Complexes of adenovirus with polycationic polymers and cationic lipids increase the efficiency of gene transfer in vitro and in vivo.

Improving the efficiency of gene transfer remains an important goal in developing new treatments for cystic fibrosis and other diseases. Adenovirus vectors and nonviral vectors each have specific advantages, but they also have limitations. Adenovirus vectors efficiently escape from the endosome and enter the nucleus, but the virus shows limited binding to airway epithelia. Nonviral cationic vectors bind efficiently to the negatively charged cell surface, but they do not catalyze subsequent steps in gene transfer. To take advantage of the unique features of the two different vector systems, we noncovalently complexed cationic molecules with recombinant adenovirus encoding a transgene. Complexes of cationic polymers and cationic lipids with adenovirus increased adenovirus uptake and transgene expression in cells that were inefficiently infected by adenovirus alone. Infection by both complexes was independent of adenovirus fiber and its receptor and occurred via a different cellular pathway than adenovirus alone. Complexes of cationic molecules and adenovirus also enhanced gene transfer to differentiated human airway epithelia in vitro and to the nasal epithelium of cystic fibrosis mice in vivo These data show that complexes of adenovirus and cationic molecules increase the efficiency of gene transfer, which may enhance the development of gene therapy.

Improving the efficiency of gene transfer remains an important goal in developing new treatments for cystic fibrosis and other diseases. Adenovirus vectors and nonviral vectors each have specific advantages, but they also have limitations. Adenovirus vectors efficiently escape from the endosome and enter the nucleus, but the virus shows limited binding to airway epithelia. Nonviral cationic vectors bind efficiently to the negatively charged cell surface, but they do not catalyze subsequent steps in gene transfer. To take advantage of the unique features of the two different vector systems, we noncovalently complexed cationic molecules with recombinant adenovirus encoding a transgene. Complexes of cationic polymers and cationic lipids with adenovirus increased adenovirus uptake and transgene expression in cells that were inefficiently infected by adenovirus alone. Infection by both complexes was independent of adenovirus fiber and its receptor and occurred via a different cellular pathway than adenovirus alone. Complexes of cationic molecules and adenovirus also enhanced gene transfer to differentiated human airway epithelia in vitro and to the nasal epithelium of cystic fibrosis mice in vivo. These data show that complexes of adenovirus and cationic molecules increase the efficiency of gene transfer, which may enhance the development of gene therapy.
Transfer of the cystic fibrosis transmembrane conductance regulator (CFTR) 1 cDNA to airway epithelia of patients with cystic fibrosis (CF) (1) could provide an important new treatment for this genetic disease. Although previous studies have demonstrated the feasibility of gene transfer to CF airway epithelia, such studies suggest that an increase in efficiency is desirable. Enhanced efficiency would allow a vector to correct the electrolyte transport abnormalities that characterize the disease and minimize toxicity by reducing the amount of vector administered. Efficiency and safety will in turn determine the therapeutic index and thus the ultimate utility of gene transfer as a potential treatment.
Adenovirus-mediated gene transfer to airway epithelia is suboptimal because binding to the apical surface of the epithelium is limited, perhaps because the apical surface does not express the fiber receptor that mediates viral attachment (7,8). 2 Limited infection in vitro and in vivo could be partially overcome when the contact time between the virus and the apical surface was increased, even with a relatively low multiplicity of infection (m.o.i.) (8). These findings suggest that if adenovirus binding or entry into the cell could be increased, then adenovirus-dependent processes subsequent to binding and entry would remain intact and would facilitate gene transfer. These functions, which include entry of virus into the cell, release of DNA from the vector, entry into the nucleus, and transcription, are influenced by specific adenovirus proteins (11).
Nonviral vector-mediated gene transfer to mature human airway epithelia (6) could also be improved. Cationic molecules, including cationic lipids complexed with DNA, bind to the cell surface, and they are often taken up into the cell. However, release of DNA from the endosome, entry into the nucleus, release of DNA from the cationic molecules, and transcription of the DNA are not specifically enhanced by the vectors and thus remain as partial barriers that limit the efficiency of transgene expression (9).
Previous reports have described gene transfer systems that combine viral and nonviral components (12)(13)(14)(15)(16)(17)(18). In most cases, adenovirus has been incorporated into gene delivery systems to take advantage of its endosomolytic properties. These studies have involved either covalent attachment of the adenovirus to a gene delivery complex or co-internalization of unbound adenovirus with cationic lipid-DNA complexes. In these formulations, the transferred gene is contained in plasmid DNA that is exogenous to the adenovirus. However, such formulations do not take advantage of adenovirus-dependent functions other than endosome disruption, and as a result, large amounts of adenovirus are required, and the increase in gene transfer has often been modest.
