Polyethylenimine (PEI) is a simple, inexpensive and effective reagent ...

Gene Therapy (1997) 4, 773–782  1997 Stockton Press All rights reserved 0969-7128/97 $12.00

Polyethylenimine (PEI) is a simple, inexpensive and effective reagent for condensing and linking plasmid DNA to adenovirus for gene delivery A Baker1, M Saltik1, H Lehrmann1, I Killisch1, V Mautner2, G Lamm1, G Christofori1 and M Cotten1 1

Institute for Molecular Pathology, Dr Bohr Gasse 7, 1030 Vienna, Austria; and 2CRC Institute for Cancer Studies, The University of Birmingham, Edgbaston, UK

A simple and inexpensive method of condensing and linking plasmid DNA to carrier adenovirus particles is described. The synthetic polycation polyethylenimine is used to condense plasmid DNA into positively charged 100 nm complexes. These PEI–DNA complexes are then bound to adenovirus particles through charge interactions with negative domains on the viral hexon. The resulting

transfection complexes deliver plasmid DNA to cells by the adenovirus infectious route without interference from virus gene expression because psoralen-inactivated virus is employed. The PEI–DNA–adenovirus complexes display DNA delivery comparable to more sophisticated DNA virus complexes employing streptavidin/biotin linkage, but require no special reagents and are much easier to prepare.

Keywords: transfection; synthetic polycations; DNA delivery; virus entry

Introduction Nearly all successful DNA delivery systems include some form of polycation. The charged, extended DNA molecule is highly susceptible to mechanical shearing and to enzymatic cleavage. Cationic agents are used to neutralize the negative charge of the DNA and condense its structure1,2 and thereby protect DNA from nuclease degradation.3 Polycations that have been used extensively for gene delivery applications include polylysine,4– 16 the cationic antibiotic gramicidin S,17–19 dendrimers or cascade polymers,20–22 or cationically modified albumin.23 Natural DNA condensation agents such as histones, protamines, and high mobility group proteins have also been investigated for gene delivery applications.24–27 Adenovirus virions added to transfection systems as free particles enhance the cellular uptake of DNA, 12,28–31 with the enhancement being a function of the cell permeabilization activity of the adenovirion.32–35 The most effective methods of employing adenoviruses for plasmid DNA delivery link the DNA to the virion in some manner. This linkage ensures that the DNA follows the efficient entry pathway of the adenovirus and lowers the quantity of virus and DNA required. Until now, most adenovirus–plasmid DNA systems have used polylysine as the DNA-binding and condensation agent because it is an effective DNA condensing agent and it is easy to modify chemically. Several methods of covalently coupling polylysine to the exterior of the adenovirus virus have been described, including an enzymatic transglutaminase method36,37 and two different chemical coupling

Correspondence: M Cotten Received 8 January 1997; accepted 14 April 1997

methods.38–40 However, methods of covalent linkage of polylysine to adenovirus virions suffer from precipitation and storage problems. The adenovirus type 5 capsid itself is negatively charged and nonspecific polylysine–adenovirus aggregates can form which limit the activity of these systems. One partial remedy involves storage of the virus–polylysine in high salt (1–2 m) but this can complicate application of these complexes. More flexible strategies have used specific but noncovalent interactions to attach antibody–polylysine (AbpL) conjugates41,42 or streptavidin–polylysine (StrpL)43 conjugates to the surface of the adenovirus. The AbpL and the StrpL methods have proven to be versatile. Because optimum gene delivery requires a certain amount of titration of the virus to DNA to polylysine ratios, the AbpL and the StrpL conjugates allow substantial titration flexibility. However, because neither the AbpL nor the StrpL conjugates are commercially available, one limit to these systems has been the synthesis of AbpL and the StrpL conjugates. It would be useful to have the flexibility of the conjugate systems without having to organize the synthesis of the conjugates. Recently the polycation polyethylenimine (PEI) has been demonstrated to be a useful reagent for gene delivery.44,45 In our efforts to develop a system for transfecting large DNA molecules such as bacterial artificial chromosomes (BACs) we encountered solubility problems with polylysine condensates of large DNA. These difficulties were solved by replacing the polylysine in the adenovirus–DNA complexes with the synthetic cation PEI.46 We have further developed the PEI–adenovirus system and we find that within certain limits it can be as useful for gene delivery as the previously described biotin-adenovirus–streptavidin-polylysine conjugate system.43 The advantages of the PEI system include the ready

PEI linkage of plasmid DNA to carrier adenovirus A Baker et al

774

availability of the reagents; the system requires only adenovirus (or psoralen-inactivated adenovirus) and commercially available PEI.

Results Titration of PEI–DNA Initial work with PEI as a DNA condensing agent for the transfection of very large DNA suggested that a positively charged DNA–PEI complex was essential for gene delivery with adenovirus carrier particles.46 We have examined this in more detail for the delivery of conventionally sized DNA plasmids (the 7 kb pCLuc used here). Boussif et al44 found that optimum DNA delivery was obtained with PEI nitrogen to DNA phosphate ratios (N:P) around 9–13, and this ratio results in DNA complexes with a calculated net positive charge. We condensed DNA with various quantities of PEI, added adenovirus particles and tested gene delivery with the resulting complexes by measuring the luciferase activity generated by successful introduction into the cell of the pCLuc plasmid. Maximum DNA delivery is obtained at PEI nitrogen to DNA phosphate ratio above 10 but drops at lower ratios (Figure 1a). The DNA–PEI complexes generated at ratios above 10 have a net positive charge and this is probably important for adenovirus capsid binding (see below). The decline in gene delivery at lower N:P ratios supports this conclusion and contrasts with the more gradual decline seen when titrating higher molecular weight PEI:DNA ratio in the absence of virus.44 With the low molecular weight PEI used here (MW 2000), gene delivery is dependent on the presence of adenovirus (sample 10, see also below). The gene delivery activity reported for PEI alone is only observed with higher molecular weight forms of PEI (25 000 MW, see Figures 5e and Table 3, line 6). However, an augmentation of gene transfer by adenovirus is also obtained with higher molecular weight PEI (compare Figure 5e with f; Table 3 line 6 with line 7). Titration of virus We next determined the quantity of adenovirus particles required for effective delivery (Figure 1b). With a constant amount of PEI–DNA complex, the presence of increasing amounts of virus per cell (from 70 to 7000) led to a steady increase in gene expression (lanes 2 to 6). Beyond this level of virus, the resulting gene expression declined slightly, probably due to toxicity associated with high virus levels (lanes 7 and 8). Comparing the PEI– DNA–adenovirus complexes with streptavidin–biotinadenovirus–transferrin-polylysine complexes that have previously been described 43 showed that at the lower virus levels (eg 7000 virus per cell and lower) the PEI complexes generated higher levels of gene expression while at higher virus levels the streptavidin–biotin system was more effective (Figure 1b, lanes 9–16). We conclude that in vitro, the PEI–DNA–adenovirus complexes can deliver DNA with efficiencies comparable to a previously established adenovirus system. Structure of PEI–DNA–adenovirus transfection complexes Light scattering was used to determine the sizes of complexes formed between PEI, DNA and adenovirus. DNA

