If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Investigator of the Howard Hughes Medical Institute. To whom correspondence should be addressed: Howard Hughes Medical Institute, University of Iowa College of Medicine, 500 EMRB, Iowa City, IA 52242. Tel.: 319-335-7619; Fax: 319-335-7623
* This work was supported by the National Institutes of Health and the Howard Hughes Medical Institute. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore by hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ Parker B. Francis Fellow.
Cationic lipids are widely used for gene transfer in vitro and show promise as a vector for in vivo gene therapy applications. However, there is limited understanding of the cellular and molecular mechanisms involved. We investigated the individual steps in cationic lipid-mediated gene transfer to cultured cell lines. We used DMRIE/DOPE (a 1:1 mixture of N-[1-(2,3-dimyristyloxy)propyl]-N,N-dimethyl-N-(2-hydroxyethyl)ammonium bromide (DMRIE) and dioleoyl phosphatidylethanolamine (DOPE)) as a model lipid because of its efficacy and because it is being used for clinical trials in humans. The data show that cationic lipid-mediated gene transfer is an inefficient process. Part of the inefficiency may result from the fact that the population of lipid-DNA complexes was very heterogeneous, even under conditions that have been optimized to produce the best transfection. Inefficiency was not due to inability of the complex to enter the cells because most cells took up the DNA. However, in contrast to previous speculation, the results indicate that endocytosis was the major mechanism of entry. After endocytosis, the lipid-DNA aggregated into large perinuclear complexes, which often showed a highly ordered tubular structure. Although much of the DNA remained aggregated in a vesicular compartment, there was at least a small amount of DNA in the cytoplasm of most cells. That observation plus results from direct injection of DNA and lipid-DNA into the nucleus and cytoplasm indicate that movement of DNA from the cytoplasm to the nucleus may be one of the most important limitations to successful gene transfer. Finally, before transcription can occur, the data show that lipid and DNA must dissociate. These results provide new insights into the physical limitations to cationic lipid-mediated gene transfer and suggest that attention to specific steps in the cellular process may further improve the efficiency of transfection and increase its use in a number of applications.
Gene transfer could represent an important advance in the treatment of both genetic and acquired diseases. Thus, there has been increasing attention focused on the development of gene transfer vectors. Viral vectors, such as recombinant adenovirus vectors, have a number of advantages for gene transfer, including their efficiency and their wide range of cell targets(
). But they also have a number of disadvantages, including the fact that they can generate several types of immune response, they often contain viral genes which could be transcribed, and there is a possibility of recombination or complementation. As a result of such limitations, there has been substantial effort focused on nonviral vectors, particularly the use of cationic lipids(
). Cationic lipids are commercially available and are widely used in the research laboratory. For gene therapy applications, cationic lipids are currently under investigation as transfer vectors for treatments focused on melanoma (
Despite their availability, wide use, and potential application as vectors for gene therapy, there is limited understanding of the cellular and molecular mechanisms involved in cationic lipid-mediated gene transfer. As a result, wide variations in formulations and protocols have been described. For example, published reports of gene transfer to airway epithelia have used lipid to DNA charge ratios of 0.03 to 3 (net positive charge from the cationic component of the lipid divided by the net negative charge of the DNA) with the same or related compounds(
). Evaluation of cationic lipids usually involves comparison of different lipids and different formulations using expression of a transgene as the end point. However, such an approach is an empiric one that may provide little understanding of intermediate steps involved in transfection. Without knowledge of the cellular mechanisms of gene transfer and the limiting barriers involved, it will be difficult to take a rational approach to develop improved methods of gene transfer and it is difficult to test specific hypotheses related to cellular and molecular mechanisms.
The goal of this work was to evaluate some of the cellular mechanisms involved in cationic lipid-mediated gene transfer and to identify the steps that may impair transfer. As a starting point, we used DMRIE/DOPE
(a 1:1 mixture of N-[1(2,3-dimyristyloxy)propyl]-N,N-dimethyl-N-(2-hydroxyethyl) ammonium bromide (DMRIE) and dioleoyl phosphatidylethanolamine (DOPE)) as a model cationic lipid. In a systematic structure-activity analysis of a large number of lipids, Felgner and colleagues (
) recently identified DMRIE/DOPE as a promising vector for gene transfer. In a comparison of several lipids, we also found DMRIE/DOPE to be effective and have optimized a series of factors and conditions required for efficient gene transfer to HeLa cells and to canine airway epithelium(
). Our results identify a number of steps at which lipid-mediated transfection is inefficient. The identification of these steps provides the opportunity to further improve the process, thereby increasing its utility and application.