We reasoned that the efficiency of gene transfer would be improved if we could combine a nonviral vector system, which could mediate cell attachment, with an adenovirus vector system, which could mediate processes in gene transfer subsequent to binding. We thought that such a hybrid system might take advantage of the unique features associated with the two individual vector systems. We initially attempted to covalently link recombinant adenovirus to various ligands in order to increase vector binding and gene transfer. However, such attempts often reduced gene expression from the adenovirus vector, probably because chemical modification of viral proteins interfered with their functions in effecting gene transfer and expression (19). 2 In the work described here, we tested the hypothesis that efficiency could be improved by using a noncovalent complex consisting of a cationic component and a recombinant adenovirus that contained the cDNA to be expressed. We postulated that a cationic component would charge-associate with adenovirus particles, which carry a net negative surface charge, and would facilitate attachment to the negatively charged cell membrane. We also postulated that if we could increase binding to the cell membrane, more adenovirus would be internalized, enhancing gene transfer and expression. To test this hypothesis and to examine the process, we used several cell systems in which the efficiency of adenovirus-mediated gene transfer varies. We then tested the gene transfer method on normal and CF airway epithelia in vitro and in vivo.
Preparation of Cationic Molecule-Adenovirus Complexes-Recombinant adenovirus was prepared by the University of Iowa Gene Transfer Vector Core at titers of ϳ1 ϫ 10 10 IU/ml. Complexes between cationic components and adenovirus particles were formed by prediluting the cationic component and the adenovirus components in Eagle's MEM in 12 ϫ 75-mm polystyrene tubes (Fisher). Ratios of cationic molecules to adenovirus particles and volumes of the dilutions are described in the figure legends. The cationic component dilution was added to the viral particle dilution, mixed by inversion or gentle pipette tip aspiration, and allowed to incubate for 15-30 min at room temperature before application to cells or tissue. We describe cationic molecule-adenovirus complexes based on the calculated average number of cationic molecules/adenovirus particle. The ratio of adenovirus particles to infectious units varied from 50 to 150.
Infection of Cells in Culture-Cultured cells were infected 18 -24 h after seeding when the cells were ϳ70% confluent unless otherwise noted. Airway epithelial cells were allowed to mature in culture for at least 10 -14 days before use so that they developed a ciliated apical surface that resembles the in vivo airway surface (8,23). The medium was replaced with 500 l of Eagle's MEM containing 1.1 ϫ 10 9 particles of adenovirus complexed with cationic component at the indicated ratios of cationic molecules to particles. The cells were incubated for 15 min to 6 h (times indicated in the figure legends) in a 5% CO 2 humidified environment at 37°C, the infection solution was removed, and  35 S-labeled Ad5RSVLacZ particles/cell or 10,000 35 S-labeled Ad5RSVLacZ particles/cell complexed with 250 PLL molecules/particle. The optimal ratios of PLL molecules to adenovirus particles were determined for each cell type as shown below in Fig. 5. At the end of the infection period, half the cultures were washed and released from the dishes, and the lysates were counted for cell-associated adenovirus. The rest of the cultures received fresh medium, and ␤-galactosidase expression was measured 24 h later (n ϭ 6). Asterisks indicate p Ͻ 0.05 compared with adenovirus alone. L.U., light units. fresh serum-containing medium was added. Cells were then incubated for an additional 24 h unless otherwise specified. In experiments where uptake of 35 S-labeled adenovirus was assessed, cells were harvested for the measurement of cell-associated radiolabeled virus at the end of the infection period, while paired groups of cells were incubated for an additional 20 -24 h (unless otherwise noted) and harvested for assessment of gene expression. Human airway epithelial cells were infected as described in the figure legends.
Gene Transfer to the Nasal Epithelium-Ad2/CFTR-8 was applied to the nasal epithelium of unanesthetized ⌬F/⌬F mice (25) as a 5-l drop containing 5 ϫ 10 7 IU/nostril adenovirus alone or PLL⅐Ad. The transepithelial electric potential difference (V t ) across the nasal epithelium was measured using techniques similar to those we previously described (25). During measurement of V t , the nasal mucosa was perfused at a rate of 50 l/min with a Ringer's solution containing 135 mM NaCl, 2.4 mM KH 2 PO 2 , 0.6 mM K 2 HPO 4 , 1. Assays-␤-Galactosidase activity was assayed using a Galacto-Lite kit (Tropix Inc., Bedford, MA) and a luminometer (Monolight 2010, Analytical Luminescence Laboratory, San Diego, CA). Cells were removed from dishes or millicell filters by incubation with 120 l of lysis buffer (100 mM potassium phosphate, pH 7.8, 0.2% Triton X-100, and 1 mM dithiothreitol) for 15 min, followed by scraping. A 4-l aliquot from each 24-well plate or millicell was used for each Galacto-Lite assay. Protein was measured using Bio-Rad protein assay reagent. Data for ␤-galactosidase activity represent total values from all cells in one well or from one millicell. All conditions were performed in triplicate on at least two different occasions. For X-gal staining, cells were washed 24 h after infection, fixed with 1.8% formaldehyde and 2% glutaraldehyde, and then incubated for 16 h in X-gal solution as described previously. Blue staining of nuclei was evaluated by light microscopy. For antihexon staining, airway epithelia cultured on permeable filter supports were studied 30 h after infection. They were fixed with acetone/methanol and stained with a polyclonal fluorescein isothiocyanate-labeled anti-hexon antibody (Chemicon International, Inc., Temecula, CA). Hexon-positive cells were counted using low magnification fluorescence

FIG. 2. Cy3-labeled PLL binding to COS-1 (A), HeLa (B), NIH-3T3 (C), and 9L gliosarcoma (D) cells.
