Nucleotide Exchange in Genomic DNA of Rat Hepatocytes Using RNA/DNA Oligonucleotides

Chimeric RNA/DNA oligonucleotides have been shown to promote single nucleotide exchange in genomic DNA. A chimeric molecule was designed to introduce an A to C nucleotide conversion at the Ser365 position of the rat factor IX gene. The oligonucleotides were encapsulated in positive, neutral, and negatively charged liposomes containing galactocerebroside or complexed with lactosylated polyethyleneimine. The formulations were evaluated for stability and efficiency in targeting hepatocytes via the asialoglycoprotein receptor. Physical characterization and electron microscopy revealed that the oligonucleotides were efficiently encapsulated within the liposomes, with the positive and negative formulations remaining stable for at least 1 month. Transfection efficiencies in isolated rat hepatocytes approached 100% with each of the formulations. However, the negative liposomes and 25-kDa lactosylated polyethyleneimine provided the most intense nuclear fluorescence with the fluorescein-labeled oligonucleotides. The lactosylated polyethyleneimine and the three different liposomal formulations resulted in A to C conversion efficiencies of 19–24%. In addition, lactosylated polyethyleneimine was also highly effective in transfecting plasmid DNA into isolated hepatocytes. The results suggest that both the liposomal and polyethyleneimine formulations are simple to prepare and stable and give reliable, reproducible results. They provide efficient delivery systems to hepatocytes for the introduction or repair of genetic mutations by the chimeric RNA/DNA oligonucleotides.

Recently, a novel technology for facilitating targeted gene correction of episomal DNA in mammalian cells has been described (1,2). The technology involves the use of a chimeric oligonucleotide (ON) 1 composed of a contiguous stretch of RNA and DNA residues in a duplex conformation with poly(T) hairpin caps at both ends and a 3Ј G:C clamp. In a typical heteroduplex molecule, two blocks of 10 2Ј-O-methyl RNA residues flank a pentameric stretch of DNA. The structure exhibits increased chemical and thermal stability as well as greater resistance to a variety of nucleases. The chimeric molecule is designed so that it aligns in perfect register with a specified genome target, with the exception of a single base pair in the region of the pentameric stretch of DNA. The single mismatched nucleotide is recognized by endogenous DNA repair systems, thus affecting alteration of the DNA sequence of the targeted gene (1,3). This technology has been shown to be successful in vitro in correcting the mutation responsible for sickle cell anemia (4) and in mutating the rat factor IX gene in rat hepatocytes in vivo (5). While the efficiency of nucleotide conversion may depend on a number of factors, delivery of DNA to the cell and its nucleus remains key. In designing an optimal delivery system, important parameters include cell toxicity and targeting specificity as well as stability of the formulation.
The delivery of intact nucleic acids from the extracellular environment to the nucleus has remained a major barrier to long term and stable gene expression both ex vivo and in vivo. After cellular uptake of nucleic acids by endocytosis, only a small fraction of the molecules reach the nucleus intact (6,7). Interestingly, a variety of studies, including direct cytoplasmic injection of nucleic acids, suggest that these molecules display significant nuclear tropism (7)(8)(9). However, the movement of DNA from the cytoplasm to the nucleus continues to be an important limitation to successful gene transfer (10). Cationic lipids and polycations have been used to complex nucleic acids, thereby protecting them from degradation, while simultaneously increasing their endocytic uptake into cells (11)(12)(13)(14). Many of these lipid formulations, although successful in cell culture, have not been useful for in vivo delivery because of short serum half-life, toxicity, and lack of tissue specificity (15). Polycations, such as poly-L-lysine, polyethyleneimine (PEI), and polyamino lipids, form water-soluble complexes that can provide simple but very efficient delivery systems. The presence of free amino groups on these agents makes them amenable to chemical modification for the attachment of ligands capable of targeting specific tissues. For example, asia-loorosomucoid (16,17) and galactose (18) have been conjugated to poly-L-lysine for targeting to the asialoglycoprotein receptor (ASGP R ) on hepatocytes. The asialoorosomucoid-poly-L-lysine system has been demonstrated to increase hepatocyte-directed gene delivery both in vitro (19) and in vivo (20). A similar strategy has been utilized for hepatocyte-specific gene delivery using lipopolyamine-condensed DNA targeted with galactose ligands (21).
Transfection of different cell types with the chimeric ONs has been accomplished with dioleoyl trimethylammoniumpropane (4), Lipofectin ® (2), and PEI (22). However, these systems lack the physicochemical stability and proper formulation for efficient in vivo delivery. The existing system of gene delivery by cationic lipids consists primarily of heterogeneous complexes and aggregates rather than liposome-encapsulated material (23). A combinatorial approach of condensing the DNA using polycationic molecules followed by liposome association or encapsulation has been demonstrated to greatly increase the efficacy of the delivery systems (24 -26). In the present study, we describe the preparation and characterization of three liposomal systems, targeted to the ASGP R on rat hepatocytes with galactocerebroside, for delivery of the chimeric molecules. In addition, we have characterized a nonliposomal targeting system using lactosylated PEI. In fact, unmodified PEI has been shown to be a useful gene delivery vehicle both in vitro and in vivo (27)(28)(29). Our results indicate that RNA/DNA ONs can be efficiently delivered to hepatocytes for gene conversion by both liposomal and nonliposomal systems without the use of viral vectors.

EXPERIMENTAL PROCEDURES
Synthesis and Fluorescent Labeling of RNA/DNA ONs-Chimeric RNA/DNA factor IX ONs were synthesized by Applied Biosystems, Inc. (Foster City, CA) as described previously (2). The crude ONs were purified by high pressure liquid chromatography and quantitated by UV absorbance. Greater than 95% of the purified ONs were determined to be full-length.