MATERIALS AND METHODS
Cells and Culture
HeLa cells were cultured in Eagle's MEM supplemented with 10% fetal calf serum, 10 mM nonessential amino acids, 100 units/ml penicillin, and 100 μg/ml streptomycin. COS-1 cells were cultured in DMEM (high glucose) supplemented with 10% fetal calf serum, 100 units/ml penicillin, and 100 μg/ml streptomycin. C127 cells were cultured in DMEM supplemented with 10% fetal calf serum and insulin (30 units/250 ml).
Plasmids and Vaccinia Virus
To assess expression in mammalian cells, we used a plasmid containing the luciferase cDNA driven by the RSV promoter, pRSV-Luc, or a plasmid in which a CMV promoter drove expression of β-galactosidase with a nuclear localization signal, pCMV-βGal. To evaluate expression in Xenopus oocytes, we used a plasmid (pMT3-SEAP) that encodes a secreted form of alkaline phosphatase (
). Plasmid DNA was purified on Qiagen columns (Qiagen Inc., Chatsworth, CA).
DMRIE/DOPE (50:50 molar ratio) was a gift from Dr. Phil Felgner, Vical Inc., San Diego, CA. Isatoic ester-labeled DMRIE was a gift from Dr. Eddy Lee, Genzyme, Inc., Cambridge, MA. The isatoic ester-labeled DMRIE/DOPE preparation was a 50:50 molar ratio of DMRIE (20 mol % isatoic ester-labeled DMRIE) with DOPE. The 4-chloro-7-nitrobenzo-2-oxa-1,3-diazole (NBD)-labeled DOPE was obtained from Avanti Polar Lipids, Alabaster. AL. The NBD-labeled DMRIE/DOPE preparation was a 50:50 molar ratio of DMRIE to DOPE (1 mol % NBD-DOPE).
To evaluate expression in the cytoplasm without the requirement for plasmid DNA to enter the nucleus, we used the vaccinia virus/T7 hybrid expression system. This system uses infection with a recombinant vaccinia virus that expresses the T7 RNA polymerase (vTF7-3) (
) and transfection with a plasmid in which the T7 promoter drives β-galactosidase expression (pTM-βGal). COS-1 cells were infected with vTF7-3 (m.o.i. of 5). After 1 h the incubation medium was removed, and the cells were transfected with increasing amounts of pTM-βGal (0.01-1 μg) complexes with DMRIE/DOPE at a 5:1 (w/w) ratio. The media was replaced 6 h after transfection, and the cells were incubated for an additional 10 h before analysis for β-galactosidase expression. As a control to test the efficiency of infection with vaccinia virus, a double infection with vTF7-3 plus vTF7-LacZ (ATCC) was performed. vTF7-LacZ encodes β-galactosidase under control of the T7 promoter.
Ethidium monoazide was coupled to DNA as described previously(
). To 200 μg of pCMV-βGal in 2 ml of H2O was added 5 μg of ethidium monoazide (Molecular Probes, Eugene, OR). This is a 50:1 molar ratio of nucleotide to probe. After a 10-min incubation period, the solution was exposed to UV light of principal wavelength 312 nm for 2 min. The solution was purified on a PD-10 column (Pharmacia Biotech, Uppsala, Sweden). To remove intercalated but not covalently bound ethidium, CsCl was added to a concentration of 1.1 g/ml and was gently mixed until it dissolved. Sodium citrate saturated isopropanol was then added and the upper phase, containing unbound ethidium, was discarded. The isopropanol washing was repeated until the upper phase appeared clear. The DNA in the bottom layer was then precipitated overnight at −20°C with 8 volumes of a 1:3 TE/absolute ethanol solution.