Cells were seeded onto 8-well slides 18 -24 h before exposure to Cy3-labeled PLL. The medium was removed and replaced with 200 l of Eagle's MEM containing 5 ng/l fluorescently labeled PLL. After a 10-min incubation period at 37°C, the cells were washed twice with Eagle's MEM and examined by confocal microscopy. Scale bar ϭ 25 m.
photomicrographs of the monolayers. This method allows for detection of infected cells by staining for the most abundant adenovirus protein, hexon, conferring better sensitivity than with reporter genes such as ␤-galactosidase.
To evaluate cell-associated 35 S-labeled adenovirus, cells were harvested at the end of the infection period. The dishes or millicells were washed four times with phosphate-buffered saline, pH 7.4; then lysis buffer was applied, the dishes were scraped, and the cell lysate was counted in a RACK BETA Model 1209 liquid scintillation counter (LKB Wallac, Gaithersburg, MD).
To evaluate transepithelial electrolyte transport by human airway epithelia, epithelia were mounted in modified Ussing chambers as described previously (26). Short-circuit current was measured under base-line conditions and after addition of amiloride (10 M) and cAMP agonists (10 M forskolin and 100 M isobutylmethylxanthine).
Fluorescence Microscopy-Cells were seeded on 8-well slides 18 -24 h before the experiment. The medium was removed and replaced with 200 l of Eagle's MEM supplemented with 5 ng/l Cy3-labeled PLL. After a 10-min incubation at 37°C, the solution was removed, and the cells were washed twice with Eagle's MEM and examined by a Bio-Rad MRC-600 confocal microscope for cell-associated fluorescence.
Transmission Electron Microscopy-Complexes of PLL and adenovirus were formed as follows. Adenovirus particles (9 ϫ 10 9 ) were prediluted in 50 l of Eagle's MEM and combined with 0, 4, 250, or 10,000 PLL molecules/particle (0, 3.3, 208, or 8320 ng, respectively) also in 50 l of Eagle's MEM. The PLL dilution was added to the virus dilution and mixed by gentle pipette tip aspiration. Complexes of cationic lipids with adenovirus particles were formed in a similar fashion. PLL-adenovirus and cationic lipid-adenovirus complexes were processed for transmission electron microscopy using a negative stain technique. Fifteen-l drops of freshly prepared samples were placed on glowdischarged collodion/carbon-coated 400-mesh copper grids for 3 min. The solution was wicked off with filter paper and replaced with 1% aqueous uranyl acetate for 30 s. After removal of this solution, grids were allowed to dry and imaged in a Hitachi H-7000 transmission electron microscope.
To evaluate the interaction of adenovirus complexes with cells, complexes of PLL or Lipofectamine with adenovirus were formed as follows. Adenovirus particles (1.5 ϫ 10 11 ) were prediluted in 150 l of Eagle's MEM and combined with 250 PLL molecules/particle (3460 ng of PLL) or 68,150 DOSPA molecules/particle (equivalent to 33 g of Lipofectamine), also prediluted in 150 l of Eagle's MEM. The cationic component was added to the virus dilution and mixed by gentle pipette tip aspiration. To follow the cellular entry of cationic molecule-adenovirus complexes, 9L gliosarcoma cells were infected for 15 min at 37°C, and then the cells were washed and fixed in 2.5% glutaraldehyde and processed using standard electron microscopic procedures. Briefly, the samples were post-fixed in 1% osmium tetroxide followed by 2.5% aqueous uranyl acetate and then dehydrated in a graded series of ethanol washes. Thin sections (70 nm) of the Eponate 12-embedded specimen were placed on 135-mesh hexagonal copper grids and poststained with uranyl acetate and Reynold's lead citrate.

RESULTS
Complexing Adenovirus with PLL Enhances Gene Transfer-As a starting point, we studied cells that varied in the efficiency of infection by adenovirus. Fig. 1A shows that application of Ad5RSVLacZ to COS-1 and HeLa cells generated high levels of ␤-galactosidase expression. In contrast, an identical experiment in NIH-3T3 and 9L gliosarcoma cells yielded much lower expression levels (Fig. 1A). Decreased expression was not likely due to cell type-specific effects on the Rous sarcoma virus promoter because we found similar effects when we used Ad2/ ␤Gal-2, which contains the cytomegalovirus promoter driving ␤-galactosidase expression (data not shown).
