector particles An RGD±AVE suspension containing 60 g of lipid was added to 15 g of plasmid DNA precomplexed with PEI, as Dynamic laser light scattering using a commercially avail- described above, and gently vortexed. The resulting AVPs able system Zetasizer 4 (Malvern Instruments, Herrenberg, are ready for transfection experiments. The indicated Germany) with PCS Software Version 1.26III (Malvern) amounts are sufficient for one 6-cm cell culture dish. was used. Autocorrelation data were analyzed by a specialized version of CONTIN.50 Zeta potential measure- DNA condensation measurement by dye exclusion ments were performed on a PALS Zeta Potential Analyzer The experimental protocol was followed as described.49 Version 3.12(Brookhaven Instruments, Holtsville, NY). Each well of a 96-well plate was filled with 95 LofTris10 Cryogenic transmission electron microscopy mM, pH 7.5. Plasmid DNA was added at concentrations of 300 to 1333 ng/mL in each well and the Picogreen2 A suspension of AVPs was applied to a holey carbon foil grid concentration was adjusted to a ratio of one molecule per and vitrified by flash-freezing in liquid ethane. The grids 3.75 bp of DNA. Fluorescence spectrophotometry was were cryo-transferred to the liquid nitrogen±cooled cryoe- performed at an excitation wavelength of 492.5 nm and an lectron microscope (Philips CM200 FEG, FEI, Germany). Images were taken at a magnification of 60,000Â under liquid nitrogen conditions at 1.5 m defocus at 160 keV.

Cell culture 50 AVE (z-av. 94 nm) 45 Human umbilical vein endothelial cells (HUVECs) were DNAPEI (z-av. 146 nm) prepared by the method of Jaffe 51 as modified by Thornton 40 DNAPEIAVE (z-av. 182 nm) et al.,52 and cultivated in EGM-2medium (BioWhittaker 35 Europe, Verviers, Belgium). The colon carcinoma cell line 30 LoVo (provided by I. Hart, London, UK), the lung 25 adenocarcinoma cell line A549 (obtained from K. Have- 20 mann, Marburg), the prostate carcinoma cell line DU-145 (obtained from G. AumuÈller, Marburg), the choriocarci- 15 noma cell line JEG-3 (obtained from A. Wellstein, 10 Georgetown University), and the osteosarcoma cell line

Fluorescence at 520 nm (a.u.) 5 Saos-2(obtained form N. La Thangue, Glasgow) were 0 cultured in Dulbecco's modified Eagle's medium supple- 50 100 150 200 250 300 350 mented with 10% fetal calf serum (FCS; BioWhittaker Diameter (nm) Europe) and 2mM L -glutamine. Cells were passaged (1:5 Figure 2. Size determination of RGD±AVEs, PEI±DNA, and RGD± for HUVECs and 1:10 for tumor cell lines) by trypsinization AVPs by photon correlation spectroscopy. The mean values for the (0.05% trypsin, 0.02% EDTA) and grown at 378Cin5% size distributions estimated by z-average calculation are indicated. CO2.

Cancer Gene Therapy, Vol 8, No 2, 2001 110 MUÈ LLER, NAHDE, FAHR, ET AL: TRANSDUCTION OF ENDOTHELIAL CELLS BY TARGETED ANIONIC LIPOSOMES

Figure 4A. Toxicity of AVPs for HUVECs compared to Lipofectamine (LA) and Superfect (SF). The analyses were performed at 6 and 24 hours after transfection (see Materials and Methods for details). A: Identification of apoptotic cells by staining with Hoechst 33258. Control, untransfected cells.

Cancer Gene Therapy, Vol 8, No 2, 2001 MUÈ LLER, NAHDE, FAHR, ET AL: TRANSDUCTION OF ENDOTHELIAL CELLS BY TARGETED ANIONIC LIPOSOMES 111

0.9

6 h 0.8 24 h

0.7

0.6

0.5

0.4

0.3

Absorbance (440nm) 0.2

0.1

0 Control RGD-AVP Lipofectamine Superfect Figure 4B. Determination of metabolic activity by WST assay. The following amounts of DNA and transfection reagents were used (per 3- cm well): AVP, 5 g DNA+20 g RGD±AVE; Lipofectamine, 2 g DNA+20 g lipid; Superfect, 2 g DNA+30 g of the transfection reagent.