To a solution of 100 μg of pCMV-βGal in H2O we added 1.63 μg of photoactivatable biotin (Pierce). This solution was exposed to ultraviolet light of principle wavelength 312 nm for 2 min. The biotinylated plasmid was purified on a PD-10 column (Pharmacia Biotech) and eluted with 20 mM Hepes, 150 mM NaCl, pH 7.4. A solution of 29 μg of the biotinylated plasmid was incubated for 30 min with 1 ml of AuroProbe EM Streptavidin G1O (Amersham, Amersham, United Kingdom) in 20 mM Hepes, 150 mM NaCl, pH 7.4. The preparation was purified on a PD-10 column and subsequently dialyzed for 24 h against 20 mM Hepes, 150 mM NaCl, pH 7.4, using a 100,000 molecular weight cut off membrane (Instrumed Inc., Union Bridge, MD). Spectroscopic measurements suggest that there was approximately 90 ng of Au-streptavidin/μg of plasmid.
Markers of Endocytosis
We used Texas Red conjugated to 10-kDa dextran (D-1863, Molecular Probes, Inc., Eugene, OR.), human transferrin (T-2875), or wheat germ agglutinin (W831). 500 μl of these tracers were applied to cells in a 20 μg/μl (transferrin and wheat germ agglutinin) or a 10 μg/μl (dextran) solution in DMEM.
HeLa, C127, and COS-1 cells were seeded at 2-3 × 105 cells/35-mm dish the day before transfection. Plasmid and lipid were each diluted in 250 μl of Eagle's MEM. Lipid was added to the plasmid, mixed by inversion, and allowed to incubate 15-30 min at room temperature before being further diluted to a final volume of 1.5 ml. Cells were washed once with Eagle's MEM, and the lipid-DNA complex was added in a volume of 1.5 ml to each 35-mm dish. The transfection complex remained on the cells for 6 h, unless otherwise noted, and was then replaced with the appropriate media for the cell type. In some experiments, the lipid-DNA complex was left on the cells for 6 h and then 1.5 ml of media containing 2 × fetal calf serum (20%) was added for an additional 14 h. Cells were assayed at the times indicated.
Luciferase activity was assayed using a kit purchased from Promega (Madison, WI) and a luminometer (Monolight 2010, Analytical Luminescence Laboratory, San Diego, CA). Cells were removed from dishes by incubation with lysis buffer (25 mM Tris phosphate, pH 7.8, 2 mM dithiothreitol, 2 mM 1,2-diaminocyclohexane-N,N,N′,N′-tetraacetic acid; 10% glycerol, and 1% Triton X-100) for 15 min followed by scraping. A 4-μl aliquot from each 35-mm dish was used for one luciferase assay. Data for luciferase activity represent total values from all cells on one dish. In each experiment, three dishes were used for each condition.
Sixteen to 36 h after transfection, cells were fixed with 1.8% formaldehyde and 2% gluteraldehyde and then incubated for 16 h. at 37°C with 0.313 μl of 40 mg/ml X-gal (5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside) in Me2SO dissolved in 12.5 ml of PBS (pH 7.8). Blue staining of nuclei was evaluated by light microscopy. Results are expressed as a percentage; at least 1000 cells were counted in each experiment.
To assess the percentage of cells in which ethidium-labeled DNA entered the cell, we used fluorescence-activated cell sorting (FACS). Transfected cells were rinsed three times with PBS, released from the substrate by incubation for 10 min with 0.05% trypsin and 0.53 mM EDTA, and then resuspended in Eagle's MEM. After centrifugation at 2000 rpm for 5 min, the supernatant was discarded and the cells were resuspended in 500 μl of PBS. Fluorescence from 10,000 individual cells was analyzed (FACScan, Lysys II software, Becton Dickinson, San Jose, CA). The percentage of cells containing labeled DNA was assessed by determining the percentage of highly fluorescent cells in each group and subtracting the fluorescence of control cells that were exposed to the labeled DNA without the cationic lipid.
Transfected cells were thoroughly washed with PBS and scraped into 200 μl PBS. Cells were freeze/thawed three times to release the DNA. The suspension was spun at 14,000 × g for 1 min, and the supernatant was applied to a nitrocellulose filter with a vacuum blotter. The filters were exposed to UV light for 2 min to link the DNA to the filter and then were probed with a 32P-labeled plasmid. Results were analyzed by autoradiography, and radioactivity was quantitated with an AMBIS (San Diego, CA) radioanalytic scanner.
Cells labeled with ethidium-DNA and the Texas Red probes were imaged using a Bio-Rad MRC-600 laser scanning confocal microscope equipped with a krypton/argon laser. The ethidium-DNA fluorescence was most efficiently excited using 488 nm illumination, while the Texas Red labeled samples were imaged using the 568 nm line. Transmitted light images of each field were also collected. The images were transferred to a UNIX workstation for merging, annotation, and printing. The DMRIE-isatoic ester labeled samples were photographed using a Zeiss IM-35 with a 4,6-diamidino-2-phenylindole filter set. The images were digitized and were also transferred to the workstation for merging, annotation, and printing.