We tested the hypothesis that decreased expression in some cells was due to decreased binding of adenovirus. We applied 35 S-labeled virus to cells; after 90 min, we removed the labeled vector, rinsed the surface four times with phosphate-buffered saline, and measured cell-associated radioactivity. Fig. 1 shows that there was a correlation between the amount of adenovirus bound to the different cell types (Fig. 1B) and the level of expression (Fig. 1A); there was substantial binding to COS-1 and HeLa cells, which showed high levels of transgene expres-
Because both the cell and the adenovirus have a net negative surface charge, we postulated that a cationic molecule might facilitate association of the vector with the cell. To learn whether a cationic molecule would bind to the cell surface, we applied Cy3-labeled PLL to different cell types. Fig. 2 shows Cy3-labeled PLL bound to the surface of the COS-1, HeLa, NIH-3T3, and 9L gliosarcoma cell lines. The surface of nearly every cell was fluorescently labeled in all cell types.
Therefore, we prepared PLL⅐Ad complexes and measured vector binding and transgene expression in the different cell lines. When we applied the PLL⅐Ad complex to NIH-3T3 and 9L gliosarcoma cells, both adenovirus binding (Fig. 1B) and transgene expression (Fig. 1A) increased to the same range as we had observed with adenovirus alone in COS-1 and HeLa cells. In contrast, complexing PLL with adenovirus had only small effects on binding and expression in COS-1 and HeLa cells. Fig.  3 shows photomicrographs of cells stained with X-gal after application of Ad2/␤Gal-2. PLL⅐Ad not only increased total ␤-galactosidase activity in NIH-3T3 and 9L gliosarcoma cells, but also increased the number of cells expressing the transgene.
Complexes of PLL⅐Ad also increased expression in primary cell cultures. In human umbilical vein endothelial cells, which are poorly infected by adenovirus alone, PLL⅐Ad increased expression 63 Ϯ 3-fold (n ϭ 3). In contrast, in primary cultures of rat hepatocytes, which are readily infected by adenovirus, PLL⅐Ad produced little enhancement (PLL⅐Ad was 1.1 Ϯ 0.1 the level of expression with adenovirus alone (n ϭ 3)).
Complexes Containing Adenovirus and Other Polycations Augment Gene Transfer-Encouraged by the results with PLL, we asked whether other cationic polymers would enhance gene transfer. We tested polyethyleneimine, DEAE-dextran, and histone (fraction V-S) and found that all facilitated gene transfer to poorly infected cell types (Fig. 4). As expected, with each polymer, the optimal cationic molecule/adenovirus particle ratio varied, producing data similar to those shown below in Fig.  5. However, not all cationic molecules were effective; spermine (a low molecular mass polyamine with a potential ϩ4 net charge) failed to enhance expression in 9L gliosarcoma cells despite testing of a range of molecule/adenovirus particle ratios (Fig. 4).
Cationic lipids complexed with adenovirus also augmented gene transfer to 9L gliosarcoma cells (Fig. 4). We obtained similar results in human umbilical vein endothelial cell cul-tures (data not shown). Although our goal was not to perform a detailed analysis of the relationship between cationic lipid structure and expression, it appeared that, in general, polyvalent lipids were more effective than monovalent lipids, e.g. compare DC-Chol (ϩ1 charge), GL-53 (ϩ2 charge), and GL-67 (ϩ3 charge). Most cationic lipid-DNA formulations include the neutral lipid DOPE; DOPE is thought to enhance gene transfer because of its fusogenic, endosome-disrupting properties. We found that inclusion of DOPE in the complex did not consistently improve expression as it usually does with cationic lipid-DNA complexes.
These data indicate that many different molecules can facilitate gene transfer to cells that are normally relatively resistant to adenovirus infection. The common feature is their cationic nature. Future studies will be required to determine more precisely the properties that generate optimal expression.
Evaluation of PLL⅐Ad Complexes-To demonstrate further the effect of PLL on adenovirus-mediated gene transfer, we show the effects of varying the ratio of PLL to adenovirus (Fig.  5). In NIH-3T3 and 9L gliosarcoma cells, as the ratio of PLL to Ad2/␤Gal-2 increased, the efficiency of gene transfer increased (Fig. 5). Maximum ␤-galactosidase activity was observed at ϳ250 PLL molecules/adenovirus particle. At high ratios of PLL to adenovirus particles, ␤-galactosidase expression tended to decrease. In contrast, increasing amounts of PLL did not in-  6. Effect of applying PLL to 9L gliosarcoma cells before application of Ad2/␤Gal-2 or complexes of PLL⅐Ad or GL-67⅐Ad. Cells were treated for 5 min under serum-free conditions with 5 ng/l PLL (solid bars) or buffer alone (hatched bars). PLL was then removed by rinsing, and cells were incubated for 1 h with 5000 particles/cell Ad2/␤Gal-2, PLL⅐Ad (250 PLL molecules/particle), or GL-67⅐Ad (2.2 ϫ 10 5 GL-67 molecules/particle). The virus-containing solution was then replaced with fresh medium, and the cells were assayed for ␤-galactosidase activity 24 h later (n ϭ at least three for each). Asterisks indicate p Ͻ 0.05 compared with control. L.U., light units. crease ␤-galactosidase expression by COS-1 cells; in fact, at a PLL molecule/particle ratio of ϳ25, we frequently noted a decrease in expression (Fig. 5). These results suggest the possibility that PLL may have interfered with the normal mechanism of adenovirus attachment and infection at intermediate ratios of PLL molecules to adenovirus particles, but then as the amount of PLL increased, PLL may have mediated adenovirus binding to the cells.