Figure 1 Titration of PEI and adenovirus. (a) Titration of PEI. The indicated quantities of 10 mm PEI (2000 MW) were diluted in 250 ml of 20 mm HEPES, pH 7.4 and mixed with 6 mg of pCLuc DNA in 250 ml of 20 mm HEPES, pH 7.4. After 20 min at room temperature, 1.4 ml of psoralen-inactivated adenovirus type 5 (dl1014, 1.5 × 109 particles per microliter) was added. Following a second 20-min incubation, 50 ml aliquots of the material were added to 2 × 104 primary human fibroblasts cells in 250 ml of DMEM (no serum) as described in the Materials and methods. At 24 h after transfection, triplicate wells were harvested, extracts were prepared and analyzed for luciferase activity. The values listed are means with the standard deviation indicated. (b) Lanes 1–8: titration of adenovirus in PEI–DNA complexes. PEI–DNA complexes were prepared as in the top panel using 30 ml of 10 mm PEI with 6 mg of DNA. After 20 min at room temperature, the various quantities of psoralen-inactivated adenovirus type 5 (dl1014) were added, resulting in the indicated virus particle:cell ratio. Following a second 20-min incubation, 50 ml aliquots of the material were added to 2 × 104 primary human fibroblasts cells in DMEM (no serum) as described above, the subsequent analysis of luciferase activity was identical. Lanes 9–16: titration of adenovirus in StrpL–TfpL–DNA complexes. StrpL–TfpL–DNA complexes were prepared as described in the Materials and methods. Following complex formation, 50 ml aliquots of the material were added to 2 × 104 primary human fibroblasts cells in DMEM (2% horse serum) as described above, the subsequent analysis of luciferase activity was identical. The resulting virus particle to cell ratios are indicated.

was condensed with PEI in buffer (20 mm HEPES, pH 7.4) containing NaCl from 0 to 500 mm. In the absence of NaCl, the DNA is condensed into 100 nm particles (Table 1). The presence of even low concentrations of NaCl results in the formation of much larger complexes, from 600–1000 nm. The larger size is likely to be due to two phenomena: at lower NaCl concentration, a charge masking effect by the salt ions could promote aggregation of initially small, positively charged PEI–DNA complexes. A second phenomenon could be due to the weak DNA binding properties of the low molecular weight PEI, with the PEI–DNA interactions impaired at the higher ionic strengths resulting in less tightly condensed DNA. At low but not at higher NaCl concentrations the addition of adenovirus type 5 or CELO virus particles results in


775 Table 1 Sizes of PEI–DNA or PEI–DNA–adenovirus complexes as a function of the ionic strength during assemblya NaCl mm 0 25 150 300 500

PEI–DNA in nm (s.d.)b 101.8 772.7 1004.1 757.6 636.4

(1.1) (97.5) (52.3) (53.5) (16.6)

AD5–PEI–DNA in nm (s.d.) 201.6 813.2 720.9 633.7 635.3

(9.0) (90.5) (14.9) (19.2) (131.5)

CELO–EPI–DNA in nm (s.d.) 199.6 1054.8 649.1 643.4 560.2

(13.4) (50.6) (30.7) (85.7) (58.1)

a

Samples were prepared in 20 mm HEPES, pH 7.4 plus NaCl at the indicated concentration. Component concentrations were as described in Materials and methods (30 ml of 10 mm PEI in 250 ml added dropwise to 6 mg of DNA (pCLuc) in 250 ml followed by 3 × 109 particles adenovirus type 5 or CELO virus in 3–6 ml of HBS/40% glycerol. b Samples sizes were measured at 1 h after addition of virus using a Brookhaven BI-90 (Brookhaven, NY, USA). Additional sizes: Ad5 (150 mm NaCl) 119.2 (3.2); CELO (150 mm NaCl) 123.6 (5.0).

complexes showing a modest increase in diameter (Table 1). This increase is due to modest changes in the ionic strength that occur when the virus (in 150 mm NaCl) is added. A smaller size of complexes prepared with CELO at 500 mm NaCl is reproducibly observed, possibly due to the lower negative charge on the CELO hexon, with CELO dissociating from PEI–DNA at this ionic strength. This dissociation of virus is consistent with the gene transfer activity of the complexes (discussed in detail below). We conclude from the light scattering results that the most discrete PEI–DNA–adenovirus complexes (200 nm) are formed in the absence of NaCl; in the presence of 150 mm NaCl, the PEI–DNA–adenovirus complexes formed are approximately 600–700 nm in diameter. To prepare the PEI–DNA–adenovirus complexes used here, we mixed 6 mg DNA (approximately 8 × 1011 molecules) with 3.5 × 109 virus particles resulting in approximately 200 DNA molecules per virus particle. Studies using a variety of sizes of DNA demonstrate that each PEI-condensed DNA complex contains approximately 10 DNA molecules of 7 kb (AB and MC, unpublished observations). The 200 DNA molecules per virus would generate, in the presence of PEI, approximately 20 DNA complexes per virus. Thus, unless there is a clustering of virus particles, only one in 20 DNA complexes should be associated with a virus particle. These conclusions are supported by electron micrographs of PEI–DNA and PEI–DNA–adenovirus complexes (Figure 2). PEI–DNA complexes associated with adenovirus are seen by electron microscopy with the expected frequency (approximately 1 in 20). Examples of complexes with and without associated virus particles are shown in Figure 2. The sizes of the complexes observed by electron microscopy are consistent with the sizes determined by light scattering (Table 1).