A 68-mer all-DNA chimeric ON was obtained from Genosys Biotechnologies, Inc. (The Woodlands, TX). The molecules were 3Ј-end-labeled using terminal transferase and fluorescein-12-dUTP from Boehringer Mannheim according to the manufacturer's recommendation. Labeled ONs were then mixed with unlabeled ONs at a 2:3 ratio.
Liposomal Formulations-Dioleoyl phosphatidylcholine, dioleoyl phosphatidylethanolamine, dioleoyl phosphatidylserine, galactocerebroside, and dioleoyl trimethylammoniumpropane were purchased from Avanti Polar Lipids Inc. (Alabaster, AL) as dried powders. Stock solutions (2 mg/ml) of the lipids and lipid films were prepared as described previously (30). Positively charged liposomes consisted of dioleoyl phosphatidylcholine/dioleoyl trimethylammoniumpropane/galactocerebroside at a 6:1:0.56 molar ratio; neutral liposomes included dioleoyl phosphatidylethanolamine/dioleoyl phosphatidylcholine/galactocerebroside at a 1:1:0.16 ratio, and negatively charged liposomes were formulated with dioleoyl phosphatidylserine/dioleoyl phosphatidylcholine/galactocerebroside at a molar ratio of 1:1:0.16. PEI (800 kDa) was purchased from Fluka Chemical Corp. (Ronkonkoma, NY). A stock solution of 0.05 M PEI monomer corresponding to ϳ50 nmol of PEI amine/l (27) was prepared in Milli Q ® (Millipore Corp., Bedford, MA) water, and its pH was adjusted to 7.6 with 6 M HCl. PEI (4 nmol of amine/nmol of RNA/DNA phosphate) and either unlabeled or fluorescently labeled chimeric molecules (100 -250 g) were each diluted into 250 l of 0.15 M NaCl. The solutions were vortexed separately for 5-10 min at room temperature followed by the dropwise addition of the PEI solution to the ON, with continual vortexing for an additional 5-10 min. The PEI complex was then added to the positively charged lipid film. For the neutral and negatively charged liposomes, naked or fluorescently labeled chimeric ONs (150 g) were diluted into 500 l of 0.15 M NaCl and added to the lipid film. The films were hydrated by alternate vortex mixing and warming until a homogeneous suspension was formed. The resulting suspension was extruded through a series of polycarbonate membranes down to a pore size of 0.05 m using a Liposofast ® miniextruder system (Avestin, Inc., Ottawa, Canada).
Analysis of the Liposome Formulations-The size of the extruded liposomes was determined by light scattering using a NICOMP 370 submicron particle size analyzer (Pacific Scientific Instruments, Santa Barbara, CA). In short, the Brownian motion of suspended particles modulates the phase of scattered light waves from a laser beam. Thus, the intensity of the detected light fluctuates in a random fashion and is detected with a photomultiplier tube. For a given particle, the average lifetime for the intensity fluctuations is roughly equal to the average time required for two particles to change their separation by one-half of the fixed laser wavelength . The spherical radius of the particle is then obtained from the diffusion coefficient of the particle over time using the Stokes-Einstein relation. The phospholipid concentrations in the liposomal preparations were determined by the method of Stewart (31). Aliquots of 20 -50 l were diluted to 500 l using Milli Q ® water, and the phospholipids were extracted twice using 500 l of chloroform: The liposomal extracts were also analyzed for PEI complexes and naked ONs following ultrafiltration to remove the nonencapsulated material. Briefly, 40 -50 l of liposome suspension after lipid extraction was diluted with 300 l of heparin (1 unit/l) and incubated at 37°C for 6 h. Following phenol/chloroform extraction, the aqueous phase was precipitated at Ϫ20°C overnight following the addition of 80 l of 3 M sodium acetate (pH 5.2), 500 l isopropyl alcohol, 10 g of tRNA, and 300 l of RNAmate ® (Intermountain Scientific, Inc., Kaysville, UT). An aliquot of the stock PEI⅐ON complex was subjected to the same extraction procedure. The ONs were analyzed by 4% low melting point agarose gel electrophoresis containing 1 g/ml ethidium bromide and visualized using UV light. The relative intensities of the bands were analyzed by densitometry using a Bio-Rad model GS-700 imaging densitometer.
Preparation of Lactosylated PEI-PEI was lactosylated by a modification of a previously described method for oligosaccharide conjugation (32). Briefly, 3-5 ml of a 0.1-0.2 M stock of either 800-or 25-kDa PEI (Aldrich) in 0.2 M ammonium acetate or Tris-HCl, pH 7.6, buffer solution was incubated with 7-8 mg of sodium cyanoborohydride (Sigma) and approximately 20 -30 mg of lactose monohydrate (Sigma) at 37°C for 10 days. The reaction mixture was dialyzed against Milli Q ® water for 48 h with 1-2 changes of water/day. The amount of sugar (as galactose) conjugated with PEI was determined by the phenol-sulfuric acid method (33). The number of moles of free secondary amines in the lactosylated PEI (L-PEI) was determined as follows. A standard curve was generated using a 0.02 M stock solution of PEI; several aliquots of the stock were diluted to 1 ml using Milli Q ® water in glass tubes, and then 50 l of ninhydrin reagent (Sigma) was added to each tube and vortexed vigorously for 10 s. Color development was allowed to proceed in the dark at room temperature for 10 -12 min, and the OD was determined at 485 nm on a Beckman DU-64 spectrophotometer.