To learn whether it was necessary to form a complex between PLL and adenovirus, we first pretreated 9L gliosarcoma cells with PLL (5 ng/l) for 5 min, removed excess PLL by washing, and then added adenovirus. Fig. 6 shows that separate application of PLL followed by adenovirus was not nearly as effective as adding the preformed PLL⅐Ad complex to the cells. Fig.  6 also shows that applying PLL to the cells before application of PLL⅐Ad or GL-67⅐Ad complexes largely prevented the augmentation of gene transfer observed with these complexes. These results suggest that PLL competed with sites that bind both PLL⅐Ad and GL-67⅐Ad, thereby attenuating gene transfer.
We did several additional studies to optimize PLL⅐Ad-mediated gene transfer. 1) We found that the size of PLL influenced gene transfer (Fig. 7A). A size of 55.8 kDa (used throughout this study) produced the greatest augmentation of gene transfer when compared with PLL of other sizes, even though the calculated ratio of positive charges to adenovirus particles remained constant. 2) Many cationic lipid-DNA complexes have a tendency to aggregate and lose their effectiveness with prolonged incubation (27). We observed a similar property with PLL⅐Ad. With a 5-h delay (at room temperature) between preparation of the complex and application to cells, expression was 28 Ϯ 10% (n ϭ 6) of the value obtained when the delay interval was only 15 min. In contrast, adenovirus alone retained its infectivity; expression after a 5-h delay at room temperature was 104 Ϯ 7% (n ϭ 6) of the value after a 15-min delay (data not shown). 3) We found that it was necessary to prepare the PLL⅐Ad complex in the absence of serum. When serum was present during complex formation, expression in 9L gliosarcoma cells was reduced to 2.8 Ϯ 0.1% (n ϭ 3) of the value obtained after preparation under serum-free conditions. 4) Serum inhibits gene transfer by some cationic lipid-DNA complexes and by some nonviral vector preparations that contain PLL. To examine the effect of serum, we used COS-1 cells so that we could compare the effect of PLL⅐Ad with adenovirus alone. Fig. 7B shows that PLL⅐Ad-mediated gene transfer to COS-1 cells was decreased in the presence of serum, but the decrease was no greater than that observed with adenovirus alone. We obtained similar results with 9L gliosarcoma cells. These data indicate that after the complex was formed, it could be used in either the presence or absence of serum. 5) We previously reported that the concentration of adenovirus was an important variable in effecting gene transfer (26). Likewise, we found that as the concentration of PLL⅐Ad increased (produced by decreasing the applied volume with a constant m.o.i.), expression increased in 9L gliosarcoma cells (Fig. 7C). 6) The enhancement produced by complexing adenovirus with PLL was not dependent on the time after seeding of 9L gliosarcoma cells (Fig. 7D). In confluent 9L gliosarcoma cells, expression with PLL⅐Ad was 32-fold greater than with adenovirus alone, and with subconfluent cells in log-phase growth, expression was 29-fold greater.
We also examined the appearance of the PLL⅐Ad complex by transmission electron microscopy. Fig. 8 shows photomicrographs of adenovirus alone and PLL⅐Ad complexes at a suboptimal infection ratio (4 PLL molecules/particle), an optimal ratio (250 PLL molecules/particle), and another suboptimal ratio (10,000 PLL molecules/particle). As the number of PLL molecules/adenovirus particle increased, we saw that the adenovirus particles tended to clump together, although we found individual adenovirus particles even at the highest ratio. These observations suggest that PLL is linking the negative surface charge on the adenovirus particle to other particles and that excessive amounts of PLL produce aggregation, which could decrease the efficiency of gene transfer, as shown in Fig. 5.
Gene Transfer by PLL⅐Ad Complexes Does Not Depend on the Fiber Receptor-In some cell types, adenovirus infection is mediated by binding of the adenovirus fiber protein to an unidentified receptor on the cell surface (28). Consistent with this conclusion, we found that excess fiber inhibited gene transfer by adenovirus alone in primary hepatocytes, which are readily infected cells (Fig. 9A). However, fiber did not inhibit expression by PLL⅐Ad and only partially inhibited expression by GL-67⅐Ad.