Ionic strength modulation of virus interactions with PEI– DNA complexes Because of viral evolution to avoid immune responses, it is expected that the exposed surfaces of the virion might be quite variable between adenovirus serotypes. For example, based on the crystal structure of the hexon,47 the EM structure of the virion,48 and the determination of antigenic sites on the hexon, 49–51 the subgroup C (Ad2 and 5) hexon residues 146–335 and 417–535 are considered to be exposed on the surface of the virion. The major antigenic determinants are dominated by two exposed loops. An approximation of the exposed hexon

Figure 2 Electron microscopic analysis of PEI–DNA–adenovirus complexes. Samples a and c were prepared using PEI (2000 MW) complexed with DNA (pCLuc) and adenovirus type 5 as described in Figure 1, with 6 mg DNA, 18 ml of 10 mm PEI (MW 2000) and 1.3 ml of adenovirus type 5 (1.5 × 109 particles per microliter). Samples b and d were prepared using five-fold higher concentrations of all components. After complex formation, the samples were incubated for 10 min on formvar/carbon-coated specimen grids which had been freshly glow discharged. Negative staining was achieved by application of 1% aqueous uranyl acetate. The samples were viewed with a JEOL 1210 transmission electron microscope operating at 80 kV. The bar indicates 80 nm.

surface charge can be made by considering the charge of amino acid residues 100–300, as this region includes these loops. The net charge of this region increases with Ad5 (−16) ,Ad3 (−9) ,CELO (−1) ,Ad40 (+2). Theoretically, the charge of the exposed fiber knob could also play a role in these interactions with PEI–DNA. Using data from Mei and Wadell52 and Chiocca et al53 the net charge of the knob regions of various adenovirus serotypes is Ad5 (−2), Ad3, (−3), CELO (−1) and Ad40 (1). Except for the reversal of Ad5 and Ad3, this is roughly the same ordering observed with hexon loop charge. However, because each virion possesses 720 hexon peptides but only 36 fiber peptides, the charge characteristics of the hexon should predominate. One might predict that if hexon surface charge interactions are important in establishing PEI–DNA contacts with the virion, then the sensitivity of gene delivery to ionic strength should follow the


776

suggesting that positively charged domains on the PEI– DNA interact with negative domains on the exposed surface of the hexon.

Figure 3 Titration of NaCl during complex formation, comparison of adenovirus type 5 to CELO virus. PEI–DNA–adenovirus complexes were prepared as in the top panel of Figure 1, sample 3, except that the NaCl concentration during assembly was varied from zero to 500 mm (at a constant 20 mm HEPES, pH 7.4). Similar particle numbers (2 × 109) of either psoralen-inactivated adenovirus type 5, or CELO virus were used, as indicated. Aliquots of the complexes (50 ml) were supplied to cells in 500 ml of DMEM containing no serum and the subsequent analysis of luciferase activity (at 48 h after transfection) were identical to Figure 1. Total luciferase activity is plotted (average of three samples, the standard deviation is indicated).

same pattern, with Ad5 and Ad3 less sensitive than Ad40 and CELO. PEI–DNA complexes were prepared at various NaCl concentrations with each of these virus types and tested for their gene delivery activity. We find that DNA delivery with Ad5 complexes varies less than 10-fold with ionic strengths from 0–500 mm (Figure 3, lanes 1–5; Table 2). The less charged CELO is predicted to be strongly influenced by salt during the complex formation. Indeed we observe a greater than 100-fold decline in gene delivery with CELO complexes formed in 500 mm NaCl compared to the zero NaCl sample (Figure 3, lanes 6–10; Table 2). Similar titrations performed with Ad3 and Ad40 revealed that Ad3 complexes at 500 mm NaCl behaved like complexes formed with the similarly charged Ad5, while the weakly charged Ad40 behaves like CELO (Table 2). Thus the response of transfection complexes to NaCl during assembly is consistent with an ionic bridge stabilizing the PEI–DNA interaction with the adenovirus. The charge character of the exposed loop region of each serotype’s hexon can largely account for the behavior of each serotype at higher ionic strengths,

Addition of a ligand or changing virus serotype can be used to enhance entry into specific cell types Adenovirus type 5 does not efficiently enter cells of hematopoietic origin.54,55 This may be due to poor expression of the virus receptor on these cells. The up-regulation of integrins, some of which can serve as receptor for the penton base, has been shown to enhance entry in lymphocytes. An alternative strategy to increase transduction of receptor-poor cells is to attach to the virus or transfection complex a ligand for the target cell; alternatively, a serotype of adenovirus that might bind the target cells with greater avidity could be used. Genetic modification of adenovirus has been used to add positively charged lysine residues to the tip of the Ad5 fiber, increasing the heparan sulfate binding potential of the resulting virus and allowing transduction of a wider range of cells.56 Another approach has been to use folate-modified antibodies to redirect adenovirus binding to the folate receptor.57 We tested the flexibility of the PEI–adenovirus system in this context by adding a ligand or changing the adenovirus serotype. The erythroleukemia cell line K562 is a hematopoietic cell type that is poorly transduced by adenovirus type 5 vectors. StrpL–pL–Ad5 DNA complexes can enter these cells but the entry is enhanced by 2 logs when the ligand transferrin is included in the complex as transferrin-polylysine (TfpL; Figure 4a, lanes 1 and 2). PEI–Ad5 complexes are also more efficiently delivered when TfpL is added to the complex (Figure 4a, lanes 3 and 4). PEI complexes formed with Ad3 however, have a basal level of transfection that is at least a log better than Ad5 complexes, suggesting that on K562 cells, the Ad3 receptor level is more favorably expressed than the Ad5 receptor. Even so, addition of TfpL to the complexes provides a further boost to transfection (Figure 4a, lanes 5 and 6). Comparing the activity of Ad5, Ad3 and CELO on A549 cells, we find that both of the human viruses work well without additional ligand (Figure 4b, lanes 9–15). The chicken adenovirus works much less efficiently than the human virus in this human cell line (Figure 4b, lanes 1–4) and addition of transferrin to the system allows the chicken virus to function nearly as well as the human virus (Figure 4b, lanes 5–8).

