Cationic liposomes coated with polyethylene glycol as carriers for oligonucleotides.

Modification of liposome surface with polyethylene glycol was used to improve oligodeoxyribonucleotide (ODN) loading, stability of the resulting complexes, and specificity of cellular delivery of ODN by cationic liposomes. Liposomes composed of a cationic lipid (DOTAP, DOGS, DDAB), a neutral lipid (DOPE), and a phospholipid derivative of polyethylene glycol (PEG-PE) formed a complex with 18-mer phosphorothioate up to ODN/lipid molar ratio of 0.25. The complexes showed intact vesicular structures similar to original liposomes and their size (100-130 nm) was unchanged after several weeks of storage, whereas complexes lacking PEG-PE showed progressive aggregation and/or precipitation. After exposure to human plasma, PEG-modified cationic liposomes retained over 60% of the originally bound ODN. PEG-coated complexes resulted in 4-13-fold enhancement of the ODN uptake by human breast cancer cells in serum-supplemented growth medium, relative to free ODN. Complexes containing conjugated anti-HER2 F(ab') fragments at the distal termini of PEG chains efficiently delivered ODN primarily into the cytoplasm and nuclei of HER2 overexpressing cancer cells and greatly enhanced the biological activity of antisense ODN. The development of PEG-modified cationic liposomes may lead to improved ODN potency in vivo.

cellular uptake and availability to hybridize to their RNA or DNA targets present in the cytoplasm and/or nucleus (1). Cellular uptake of ODN, as well as their activity in cell cultures, can be improved by cationic liposomes (2)(3)(4), first used to improve cell transfection with plasmid DNA (5) and RNA (6). Such formulations, however, require substantial excess of cationic lipid over ODN (4) and are unsuitable for in vivo use because of inadequate particle size stability and increased risk of toxicity associated with the high dose of cationic lipid coadministered with the therapeutic dose of ODN. Furthermore, positively charged particulates are prone to nonspecific interactions with plasma proteins and negatively charged cell surfaces. This may lead to destabilization, dissociation, and rapid clearance of the liposome-ODN complexes by the mononuclear phagocyte system before reaching the diseased tissue.
PEG-PE, a phospholipid derivative of the hydrophilic polymer polyethylene glycol (M r ϭ 2,000), increases particle stability of small unilamellar phosphatidylcholine liposomes in plasma and reduces their interaction with biological macromolecules and/or cell surfaces (7). PEG-PE containing sterically stabilized liposomes have long circulation lifetimes, show high accumulation in pathological foci, such as tumors (7,8), and are successfully being used as a delivery system for the cytotoxic drug doxorubicin in human patients (9,10). More recently, studies have shown accumulation in mice tumor of encapsulated antisense ODN using sterically stabilized liposomes (11). Specific delivery of sterically stabilized liposomes to target cells can be achieved by attachment of target-seeking molecules, such as antibodies or their fragments, preferably at the distal ends of liposome-associated polyethylene glycol chains (12). However, the use of this system for the delivery of ODN may be limited by its low entrapment efficiency and difficulty for the entrapped ODN to leave the liposome interior.
Recently, it was found that incorporation of PEG-PE can stabilize the particles of cationic liposome-plasmid DNA complexes (13). Therefore, we hypothesized that incorporation of PEG-PE into cationic liposomes will increase particle stability of liposome-ODN complexes and make them more attractive for medical use. Using an 18-mer phosphorothioate as a model for any type of biologically active ODN, we show here that modification with PEG-PE not only prevents aggregation and dramatically increases particle stability of cationic liposome-ODN complexes but also allows extraordinarily high loading of ODN, resulting in a particle with overall neutral or negative charge. We further demonstrate that complexes of ODN with PEGmodified cationic liposomes have good stability in human plasma, and retain ultrastructural characteristics of original liposomes. We also show that PEG-coated cationic liposomes enhance the cellular uptake of ODN in serum-supplemented cell growth medium and that conjugation of antibody frag-ments recognizing the HER2 receptor on such cationic liposome-ODN complexes triggers efficient delivery of ODN to their sites of action (cytoplasm/nucleus) resulting in enhanced biological activity of antisense ODN in a target-specific manner.