We also examined the effect of neutralizing antibodies. Fig.  9B shows that neutralizing antibodies directed against fiber did not interfere with expression by PLL⅐Ad when we used optimal PLL molecule/particle ratios. However at a suboptimal ratio (25 PLL molecules/particle), anti-fiber antibody produced some inhibition of expression. These data, together with the adenovirus binding experiments, the effect of fiber, and the enhancement of gene transfer to poorly infected cells, suggest that PLL causes adenovirus to bind to and then infect cells through pathways other than the fiber receptor-mediated pathway.
In contrast to the effect of anti-fiber antibody, anti-hexon antibody inhibited infection by both adenovirus and PLL⅐Ad (Fig. 9B). Chloroquine, which inhibits adenovirus infection by raising the pH of the endosomes and by preventing release from the endosomes (29), also inhibited gene transfer by PLL⅐Ad. Treatment of 9L gliosarcoma cells with 100 M chloroquine for 60 min decreased ␤-galactosidase expression following infection with PLL⅐Ad by 59 Ϯ 10% compared with expression obtained in the absence of chloroquine (n ϭ 6). These results suggest that adenovirus-dependent steps subsequent to binding and uptake into the cell are required for infection.
Because the functional data suggested that the mechanism for binding and uptake of PLL⅐Ad was different from that of adenovirus alone, we looked for a structural correlate using transmission electron microscopy. We used 9L gliosarcoma cells because they are poorly infected by adenovirus due to poor binding of virus. We also used a very large number of particles/ cell in order to observe the vector-cell interaction. At high m.o.i. values, there were a substantial number of adenovirus particles in the cells, even in 9L gliosarcoma cells, which show limited binding of adenovirus (Fig. 10, A and B). The electron photomicrographs show that adenovirus was usually taken up into the cells as single particles or occasionally as a small (two or three) number of particles. In the cells treated with PLL⅐Ad (Fig. 10, C and D) or Lipofectamine-adenovirus (Fig. 10, E and  F), there were a greatly increased number of adenovirus particles in the cells, consistent with the data shown in Fig. 1. It is also apparent that with the PLL⅐Ad and cationic lipid-adenovirus complexes, multiple adenovirus particles were present in the endosomes and that the endosomes were larger than those containing adenovirus alone. In addition, with cationic lipidadenovirus complexes, we sometimes saw a lamellar appear- FIG. 9. Effect of fiber and anti-adenovirus antibodies on gene transfer by adenovirus alone or by PLL⅐Ad. A, fiber protein (70 g/ml) (hatched bars) or solution alone (solid bars) was applied to primary cultures of rat hepatocytes in Eagle's MEM, and 10 min later, Ad2/␤Gal-2 alone, PLL⅐Ad (250 PLL molecules/particle), or GL-67⅐Ad (2.2 ϫ 10 5 GL-67 molecules/particle) was added. Vector was removed 2 h later, and ␤-galactosidase activity was measured 20 -24 h later (n ϭ 4). B, anti-fiber antibody (1:100 dilution) (hatched bars), anti-hexon antibody (1:100 dilution) (shaded bars), or solution alone (solid bars) was mixed with Ad2/␤Gal-2 or PLL⅐Ad at a ratio of either 25 or 250 PLL molecules/particle. The solution was incubated for 10 min at 25°C and then applied to the cells. Vector was removed from the cells by rinsing 30 min later, and ␤-galactosidase activity was measured 20 -24 h later. Data are presented relative to expression in the absence of antibody (n ϭ 4). Asterisks indicate p Ͻ 0.05 compared with control. L.U., light units.
ance of the lipid around and associated with the virus (Fig.  10E, inset).
Gene Transfer by PLL⅐Ad Complexes to Airway Epithelia in Vitro and in Vivo-We also assessed the effect of complexes on airway epithelia. Primary cultures of human airway epithelia grown at the air-liquid interface differentiate into a respiratory epithelium that has a cilium-covered surface and many characteristics of the native epithelium (8,23). To test infection of mature epithelia, we applied adenovirus alone or complexed with PLL to the apical surface for 30 min. Then the virus was removed, and the surface was washed twice to remove unattached virus. Fig. 11 shows that more cells were infected with PLL⅐Ad than with adenovirus alone; there were 94 Ϯ 48 hexonpositive cells/low power field with adenovirus alone versus 797 Ϯ 115 with PLL⅐Ad (p ϭ 0.001; n ϭ 4).
We also applied 20 m.o.i. of Ad2/␤Gal-2 alone, PLL⅐Ad, or GL-67⅐Ad to the mucosal surface of mature epithelial monolayers for 30 min. Fig. 12A shows that epithelia treated with PLL⅐Ad or GL-67⅐Ad generated more ␤-galactosidase activity than those treated with adenovirus alone. Fig. 12B shows that, as we previously reported (8), application of adenovirus expressing CFTR to the apical surface of CF airway epithelia for 30 min had little effect on cAMP-stimulated Cl Ϫ transport. However, treatment with PLL⅐Ad complexes restored a cAMPstimulated Cl Ϫ current to CF epithelia. These data suggest that the complexes are more efficient than adenovirus alone at gene transfer to mature human airway epithelia.