Table 2 Gene transfer as a function of ionic strength during complex assembly NaCl a

Ad5 (−16)b

0 25 150 300 500

100 61.5 43.5 30.7 21.1

mm mm mm mm mm

(6.4) (3.6) (3.4) (1.4)

Ad3 (−9) 100 43.5 (3.8) 24.4 (2.4) 25.5 (0.11) 14.5 (1.8)

Ad40 (+2) 100 99.1 44.8 20 0.6

(4.2) (6.1) (5.2) (0.08)

CELO (−1) 100 15.7 8.7 4.1 1.2

(3.0) (1.6) (0.3) (0.4)

Gene transfer into A549 cells using PEI–DNA–adenovirus complexes. Transfections were performed as described in Figure 3. Values are expressed as the percentage of the 0 mm NaCl luciferase expression for each virus type. Each value represents the average of three transfections, with a standard deviation indicated. a NaCl was present at the indicated concentrations during transfection complex assembly. b The net charge of the hexon surface loops is indicated in parentheses.


while inhibiting the Ad5 complexes to 50% of the no serum value. The presence of 10% fetal calf serum results in a nearly 100-fold drop in the delivery with Ad5 complex; while the Ad3 complex declines to 14% of the no serum value. These results indicate that the delivery with PEI–DNA– adenovirus complexes is mildly sensitive to serum, with optimum results obtained in the absence of serum. Ad3 complexes are less sensitive to the presence of serum suggesting that either the serum impairs Ad5–cell interactions greater than Ad3–cell interactions. Alternatively, the serum may influence the PEI–DNA interactions with the virus, and the Ad3 capsid provides a more stable PEI–DNA complex.

Figure 4 Addition of a ligand or changing virus serotype can be used to enhance entry into specific cell types. (a) K562 cells were grown for 18 h in the presence of desferroxamine to stimulate transferrin receptor levels.7 Just before transfection, the cells were suspended in fresh medium and plated in a 24-well plate, 500 000 cells per 250 ml per well. Lanes 1 and 2, StrpL–TfpL–adenovirus complexes were formed as described in the Materials and methods except that in the lane 1 sample, the 4 mg of TfpL was replaced with 2 mg of polylysine (equivalent amount of cation as supplied by the TfpL). Lanes 3–6: standard PEI–DNA–adenovirus complexes were prepared with 30 ml of 10 mm PEI, 6 mg of DNA, and 2 × 109 virus particles, with the following modification. Lane 4 and 6 samples: 0.4 mg of TfpL was added to the PEI solution before it was added to the DNA solution; lane 5 and 6 samples contained Ad3 rather than Ad5 particles. Each well (500 000 cells per 250 ml of RPMI, no serum) was transfected with 250 ml of DNA complex. Total luciferase activity is plotted (average of three samples, the standard deviation indicated). (b) Standard PEI–DNA–adenovirus complexes (with 30 ml of 10 mm PEI, 6 mg of DNA) were prepared with the following modifications. Samples 2–8 contained CELO virus at the indicated virus to cell ratios. The PEI solutions used to form samples 5–8 contained 0.4 mg of TfpL. Samples 9–12 contained Ad5 and samples 13–15 contained Ad3 at the indicated virus to cell ratios. Aliquots of the transfection complexes (50 ml) were added to A549 cells (50 000 cells per well ) in 250 ml of DMEM, no serum. Subsequent analysis of luciferase activity at 24 h after transfection was as described in Figure 1. Total luciferase activity is plotted (average of three samples, the standard deviation indicated).

Limitations of the PEI system Gene delivery is relatively stable with quantities of complex down to 30 ml of complex in 250 ml of medium for 20 000 cells in a 2.4 cm diameter well (24-well plate). However, 20 ml of complex in 500 ml of medium is completely negative for gene delivery (results not shown). This rapid fall-off of activity suggests that a dissociation of the effective gene delivery occurs when the complex is diluted. Many transfection methods are influenced by the presence of serum. The influence of serum on transfection efficiency was tested with PEI–DNA–adenovirus complexes, using complexes formed with either adenovirus type 5 or adenovirus type 3. With both virus serotypes, maximum delivery is obtained when the transfection occurs in the absence of serum. The presence of 2% horse serum has very little effect on the Ad3 complexes,

Transfection efficiency It is useful to determine if the PEI transfection method produces many transfected cells expressing the transgene at a low level or a few cells expressing at high level. This analysis was performed using a green fluorescent protein marker gene that allowed analysis of transfection at the cellular level by either microscopy or by FACS analysis. Cells were transfected with a plasmid encoding GFP using lipofectamine, the streptavidin-polylysine–biotinadenovirus system, PEI (2000 MW) in the absence or presence of adenovirus and PEI (25 000 MW) in the absence or presence of adenovirus. With all reagents, preliminary experiments with a luciferase plasmid identified optimum transfection conditions. Examples of typical transfected populations are shown in Figure 5. It is clear that the adenovirus–PEI systems are as efficient as the previously described StrpL–adenovirus complexes and largely superior to a cationic lipid. A quantification of the percentage of positively transfected cells is presented in Table 3. A more sensitive FACS analysis (Table 3) was performed to quantify the transfection efficiencies with greater precision, as the microscope reflects only the highest expressing cells. The transfection efficiencies of both forms of PEI are markedly enhanced by the presence of adenovirus, giving 25–40% of the cells expressing the transgene. The levels are higher than the values determined by microscopy but reflect an increased sensitivity of the FACS method (eg PEI 2000–Ad gives 9% by microscopy and 25–30% by FACS, PEI 25 000–Ad gives approximately 20% by microscopy and 35–40% by FACS (Table 3). The FACS analysis also provides information on the apoptotic status of the transfected cells. The cellular changes associated with apoptosis result in cells bearing sub-2 N quantities of chromosomal DNA which can be detected by propidium iodide staining. The percentage of GFP-positive cells displaying an apoptotic morphology (sub-2 N DNA content) is also presented in Table 3. Apoptosis in nontransfected cells under the same growth conditions is 0.5%. We find that many methods of transfection induce a cell death response and this is consistent with previous work on transfection-induced apoptosis.69 Importantly, the presence of adenovirus in the transfection complexes does not result in an increase in apoptosis rate compared to transfection with PEI– DNA complexes alone (compare sample 6, no virus, to samples 7 and 8, PEI with virus). PEI–DNA complexes enhance recombinant adenovirus entry PEI–DNA complexes with excess positive charge will bind and accumulate adenovirus particles into larger

777


778

Figure 5 Analysis of transfection efficiency by microscopy. At 24 h before transfection coverslips (24 × 24 mm glass, La Fontaine, Germany) were placed in 3.5 cm wells (six-well plate) and A549 cells were plated (400 000 per well). Transfection complexes were prepared with pEGFP-C1. Lipofectamine samples used 30 ml of Lipofectamine (Life Technologies, Rockville, MD, USA) in 250 ml of DMEM, mixed with 6 mg of DNA in 250 ml of DMEM; after 30 min at room temperature, 150 ml transfection mixture was added to cells. The pL–adenovirus sample was prepared as described in Figure 1, bottom panel sample 8. The PEI transfections were performed as in Figure 1, top panel. PEI 2000 sample used 18 ml of PEI 2000 (10 mm), the PEI 25 000 sample used 6 ml of 10 mm PEI 25 000, in both cases with 6 mg of DNA. Where indicated, 7 × 109 adenovirus particles were included in the complex. At 24 h after transfection, the cell-bearing coverslips were recovered, washed once in PBS and the cells were fixed in 1% formaldehyde in PBS for 10 min. The cells were washed 5 × 5 min in PBS, in the final wash Hoechst 33258 (Sigma, St Louis, MO, USA) was included at 100 ng/ml. Coverslips were dried, mounted under Mowiol (Hoechst) and examined by fluorescence microscopy. Images were captured by CCD and assembled using Adobe Photoshop. GFP expression generates a green signal, both cellular and transfected DNA generate a blue signal.