WST assay FACS analysis of v 3 expression One day before transfection, cells were seeded in six-well Cells from a 10-cm dish were detached with 0.2% EDTA plates (800,000 cells/well). Cells were incubated for 6 (Roche), washed once with PBS, and incubated with the hours with the transfection reagents. Lipofectamine and monoclonal anti± v 3 antibody LM609 (Chemicon Int., Lipofectin were purchased from Gibco-BRL (Eggenstein, Hofheim, Germany) diluted 1/100 in PBS with 1% FCS. Germany). Effectene and Superfect were obtained from After washing once with PBS, cells were stained with a Qiagen (Hilden, Germany). The concentrations of the Cy3-labeled secondary antibody (goat anti±mouse-Cy3, 0 transfection reagents were optimized for transduction F(ab)2 fragment; Dianova, Hamburg, Germany) diluted efficiencies and low toxicities (see legend to Figure 4 1/200 in PBS with 1% FCS. All incubations were for details). The medium was replaced with normal culture performed on ice for 30 minutes. PBS-washed cells were medium, and WST assays (Roche, Mannheim, Germany) analyzed with a FACSCalibur using an excitation at 550 were performed either directly or after another 18 hours. nm. The fluorescence was amplified linearly. A minimum The supernatant was withdrawn; 500 L of WST diluted of 10,000 cells was analyzed. As control, unlabeled cells 1:10 in medium was added to each well and incubated for and cells stained with the secondary antibody only were 30 minutes at 378C. Aliquots of 100 L were transferred measured. to 96-well plates, and the plates were read at 440 nm. All measurements were performed as triplicates, and each experiment was repeated at least three times. RESULTS Construction and physico-chemical properties of AVEs FACS analysis of GFP expression Unilamellar vesicles were prepared from dry lipid films of Cells were grown on 10-cm dishes to 70% confluence and phosphatidylethanolamine, phosphatidylserine, and choles- transfected for 1 hour with AVPs made up of 45 g of DNA terol, which represent major constituents of the HIV and 180 g of AVE. The cells were trypsinized 23 hours envelope, and extruded using a LiposoFast2 extruder. The later, washed once with PBS, fixed in ice-cold 75% ethanol cyclic ligand with the RGD motif 27 was coupled to the overnight at 48C, resuspended in PBS, treated with RNase A liposomes via an anchor lipid containing a glutaric acid (Roche, Mannheim, Germany; 400 g/mL) overnight at group. Coupling efficiency was 1 mol%, corresponding to 48C, and stained with propidium iodide (20 g/mL) for at 1000 peptide ligands per liposome, as judged by coupling a least 10 minutes. The cells were analyzed by flow cytometry fluorescent derivative of the peptide to the liposomal surface. (FACSCalibur; Becton Dickinson, Heidelberg, Germany) The average size of these RGD±AVEs was 94 nm according using a laser excitation at 488 nm. to photon correlation spectroscopy analysis (Fig 1). The