Previous studies have administered adenovirus vectors encoding CFTR to the nasal epithelium of patients with CF and tested for correction of the electrophysiologic defect. We initially reported correction of the CF electrophysiologic abnor-malities following application of Ad2/CFTR-1 to nasal epithelium that was injured during the application procedure (2). However, when adenovirus vector was applied to intact respiratory epithelium, we saw only limited evidence of gene transfer (5). Hay et al. (4) found evidence of partial electrophysiologic correction, and Knowles et al. (3) found no evidence of correction. These data suggest that more efficient gene transfer to human ciliated respiratory epithelia in vivo would be valuable. We (8) and others (7) have also found that a single application of adenovirus to the nasal epithelium of CF mice is unable to correct the CF Cl Ϫ transport defect, although if the duration of contact with the epithelium is prolonged, some correction does occur (8). Therefore, we tested the ability of PLL⅐Ad to correct the electrophysiologic defect in the nasal epithelium of CF mice. Fig. 13 shows that CF mice bearing the ⌬F508 mutation have a basal voltage that was more electrically negative than that of wild-type mice, and the voltage failed to hyperpolarize in response to perfusion with a solution containing a low Cl Ϫ concentration. As previously reported (8), administration of Ad2/CFTR-8 failed to correct either of these defects (Fig. 13). However, after addition of PLL⅐Ad, both electrophysiologic properties were corrected to the normal range. These data indicate that functional CFTR Cl Ϫ channels were restored in the nasal epithelium. DISCUSSION In this study, we combined adenovirus with cationic molecules to generate complexes that increased the efficiency of gene transfer. Expression of reporter genes was increased in a number of cultured cells, and gene transfer to primary cultures of mature human airway epithelia was enhanced. In vivo, the  C and D), or Lipofectamine-adenovirus (6.9 ؋ 10 4 molecules/particle) (E and F) applied to 9L gliosarcoma cells. The inset in E shows a higher magnification of the lamellar structure of the lipid surrounding the adenovirus particles. Vector (2 ϫ 10 6 particles/cell) was applied for 15 min and then washed and prepared for microscopy as described under "Materials and Methods." Scale bars ϭ 100 nm.
complexes corrected the electrophysiologic abnormalities that characterize CF epithelia.
By including PLL or cationic lipids in the complex, we were able to increase binding and cellular uptake. These are important barriers to adenovirus-mediated gene transfer in cells that do not express a cell-surface receptor that binds adenovirus fiber (28,30,31). Binding and uptake of the complex did not appear to be dependent on fiber receptors because expression was not blocked by addition of excess fiber protein or a neutralizing antibody to fiber. Transmission electron microscopy also suggested that the complex enters the cell via a mechanism different from that of adenovirus alone.
By including adenovirus in the complex, we were able to take advantage of adenovirus-dependent processes that facilitate expression such as escape from the endosome and entry into the nucleus (11); these processes are significant barriers to nonviral vector-mediated gene transfer. Our data indicate that adenovirus retains its ability to facilitate DNA delivery, even when entry into the cell is via a pathway different from the usual receptor-mediated mechanism. But the finding that a neutralizing antibody against hexon and chloroquine pretreatment inhibited expression suggests that when viral processes after uptake are compromised, the efficiency of infection falls.
These processes may include escape from the endosome and transport to the nucleus.
We found that we could combine many different cationic molecules with adenovirus to facilitate gene transfer. However, there were important differences in the ability of various cationic molecules to enhance gene transfer, and not all cationic molecules were effective. The receptors for complex binding are most likely negatively charged molecules on the cell surface. Although we have not identified specific receptors, it seems likely that many different negatively charged molecules, e.g. sialic acids, may bind the complexes. This conclusion is consistent with our observation that pretreatment of cells with PLL blocked gene transfer by complexes containing either PLL or GL-67.
Although we did many of our studies with PLL to test the concept, PLL may not be the cationic component of choice for gene therapy applications. In this regard, cationic lipids might offer advantages because they have already been administered to humans (6) and function as well or better than PLL in airway epithelia. Cationic lipids might also offer other advantages by providing additional components that facilitate gene transfer. For example, the cholesterol in GL-67 could possibly facilitate cell binding. Other cationic molecules, such as dendrimers, might also prove to be effective. Future studies will be required to evaluate cell type-specific effects and to identify the relationship between the structure of the cationic molecule, its function in the complex, and its interaction with the cell.