EXPERIMENTAL PROCEDURES
Preparation of PEG-modified Cationic Liposomes and Liposome-ODN Complexes-DOPE, DDAB, and DOTAP were from Avanti Polar Lipids (Alabaster, AL). DOGS was a gift from Dr. Francis Szoka, Jr., UCSF). PEG-PE was a gift from Sequus Pharmaceuticals, Inc. (Menlo Park, CA), and all phosphorothioate 18-mer bcl-2 antisense ODN (5Јtct-ccc-agc-gtg-cgc-cat-3Ј) and corresponding FITC-labeled and 35 S-labeled ODN were provided by Genta, Inc. (San Diego, CA). Other reagents were from Sigma unless indicated otherwise. Liposomes containing cationic lipid, DOPE, PEG-PE, and Rh-PE (molar ratio 25: 25:3:0.1) were prepared by repeated freeze-thawing followed by extrusion through 0.05-m pore membranes (14). Liposome-ODN complexes were produced by overnight incubation of the liposomes with ODN in HBS (20 mM HEPES-Na, 144 mM NaCl, pH 7.2) at 4°C with gentle stirring. Unbound ODN was removed by gel-exclusion chromatography on Sepharose 4B (Amersham Pharmacia Biotech). For cell culture studies, the preparations were filter sterilized through 0.45-m pore membrane (Corning, NY). Particle size was determined by dynamic laser light scattering using unimodal and differential size distribution processor analysis (Coulter Model N4 particle size analyzer).
Preparation of HER2-targeted PEG-modified Cationic Liposome-ODN Complexes-PEG-modified cationic liposomes bearing anti-HER2 F(abЈ) at the distal ends of PEG chains were prepared by a modification of our earlier method (15). DOTAP/DOPE/PEG-PE/MP-PEG-PE/Rh-PE liposomes (25:25:2:1:0.1 molar ratio) were prepared as above except that all procedures were performed in MES-buffer (20 mM MES, 2 mM EDTA, 144 mM NaCl, pH 6). F(abЈ) fragments of the recombinant anti-HER2 monoclonal antibody (rhuMAbHER2) (Genentech, Inc, South San Francisco, CA), obtained as reported previously (16), were added to the liposomes at 0.13 mg of protein per each 10 mol of lipid, and incubated at pH 7.3-7.4 (NaOH) for 2 h at room temperature under inert atmosphere. The reaction was stopped by blocking excess maleimido groups with 2 mM ␤-mercaptoethanol. Unconjugated protein was removed by gel-exclusion chromatography on Sepharose 4B, and liposomes with conjugated F(abЈ) fragments were loaded with ODN as above.
Quantitation of Liposome Lipids, ODN, and Liposome-conjugated F(abЈ)-Liposome concentration in stock preparations and ODN concentration in stock solutions were determined by phosphorus assay (17) and spectrophotometry (⑀ 260 ϭ 159, 160 M Ϫ1 ⅐cm Ϫ1 ), respectively. The amount of liposome-conjugated F(abЈ) fragments per liposome was determined after separation of the conjugates by SDS-polyacrylamide gel electrophoresis followed by Coomassie Blue staining. The bands (46 kDa for intact F(abЈ) and 48 -49 kDa for PEG-conjugated F(abЈ)) were cut out and extracted with 50% dimethylformamide, and the amount of protein per band was determined by spectrophotometry using simultaneously run F(abЈ) standards. In diluted samples containing octaethylene glycol dodecyl ether (C 12 E 8 ) (0.05% (w/v)), liposome lipids and ODN were assayed by fluorescence spectroscopy using Rh-PE and FITC-ODN, respectively.
Zeta Potentials-Electrophoretic mobility of liposomes and liposome-ODN complexes were measured in 2 mM Tris-HCl, pH 8, containing 10% sucrose (w/v), using a Zetasizer 4 (Malvern Instruments, Ltd., UK). Zeta potentials were calculated from the mean of three runs after adjustment to a negatively charged standard (AZ55, Malvern).
Freeze-fracture Electron Microscopy-Specimens were quenched rapidly for freeze-fracture electron microscopy using the sandwich technique and liquid propane (cooling rate Ͼ10 4 K/s). Cryo-fixed samples were fractured and shadowed in a Balzers BAF 400D freeze-fracture device at -120°C and 2 ϫ 10 Ϫ6 torr. The cleaned replicas were examined in a Zeiss CEM 902A transmission electron microscope (18).