complexes. The initial adenovirus–cell interactions are limited by diffusion and the inclusion of an inert polymer that limits viral diffusion has been shown to improve viral transduction efficiency.58 Furthermore, the simple addition of charged amino acids to the Ad5 fiber has also improved transduction efficiencies in cells lacking abundant fiber receptors56 and the inclusion of polycations can also enhance Ad2 transduction efficiencies.70 We asked if including the adenovirus in a charged PEI–DNA complex could modulate virus entry into cells. A recombinant adenovirus type 5 bearing a luciferase gene in place of the E1 region (AdLuc1) was assembled into PEI–empty plasmid complexes and the material was supplied to pri-

mary human fibroblasts. In these complexes, the plasmid DNA, serves a structural role rather than encoding a gene to be delivered. For comparison, the same quantity of AdLuc1 was used in the absence of PEI–DNA. The resulting luciferase activity is plotted as a function of virus:cell ratio (Figure 6). At all ratios except the highest, the inclusion of the virus in a PEI complex resulted in a 2log increase in the quantity of luciferase gene expression. Thus, including the virus in a larger, charged complex results in more efficient transduction. This may have useful applications in vitro when attempting to modify cells that lack abundant viral receptors or in vivo when modifying cells locally, for example in vascular or


Table 3 Transfection efficiency Transfection conditions

Microscopy % of positive cells (s.d.)a,b

1. 2. 3. 4. 5. 6. 7. 8.

3.0% 15.8% 0% 9.1% ND 1.1% 19.0% ND

Lipofectamine StrpL/Ad PEI 2000 PEI 2000/Ad5 PEI 2000/Ad3 PEI 25,000 PEI 25,000/Ad5 PEI 25,000/Ad3

(1.0) (4.2) (2.7) (0.8) (7)

FACS % of positive cells (s.d.)c

ND ND 0% 26.4 (2.6) 27.2 (1.7) 1.9 (0.67) 40.0 (2.5) 39.7 (0.93)

Apoptotic cells % of GFP-positive (s.d.)d ND ND 0% 6.1 (0.92) 1.9 (0.17) 4.3 (0.50) 2.0 (0.23) 2.3 (0.21)

a

Transfections were performed as described in the legend to Figure 5. b For each condition, seven fields of 130–200 cells were scored. The value represents the average with a standard deviation in parentheses. c FACS analysis of cells transfected with a GFP expression plasmid (pEGFP-C1) using various conditions. At 72 h after transfection cells were harvested and processed for FACS analysis as described in the Materials and methods. The values represent the percentage of the population expressing GFP (total GFPpositive cells divided by the total number of cells analyzed); each value represents the average of three separate transfections with a standard deviation indicated. Each measurement included 50 000 total events gated for size and single cells. d The percentage of GFP-positive cells that are apoptotic as determined by sub-2 N DNA content. The frequency of apoptotic cells in nontransfected cells was 0.5%. ND, not determined.

intratumoral applications. Of course in many in vivo applications, assembling the virus in a large, charged complex may result in reduced performance due to nonspecific binding or altered routing of the complex.

Discussion One can propose a fairly clear mechanism to account for the transfection function of the PEI–DNA–adenovirus complexes. PEI binds and condenses the plasmid DNA into positively charged bundles of 100–200 nm (in low salt) or 800–1000 nm (in 25–500 mm NaCl). The positive charge on the condensed DNA allows the bundle to bind adenovirus, possibly through interactions with negative charged domains on the hexon. A correlation between the charge of hexon possessed by various adenovirus serotypes and the function of that serotype to form productive transfection complexes in 500 mm NaCl supports this idea. As a consequence, adenovirus serotypes with acidic motifs on the surface of their hexons (eg Ad 2, 5, 3, 7) work best in this application. Upon exposure to target cells, the adenovirus can function as a ligand binding to their natural receptors. In this regard, Ad5 works better in a cell line of lung epithelial origin (A549) than in fibroblasts or erythroleukemia cells (K562) while Ad3 performs better than Ad5 in fibroblasts and K562 cells. The inclusion of an additional ligand (transferrin– polylysine) can improve the delivery with some cell types or viruses (eg with the avian adenovirus CELO in human cells). It is also possible that positive charge from the PEI– DNA can also help in cell binding. Clearly under some conditions PEI–DNA complexes alone can deliver DNA to cells and these properties may also be involved in the function of the PEI–DNA–adenovirus transfection complexes. We presume that these complexes follow some form of the natural adenovirus entry pathway, and that the virus may facilitate the delivery of the PEI–DNA complex to the nucleus of the target cell. Finally, within the nucleus the more moderate PEI to DNA interactions (relative polylysine to DNA interactions) might facilitate dissociation of the polycation and assembly of the DNA into transcriptionally active material. There are, of course limitations to this system. The complex does not withstand severe dilution, and gene delivery is somewhat sensitive to serum during transfection. On a per virus basis, the PEI–DNA–adenovirus complexes are comparable to the StrpL–TfpL–DNA–adenovirus system.43 There are however several advantages to this system. It is very simple to establish, different serotypes of adenovirus can be readily incorporated, and it generates high levels of transient gene expression.

Materials and methods Figure 6 Use of PEI to enhance recombinant adenovirus entry. Ad/PEI: standard PEI–DNA complexes were formed with 6 mg of the empty vector pSP65 and 30 ml of 10 mm PEI (2000 MW). After complex formation for 20 min, aliquots of the luciferase-encoding adenovirus AdLuc1 were added for an additional 20 min. Aliquots of the complexes (50 ml) were supplied to primary human fibroblasts in 500 ml of DMEM containing no serum and the subsequent analysis of luciferase activity (at 48 h after transfection) was identical to Figure 1. Total luciferase activity is plotted (average of three samples, the standard deviation indicated) as a function of the virus particles per cell. Ad: aliquots of AdLuc1 were diluted in HBS to generate the samples containing identical virus concentrations as in the Ad–PEI samples, but lacking PEI and plasmid DNA. Treatment of cells and analysis of luciferase activity was identical.