Previous reports have described the incorporation of adenovirus into gene transfer complexes (12-14, 16, 17, 32-37). Such complexes took advantage of the endosome-disrupting properties of adenovirus to enhance gene transfer. Several investigators have chemically (12-14, 17, 32, 36) or immunologically (13,36,37) coupled adenovirus to PLL alone, to PLL conjugated to a receptor ligand (12)(13)(14)32), or to PLL plus PLL conjugated to a receptor ligand (17,32,35,36). This complex was then charge-associated with plasmid DNA, and expression of plasmid DNA was used to evaluate gene transfer. In addition, unbound adenovirus has been co-internalized with PLL-plasmid DNA complexes with and without an associated receptor ligand (16,18,38) and with cationic lipid-DNA complexes (9,15). Many of these reports describe an increase in transgene expression. However, to our knowledge, previous reports have not used cationic molecule-adenovirus complexes in which the transgene is encoded in adenovirus DNA. This approach uses the cationic molecule to enhance adenovirus uptake into poorly infected cells and then utilizes adenovirus proteins to facilitate steps in gene transfer subsequent to uptake.
The system we describe has a number of advantages. It may be of value in gene transfer for cells and tissues in which infection is limited because the cells lack fiber receptors. On the other hand, it probably offers no advantage for cells that are already easily infected by adenovirus, such as hepatocytes. Because of its nonspecific nature, it may be useful for tissues in which the vector can be applied selectively to the target cells, e.g. to the apical surface of airway epithelia. The system also has the advantage that many different cationic molecules can be evaluated readily; because covalent linkage is not required, less time may be needed to evaluate specific conditions, and there may be less danger of inactivating important viral functions. The system we describe also has some disadvantages. The complex needs to be used shortly after preparation to ensure optimal gene transfer. The system is more complex than either adenovirus or cationic lipid-DNA alone. Moreover, after in vivo delivery, there is the possibility of toxicity from each of the individual components. However, future studies may solve some of these problems, and their importance may be dimin- FIG. 11. Infection of primary cultures of mature human airway epithelia by wild-type Ad2 alone (A) or complexed with PLL⅐Ad (B). Epithelia were studied 21 days after they were seeded when they were confluent and differentiated. Ad2 or PLL⅐Ad (m.o.i. ϭ 50, particle/IU ratio ϭ 12) was applied to the apical surface of epithelia (ϳ5 ϫ 10 5 cells/epithelium) for 30 min. Vector was then removed, and the epithelium was rinsed twice. Thirty h later, the epithelium was fixed and stained with anti-hexon antibody.
ished by increased efficiency.
We also speculate that the system we describe could be adapted to include additional functions. For example, specific ligands might be included, and the amount of cationic molecule available for binding to the cell might be decreased in an attempt to target the vector to specific cells. Using such a system, perhaps adenovirus infection of its normal targets, fiber receptor-bearing cells, could be minimized. Finally, it is conceivable that adenovirus could be coated with cationic molecules that replace the cell binding and internalization functions of the virus, yet shield it from neutralizing antibodies that can prevent repeat administration.
This work shows that complexes of adenovirus with cationic molecules enhance gene transfer in vitro and in vivo. Because they interact with the cell surface through charge association, these and related complexes may increase gene transfer to a number of different cell types. With improvement in the efficiency of gene transfer, it may be possible to deliver less vector and, as a result, to reduce toxicity and the immune response. A consequent increase in the therapeutic index would further enhance the development of successful gene therapy. FIG. 12. Effect of adenovirus alone, PLL⅐Ad, and GL-67⅐Ad on gene transfer to human airway epithelia grown at the air-liquid interface on permeable filter supports. A, Ad2/␤Gal-2 alone, PLL⅐Ad (250 molecules/particle), or GL-67⅐Ad (2 ϫ 10 5 molecules/particle) was applied in 50 l (m.o.i. ϳ 20) to the apical surface. Vector was applied 12 days after cells were seeded, and ␤-galactosidase activity was measured 4 days later (n ϭ 3). Asterisks indicate p Ͻ 0.01 compared with adenovirus alone. B, Ad2/CFTR-8 were prepared as described for A and applied to CF airway epithelia 10 -14 days after they were seeded. The change in short-circuit current in response to cAMP agonists (Isc cAMP ) was measured 4 -5 days later (n ϭ at least eight in each group). Asterisks indicate p Ͻ 0.05 compared with adenovirus. L.U., light units.

FIG. 13. Effect of adenovirus and PLL⅐Ad complexes on basal voltage (Vt) (A) and change in voltage produced by perfusion with a solution containing a low Cl ؊ concentration (⌬Vt Low Cl ) and cAMP agonists in the presence of amiloride (B).
Either Ad2/ CFTR-8 alone or complexed with PLL (250 PLL molecules/particle) was applied to the nasal epithelium of mice containing the ⌬F508 mutation in the mouse CFTR gene. Vector (5 ϫ 10 7 IU/nostril) was applied to the nasal mucosa in 5 l. Electrophysiologic measurements were obtained 2 days after vector administration. Data for wild-type (ϩ/ϩ) and CFTR-⌬F508 (⌬F/⌬F) mice that were not exposed to adenovirus are from Zeiher et al. (25). Asterisks indicate p Ͻ 0.01 compared with ⌬F/⌬F mice treated with adenovirus alone (p Ͻ 0.01; n ϭ 12 for adenovirus and 13 for PLL⅐Ad).