Treatment with Plasma-Liposome-ODN complexes were incubated for 4 h at 37°C in citrate phosphate adenine-preserved recovered human plasma (Blood Transfusion Center, Moffit Hospital, UCSF) diluted with an equal volume of HBS. The samples were then chromatographed on a Sepharose 4B column, and collected fractions were assayed by fluorometry for liposome lipids and ODN. Proteins were determined by dye binding assay (Bio-Rad). Biological Activity of bcl-2 Antisense-Cellular expression of the bcl-2 protein was determined by ELISA (Endogen, Inc., Woburn, MA). Human breast carcinoma BT-474 cells were grown to subconfluency in 8-well plates and treated for 4 h with bcl-2 antisense either free or complexed with PEG-coated cationic liposomes, in 10% FBS-supplemented cell growth media, at a final antisense concentration of 2 M. At the end of incubation time, cells were washed, fresh medium was added, and cells were allowed to grow for an additional 44 h before bcl-2 expression determination. straight line, consistent with the model of simple saturation binding with independent binding sites, apparent K d ϭ 0.103 Ϯ 0.02 mol of ODN/mol of lipid, and maximum binding corresponding to charge equivalent complex. In the case of DOTAPand DDAB-containing liposomes, the slope decreased with increasing [ODN/LIP] bound , indicating the possibility of negative cooperativity of the binding or of multiple binding sites. Analysis of the curves according to Hunston (19) showed good agreement with the mechanism of negative cooperativity (K d,2 ϭ 0.51-0.56 mol of ODN/mol of total lipid, cooperativity parameter ϭ 4 -19) with maximum binding equivalent to approximately one molecule of ODN per molecule of a cationic lipid. Using DOTAP, the complexes with [ODN/LIP] bound ϭ 0.25 could be produced, which is equivalent to ODN content of approximately 70% by mass. Conjugation of antibody F(abЈ) fragments to PEG-modified cationic liposomes had no effect on the ODN binding.

Characterization of ODN Complexes with PEG-modified Cat
PEG-PE, as well as ODN loading substantially affected the surface charge of cationic liposomes. Incorporation of 6 mol% PEG-PE reduced the positive zeta potential of cationic liposomes from 25-30 mV to 7-10 mV. Increased loading of liposomes with ODN further decreased and eventually changed their zeta potential to negative values (Fig. 2). When [ODN/ LIP] bound increased from 0 to 0.03-0.04, zeta potentials decreased, passed neutrality, and reached a plateau at about -20 mV.
Freeze-fracture electron micrographs of DOTAP/DOPE/ PEG-PE liposomes (Figs. 3A and 4A) showed spherical vesicles with smooth fracture faces. About 80% of the vesicles were less than 100 nm in diameter while 15% were larger than 100 nm and smaller than 300 nm, with the rest slightly larger. The presence of conjugated F(abЈ) fragments had no effect on the structure and size of PEG-modified cationic liposomes or their complexes with ODN (Fig. 3, B and D). Vesicular structures and size were well preserved after the complex formation with ODN both at lower ODN load (Fig. 3C) or higher load (Fig. 4B). In all PEG-PE containing preparations, particles of the complexes were much smaller than in similar preparations without PEG-PE (data not shown). At low ODN loads, the presence of ODN seemed to transform small liposomes into an involuted shape (two of such structures are marked by arrowheads in Fig.  3C). In the preparations containing conjugated F(abЈ) (Fig. 3D) as well as at higher ODN/lipid ratio (Fig. 4B), some of the vesicles were aggregated (seldom more than three), forming slightly larger particles (up to 300 nm in diameter). However, the particle size of complexes was similar to that of the corresponding control liposomes within each preparation. For high ODN to lipid ratio, the size distribution of complexes seen on electron micrographs corresponds to the values of particle size obtained by differential (multimodal) dynamic light scattering analysis (Table I,  loads, the boundaries of the vesicles were smooth, continuous, and intact (Fig. 3, C and D), whereas at high ODN loads, the boundaries were often dotted and discontinuous (some of these areas are marked with arrowheads on Fig. 4B). Thus, the high amount of interacting ODN seemed to distort the normally smooth boundaries of the vesicles. These distortions resemble micellar structures attached to the surface of the lipid bilayer. However, the individual ODN strands are not visible on freezefracture electron micrographs because of the resolution limit (Ͼ2 nm) (18).