PEI solution All transfections were performed with PEI MW 2000 (Aldrich, (Milwaukee, WI, USA) cat No. 40 870–0; 50% solution) or PEI MW 25 000 (Aldrich cat No. 40 872–7). As described by Boussif et al,44 a 10 mm stock solution was made by mixing 9 mg of PEI in 10 ml of water, adjusting the pH to 7 with HCl, and passing the solution through a 0.2-micron filter. The filtered stock solution was stored at 4°C. Before use, the 10 mm PEI stock solution was vortexed vigorously.

779


780

DNA Plasmid DNA was prepared by ion exchange chromatography and treated to remove LPS as previously described.59 The luciferase marker gene plasmid used bears a cytomegalovirus immediate–early enhancer/ promoter-driven luciferase cDNA (pCLuc).60 The green fluorescent protein (GFP) plasmid used was the enhanced GFP plasmid pEGFP-C1 from Clontech (Palo Alto, CA, USA). Adenovirus The Ad5 virus used here was the E4-defective dl101461 grown in W162 cells.62 For several experiments (where indicated) adenovirus type 3 and type 40 were used. These viruses were grown in 293 cells (ATCC CRL1573). 63 Virus growth, purification (banded twice in CsCl) and inactivation with psoralen was as previously described.37,64 The CELO virus was grown and purified as previously described.65 For the experiments in this manuscript, biotinylated virus43 was employed, subsequent experiments demonstrated that the biotin modification of the Ad5 or CELO virion had no effect on the performance of viruses in PEI–DNA complexes. The luciferase-encoding, E1-defective Ad5 virus AdLuc1 was prepared using the two plasmid system described by Bett et al66 and will be described elsewhere. AdLuc1 was grown on 293 cells and purified by double banding in CsCl. Cells Primary human skin fibroblasts were used between passage 5 and 15 and grown in DMEM plus 10% FCS (DMEM, 2 mm glutamine, 100 IU penicillin, 100 mg/ml streptomycin and 10% (v/v) fetal calf serum). A549 (ATCC CCL-185) were grown in DMEM plus 10% FCS, K562 (ATCC CCL-243) were grown in RPMI 1640 plus 10% FCS (RPMI with 2 mm glutamine, 100 IU penicillin, 100 mg/ml streptomycin and 10% (v/v) fetal calf serum). All sera were heat inactivated at 56°C for 60 min. Standard PEI–DNA–adenovirus transfection conditions Unless indicated otherwise in the particular experiment, the standard method of forming the PEI complex was as follows. DNA (6 mg) was diluted in 250 ml of 20 mm HEPES, pH 7.4; PEI (24–30 ml of 10 mm PEI MW 2000, pH 7) was diluted in 250 ml of 20 mm HEPES, pH 7.4. The diluted PEI was then added dropwise to the DNA solution, with gentle agitation after the addition of each drop. The sample was allowed to stand at room temperature for 20 min and then an aliquot of psoralen-inactivated adenovirus (0.15 to 5 ml, 2.5 × 109 particles per microliter in HBS (150 mm NaCl, 20 mm HEPES pH 7.4) containing 40% glycerol) was added to the PEI–DNA. After an additional 20 min at room temperature, 50 ml aliquots of the complex were added to cell culture medium without serum (250 ml, covering 20 000–50 000 cells in a 1.5 cm diameter well, eg the well of a 24-well dish). The transfection medium was replaced with fresh, serum-containing medium after 4 h. Streptavidin-polylysine–transferrin-polylysine system Biotinylated, psoralen-inactivated adenovirus was diluted in 150 ml of HBS and mixed with 1 mg of streptavidin-polylysine37,43 in 150 ml of HBS. After a 30 min incubation at room temperature, this was mixed with 100 ml

of HBS containing 6 mg of DNA. After a further 30-min incubation, 4 mg of transferrin-polylysine37,67 in 100 ml of HBS was added. After a final 30-min incubation, aliquots of the mixture were added to the cell cultures as indicated.

Determination of luciferase activity Luciferase assays were performed as previously described.68 Briefly, cells were washed once with PBS and then lysed with 150 ml of 0.25 m Tris, pH 7.4, 0.1% Triton X-100 (per 2.4-cm well). The lysate was transferred to a round-bottomed 96-well plate, centrifuged for 5 min at 1500 r.p.m. (Beckman rotor number GH 3.7, approximately 450 g) and the luciferase activity in the supernatant was determined. Extracts were standardized for protein content. All values presented are the average of three transfection samples with the standard deviation indicated. Flow cytometric analysis to determine GFP expression At 72 h after transfection, adherent cells were trypsinized, washed with 2 × 4 ml PBS and fixed in 2% paraformaldehyde, 100 mm NaCl, 300 mm sucrose, 3 mm MgCl2 1 mm EGTA, 10 mm PIPES pH 6.8 at room temperature for 30 min, washed with 2 × 4 ml of PBS and post-fixed in ice-cold 70% ethanol for 10 h. Subsequently, cells were washed with 2 × 4 ml PBS, treated with RNase A (50 mg/ml) in PBS for 30 min, washed with 2 × 4 ml PBS and stained with propidium iodide (50 mg/ml) for 30 min before FACS analysis. Flow cytometric analysis of the cells was carried out on a Becton Dickinson (Mountain View, CA, USA) FACScan equipped with a doublet discrimination module.

Acknowledgements We thank Ernst Wagner for streptavidin–polylysine and transferrin–polylysine conjugates.

References 1 Chattoraj DK, Gosule LC, Schellman JA. DNA condensation with polyamines. J Mol Biol 1978; 121: 327–337. 2 Arscott PG, Li AZ, Bloomfield VA. Condensation of DNA by trivalent cations. 1. Effects of DNA length and topology on the size and shape of condensed particles. Biopolymers 1990; 30: 619–630. 3 Chiou HC et al. Enhanced resistance to nuclease degradation of nucleic acids complexed to asialoglycoprotein-polylysine carriers. Nucleic Acids Res 1994; 24: 5439–5446. 4 Wu GY, Wu CH. Receptor-mediated in vitro gene transformation by a soluble DNA carrier system. J Biol Chem 1987; 262: 4429–4432. 5 Wu C, Wilson J, Wu G. Targeting genes: delivery and persistent expression of a foreign gene driven by mammalian regulatory elements in vivo. J Biol Chem 1989; 264: 16985–16987. 6 Wagner E et al. Transferrin–polycation conjugates as carriers for DNA uptake into cells. Proc Natl Acad Sci USA 1990; 87: 3410– 3414. 7 Cotten M et al. Transferrin–polycation-mediated introduction of DNA into human leukemic cells: stimulation by agents that affect the survival of transfected DNA or modulate transferrin receptor levels. Proc Natl Acad Sci USA 1990; 87: 4033–4037. 8 Zenke M et al. Receptor-mediated endocytosis of transferrin polycation conjugates: an efficient way to introduce DNA into hematopoietic cells. Proc Natl Acad Sci USA 1990; 87: 3655–3659.