Preparation of L-PEI⅐ON Complexes-An equivalent of 5 nmol of amine as L-PEI and 5 nmol as PEI per nmol of RNA/DNA phosphate were mixed, diluted in 0.15 M NaCl as required, and used for transfection as described previously (5). For the in vitro fluorescently labeled ON uptake experiments, the 25-kDa PEI plus L-PEI mix (1:1 molar ratio of amines) was complexed with the labeled chimeric molecules at a ratio of 10 or 6 nmol of PEI amines per nmol of DNA/RNA phosphate in 5% dextrose. For the in vivo studies, the 25-kDa PEI plus L-PEI mix (1:1 molar ratio of amines) was complexed with the RNA/DNA ONs at the 6:1 amine:phosphate ratio in 5% dextrose.
Nuclease resistance of the complexed RNA/DNA molecules was determined by treating the samples with 40 units of RNase A and 40 units of RQ1 RNase free DNase (Promega Corp., Madison, WI) for 40 min at 37°C. The reaction was terminated using 2 l each of 0.5 M EGTA and 0.5 M EDTA. Following two phenol/chloroform extractions, the complexes were dissociated using heparin (50 units/g of nucleic acid) at 37°C for 60 min, and the products were analyzed on a 4% low melting point agarose gel.
Size Analysis of the ON and L-PEI⅐ON Complexes-The size of the chimeric molecules alone or complexed to PEI was determined by gas phase electrophoretic mobility molecular analysis (GEMMA) (TSI Inc., Minneapolis, MN) (34) or by light scattering measurements. For GEMMA, chimeric ONs (4 mg/ml) were diluted at a 1:3,000 ratio with 0.02 M ammonium acetate buffer. The diluted suspension was then transformed into an aerosol by electrospray drying. The resulting high charge on the molecules was neutralized by a radioactive ␣-emitter, and the singly charged chimeric molecules were size-separated according to their mobility in air.
Electron Microscopy-Liposome suspensions were applied as a drop on glow-discharged formvar carbon-coated grids (300 or 400 mesh, Polysciences Inc., Warrington, PA) and negatively stained using 2% ammonium molybdate. PEI complexes were stained with either a 1% uranyl acetate or a 2% ammonium molybdate solution, and the samples were visualized using a JEOL100-CX electron microscope. For scanning electron microscopy, samples were mounted in drops onto borosilicate chips and fixed with 2% osmium tetroxide vapors for 2 h at 25°C, followed by stepwise dehydration in ethanol (35). The mounted samples were stored in 100% ethanol until critical point drying in liquid carbon dioxide. Immediately following drying, samples were sputter-coated with platinum metal for 6 min at 4.2 V. Samples were viewed the following day on a Hitachi S-900 field emission scanning electron microscope at 1.5 kV at magnifications of ϫ 1,200 -200,000.
For freeze-fracture analysis, anionic liposomes were made using fluorescein-, Cy3-labeled, or naked ONs, frozen in freon, and then stored in liquid nitrogen until fracture. Fracture was performed on a BAF 060 apparatus. The fractured surfaces were coated with 6 nm of platinum at a 45°angle to the surface, followed by 20 nm of carbon at a 90°angle to the surface. Replicas were cleaned overnight by flotation in 1 M sodium hypochlorite and then rinsed for 1 h in water. After transfer to 200 mesh carbon grids, the specimens were viewed at 80 kV on a JE0L100-CX electron microscope.
Transfections-Positive, neutral, and negative liposomes were diluted with Hepes-buffered saline, pH 7.4, to a final concentration of 2-6 g of ON/100 l and used for transfection as described previously (30). Nonencapsulated PEI⅐ON complexes at identical amine:phosphate ratios were used as positive controls and diluted as described above. The cells were washed twice with William's E medium further supplemented with an additional 2.5 mM CaCl 2 and then incubated with 1 ml of the same medium. An aliquot of 100 l of transfecting solution was then added to each 35-mm dish. After 18 h of transfection, an additional 2 ml of medium containing 10% heat-inactivated FBS was added, and the cells were maintained for an additional 6 -30 h.
Transfection of Plasmids with L-PEI-The reporter plasmid PGL3 (Promega Corp.) encoding the luciferase gene was prepared in 0.15 M NaCl or 5% dextrose using various ratios of L-PEI to 25-kDa PEI amines as well as nucleic acid phosphate to PEI amines. Hepatocytes were harvested 48 h after transfection using 200 l of reporter lysis buffer, and the luciferase activity was determined by the luciferase assay reagent (Promega) according to the manufacturer's protocol. Protein was quantitated with the Bio-Rad protein assay reagent as specified by the manufacturer.
Nuclear Uptake of Fluorescently Labeled ONs-Hepatocytes transfected with the fluorescently labeled ONs were fixed 24 h post-transfection in phosphate-buffered saline, pH 7.4, containing 4% paraformaldehyde (w/v) for 10 min at room temperature. Following fixation, the cells were counterstained using DiI (Molecular Probes, Inc., Eugene, OR) and coverslipped using SlowFade™ (Molecular Probes) antifade mounting medium in phosphate-buffered saline and examined using a MRC1000 confocal microscope (Bio-Rad) (22).