To study endocytosis of other markers, COS-1 cells were cultured on collagen-coated four-well (2 cm2 each) glass slides. Following a DMEM wash, 500 μl of each fluorescent probe solution was placed on the cells for 4-6 h. The cultures were then washed in DMEM, covered with COS media, and returned to the incubator. At 24 h. the cells were washed in PBS, pH 7.4, and fixed in 2% freshly prepared formaldehyde in PBS for 15 min. The slides were then rinsed three times in PBS and once in double-distilled H2O and, following removal of the wells, mounted in Gel/Mount (Biomeda Corp., Foster City, CA).
Lipid-DNA complexes were processed for transmission electron microscopy (TEM) using a negative stain/rotary shadow technique. Fifteen-μl drops of freshly prepared samples were placed on glow-discharged collodion/carbon-coated 400-mesh copper grids for 3 min. Solution was wicked off with filter paper and replaced with 1% aqueous uranyl acetate for 30 s. After removal of the solution, grids were rinsed in double-distilled H2O and allowed to dry. Rotary shadowing was performed using 1 inch of 0.008-inch Pt/Pd 60/40 wire at a 7° angle. Grids were imaged in a Hitachi H-7000 TEM.
To follow the cellular entry and fate of DNA, COS cells were transfected with gold-labeled DNA complexed with DMRIE/DOPE at a 1:5 (w/w) ratio. Cells were fixed at various times in 2.5% gluteraldehyde and processed using standard EM 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 stained with uranyl acetate and Reynold's lead citrate.
To identify lysosomes, we localized acid phosphatase(
). After treatment with the lipid-DNA-gold complex, cells were fixed in 2% gluteraldehyde in 0.1 M cacodylate buffer, pH 7.2, at 4°C for 1 h. Cultures were then rinsed three times for 10 min in 0.1 M cacodylate buffer at room temperature, rinsed three times for 10 min in Tris-maleic buffer, pH 5.0, at 37°C, and incubated in reaction solution consisting of 0.25% sodium β-glycerophosphate and 0.08% lead nitrate in Tris-maleic buffer for 1 h at 37°C. Cells were rinsed in Tris-maleic buffer and cacodylate buffer as above, post-fixed in 2% osmium tetroxide for 1 h, and processed for TEM as described previously.
Studies in Oocytes
Ovarian lobes were surgically removed from adult albino Xenopus laevis females. The isolated oocytes were rinsed in Ca2+-free modified Barth's solution (MBS), and the follicles were removed using collagenase at 2 mg/ml (Sigma, Type 1A). The defolliculated oocytes were maintained in MBS at 17°C prior to injection(
). Healthy oocytes were injected with 1.5-10 nl of either plasmid (pMT3-SEAP) alone or plasmid complexed with DMRIE/DOPE using a pressure microinjector (Narishige USA, IM-200). The concentration of plasmid injected was constant at 0.03 μg/μl for all nuclear and cytoplasmic injections. The concentration of DMRIE/DOPE was varied as indicated. The lipid and plasmid were mixed in TE buffer and incubated at room temperature for a minimum of 15 min before injection. Each injected oocyte was placed into a 96-well flat-bottom culture plate containing 0.2 ml of MBS and incubated at 17°C for 3 days. The amount of alkaline phosphatase secreted into the culture medium for each injected oocyte was assayed as described elsewhere(
), transfection was optimal when the charge ratio of DMRIE/DOPE to DNA was slightly positive, i.e. when the charge ratio was approximately 1 to 1.2, corresponding to a weight:weight ratio of total lipid-DNA of approximately 5:1. To assess the structure of this complex, we used electron microscopy with negative staining and rotary shadowing of the complexes. This method has been modified from the Kleinschmidt method that is classically used for electron microscopic imaging of DNA (
) in order to avoid excessive shearing, to prevent changes in lipid configuration that could result from exposure to alcohol, and to avoid changes in the complex that might result from adding high concentrations (0.4 mg/ml) of another cationic molecule, cytochrome c.