Liposomal vehicles for drug delivery must be sufficiently stable against the loss of drug load and particle aggregation. The stability against dissociation was defined as the ability of PEG-modified cationic liposomes to retain ODN molecules in a physiological medium, such as blood plasma. After incubation with 50% human plasma, the complex of ODN with PEGmodified DOTAP/DOPE-containing liposomes was chromato-graphed on Sepharose 4B. About 65% of the ODN and 100% of the lipid were co-eluted in the void volume fraction (as determined by the peak area), whereas the rest of ODN was coeluted with the broad peak of plasma proteins (Fig. 5).
Modification of cationic liposomes with PEG-PE remarkably increased the stability of ODN-liposome complexes against aggregation (Table I). Without PEG-PE, cationic liposomes mixed with ODN in HBS at ODN/lipid molar ratio of 0.007 formed large particles which further aggregated and completely precipitated within 4 days at 4°C. When ODN was in substantial excess over the cationic lipid ([ODN/LIP] bound ϭ 0.2), the particles were smaller and more stable to aggregation, although ultimately there was complete precipitation of the complex in 4 weeks. In contrast, the size of complexes of ODN with PEGmodified cationic liposomes did not change from its initial value of 110 -125 nm for at least 4 weeks at both low and high ODN/lipid ratios, according to unimodal cumulant results from dynamic light scattering measurements.

Interaction of Liposome-ODN Complexes with Cancer Cells-
The effect of PEG-modified cationic liposomes on the cellular uptake of ODN was studied in two human breast cancer cell lines: SK-BR-3, which overexpresses the HER2 oncoprotein (approx. 10 6 copies/cell), and MCF-7, a line with low expression of HER2 (approx. 10 4 copies/cell) (20). The uptake data are summarized in Table II   d Differential size distribution processor analysis based on particle number, particle weight, and light intensity gave two major distinct size populations of 65-80 and 200 -250 nm in diameter. excess of free anti-HER2 F(abЈ), the uptake of ODN decreased slightly (by 25%) and reached the level of target-negative (MCF-7) cells. The amount of cell-associated liposome lipid followed the same trend as that of ODN.
While immunotargeting of the liposome-ODN complex to HER2-overexpressing cells only slightly affected the total amount of cell-associated ODN, there was a pronounced effect on the intracellular distribution of ODN. Fluorescence microscopy of SK-BR-3 cells incubated with HER2-targeted liposome-ODN complexes revealed a large diffuse cytoplasmic localization of FITC-ODN as well as a high accumulation in the nucleus of every cell, with few additional cytoplasmic punctate patterns (Fig. 6, bottom left panel). Lipid distribution was different from that of ODN and was only seen at the cell surface and within intracellular organelles (Fig. 6, top left panel). Flu-orescent punctate cytoplasmic patterns are typical for endocytic vesicles (15,16). When SK-BR-3 cells incubated with HER2-targeted complexes in the presence of an excess of free F(abЈ), FITC-ODN was exclusively concentrated inside perinuclear vesicles (Fig. 6, bottom right panel), together with colocalized lipid (Fig. 6, top right panel). The same feature was observed after incubation of non-targeted complexes with SK-BR-3 cells or after incubation of HER2-targeted complexes with MCF-7 cells (not shown).
Effect of Liposome-Antisense Complexes on bcl-2 Expression- Fig. 7 shows the effect of cellular treatment with free and formulated bcl-2 antisense on the level of expression of bcl-2 protein in BT-474 cells. No antisense effect could be detected when ODN was associated with PEG-coated cationic liposomes (bar a). On the contrary, when the corresponding targeted ODN delivery system was used, a 46% inhibition of bcl-2 expression was observed (bar b). In cases where ODN antisense was delivered by HER2-targeted liposomes after pre-incubation with a 10-fold excess of free F(abЈ) fragments (bar c) or where free antisense molecules were used (bar d), only a 20% inhibition of bcl-2 expression was obtained.