In Vivo Delivery of L-PEI Complexes-Male Sprague-Dawley rats (Harlan Sprague-Dawley) (ϳ65 g) were maintained on a standard 12-h light-dark cycle and fed ad libitum standard laboratory chow. The rats were restrained by hand, and the chimeric molecules, complexed with 25-kDa PEI/L-PEI (1:1 molar ratio of amines) at a ratio of 6 equivalents of PEI nitrogen per RNA/DNA phosphate in 500 l of 5% dextrose (28) were administered in vivo by tail vein injection. Vehicle controls received an equal volume of PEI in Tris-HCl, pH 7.6. The animals were anesthesized with ether 1 and 18 weeks postinjection and underwent a midline incision. The liver was exposed, and random samples were excised for DNA isolation. After 3 weeks, two of the animals received an additional injection of 500 g of the L-PEI/chimeric complexes followed 24 h later by 70% partial hepatectomy (37) to determine the replicative stability of the genomic nucleotide conversion. Genomic DNA larger than 100 -150 base pairs was isolated from the tissue samples using the high pure PCR template preparation kit (Boehringer Mannheim) for PCR amplification of exon 8 of the rat factor IX gene.
Factor IX Activity-Blood samples were obtained from the test groups at varying times up to 52 weeks after the final tail vein injection and mixed in 0.1 volume of 0.105 M sodium citrate/citric acid. The blood samples were centrifuged at 2,500 ϫ g, followed by 15,000 ϫ g centrifugation, and the resulting plasma was stored at Ϫ70°C. The factor IX activity was determined from activated partial thromboplastin time assays as described previously (5). The factor IX activity of duplicate samples was determined from a log-log standard curve constructed from the activated partial thromboplastin time results of pooled plasma from 12 similarly aged normal male rats through multiple dilutions.
DNA Isolation and Cloning-The chimeric ON-and vehicle-transfected cells were harvested by scraping 48 h after transfection. Genomic DNA larger than 100 -150 base pairs was isolated using the high pure PCR template preparation kit (Boehringer Mannheim). The isolated DNA (500 ng) from either primary hepatocytes or liver was used for PCR amplification of a 374-nucleotide fragment of the rat factor IX gene as described previously (5). The primers were designed as 5Ј-ATTGC-CTTGCTGGAACTGGATAAAC-3Ј and 5Ј-TTGCCTTTCATTGCACAT-TCTTCAC-3Ј (Oligos Etc., Wilsonville, OR) corresponding to nucleotides 433-457 and 782-806, respectively, of the rat factor IX cDNA (38). The PCR amplicons were subcloned into the TA cloning vector pCR™2.1 (Invitrogen, San Diego, CA), and the ligated material was used to transform frozen competent E. coli.
Colony Lift Hybridization and Sequencing-18 -20 h after plating, the colonies were lifted onto MSI MagnaGraph nylon filters, replicated, and processed for hybridization according to the manufacturer's recommendation. The filters were hybridized with 17-mer ON probes 365 A (5Ј-AAGGAGATAGTGGGGGA-3Ј) or 365 C (5Ј-AAGGAGATCGTGG-GGGA-3Ј) (Life Technologies), where the targeted nucleotide for mutagenesis is underlined. The probes were 32 P-end-labeled using [␥-32 P]ATP (Ͼ7,000 Ci/mmol) and T4 polynucleotide kinase (New England Biolabs, Inc., Beverly, MA). Hybridizations were performed at 37°C, and the filters were processed posthybridization as described previously (5,39). Autoradiography was performed with NEN ® Reflection film at Ϫ70°C using an intensifying screen. Plasmid DNA was prepared from colonies hybridizing to either 365 A or 365 C using a Qiagen miniprep kit (Chatsworth, CA). Sequencing was performed with an ABI 370A sequencer (Perkin-Elmer, Corp., Foster City, CA) using a gene-specific primer, 5Ј-GTTGACCGAGCCACATGCCTTAG-3Ј, corresponding to nucleotides 616 -638 of the rat factor IX cDNA (38), and the mp13 forward and reverse primers.
Statistical Analysis-Data were analyzed using InStat version 2.01 (GraphPad Software, San Diego, CA) to calculate analysis of variance, and probability (p) values were determined using Bonferroni multiple comparisons. Two column comparisons were analyzed using Welch's alternate t test.

Characterization of the Liposomal Formulations and PEI/L-PEI Complexes-By
GEMMA the chimeric ONs ranged in size from 4 to 7 nm, with major peaks at 4.6 and 5.8 nm (Fig. 1a). After heating to 95°C and then cooling to room temperature, the chimeric molecules exhibited a unimodal distribution with an average diameter of 4.4 nm (Fig. 1b). ONs complexed with PEI in 0.15 M NaCl ranged in size from 20 to 200 nm. Electron microscopy revealed that the complexes were often aggregated into large clusters (Fig. 2, a and c), thus explaining the wide range in their size distribution. However, when the complexes were prepared in 5% dextrose, they appeared as smaller discrete particles with a mean diameter of ϳ20 nm (Fig. 2, b and  d). Following size-selected extrusion, the liposomes showed a unimodal size distribution with a volume-weighted mean diameter of Ͻ50 nm by light scattering. Neither the anionic nor cationic liposomes changed in size after overnight incubation at 37°C or storage for 1 month at 4°C. In contrast, the neutral liposomes aggregated and fused into larger vesicles under these conditions. The uniformity of anionic and cationic liposome suspensions was confirmed by electron microscopy. The PEI⅐ON complexes were either situated within the bilayer of the cationic liposomes or more typically entirely encapsulated within the liposome core (Fig. 3a). Negative liposomes prepared with the naked chimeric molecules appeared as uniform vesicles with small electron dense particles randomly distributed within the vesicles (Fig. 3b). The encapsulation of ONs was confirmed by freeze-fracture analysis of the anionic liposomes (Fig. 3, c and d).