Cancer Gene Therapy, Vol 8, No 2, 2001 112 MUÈ LLER, NAHDE, FAHR, ET AL: TRANSDUCTION OF ENDOTHELIAL CELLS BY TARGETED ANIONIC LIPOSOMES zeta potential was À45 mV. The AVEs could be stored for Ligand-dependent interaction with ECs >2months without loss of activity and did not exhibit any HUVECs were exposed to AVPs for 1 hour with or size increase. Plasmid DNA was condensed by mixing with without the RGD ligand that was labeled with a fluore small molecular mass nonlinear PEI (or protamine sulfate scent lipid (N- (7-nitrobenz-2-oxa-1,3-diazol-4-yl) - for comparison). DNA condensation was already achieved 1,2-dihexadecanoyl- sn-glycero-3-phospho-ethanolamine, at an N/P ratio of 8, as judged by dye exclusion (Fig 2). NBD-PE). Cells were observed directly 1 hour after At an N/P ratio of 20.7, which was chosen for all subsequent incubation with the labeled AVPs (Fig 5; 1 hour) or after experiments, DNA condensation was complete (Fig 2). The another 2hours in normal culture medium (Fig 5; 3 hours) mean size of this complex was 146 nm (Fig 1). Packaging of under a fluorescence microscope. At both time points, a this PEI±DNA plasmid complex into the RGD±AVEs gave considerably stronger staining was detectable in cultures that rise to a moderate size increase and yielded particles with a had received the RGD±AVPs. More than 99% of these cells mean size of 182nm (Fig 2). These fully assembled were stained. Confocal microscopy also showed that the particles will subsequently be referred to as artificial viral staining was not confined to the cell surface, but also particles (AVPs). The expected structure of the AVPs, i.e., occurred intracellularly, indicating endosomal uptake of the the encapsulation of condensed DNA in a lipid layer within RGD±AVPs (data not shown). These observations demon- discrete particles, was confirmed by cryoelectron micro- strate a highly efficient and selective interaction of the scopy (Fig 3). The AVPs were stable for at least 350 hours RGD±AVPs with ECs. in solution as measured by photon correlation spectroscopy (data not shown). Highly efficient transduction of ECs Toxicity for ECs Next, we evaluated the transduction efficiency of the To assess potential toxicities, the RGD±AVPs were RGD±AVPs with respect to (i) transduction efficiency, incubated with cultures of HUVECs for 6 hours and (ii) contribution of the ligand to EC transduction, and observed for another 18 hours. No indication for RGD± (iii) cell type specificity. For this purpose, a plasmid AVP±induced cell death could be detected, as suggested by carrying a CMV promoter-driven nuclear green fluorescent the absence of apoptotic cells identified by staining with protein (H2BGFP) gene 28 was packaged into RGD± Hoechst 33258 (Fig 4A). In addition, no influence on AVPs. Cells were exposed for 1 hour to the RGD±AVPs, metabolic activity of the cells in the presence or absence of and the fraction of green fluorescent cells was determined the RGD±AVPs was seen, as evidenced by WST assay (Fig 23 hours later (Fig 6). A quantitative evaluation of the 4B). In contrast, complexes of DNA with commercially experiment showed that the majority of cells were available cationic lipids (Lipofectamine, Lipofectin) or successfully transduced by RGD±AVPs and expressed polymers (Effectene, Superfect) killed the majority of the GFP. These experiments were performed in complete cell cells within the same period of time (Fig 4 and data not culture medium containing 10% FCS, indicating that the shown). RGD±AVPs were not inhibited by serum proteins. When

- RGD + RGD

1h

Figure 5. Binding of AVPs to HUVECs. HUVECs were exposed for 1 hour to 3h NBD-PE±labeled AVPs with and without RGD ligand and analyzed by fluores- cence microscopy either directly ( 1 hour) or after another 2 hours in normal culture medium (3 hours).

Cancer Gene Therapy, Vol 8, No 2, 2001 MUÈ LLER, NAHDE, FAHR, ET AL: TRANSDUCTION OF ENDOTHELIAL CELLS BY TARGETED ANIONIC LIPOSOMES 113

Figure 6. Transduction of HUVECs by RGD±AVPs carrying a CMV-H2BGFP plasmid. Cells were exposed for 1 hour to the AVPs and fluorescence microscopy was performed after another 23 hours in normal medium (left panel). The right panel shows the same cells stained with Hoechst 33258. Representative pictures are shown.