Although the preparation had been optimized for transfection, we were surprised to find a very heterogeneous population of complexes. Fig. 1A shows free DNA that has not been complexed with lipid, and Fig. 1(B-F) show examples of the type of lipid-DNA particles we observed. In some cases the DNA appeared to be compacted into relatively dense particles, but as shown in Fig. 1B, compacted and free DNA could be observed in the same field. Often when dense aggregates were observed, DNA also appeared to extend from the complex, forming looped structures (Fig. 1E). In some cases the DNA appeared to be free (Fig. 1D), and in other cases it may have been coated with lipid to form an extended structure (Fig. 1E). In some cases, relatively large aggregates appeared to form (Fig. 1C), and less frequently we saw strands of complexes (Fig. 1F). The complexes were quite heterogeneous, although the most frequently observed complexes resembled those in Fig. 1(B and E). In all cases the complexes appeared to be at least 100 nm or larger, at least in one dimension.
) have also imaged lipid-DNA complexes formed from calf thymus DNA or plasmid DNA complexed with an equimolar mixture of N-1-[2,3-bis(oleoyloxy)]propyl]-N,N,N-trimethylammonium chloride (DOTMA) and phosphatidylethanolamine. At a lipid-DNA charge ratio of 1.0, they showed an electron photomicrograph of a rod-like complex approximately 700 nm long, which most closely resembled the structures indicated by solidarrows in Fig. 1B of the present study. However, there was no indication of the substantial heterogeneity that we observed. Besides the difference in lipid, one additional difference between our study and theirs is that they used the Kleinschmidt method(
), whereas we used a much different method to place the sample on the grid.
At present we do not know whether one or all of the various forms of complex shown in Fig. 1 is the most efficient transfection particle. However, methods designed to produce homogeneous complexes and to identify the complex(s) that is most effective at transfection could provide an important advance in improving gene transfer.
Entry of Cationic Lipid-DNA Complex into Cells
The first step of transfection is entry of the DNA into the cells. To evaluate this step, we complexed DMRIE/DOPE with DNA that had been covalently labeled with ethidium monoazide. We exposed cells to the complex for varying intervals, removed the complex by rinsing, released the cells from the substrate, and then used FACS to determine the percentage of cells that had taken up the labeled DNA. Fig. 2 shows histograms of fluorescence intensity versus number of COS cells following increasing durations of exposure to DMRIE/DOPE•DNA (5:1, w/w ratio). After 30 min of exposure, less than 5% of the cells showed fluorescence above background. However, with increasing duration of exposure, the percentage of fluorescent cells increased, suggesting that the process of DNA entry into the cells is relatively slow. The low frequency of highly fluorescent cells at this short time point suggests that the rinsing procedure was successful in removing extracellular or membrane-bound DNA. In these experiments cells were washed with PBS to remove lipid-DNA complex. In additional experiments in which cells were incubated with lipid-DNA complex for 6 h and then rinsed with PBS plus phospholipase D and DNase, we did not remove additional DNA. Attempts to remove any extracellular complex by acid washing or to prevent entry by incubation of cells with lipid-DNA for 6 h at 4°C led to detachment of cells from the dish. These data plus the experiments described below suggest that most of the complex was internalized by 6 h. These results are consistent with data showing that total cell-associated plasmid increases with time of exposure to cationic lipid(
). The findings are also consistent with previous observations that prolonged exposure to lipid-DNA complex increased the level of transgene expression, and that there was little expression following a 30-min incubation(
To further evaluate efficiency of uptake, we used three different cell types: COS, HeLa, and C127. Fig. 3 shows that more of the COS and HeLa cells took up ethidium-labeled DNA than did C127 cells, suggesting cell type-dependent variability in lipid-DNA uptake. To learn whether the lipid-DNA uptake by different cells paralleled expression of transgene, we measured the percentage of cells showing nuclear localized β-galactosidase activity after transfection with pCMV-βGal and we measured luciferase activity after transfection with pRSV-Luc. Fig. 3 shows that both measures of expression paralleled the uptake of DNA; COS and HeLa cells showed more expression than did C127 cells. These data are consistent with previous observations that the efficiency of cationic lipid-mediated transfection varies for different cell types(
). More importantly, the correlation between the percentage of cells taking up DNA and the percentage of cells expressing transgene indicates that in some cells lipid-DNA uptake may be an important barrier to transfection. However, they also suggest that additional barriers to transfection may be responsible for differences in efficiency observed with different cell types.