L-PEI was analyzed for sugar content as well as for unconjugated secondary amines. The conjugation of lactose and PEI amines was reproducible and increased with incubation time and lactose concentration (Fig. 4). The percentage of lactose covalently linked to PEI ranged from 24.1 to 25.8% (Table I), and this agreed with the detectable levels of free amines. The optimal amounts of L-PEI or PEI per g of nucleic acid were calculated based on the nmol of secondary amines required to completely inhibit nucleic acid detection by ethidium bromide on an agarose gel (16). The complexed ONs were resistant to nuclease degradation in contrast to free RNA/DNA ONs, which were completely degraded (Fig. 5). Chimeric molecules encapsulated in the different liposome formulations showed a similar protection from nuclease degradation (data not shown).
In Vitro Luc Expression in Hepatocytes Using L-PEI⅐PGL3 Plasmid Complexes-L-PEI (25 kDa) was evaluated for transfection efficiency using the PGL3 plasmid encoding luciferase under control of the SV40 promoter. Interestingly, when the sugar:amine ratio in L-PEI ranged from approximately 0.1 to 1.7, no significant difference was observed in the level of luciferase expression at plasmid doses of Ն2 g (Table II). Significant differences in luciferase expression occurred only at the 1-g plasmid dose (p Ͻ 0.03) in comparing low level (sugar: amine ϭ 0.051) to high level (sugar:amine ϭ 1.19) lactosylation. Only the low level L-PEI displayed a clear dose response, even at the 1-g dose (Table III). Combining the L-PEI with naked PEI at a 1:1 molar amine ratio significantly improved (p Ͻ 0.05) the transfection efficiency over the entire range of lactosylation (Table II).
Complexes were also prepared at different DNA phosphate to L-PEI/PEI amine ratios using the 25% L-PEI. The transfection efficiency of PEI with 4 g of plasmid compared favorably with the L-PEI/PEI mix at the 1:10 phosphate:amine ratio (Table IV). That represented an effective mean cation:anion charge ratio of ϳ2 (28). However, using 2 g of plasmid and/or other ratios (1:6 and 1:4, representing cationic to anionic charge ratios of ϳ1.2 and neutral, respectively), the L-PEI/PEI mixes were significantly more efficient (p Ͻ 0.05). At both plasmid doses, the transfection efficiency of the L-PEI/PEI at a 1:10 ratio was Ͼ1:6, which was Ͼ1:4, as previously observed with the nontargeted PEI (28,29). The molar ratio of amine derived from either PEI or L-PEI was also varied to determine the effect on transfection efficiency with 2 g of plasmid. Increasing the molar ratio of amines to 2:1 L-PEI:PEI decreased the luciferase expression Ͼ90% at either the 1:10 or 1:6 phosphate to amine ratio (p Ͻ 0.001). Alteration of the molar amine ratio to 1:2 L-PEI:PEI also resulted in a significant loss of luciferase expression (p Ͻ 0.05); however, the extent of the decrease was substantially less at the 1:10 ratio than at the 1:6 ratio, suggesting nonspecific cationic uptake of the plasmid at this ratio. In fact, when 100 mM D-galactose was included in the transfections to inhibit receptor uptake, at the 1:10 ratio luciferase expression was decreased only 20% using 2 g of plasmid.
Nuclear Localization of the Chimeric Molecules-There was almost no detectable cytoplasmic or nuclear fluorescence with the 1:4 ON-PEI complexes in the primary hepatocytes after 24 h of transfection. In contrast, both the cationic and anionic liposome formulations exhibited cytoplasmic and even greater nuclear labeling with the fluorescent ONs. Furthermore, the nuclear fluorescence with the anionic liposomes (Fig. 6b) was more intense than with the cationic liposomes (Fig. 6a). Interestingly, the neutral liposomes displayed somewhat greater cytoplasmic than nuclear fluorescence. Co-incubation of all three targeted liposomes with 50 mM D-galactose resulted in significant inhibition in uptake of the fluorescently labeled ONs (Fig. 6c). Studies with L-PEI indicated that the targeted system was much more efficient in the intracellular delivery of the fluorescent chimeric molecules than the unmodified carrier. This was apparent for both the 800-and 25-kDa polymers (data not shown). In particular, for the 25-kDa derivative, a 1:6 ratio resulted in significant nuclear labeling as well as some punctate staining primarily in the perinuclear regions (Fig. 7c). This was also observed with the 1:10 phosphate:amine ratio; however, the abundance of punctate fluorescence in the cytosol and especially in the perinuclear region was increased (Fig.  7a). Galactose at 100 mM almost completely inhibited the appearance of the fluorescein label in cells incubated with the 1:6 phosphate:amine complexes (Fig. 7d). Although some detectable nuclear labeling was still observed with the 1:10 ratio, almost no fluorescence was present in the cytosol or perinuclear regions (Fig. 7b).
Mutant Conversion of the Rat Factor IX Gene-Chimeric ONs were designed that targeted the nontranscribed factor IX genomic DNA strands (Fig. 8). The RNA/DNA sequence of the molecules was identical in sequence to the wild-type gene except for the indicated substitutions in the pentameric stretch of DNA residues corresponding to the targeted nucleotide at Ser 365 . Isolated rat hepatocytes were transfected with the fac- tor IX chimeric molecules to induce targeted nucleotide exchange of Ser 365 3 Arg 365 in the rat genomic sequence. Conversion results in a gene sequence change from the wild-type A:T pairing to a mutant G:C pairing, which encodes an inactive form of factor IX.