9 Trubetskoy VS, Torchilin VP, Kennel S, Huang L. Cationic liposomes enhance targeted delivery and expression of exogenous DNA mediated by N-terminal modified poly(L-lysine)-antibody conjugate in mouse lung endothelial cells. Biochim Biophys Acta 1992; 1131: 311–313. 10 Trubetskoy VS, Torchilin VP, Kennel S, Huang L. Use of N-terminal modified poly(L-lysine)-antibody conjugates as a carrier for targeted gene delivery in mouse lung endothelial cells. Bioconj Chem 1992; 3: 323–327. 11 Midoux P et al. Specific gene transfer mediated by lactosylated poly-L-lysine into hepatoma cells. Nucleic Acids Res 1993; 21: 871–878. 12 Baatz JE et al. Utilization of modified surfactant-associated protein B for delivery of DNA to airway cells in culture. Proc Natl Acad Sci USA 1994; 91: 2547–2551. 13 Perales JC et al. Gene transfer in vivo: sustained expression and regulation of genes introduced into the liver by receptor-targeted uptake. Proc Natl Acad Sci USA 1994; 91: 4086–4090. 14 Ferkol T, Kaetzel CS, Davis PB. Gene transfer into respiratory epithelial cells by targeting the polymeric immunoglobulin receptor. J Clin Invest 1993; 92: 2394–2400. 15 Ferkol T et al. Gene transfer into the airway epithelium of animals by targeting the polymeric immunoglobulin receptor. J Clin Invest 1995; 95: 493–502. 16 Gao X, Huang L. Potentiation of cationic liposome-mediated gene delivery by polycations. Biochemistry 1996; 35: 1027–1036. 17 Legendre JY, Szoka FC. Cyclic amphipathic peptide–DNA complexes mediate high-efficiency transfection of adherent mammalian cells. Proc Natl Acad Sci USA 1993; 90: 893–897. 18 Legendre JY, Supersaxo A. Short-chain phospholipids enhance amphipathic peptide-mediated gene transfer. Biochem Biophys Res Commun 1995; 217: 179–185. 19 Hara T et al. Effects of fusogenic and DNA-binding amphiphilic compounds on the receptor-mediated gene transfer into hepatic cells by asialofetuin-labeled liposomes. Biochim Biophys Acta, Gene Struct Expr 1996; 1278: 51–58. 20 Haensler J, Szoka FC. Polyamidoamine cascade polymers mediate efficient transfection of cells in culture. Bioconj Chem 1993; 4: 372–379. 21 Bielinska A et al. Regulation of in vitro gene expression using antisense oligonucleotides or antisense expression plasmids transfected using starburst PAMAM dendrimers. Nucleic Acids Res 1996; 24: 2176–2182. 22 Kukowska-Latallo JF et al. Efficient transfer of genetic material into mammalian cells using Starburst polyamidoamine dendrimers. Proc Natl Acad Sci USA 1996; 93: 4897–4902. 23 Huckett B, Ariatti M, Hawtrey AO. Evidence for targeted gene transfer by receptor-mediated endocytosis: stable expression following insulin-directed entry of neo into HepG2 cells. Biochem Pharmacol 1990; 40: 253–263. 24 Bo¨ttger M et al. Condensation of vector DNA by the chromosomal protein HMG1 results in efficient transfection. Biochim Biophys Acta 1988; 950: 221–228. 25 Kaneda Y, Iwai K, Uchida T. Increased expression of DNA cointroduced with nuclear protein in adult rat liver. Science 1989; 243: 375–378. 26 Tomita N et al. Direct in vivo gene introduction into rat kidney. Biochem Biophys Res Commun 1992; 186: 129–134. 27 Wagner E, Cotten M, Foisner R, Birnstiel ML. Transferrin–polycation–DNA complexes: the effect of polycations on the structure of the complex and DNA delivery to cells. Proc Natl Acad Sci USA 1991; 88: 4255–4259. 28 Curiel DT, Agarwal S, Wagner E, Cotten M. Adenovirus enhancement of transferrin-polylysine mediated gene delivery. Proc Natl Acad Sci USA 1991; 88: 8850–8854. 29 Cotten M et al. High-efficiency receptor-mediated delivery of small and large (48 kb) gene constructs using the endosomedisruption activity of defective or chemically inactivated adenovirus particles. Proc Natl Acad Sci USA 1992; 89: 6094–6098. 30 Cristiano RJ, Smith LC, Woo SLC. Hepatic gene therapy: adenovirus enhancement of receptor-mediated gene delivery and expression in primary hepatocytes. Proc Natl Acad Sci USA 1993; 90: 2122–2126.