The fact that a large percentage of cells took up some DNA seemed encouraging. To more accurately assess the efficiency of the uptake step, we measured the amount of DNA that was in the cells. To do this we exposed cells to the lipid-DNA complex for varying intervals of time, then removed the complex by rinsing and measured the amount of intracellular DNA by dot blot with a probe to the plasmid DNA. Fig. 4 shows the results in COS cells. At 5 min the amount of DNA in the cells was very small, but the amount increased progressively with time. These results are consistent with our data using ethidium-labeled DNA and suggest that covalent modification with ethidium monoazide did not substantially alter the ability of DNA to enter the cell. The dot blots show that after 6 h of exposure the cells had taken up approximately 60% of the DNA that was added (n = 6). These data are consistent with the previous observation that when NIH 3T3 cells were exposed to DOTMA/DOPE•RNA, 20-30% of the RNA became RNase-resistant(
) makes interpretation of those results less clear.) By standardizing the dot blots to known amounts of DNA, we estimated the absolute amount of DNA that entered the cells. At the start of the experiment we added 2 μg of DNA to COS cells, and 6 h later we found that the cells had taken up 1.2 ± 0.1 μg (n = 3). We calculate that on average each cell took up 2.95 × 105 plasmids. It is likely that a large percentage of the cells contained a very large number of plasmids. However, our expression data obtained under similar conditions showed that less than 50% of the cells expressed the transgene. Thus, steps subsequent to uptake may be important impediments to transfection.
Mechanism of Entry into the Cell
Although a large amount of the lipid-DNA complex entered the cell, the mechanisms by which it did so are not well understood. To evaluate this process, we used electron microscopy. In order to identify the DNA, the plasmid was labeled with gold particles before it was complexed with lipid. Fig. 5(A-F) shows representative examples of the entry process. At early times (Fig. 5, A and B), the DNA-lipid complex appeared as an electron-dense particle at the cell surface. Then, as the duration of incubation increased, the complex was taken up into the cell by an endocytic process. Once in the cytoplasm, the labeled complex appeared to be contained within vesicles or endosomes. We did not find gold or electron dense complexes free in the cytoplasm. We obtained similar results with DNA that was not gold-labeled (Fig. 5F), suggesting that the labeling itself did not influence the process of cell uptake and disposition. We also obtained similar results in HeLa cells.
Uptake of complexes predominantly by endocytosis is not what we had expected. It has often been assumed that cationic lipid-mediated transfection results from fusion of the positively charged complex with the plasma membrane resulting in direct entry of the lipid into the cytoplasm(
). However, several observations have suggested that endocytosis may be involved. The ability of chloroquine to enhance cationic lipid-mediated transfection in some cases has been interpreted to suggest that this agent, which increases endosome pH and prevents endosome-lysosome fusion, may aid escape of the complex from the endosome(
). Although our data do not allow us to exclude the possibility that some of the DNA may have entered the cytoplasm directly or that a different cationic lipid might produce a different mechanism of entry, these results suggest that DMRIE/DOPE•DNA complexes enter COS and HeLa cells primarily through endocytosis.
Intracellular Disposition of Lipid-DNA Complexes
Evidence that the lipid-DNA complex enters the cell via endocytosis immediately raises questions about its intracellular fate. To investigate this issue and to use an independent method to support what we had observed by electron microscopy, we labeled each component of the complex and evaluated the cellular location with confocal microscopy. Twenty four h after adding the complex, we examined fluorescence from ethidium-labeled DNA (Fig. 6A), isatoic ester-labeled DMRIE (Fig. 6B), or NBD-labeled DOPE (Fig. 6C). In each case, by 24 h the fluorescence had coalesced and was observed predominantly in discrete foci in the perinuclear area. Although we were not able to evaluate fluorescence from labeled lipid and labeled DNA in the same experiment (because fluorescence from ethidium-labeled DNA was too weak for colocalization), the same perinuclear pattern and the same time course of accumulation suggested that it was the lipid-DNA complex that was accumulated and not just a single component of the complex. We obtained similar results in HeLa cells. When we exposed cells to fluorescently labeled transferrin and dextran, agents that are endocytosed and delivered to the lysosomal compartment(
), we found a similar pattern of fluorescence at 24 h (Fig. 6, D and E). This pattern suggests that the complex was endocytosed and delivered to a perinuclear compartment where the endosomes or vesicles fused to generate large aggregates. We observed a similar pattern in HeLa cells. Of note, we did not observe fluorescent DNA in the nucleus.