protamine sulfate was used as the DNA-compacting agent targeting contributes to the observed EC selectivity of instead of PEI, we observed a >50-fold decrease in the RGD±AVPs. transduction efficiency. Likewise, the PEI±DNA complex, on its own (without lipid envelope), gave poor transduc- tion efficiencies ( <5% transduced cells). These results underscore the importance of the precise composition of the RGD±AVPs established in this study and strongly suggest that PEI fulfils other roles than merely condensing the DNA. To quantitate the results, we performed FACS analyses and, for comparison, included AVPs lacking the RGD ligand. The data in Figure 7 clearly confirm the dramatic transduction efficiency already seen in the experiment shown in Figure 6. Approximately 99% of the cells receiving RGD±AVPs showed GFP expression. In contrast, GFP expression was seen in only 40% of cells transduced with AVPs lacking the ligand, and the level of this expression was considerably lower than that seen with the ligand-carrying AVPs. Although these results clearly indicate a role for the RGD ligand in the transduction of HUVECs, the significant extent of transduction seen with the RGD-less AVPs points to an additional mechanism of EC targeting by the AVPs.

Cell type selectivity of transduction Finally, we analyzed the transduction of HUVECs relative to other cell types, which lack v 3 integrin expression, as determined by immunostaining analysis using antibodies recognizing the heterodimeric v 3 integrin complex. The results of this experiment are summarized in Figures 8 and 9. The data obtained with the RGD±AVPs clearly show that the fraction of GFP- positive cells was approximately 10-fold higher with HUVECs compared to cells expressing no or low levels of v 3 integrin, including osteosarcoma cells (Saos-2), prostate carcinoma cells (DU-145), colon carcinoma cells (LoVo), lung adenocarcinoma cells (A549), and choriocarcinoma cells (JEG-3). However, we also found a preferential transduction of ECs by the RGD-less Figure 7. FACS analysis of HUVECs 24 hours after transduction with AVPs (Fig 8). In this case, the fraction of transduced AVPs ( ‹RGD ligand) carrying a CMV-H2BGFP plasmid compared HUVECs clearly exceeded that seen with the other cell to untreated cells. Cells were exposed for 1 hour to the AVPs and types by a factor of 5. This observation supports the FACS analysis was performed after another 23 hours in normal conclusion that a mechanism other than RGD-mediated medium.

Cancer Gene Therapy, Vol 8, No 2, 2001 114 MUÈ LLER, NAHDE, FAHR, ET AL: TRANSDUCTION OF ENDOTHELIAL CELLS BY TARGETED ANIONIC LIPOSOMES

100 +RGD -RGD 80

60

40

20

0 Fraction of positive cells (% of of positive RGD-AVP) Fraction HUVEC + HUVEC Saos-2 DU-145 LoVo A549 JEG-3 Figure 8. Transduction of different cell types with AVPs (A) and RGD±AVPs (B) carrying a CMV promoter H2BGFP plasmid: summary of FACS analyses determining the fraction of GFP-positive cells. Cells were exposed for 1 hour to the AVPs and FACS analysis was performed after another 23 hours in normal medium. Values were calculated relative to the transduction efficiency obtained with RGD-AVP±transfected HUVECs (100%).

HUVEC Saos-2 DU-145

4 4 4 100 101 102 103 10 100 101 102 103 10 100 101 102 103 10

LoVo A549 JEG-3

4 2 3 4 4 100 101 102 103 10 100 101 10 10 10 100 101 102 103 10

Figure 9. FACS analysis of v 3 expression in HUVECs and different tumor cell lines after indirect immunostaining as described in Materials and Methods. Bold lines: specific staining (first antibody: v 3 -specific; second antibody: Cy3-labeled); thin line: unspecific staining (second antibody only); dotted line: unlabeled cells. The relative fraction of cells is plotted against the intensity of fluorescence.