DISCUSSION
In this study, we explored the possibility of forming PEGmodified cationic liposome-ODN complexes. We studied their physico-chemical characteristics, their stability in biological fluids, and the applicability of the above-cited targeting method in vitro in order to develop a suitable ODN delivery system for future in vivo applications.
Binding of ODN to cationic liposomes was very efficient even when containing a relatively high amount of PEG-PE (5.7 mol%) (Fig. 1). Evidently, PEG coating does not shield the positively charged liposome surface from interaction with ODN. Furthermore, such liposomes can bind an impressive amount of ODN without the loss of structural stability in a wide range of ODN/lipid charge ratios. When liposomes were composed of DOTAP or DDAB, cationic lipids bearing single positive charge, the amounts of bound ODN exceeded the charge equivalent and, with gradually decreased affinity, the binding continued until the physical limit of one ODN molecule per one positively charged lipid group. This mechanism af- forded exceptionally high loads of ODN in the complex without aggregation. Thus, after overnight incubation of PEG-coated DOTAP/DOPE liposomes with excess ODN, the resulting complex contained up to 0.25 mol of ODN per mol of total lipid (Fig.  1). Obviously, at such high ODN content the number of negatively charged groups in the complex far exceeded the number of available positive charges. This property of PEG-modified cationic liposomes is remarkable, since it has been shown (21) and confirmed by our present data (Table II) that ODN induce aggregation and fusion of cationic liposomes as a function of ODN/lipid charge ratio. Modification of cationic liposomes with PEG-PE allowed the formation of ODN complexes that are stable in size over prolonged storage (Table II). The stabilizing effect of PEG coating on the cationic liposome-ODN complex is likely to result from preventing cross-linking (and thus aggregation and precipitation) of the vesicles by ODN molecules whose stretched length approximates the extension length of the liposome PEG coating in "brush" regime (5-7 nm at Ն 4 mol% PEG-PE) (22,23), as well as avoiding bilayer contact needed for the aggregation to occur. Stabilization of cationic liposomes by PEG coating also contributed to high ODN-binding capacity by allowing the complex to pass through the charge neutralization point at increasing ODN/lipid ratio without aggregation.
Modification of cationic liposomes with PEG-PE not only increased particle stability after complexation with ODN, but also resulted in practically complete preservation of the original vesicles in the complex, as shown by freeze-fracture electron microscopy (Figs. 3 and 4). Preservation of the original small size of the vesicles after complexation with ODN may favor their better extravasation into the target tissue in vivo since the size limit for liposome extravasation ranges between 400 and 600 nm (24).
Addition of ODN even at low concentrations sharply modified the zeta potential of PEG-modified cationic liposomes to the saturation value of -22 to -25 mV, reached at approximately stoichiometric charge ratio of ODN to cationic lipid. The existence of this plateau suggests shielding effects of PEG on the excess negative charge carried by ODN nucleotides which are not associated with lipid positive charges. Although the pharmacokinetic behavior of PEG-modified cationic liposome-ODN complexes has not been completely characterized as yet (the study is in progress), the shielding effect of PEG on excess negative charges of liposome-bound ODN may have a beneficial effect on their ability to evade mononuclear phagocyte clearance.
PEG-modification increased the stability of cationic liposome-ODN complexes not only against aggregation, but also against dissociation and disintegration in the presence of plasma proteins. After incubation with plasma, practically all of the lipid and the major portion of the originally bound ODN was retained in the complex (Fig. 5). Due to the high initial load of ODN ([ODN/LIP] bound ϭ 0.2), the remaining ODN still constituted an impressive proportion of the complex, about 50% by mass.