We also used electron microscopy and gold-labeled DNA to assess the fate of lipid-DNA complexes 24 h after application to the cells. Fig. 7(A-C) shows that the gold-labeled DNA remained in vesicles that were often found in a perinuclear location. Moreover, the vesicles were often very large, much larger than anything we observed at 1-6 h (Fig. 5). This result is consistent with the appearance of discrete areas of perinuclear fluorescence observed with fluorescently labeled DNA and lipid (Fig. 6, A-C). These results suggest that the lipid-DNA complex was endocytosed and moved toward the nucleus where the endosomes fused, and coalesced into large membrane-bound vesicles. Of note, all of the gold-labeled DNA was observed within the membrane-bound vesicular complexes; none was observed free in the cytoplasm or in the nucleus.
The appearance of the lipid-DNA was interesting in that it often developed a highly ordered pattern. Fig. 7D shows a frequently observed regular lamellar pattern with a periodicity of approximately 3.2-4.5 nm. Fig. 7E shows an example of what may represent this pattern in cross-section. The appearance is one of a series of regularly packed tubules. The lumen of the tubule was approximately 6.5 nm in diameter, and the center-to-center distance between tubules was approximately 17.5 nm. One way to explain the regular appearance would be to assume that a strand of DNA is surrounded by a bilayer or in some cases a tubular monolayer of lipid. Such arrangements could give the regularly shaped appearance described above and could account for the interaction of lipid and DNA.
We considered the possibility that the lipid-DNA complex is contained within lysosomes. To identify lysosomes, we used acid phosphatase enzyme-cytochemistry. Fig. 8 shows an example of an electron photomicrograph taken 24 h after application of the lipid-DNA complex. The figure shows that the gold-labeled DNA and the reaction product were present in different cellular compartments. This result suggests that the lipid-DNA was present in endosomes that did not fuse with the lysosomes. It is also possible that the presence of the lipid-DNA complex prevented fusion.
These results indicate that most of the lipid-DNA complex is endocytosed and retained in the perinuclear area. However, because treatment of cells with lipid-DNA complex can lead to transgene expression, at least some of the DNA must escape from the endosomal compartment. Our inability to detect gold-labeled DNA free in the cytoplasm or nucleus might reflect the fact that not all of the DNA was labeled and some unlabeled DNA was able to escape from endosomes; it could be that labeling with gold prevented DNA from escaping from the endosome, or it could reflect the fact that very little DNA escaped from the endosome and the sensitivity of electron microscopic detection is not high enough to detect it. The fact that we are led to similar conclusions with results with three different techniques (light microscopic evaluation of fluorescently labeled DNA and fluorescently labeled lipid, EM evaluation of gold-labeled DNA, and quantitation of DNA uptake by dot blot) serves to validate the methods and strengthen the conclusions. Thus, escape of DNA from endosomes is an important barrier to transfection.
It is interesting that escape of DNA from the endosome is also a major barrier for DNA delivery via transferrin-coupled to DNA-polylysine complexes(
). In that system, addition of adenovirus to enhance escape from the endosomal compartment improved transfection. Likewise with cationic lipid-mediated DNA transfer, treatment of cells with adenovirus (200 plaque-forming units/cell) increased the efficiency of transfection 2-7-fold(
J. Zabner, A. J. Fasbender, T. Moninger, K. A. Poellinger, and M. J. Welsh, unpublished observation.
With only this modest increase in expression, the qualitative methods we have used are not able to identify the mechanisms by which adenovirus enhanced expression. It may be that adenovirus enhanced escape from the endosomes, but other mechanisms are possible, including binding of lipid-DNA complex to the adenovirus or enhancement of DNA entry into the nucleus (see below).
Percentage of Cells with Cytoplasmic DNA
Our FACS analysis of cells exposed to ethidium-labeled DNA complexed to lipid indicated that most COS cells contain DNA. However, less than 50% of cells were positive for β-galactosidase activity as assessed by X-gal staining. The studies described above indicate that much of the DNA remains in a vesicular compartment (Figure 6:, Figure 7:). To determine what percentage of the cells contain cytoplasmic DNA that is capable of expressing a transgene, we transfected cells with a plasmid containing the T7 promoter driving β-galactosidase expression (pTM-βGal) and infected them with a recombinant vaccinia virus that expresses the T7 polymerase(
). This system allows transcription of plasmid DNA in the cytoplasm without the requirement for transfer of DNA to the nucleus.