Cancer Gene Therapy, Vol 8, No 2, 2001 MUÈ LLER, NAHDE, FAHR, ET AL: TRANSDUCTION OF ENDOTHELIAL CELLS BY TARGETED ANIONIC LIPOSOMES 115

DISCUSSION cyclic RGD-containing peptide has previously been identified by phage display techniques as ligand for tumor Nonviral, lipid-based vectors have attracted particular EC targeting, presumably through interaction with vû3 attention in the field of gene therapy for a number of integrin.27 Our data show that the presence of this peptide on reasons, including safety, lack of antigenicity, versatility, and the surface of the AVPs significantly increases their ability to ease of handling. Most studies pertaining to the development interact with ECs (Fig 4) as well as their transduction of liposomal vectors have made use of cationic lipids, mostly efficiency (Figs 6±8). That the RGD peptide is instrumental because of their intrinsic property to complex with, and thus in EC targeting is also suggested by the observation that cell 29 condense, the negatively charged DNA. However, lipo- lines lacking vû3 integrin expression (Fig 9) showed a somes with cationic lipids undergo intense reorganization clearly reduced transduction (Fig 8). However, our data also after contact with DNA, and the resulting structures barely suggest that RGD-mediated targeting in not the only way by resemble liposomes.30 However, liposomes exhibiting a which EC selectivity is achieved by the RGD±AVPs. Thus, neutral or negatively charged surface are unable to AVPs lacking the RGD ligand were also able to transduce encapsulate sufficient amounts of plasmid DNA which has ECs to a significant extent (Fig 7), and this transduction a comparable hydrodynamic diameter as the liposomes efficiency was clearly greater than that seen with cell type 29 themselves. This problem can be circumvented by lacking vû3 integrin expression (Fig 8). It has previously precondensing the plasmid DNA with a synthetic polycation been shown that anionic lipids, such as phosphatidylserine, and coating this complex with negatively charged liposomal can bind to lipoprotein receptors.40 Such receptors are also membranes. The generation of transduction-competent present in ECs, and some may be expressed in a cell type± particles, according to this principle, has been accomplished specific fashion. Interaction of phosphatidylserine in liposo- by using polylysine-complexed DNA in conjunction with mal membranes with ECs has indeed been described,41 anionic and pH-sensitive lipids, the latter being required for and the adhesion of human erythrocytes to vascular endosomal release of the DNA.31 There are, however, a endothelium has been reported to be dramatically increased number of drawbacks associated with this approach, such as by phosphatidylserine present on the erythrocyte surface.42 the high toxicity of the polylysine and the serum sensitivity It is, therefore, not unlikely that a similar mechanism of the particles resulting from the presence of the pH- contributes to the selective transduction of ECs by RGD± sensitive lipid CHEMS. AVPs which contain phosphatidylserine as a major consti- We have used a different strategy for the construction of tuent. This hypothesis will be explored in more detail in anionic lipid-based vectors, i.e., condensation of DNA future experiments. with PEI and packing of these complexes into anionic The specific composition of the targeted AVPs, in liposomes whose composition is based on that of human particular the use of physiologic, anionic lipids, suggests immunodeficiency virus, referred to as AVEs.32 PEI that the vector system should also be applicable in vivo and appeared to be particularly suitable for this approach might have some advantages over previously tested because of its efficient DNA-packing properties,33 its liposomal vectors. With most cationic gene delivery systems, known endosomolytic property,34 its ability to protect the a net positive charge promotes the unspecific interaction of complexed DNA from cytoplasmic nucleases,35,36 and its the cationic complex with cell surface and plasma claimed potential to promote nuclear entry.37,38 In the proteins.24,43 Opsonization, destabilization, and rapid uptake present study, we show that condensation of plasmid DNA of these complexes by phagocytic cells lead to diminished with low-molecular-weight, branched PEI (Fig 1) and transduction efficiencies in vivo. Negatively charged lipo- packaging of this complex into AVEs gives rise to somes also interact with the biological environment in a structures (AVPs) resembling viral particles (Fig 3) that nonspecific manner, but a liposomal gene delivery system are nontoxic (Fig 4) and exhibit a very high transduction with a net negative surface potential should exhibit less efficiency for ECs in the presence of serum (Figs 6±8). nonspecific tissue uptake and a better overall biocompat- One of the most crucial components of these AVPs is, ibility than cationic carrier systems.43 indeed, the nature of the condensing agent. Thus, the Taken together, our data indicate that the RGD±AVPs exchange of the low-molecular-weight, branched PEI with are nontoxic, serum-resistant, selective for ECs, and protamine sulfate decreased the transduction efficiency by exhibit an unprecedented transduction efficiency for ECs. >50-fold. Likewise, the use of high-molecular-weight In principle, the same concept should also work with other PEI basically abrogated the transduction of ECs (data not ligands and cell types, and the use of tissue-specific shown). This observation is in agreement with the finding promoters 44 could increase the selectivity of the vector that low-molecular-weight, branched PEI has a low system even further. We, therefore, believe that these toxicity for cultured cells compared to the strongly toxic targeted liposomal vectors represent a true advance in the high-molecular-weight PEI.39 field of vector development. The AVPs designed in the present study were also endowed with a targeting device for vû3 integrins which is overexpressed on tumor ECs.8 Because cultured ECs also express this integrin (see Fig 9), probably because ACKNOWLEDGMENTS they are in a partially activated state due to the artificial culture conditions (our unpublished observations), these The authors thank Dr. H. Schreier (Sebastopol, CA) for cells represent a suitable model for targeting studies. A useful discussions and critical reading of the manuscript;