Conversion of the Targeted Nucleotide at Ser 365 -The nuclear localization of the fluorescently labeled chimeric molecules indicated efficient transfection in the isolated rat hepatocytes. The cells were then transfected with the unlabeled factor IX chimeric molecules at comparable concentrations using PEI, L-PEI, or liposomes as the carrier. Additionally, vehicle control transfections were performed simultaneously. The cells were harvested 48 h after transfection and the DNA isolated and processed for hybridization. The frequency of the targeted A 3 C nucleotide conversion at Ser 365 was determined by hybridization of duplicate colony lifts of the PCR-amplified and cloned 374-nucleotide stretch of exon 8 of the factor IX gene (5, 38). ON probes corresponding to nucleotides 710 -726 of the cDNA sequence were used to distinguish between the wild-type 365 A (5Ј-AAGGAGATAGTGGGGGA-3Ј) or the converted 365 C (5Ј-AAGGAGATCGTGGGGGA-3Ј) sequence.
The overall frequency of conversion of the targeted nucleotide was calculated by dividing the number of clones hybridiz- Chimeric ONs with or without PEI were encapsulated in liposomes as described under "Experimental Procedures." The liposomal formulations were applied to glow-discharged formvar carbon-coated grids and negatively stained with a 2% ammonium molybdate solution. a, positive liposomes were prepared with PEI⅐ON complexes and extruded to a mean diameter of 50 nm. The chimeric ON-PEI complexes were 15-20 nm in size as determined by GEMMA and light scattering. Magnification was ϫ 225,000. b, negative liposomes containing encapsulated naked chimeric molecules appearing as punctate irregularities. Magnification was ϫ 112,500; inset magnification was ϫ 200,000. Freeze-fracture analysis of the anionic liposomes (c) alone or (d) containing encapsulated chimeric ONs. The mean diameter of the chimeric ON determined by GEMMA was ϳ4 nm. Magnification was ϫ 112,500; inset magnification was ϫ 225,000.
FIG. 4. Lactose conjugation to PEI. 25-kDa PEI and lactose were conjugated with sodium cyanoborohydride as described under "Experimental Procedures." After 10 days, the samples were dialyzed against Milli Q ® water for 48 h with 1-2 changes/day. Lactosylation, determined as galactose remaining in the dialyzed product, increased almost linearly with the concentration of lactose. a Values of secondary amines were determined using the ninhydrin assay as described under "Experimental Procedures." b Values of lactose were determined using the phenol-sulfuric acid method as outlined under "Experimental Procedures." c The mean and S.D. for six independent 10-day reaction products of 25-kDa PEI, which was lactosylated using 0.2 M PEI as a monomer in ammonium acetate buffer, pH 7.6, containing 10 mg of lactose and 2.67 mg of sodium cyanoborohydride/ml as described under "Experimental Procedures." ing with the 365 C ON by the total number of clones hybridizing with both ON probes. Vehicle-transfected cells and those transfected with unrelated chimeric ONs did not yield any clones that hybridized with the 365 C probe. The A 3 C conversion in the hepatocytes transfected with the 25-kDa L-PEI/factor IX chimeric molecules was dose-dependent, with a maximum exchange frequency of 20.2 Ϯ 1.2%. All three liposomal delivery systems also resulted in a dose-dependent targeted factor IX conversion in the primary hepatocytes, with significant increases at 270 nM relative to 180 nM (p Ͻ 0.05) and 90 nM (p Ͻ 0.001) (Table V). However, only the cationic and anionic formulations showed significantly increased conversion at the 180 nM dose (p Ͻ 0.05 and p Ͻ 0.01, respectively). Both the negative and positive liposomal delivery systems were efficient in promoting conversion; however, at the 180 and 270 nM concentrations, the neutral liposomes were significantly less efficient for conversion than the negative liposomes (p Ͻ 0.001).
In Vivo Nucleotide Exchange in Rat Liver Using L-PEI-In vivo administration via tail vein injection of the 1:6 ON:PEI complexes was associated with significant conversion (48.7 Ϯ 10.4%) of the targeted nucleotide in the intact liver. Additionally, no difference was noted between animals that received the chimeric ONs, which targeted the actively transcribed strand versus nontranscribed strand of the factor IX gene. Moreover, when the frequency of genomic A 3 C conversion was determined 18 weeks after the final injection, no significant difference was noted from that determined at 1 week.
Phenotypic changes in genomic conversion were determined by the activated partial thromboplastin time at 3, 18, and 46 weeks after the final injections. At 3 weeks, the treated rats (n ϭ 9) exhibited factor IX activity ranging from 38.2 to 70.8% (mean, 55.8 Ϯ 11.9%). This was significantly decreased     negative liposomeencapsulated fluorescent chimeric ONs as described under "Experimental Procedures." After 24 h of transfection, the hepatocytes were fixed using 4% paraformaldehyde in phosphate-buffered saline. Postfixation, the cells were counterstained using the cationic membrane probe DiI to visualize the cytoplasm. Double-label detection of the DiI fluorescence (orange) and fluorescein (green) was performed using a Bio-Rad MRC1000 confocal microscope. c, negative liposome-mediated uptake of fluorescent ONs was inhibited in the presence of 50 mM D-galactose.