31 Yoshimura K, Rosenfeld M, Seth P, Crystal R. Adenovirusmediated augmentation of cell transfection with unmodified plasmid vectors. J Biol Chem 1993; 268: 2300–2303. 32 Ferna´ndez-Puentes C, Carrasco L. Viral infection permeabilizes mammalian cells to protein toxins. Cell 1980; 20: 769–775. 33 Fitzgerald D, Padmanabhan R, Pastan I, Willingham M. Adenovirus-induced release of epidermal growth factor and Pseudomonas toxin into the cytosol of KB cells during receptormediated endocytosis. Cell 1983; 32: 607–617. 34 Carrasco L. Entry of animal viruses and macromolecules into cells. FEBS Lett 1994; 350: 151–154. 35 Carrasco L. Modification of membrane permeability by animal viruses. Adv Virus Res 1995; 45: 61–112. 36 Zatloukal K et al. Transferrinfection: a highly efficient way to express gene constructs in eukaryotic cells. Ann NY Acad Sci 1992; 660: 136–153. 37 Cotten M et al. Adenovirus polylysine DNA conjugates. Curr Protocols Hum Genet 1997; 12.3.1–12.3.33. 38 Cristiano RJ et al. Hepatic gene therapy: efficient gene delivery and expression in primary hepatocytes utilizing a conjugated adenovirus–DNA complex. Proc Natl Acad Sci USA 1993; 90: 11548–11552. 39 Fisher KJ, Wilson JM. Biochemical and functional analysis of an adenovirus-based ligand complex for gene transfer. Biochem J 1994; 299: 49–58. 40 Fisher KJ, Kelley WM, Burda JF, Wilson JM. A novel adenovirus-adeno-associated virus hybrid vector that displays efficient rescue and delivery of the AAV genome. Hum Gene Ther 1996; 7: 2079–2087. 41 Curiel D et al. High efficiency gene transfer mediated by adenovirus coupled to DNA polylysine complexes via an antibody bridge. Hum Gene Ther 1992; 3: 147–154. 42 Scaria A, Curiel DT, Kay MA. Complementation of a human adenovirus early region 4 deletion mutant in 293 cells using adenovirus–polylysine–DNA complexes. Gene Therapy 1995; 2: 295–298. 43 Wagner E et al. Coupling of adenovirus to transferrin–polylysine–DNA complexes greatly enhances receptor-mediated gene delivery and expression of transfected cells. Proc Natl Acad Sci USA 1992; 89: 6099–6103. 44 Boussif O et al. A versatile vector for gene and oligonucleotide transfer into cells in culture and in vivo: polyethylenimine. Proc Natl Acad Sci USA 1995; 92: 7297–7301. 45 Abdallah B et al. A powerful nonviral vector for in vivo gene transfer into the adult mammalian brain: polyethylenimine. Hum Gene Ther 1996; 7: 1947–1954. 46 Baker A, Cotten M. Delivery of bacterial artificial chromosomes into mammalian cells with psoralen-inactivated adenovirus carrier. Nucleic Acids Res 1997; 25: 1950–1956. 47 Roberts MM, White JL, Gru¨tter MG, Burnett RM. Three-dimensional structure of the adenovirus major coat protein hexon. Science 1986; 232: 1148–1151. 48 Stewart PL, Burnett RM, Cyrklaff M, Fuller SD. Image reconstruction reveals the complex molecular organization of adenovirus. Cell 1991; 67: 145–154. 49 Mautner V, Willcox HNA. Adenovirus antigens, a model system in mice for subunit vaccination. J Gen Virol 1974; 25: 325–336. 50 Toogood CIA, Crompton J, Hay RT. Antipeptide antisera define neutralizing epitopes on the adenovirus hexon. J Gen Virol 1992; 73: 1429–1435. 51 Crompton J, Toogood CIA, Wallis N, Hay RT. Expression of a foreign epitope on the surface of the adenovirus hexon. J Gen Virol 1994; 75: 133–139. 52 Mei Y-F, Wadell G. Molecular determinants of adenovirus tropism. In: Doerfler W, Bo¨hm P (eds). The Molecular Repertoire of Adenoviruses. Current Topics in Microbiology and Immunology 199/III. Springer-Verlag: Berlin, 1995, pp 213–228. 53 Chiocca S et al. The complete DNA sequence and genomic organization of the avian adenovirus CELO. J Virol 1996; 70: 2939–2949. 54 Silver L, Anderson C. Interaction of human adenovirus serotype 2 with human lymphoid cells. Virology 1988; 165: 377–387.

781


782

55 Horvath J, Weber J. Nonpermissivity of human peripheral blood lymphocytes to adenovirus type 2 infection. J Virol 1988; 62: 341–345. 56 Wickham TJ, Roelvink PW, Brough DE, Kovesdi I. Adenovirus targeted to heparan-containing receptors increases its gene delivery efficiency to multiple cell types. Nat Biotechnol 1996; 14: 1570–1573. 57 Douglas JT et al. Targeted gene delivery by tropism-modified adenoviral vectors. Nat Biotechnol 1996; 14: 1574–1578. 58 March KL, Madison JE, Trapnell BC. Pharmokinetics of adenoviral vector-mediated gene delivery to vascular smooth muscle cells: modulation by Poloxamer 407 and implication for cardiovascular gene therapy. Hum Gene Ther 1995; 6: 41–53. 59 Cotten M et al. Lipopolysaccharide is a frequent contaminant of plasmid DNA preparations and can be toxic to primary cells in the presence of adenovirus. Gene Therapy 1994; 1: 239–246. 60 Plank C et al. Gene transfer into hepatocytes using asialoglycoprotein receptor mediated endocytosis of DNA complexed with an artificial tetra-antennary galactose ligand. Bioconj Chem 1992; 3: 533–539. 61 Bridge E, Ketner G. Redundant control of adenovirus late gene expression by early region 4. J Virol 1989; 63: 631–638. 62 Weinberg DH, Ketner G. A cell line that supports the growth of a defective early region 4 deletion mutant of human adenovirus type 2. Proc Natl Acad Sci USA 1983; 80: 5383–5386. 63 Graham F, Smiley J, Russel W, Nairn R. Characteristics of a

64

65

66

67

68 69

70

human cell line transformed by DNA from human adenovirus type 5. J Gen Virol 1977; 36: 59–94. Cotten M et al. Psoralen treatment of adenovirus particles eliminates virus replication and transcription while maintaining the endosomolytic activity of the virus capsid. Virology 1994; 205: 254–261. Cotten M, Wagner E, Zatloukal K, Birnstiel ML. Chicken adenovirus (CELO virus) particles augment receptor-mediated DNA delivery to mammalian cells and yield exceptional levels of stable transformants. J Virol 1993; 67: 3777–3785. Bett AJ, Haddara W, Prevec L, Graham FL. An efficient and flexible system for construction of adenovirus vectors with insertions or deletions in early region 1 and 3. Proc Natl Acad Sci USA 1994; 91: 8802–8806. Wagner E et al. DNA-binding transferrin conjugates as functional gene delivery agents: synthesis by linkage of polylysine or ethidium homodimer to the transferrin carbohydrate moiety. Bioconj Chem 1991; 2: 226–231. Cotten M, Wagner E, Birnstiel ML. Receptor-mediated transport of DNA into eukaryotic cells. Meth Enzymol 1993; 217: 618–644. Chiocca S, Baker A, Cotten M. Identification of a novel antiapoptotic gene in the avian adenovirus CELO. J Virol 1997; 71: 3168–3177. Arcasoy SM et al. Polycations increase the efficiency of adenovirus-mediated gene transfer to epithelial and endothelial cells in vitro. Gene Therapy 1997; 4: 32–38.