Cancer Gene Therapy, Vol 8, No 2, 2001 116 MUÈ LLER, NAHDE, FAHR, ET AL: TRANSDUCTION OF ENDOTHELIAL CELLS BY TARGETED ANIONIC LIPOSOMES

Drs. T. Kanda and G. Wahl (La Jolla, CA) for the H2BGFP knob-independent pathway for infection. JVirol. 1999;73: plasmid; Prof. T. Kissel (Marburg) for low-molecular- 5156±5161. weight, branched PEI used in initial studies; Dr. Stark (IMT, 19. Reynolds P, Dmitriev I, Curiel D. Insertion of an RGD motif Marburg) for Cryo-TEM analysis; Dr. Krause (IMT, into the HI loop of adenovirus fiber protein alters the Marburg) for peptide synthesis; and Dr. G. Bantus (Halle) distribution of transgene expression of the systemically administered vector. Gene Ther. 1999;6:1336±1339. for valuable suggestions and help during the optimization of 20. Felgner PL, et al. Lipofection: a highly efficient, lipid- the lipid anchor synthesis and peptide conjugation. mediated DNA transfection procedure. Proc Natl Acad Sci USA. 1987;84:7413±7417. 21. Wagner E, et al. Influenza virus hemagglutinin HA-2 N- terminal fusogenic peptides augment gene transfer by transfer- REFERENCES rin±polylysine±DNA complexes: toward a synthetic virus± like gene-transfer vehicle. Proc Natl Acad Sci USA. 1992;89: 1. Jain RK. Delivery of molecular and cellular medicine to solid 7934±7938. tumors. J Controlled Release. 1998;53:49±67. 22. Boussif O, et al. A versatile vector for gene and oligonucleotide 2. Folkman J. Fighting cancer by attacking its blood supply. Sci transfer into cells in culture and in vivo: polyethyleneimine. Am. 1996;275:150±154. Proc Natl Acad Sci USA. 1995;92:7297±7301. 3. Boehm T, Folkman J, Browder T, O'Reilly MS. Antiangio- 23. Lee RJ, Huang L. Lipidic vector systems for gene transfer. Crit genic therapy of experimental cancer does not induce acquired Rev Ther Drug Carrier Syst. 1997;14:173±206. drug resistance. Nature. 1997;390:404±407. 24. Szebeni J. The interaction of liposomes with the complement 4. Arap W, Pasqualini R, Ruoslahti E. Chemotherapy targeted to system. Crit Rev Ther Drug Carrier Syst. 1998;15:57±88. tumor vasculature. Curr Opin Oncol. 1998;10:560±565. 25. Kaneda Y, Saeki Y, Morishita, R. Gene therapy using HVJ 5. Plate KH, et al. Vascular endothelial growth factor and glioma liposomes: the best of both worlds? Mol Med Today. angiogenesis: coordinate induction of VEGF receptors, dis- 1999;5:298±303. tribution of VEGF protein and possible in vivo regulatory 26. Chander, R, Schreier H. Artificial viral envelopes containing mechanisms. Int J Cancer. 1994;59:520±529. recombinant human immunodeficiency virus (HIV) gp160. 6. Burrows FJ, et al. Up-regulation of endoglin on vascular Life Sci. 1992;50:481±489. endothelial cells in human solid tumors: implications for 27. Pasqualini R, Koivunen E, Ruoslahti E. Alpha v integrins as diagnosis and therapy. Clin Cancer Res. 1995;1:1623±1634. receptors for tumor targeting by circulating ligands. Nat 7. Miller DW, et al. Elevated expression of endoglin, a component Biotechnol. 1997;15:542±546. of the TGF±beta-receptor complex, correlates with prolifera- 28. Kanda T, Sullivan KF, Wahl GM. Histone±GFP fusion protein tion of tumor endothelial cells. Int J Cancer. 1999;81:568±572. enables sensitive analysis of chromosome dynamics in living 8. Ruoslahti E. Fibronectin and its integrin receptors in cancer. mammalian cells. Curr Biol. 1998;8:377±385. Adv Cancer Res. 1999;76:1±20. 