(p Ͻ 0.001) from the vehicle control group (n ϭ 6), which ranged from 100.1 to 117.4% (mean, 112.7 Ϯ 8.7%). At 18 and 46 weeks, no significant changes in factor IX activity were detected in the experimental group (mean, 56.2 Ϯ 8.2 versus 57.9 Ϯ 6.3%, respectively) relative to that of the controls (mean, 111.3 Ϯ 7.9 versus 105.4 Ϯ 5.5%, respectively). At each time point, the decreased factor IX activity was associated with an ϳ30-s increase in activated partial thromboplastin time relative to controls. Treated rats were also subjected to 70% partial hepatectomy of the liver and analyzed for alterations in factor IX activity. During the regenerative period in which the liver mass is restored within ϳ10 days, at least 95% of remaining hepatocytes replicate in two synchronous waves (36). The factor IX activity had decreased from a mean of 53.8% observed at 3 weeks to 37.0% of control values at 49 weeks post-injection/ partial hepatectomy. The results suggested that in the intact liver, the site-directed genomic nucleotide exchange is replicatively stable and results in a long term phenotypic alteration in the coagulant activity as measured by the activated partial thromboplastin time.
Sequence Analysis of the Mutated Factor IX Gene-Sequencing of the wild-type and mutated factor IX genes was performed to confirm the results from the filter lift hybridizations. A  b Isolated rat hepatocytes were transfected with the L-PEI⅐ON complexes at a 1:10 phosphate:amine ratio in 0.15 M NaCl as described under "Experimental Procedures." minimum of 10 independent clones hybridizing to either 365 A or 365 C from the intact liver or isolated hepatocytes were analyzed. The results of the sequencing indicated that colonies derived from the factor IX-transfected primary hepatocytes hybridizing to the 365 C ON probe converted to a C at Ser 365 . The same A 3 C conversion at Ser 365 was observed in the clones derived from the transfected rat liver that hybridized with the 17-mer 365 C probe. In contrast, those colonies hybridizing to 365 A exhibited the wild-type factor IX sequence. No base substitution other than the targeted change at Ser 365 was detected in the clones sequenced. Additionally, both the start and end points of the cloned 374-nucleotide PCR-amplified genomic DNA from the isolated hepatocytes and intact liver corresponded precisely to those of the primers used for the amplification process, indicating that the sequenced DNA was derived from genomic DNA rather than nondegraded chimeric ONs (data not shown).

DISCUSSION
Chimeric RNA/DNA molecules have been successful in promoting single nucleotide exchange in the rat factor IX gene (5) and in the ␤-globin gene responsible for sickle cell anemia (4). These ONs, when combined with a versatile delivery system, could potentially be useful therapeutic agents in treating certain genetic diseases. Liposomal delivery systems have long been identified as safe and effective drug carriers capable of being modified to target various cells and tissues. The most common liposomes used for gene delivery are those prepared with various cationic lipids in combination with a fusogenic lipid (15,40). However, the majority of the cationic lipid formulations tend to be entrapped in the pulmonary capillaries, thus limiting their access to other organs (41,42).
In this study, cationic, neutral, and anionic targeted liposomal formulations were prepared and characterized, and the efficacy of each system was assessed for potential use as a delivery system for RNA/DNA ONs. In contrast to previously reported liposomal formulations of nucleic acids, the chimeric molecules were directly encapsulated within the liposomes rather than relying on nonspecific interaction or association. Dioleoyl trimethylammoniumpropane was chosen as the cationic lipid because of its decreased toxicity (43) and its more efficient uptake by liver cells as compared with other cationic lipids (44 -46). Dioleoyl phosphatidylserine was selected as the anionic lipid due to its efficiency in the delivery of DNA (47) and its somewhat greater intrahepatic distribution (48,49). Galactocerebroside was included in all the formulations, since it has been used successfully as a targeting ligand for the hepatocyte ASGP R (50).
Light scattering measurements and electron microscopy studies revealed that the cationic and anionic liposomes remained stable after encapsulation of the chimeric ONs. Neutral liposomes, however, were thermodynamically unstable and tended to aggregate within hours of preparation. Efficient and stable sizing of the anionic and cationic liposomes to 50 nm was achieved using membrane extrusion. The small size of the liposomal delivery vehicle is critical, since the access to the hepatocyte through the fenestrated epithelium of the liver sinusoid is greatly enhanced when particles are Ͻ100 nm in size (51). The smaller size of the liposome also significantly decreases nonspecific trapping of the particles in the lungs and spleen (18,42), while including the targeting ligand provides hepatocyte-specific uptake via the ASGP R (18 -20, 50). The incorporation of dioleoyl phosphatidylserine in the anionic formulation may have promoted hepatocyte uptake, since this lipid constituent appears to preferentially direct liposomes to hepatocytes in vivo (49). The ability to encapsulate the naked chimeric ONs in the anionic and neutral liposomes may have resulted, in part, from their small and very uniform size (ϳ4.4 nm). Interestingly, the heating and cooling of the chimeric molecules altered their GEMMA distribution from bimodal to unimodal. The observed bimodal distribution, with the peaks differing by the cube root of 2 in the unheated population, resulted from a dimeric form of the chimeric ON (52). In fact, both monomeric and dimeric bands were detected when analyzed by gel electrophoresis (data not shown).
All three liposome preparations were efficient in delivering the chimeric molecules intracellularly and promoting a significant A 3 C conversion at the Ser 365 position in the rat factor IX gene. The in vitro conversion efficiencies obtained with the different liposomal formulations were consistently greater than non-L-PEI using similar concentrations of ONs (5). As previously reported, when targeted to a specific receptor, such as that for folate, the negatively charged liposomes exhibited better cellular uptake than their positively charged counterparts (26). This difference in cellular uptake may account for the increased conversion efficiency of the targeted site observed with the negative liposome-encapsulated chimeric ONs, since transfection efficiency and conversion rate correlate in vitro (22). Additionally, by confocal microscopy, cells transfected with the cationic liposomes displayed more punctate cytoplasmic localization of the labeled chimeric molecules than the anionic transfected cells, suggesting less efficient release of the ON complexes from the endocytic compartment. While the cationic and anionic liposome formulations were efficient, nontoxic, and stable, extrusion of the vesicles to ϳ50 nm in size often resulted in encapsulation of 60% or less of the initial concentration of chimeric molecules.