Complexation of ODN with PEG-modified cationic liposomes enhanced the cellular uptake of ODN severalfold, similar to that observed with nonmodified cationic liposomes (2,4). In an attempt to increase the cellular ODN uptake in a targeted manner, we modified the complex with a targeting device previously developed for neutral liposomes (15). In the case of PEG-modified complexes, conjugated anti-HER2 F(abЈ) fragments further increased the cellular uptake of both lipid and ODN by a factor of 2.3-2.6, regardless of the cellular expression level of the target antigen. Since this enhancement was not completely abolished by preincubation of cells with free F(abЈ) that blocked the surface antigen, it must be attributed to additional nonspecific interactions which at present are poorly understood. The targeting effect, however, revealed itself in an unexpected way, namely, as dramatically increased diffuse cytoplasmic and nuclear localization of ODN in target-positive (SK-BR-3) cells. The green fluorescence has been attributed exclusively to the presence of ODN since it was shown that fluorescent ODN metabolites are quickly discarded from cells (25). Depending on their mode of action, ODN must either accumulate free within the cytoplasm to hybridize to their target mRNA, or in the nucleus when they are designed to be complementary to sites of transcription factor binding, splicing sites, nascent transcripts or to form a triple helix with DNA (26). Herein, we show that PEG-coated cationic liposome-ODN complexes conjugated to anti-HER2 F(abЈ) at their surface efficiently delivered ODN diffusely in the cytosol and to a larger extent in the nucleus of HER2-overexpressing breast cancer cells. Because there was no co-localization of ODN and lipid in these regions, ODN was presumably free of lipid, which is a requirement for their biological activity. When the specific interaction between the targeted complex and cell surface p185 HER2 was blocked or missing, FITC-ODN was exclusively localized in the endosomal/lysosomal compartment, evidence of the importance of antibody-antigen interactions for successful delivery of ODN to their sites of action. We address the dependence of the cellular disposition of ODN upon the targetability of the delivery system used, in terms of the ability for ODN to evade endocytic compartments. It has been shown that after endocytosis of cationic liposome-ODN complexes, the escape of ODN from early endosomes into the cytoplasm is a critical step for their biological activity (4). Therefore, HER2 receptor-mediated endocytosis could direct the targeted complex into early endosomes that may be more permissive for the escape of ODN into the cytoplasm than the early endosomes formed in response to non-target specific interactions of the cell membrane with the complex itself. Confirmation of this hypothesis as well as the determination of the exact mechanism by which HER2targeted complexes deliver ODN into the cytoplasm and nucleus of target-positive cells require further study.
Bcl-2 proto-oncogene encodes a 26-kDa protein that promotes cell survival by blocking programmed cell death (also termed apoptosis) (27). Bcl-2 protein has been detected in a wide variety of cancers including breast carcinomas (28). To test the cell specificity of our ODN delivery systems in terms of biological antisense activity, we took advantage of the human breast cancer cell line BT-474, which both expresses the bcl-2 protein (29) and overexpresses the HER2 receptor (20). Importantly, in a nontargeted complex bcl-2 antisense did not promote any detectable inhibition of the level of expression of the bcl-2 protein in BT-474 breast cancer cells, whereas when the targeted system was employed, a 46% inhibition of cellular bcl-2 level was found over untreated cells. However, in accordance to cell uptake experiments (see Table II), presaturation of the HER2 receptor with free F(abЈ) fragments did not block entirely the antisense activity of ODN when delivered by the HER2-targeted liposomes, where the level of bcl-2 expression reached about the same as when free antisense molecules were used. In view of these data, it seems that cell binding/internalization of positively charged particles, which involve both nonspecific (electrostatic) and specific (ligand-receptor) interactions (the complexes used had a net positive charge, [ϩ/Ϫ] ϭ 4:1) can only be rendered partially tissue-specific. Therefore, despite a drastic enhancement in uptake by target cells, in intracytosolic/intranuclear ODN accumulation, and in antisense activity observed when ODN was delivered by HER2targeted liposomes compared with non-targeted liposomes, we believe that there is still a need for additional developments (currently in progress in our laboratory) to further optimize the ODN delivery system in terms of cell specificity.
PEG-modified cationic liposomes described in this study offer a number of advantages as carriers for cellular delivery of ODN. They allow exceptionally high loads of ODN without loss of structural stability. The resulting complexes are stable against aggregation and retain both their ODN load in blood plasma and the ability to enhance cellular delivery of ODN in the presence of serum. The presence of PEG also limits the negative zeta potential observed on the surface of the complexes. Finally, PEG-modified cationic liposome-ODN complexes can be made to incorporate the previously developed liposome targeting technology, resulting in efficient delivery of ODN into the cytoplasm and nuclei of target cells as well as enhanced antisense activity. In conclusion, PEG-modified cationic liposomes may provide a useful step in the development of an efficient pharmaceutical carrier for systemic in vivo delivery of ODN.