29. Ledley FD. Nonviral gene therapy: the promise of ge- 9. Pasqualini R, et al. Aminopeptidase N is a receptor for tumor- nes as pharmaceutical products. Hum Gene Ther. 1995;6: homing peptides and a target for inhibiting angiogenesis. 1129±1144. Cancer Res. 2000;60:722±727. 30. Sternberg B, Sorgi FL, Huang L. New structures in complex 10. Witte L, et al. Monoclonal antibodies targeting the VEGF formation between DNA and cationic liposomes visualized receptor-2(Flk1/KDR) as an antiangiogenic therapeutic by freeze-fracture electron microscopy. FEBS Lett. 1994;356: strategy. Cancer Metastasis Rev. 1998;17:155±161. 361±366. 11. Ramakrishnan S, Olson TA, Bautch VL, Mohanraj D. Vascular 31. Lee RJ, Huang L. Folate-targeted, anionic liposome-entrapped endothelial growth factor±toxin conjugate specifically inhibits polylysine-condensed DNA for tumor cell±specific gene KDR/flk-1±positive endothelial cell proliferation in vitro and transfer. J Biol Chem. 1996;271:8481±8487. angiogenesis in vivo. Cancer Res. 1996;56:1324±1330. 32. Schreier H, et al. (Patho)physiologic pathways to drug tar- 12. Klement G, et al. Continuous low-dose therapy with vinblas- geting: artificial viral envelopes. J Mol Recognit. 1995;8: tine and VEGF receptor-2antibody induces sustained tumor 59±62. regression without overt toxicity. J Clin Invest. 2000;105:R15± 33. Godbey WT, Wu KK, Hirasaki GJ, Mikos AG. Improved R24. packing of poly(ethyleneimine) /DNA complexes increases 13. Matsuno F, et al. Induction of lasting complete regression of transfection efficiency. Gene Ther. 1999;6:1380±1388. preformed distinct solid tumors by targeting the tumor 34. Behr J. The proton sponge: a trick to enter cells the viruses did vasculature using two new antiendoglin monoclonal antibodies. not exploit. Chimia. 1997;51:34±36. Clin Cancer Res. 1999;5:371±382. 35. Liu Y, et al. Cationic liposome-mediated intravenous gene 14. Arap W, Pasqualini, R, Ruoslahti E. Cancer treatment by delivery. J Biol Chem. 1995;270:24864±24870. targeted drug delivery to tumor vasculature in a mouse model. 36. Liu Y, et al. Factors influencing the efficiency of cationic Science. 1998;279:377±380. liposome-mediated intravenous gene delivery. Nat Biotechnol. 15. Rajotte D, et al. Molecular heterogeneity of the vascular 1997;15:167±173. endothelium revealed by in vivo phage display. J Clin Invest. 37. Pollard H, et al. Polyethyleneimine but not cationic lipids 1998;102:430±437. promotes transgene delivery to the nucleus in mammalian cells. 16. Colin M, et al. Liposomes enhance delivery and expression of J Biol Chem. 1998;273:7507±7511. an RGD-oligolysine gene transfer vector in human tracheal 38. Godbey WT, Wu KK, Mikos AG. Tracking the intracellular cells. Gene Ther. 1998;5:1488±1498. path of poly(ethyleneimine) /DNA complexes for gene 17. Erbacher P, Remy JS, Behr JP. Gene transfer with synthetic delivery. Proc Natl Acad Sci USA. 1999;96:5177±5181. virus±like particles via the integrin-mediated endocytosis 39. Fischer D, et al. A novel nonviral vector for DNA delivery pathway. Gene Ther. 1999;6:138±145. based on low-molecular-weight, branched polyethyleneimine: 18. Vigne E, et al. RGD inclusion in the hexon monomer effect of molecular weight on transfection efficiency and provides adenovirus type 5±based vectors with a fiber cytotoxicity. Pharm Res. 1999;16:1273±1279.