A number of nonliposomal delivery systems have been developed and characterized. These formulations typically involved condensing the nucleic acid with a polycation, thereby compacting the molecules as well as protecting them from degradation by various nucleases (19,20,27,54). Additionally, the polycations have been modified to promote receptor-specific targeting. For example, poly-L-lysine coupled to asialoorosomucoid (19) and galactose (17,18,53) markedly improved delivery of a number of transgenes to hepatocytes by targeting to the AS-GP R . In this study, we modified PEI by attaching lactose to its amine nitrogens by a simple procedure, thereby targeting the L-PEI to the ASGP R . This modification resulted in a significantly improved delivery to the cells. Interestingly, the 25-kDa L-PEI was more effective than the 800-kDa L-PEI for in vitro delivery of the chimeric molecules (5), compared with the unmodified carrier (27). The observed aggregation of the ON-PEI and L-PEI complexes at high loading doses was readily inhibited by complexing the molecules in 5% dextrose instead of 0.15 M NaCl. This simple modification maintains the complexes as small, discrete particles rather than the network-like clusters observed in NaCl.
Several ratios of the RNA/DNA phosphate to 25-kDa L-PEI plus PEI amine were tested for hepatocyte uptake in vitro. Both a 1:6 and a 1:10 ratio demonstrated intense nuclear and some cytoplasmic labeling with the fluorescent ONs. However, uptake of the molecules by the ASGP R was completely inhibited in the presence of 100 mM D-galactose only at the 1:6 ratio. These data suggest that the decreased positive charge ratio of the 1:6 ON phosphate to amine induces binding by the ASGP R , while the increased cationic nature of the 1:10 complexes promotes internalization of the labeled ON complexes by both receptor-mediated as well as nonspecific pathways. This is supported by the data from the PGL3 plasmid studies, since the luciferase activity was markedly decreased only with the 1:6 phosphate:amine ratio in the presence of 100 mM D-galactose. Additionally, PEI lactosylation (sugar:amine ratio) of 0.1 or higher significantly increased delivery of the luciferase reporter plasmid PGL3 to isolated rat hepatocytes. Combining the L-PEI and PEI improved the transfection efficiency significantly over L-PEI alone, and a 1:1 molar ratio of amines from each form of PEI was determined to be optimal. The decreased transfection efficiency observed with the L-PEI alone may be due to slightly inefficient compaction of the plasmid DNA by the L-PEI. In fact, the degree of compaction of plasmid DNA by PEI, rather than ionic charge, directly affects the cytoplasmic to nuclear translocation of plasmid DNA (54). The efficiency of the 25-kDa L-PEI and PEI mix to promote expression of the PGL3 plasmid encoded luciferase in the nonreplicating rat hepatocytes may be due, in part, to the ability of this cation to promote efficient delivery of the plasmid DNA from the cytoplasm to the nucleus. When injected directly into the nucleus, the complex of PEI with a plasmid did not prevent expression of the encoded DNA (54). However, our preliminary studies using fluorescently labeled L-PEI and the chimeric molecules suggest that PEI does not enter the nucleus with the RNA/DNA ONs (data not shown).
The ultimate success of any gene therapy approach requires prolonged systemic expression of therapeutic levels of the delivered transgene. Although modification of hepatic gene expression using replication-deficient adenoviruses has been successful, there continues to be significant immunogenicity and lack of sustained expression of the transgene in the absence of repetitive treatments (55)(56)(57). The alternative approach provided by the RNA/DNA technology, i.e. repairing a liver genetic defect in situ by site-specific nucleotide exchange, may significantly overcome some of these challenges. To this end, the in vivo results with the 25-kDa L-PEI⅐ON complexes demonstrated that a mean 48% A to C conversion in the genomic DNA for the rat factor IX gene in liver could be achieved using this system. This genomic sequence alteration produced a phenotypic change resulting in a ϳ45% decrease in factor IX coagulant activity. Moreover, the phenotypic expression of this targeted genomic base exchange was still observed 46 weeks after the injection of the chimeric molecules. Finally, the genomic A to C conversion induced by the chimeric molecules is replicatively stable as demonstrated by the lack of significant change in the phenotypic expression of this trait following partial hepatectomy. In support of these observations, RNA/DNA ONs were recently shown to correct a point mutation in the tyrosinase gene of melanocytes derived from albino mice (58). Those results demonstrated a permanent and stable gene correction by clonal analysis, restoring enzyme activity and melanin production.
In conclusion, three liposomal systems and a novel ligandbased carrier were developed for targeted gene delivery to the ASGP R in liver. Each system was effective in mediating nucleotide exchange in genomic DNA with the RNA/DNA chimeric molecules. Anionic liposomes and a 25-kDa lactosylated PEI product were the most efficient vehicles in vitro. The latter was also shown to be highly effective for in vivo delivery, which may in part be due to its ability to promote cytoplasmic to nuclear translocation of nucleic acids in nonreplicating cells (54). In a separate study, we have corrected the genetic mutation for the UDP-glucuronosyltransferase-1 gene in the Gunn rat model of Crigler-Najjar syndrome type I. The results obtained in vivo with both L-PEI and anionic liposomes were essentially identical, indicating that both systems are equally effective for delivery of the chimeric ONs to hepatocytes (59). The ease with which these systems can be prepared and characterized makes them attractive formulations for therapeutic delivery of gene products.