Cancer Gene Therapy, Vol 8, No 2, 2001 MUÈ LLER, NAHDE, FAHR, ET AL: TRANSDUCTION OF ENDOTHELIAL CELLS BY TARGETED ANIONIC LIPOSOMES 117

40. Fadok VA, et al. The role of phosphatidylserine in recognition 46. MacDonald RC, et al. Small-volume extrusion apparatus for of apoptotic cells by phagocytes. Cell Death Differ. 1998;5: preparation of large, unilamellar vesicles. Biochim Biophys 551±562. Acta. 1991;1061:297±303. 41. Kamps JA, Morselt HW, Scherphof GL. Uptake of liposomes 47. Weissig V, Lasch J, Klibanov AL, Torchilin VP. A new containing phosphatidylserine by liver cells in vivo and by hydrophobic anchor for the attachment of proteins to liposomal sinusoidal liver cells in primary culture: in vivo±in vitro membranes. FEBS Lett. 1986;202:86±90. differences. Biochem Biophys Res Commun. 1999;256:57±62. 48. Blakeslee D, Baines MG. Immunofluorescence using dichlor- 42. Closse C, Dachary-Prigent J, Boisseau MR. Phosphatidylser- otriazinylaminofluorescein (DTAF): I. Preparation and frac- ine-related adhesion of human erythrocytes to vascular tionation of labelled IgG. J Immunol Methods. 1976;13:305. endothelium. Br J Haematol. 1999;107:300±302. 49. Fahr A, Butt H-J, Nelles G, Welz C. Biophysical aspects 43. Roerdink F, Wassef NM, Richardson EC, Alving CR. Effects of nonviral gene delivery systems. J Liposome Res. 1998;8: of negatively charged lipids on phagocytosis of liposomes 9±12. opsonized by complement. Biochim Biophys Acta. 1983;734: 50. Ruf H, Georgalis Y, Grell E. Dynamic laser light scattering to 33±39. determine size distributions of vesicles. Methods Enzymol. 44. Nettelbeck DM, Jerome V, Muller R. Gene therapy: designer 1989;172:364±390. promotersfortumourtargeting.TrendsGenet.2000;16:174±181. 51. Jaffe E. In: Jaffe E, ed. Biology of Endothelial Cells. Boston: 45. Ahl PL, et al. Enhancement of the in vivo circulation lifetime of Martinus Nijhoff; 1986:1±13. L -alpha-distearoylphosphatidylcholine liposomes: importance 52. Thornton SC, Mueller SN, Levine EM. Human endothelial of liposomal aggregation versus complement opsonization. cells: use of heparin in cloning and long-term serial Biochim Biophys Acta. 1997;1329:370±382. cultivation. Science. 1983;222:623±625.

Cancer Gene Therapy, Vol 8, No 2, 2001