Fig. 9 shows that even when 0.01 μg of pTM-βGal was transfected with DMRIE/DOPE and cytoplasmic transcription was driven by recombinant vaccinia virus, most cells were positive for X-gal staining. By comparison, with 100-fold more of a plasmid (pCMV-βGal) that required nuclear delivery, only 10% of the cells were positive. We considered the possibility that vaccinia virus infection might have disrupted the endosomes, thereby releasing DNA into the cytoplasm or making it available for transfer to the nucleus. However, vaccinia virus infection did not increase expression from cells transfected with pCMV-LacZ (not shown). Moreover, evaluation of the cells by electron microscopy 16 h after the transfection/infection procedure revealed gold labeled DNA complexed to lipid in intact vesicles and endosomes, in addition to numerous intracellular viral particles (Fig. 10). After virus infection, we observed no free gold particles in the cytoplasm.
These data indicate that after transfection most cells contain at least some DNA in the cytoplasm. This conclusion is consistent with the observation of Gao and Huang(
), who delivered T7 RNA polymerase to cells along with cationic lipid and plasmid containing a T7 promoter driving CAT expression. They found that total transgene expression was greater than that observed with a plasmid that required nuclear expression; however, they made no assessment of the percentage of cells transfected. These data suggest that one of the most important barriers to transfection may be movement of DNA from the cytoplasm to the nucleus.
Entry of DNA into the Nucleus and Nuclear Transcription
To evaluate further the movement of DNA from the cytoplasm to the nucleus, we used the Xenopus oocyte model system in which we injected DNA directly into the nucleus or into the cytoplasm. We used a plasmid (pMT3-SEAP) encoding a secreted form of alkaline phosphatase, so that we could readily assay recombinant protein production by sampling the extracellular media and measuring alkaline phosphatase activity(
). Fig. 11A shows that when the plasmid was injected into the nucleus, alkaline phosphatase was secreted into the medium. However, when the same amount of DNA was injected into the cytoplasm, no expression of the reporter gene was observed.
This result indicates that DNA traffic from the cytoplasm to the nucleus is an inefficient process. This conclusion is consistent with Capecchi's observation (
) that injection of plasmid into the nucleus of a mouse cell line led to protein expression in over 50% of cells, whereas injection into the cytoplasm led to expression in <0.01% of cells. The fact that we observed transfection in mammalian cells treated with lipid-DNA, but not the oocyte may be in part due to the fact that the mammalian cells are dividing whereas the oocyte is stationary. In contrast to our data, an 18-bp oligonucleotide delivered to endothelial cells with DOTMA/DOPE accumulated in the nucleus(
). The reason for the difference between our results using a plasmid and the results with oligonucleotides most likely relate to the size of the DNA. Oligonucleotides may readily pass through nuclear pores, which have a diffusion limit of approximately 40,000 Da(
) suggested that cationic lipids may alter the intracellular distribution of oligonucleotides by increasing delivery to the nucleus, we asked whether the lipid-DNA complex might improve nuclear targeting of the plasmid and thereby increase expression. However, when we injected the complex into the cytoplasm at varying lipid-DNA ratios, alkaline phosphatase production was not substantially increased (Fig. 11B), suggesting that complexing plasmid with lipid did not improve transport to the nucleus.
We also asked whether complexing the DNA with the lipid would impair transcription when the complex is injected into the nucleus. Fig. 11C shows that when we used a lipid-DNA ratio of 5:1 (w/w), which was optimal for transfection, the production of secreted alkaline phosphatase was inhibited compared to injection of DNA alone. In contrast, when we used lipid-DNA (w/w) ratio of 1:1, which is suboptimal for transfection and in which there is much more uncomplexed DNA, the cells produced alkaline phosphatase. This result suggests that complexing DNA with lipid prevents expression of the encoded protein, probably because the DNA does not dissociate from the lipid and is not available for transcription. This observation is consistent with the finding that DNA complexed with cationic lipid is protected from DNase in vitro(
). Thus, dissociation of DNA from the lipid complex would appear to be another important variable limiting gene transfer and expression.
Summary of Barriers to Cationic Lipid-mediated Transfection
Our results suggest that the process of gene transfer by cationic lipids is inefficient. It was striking to see that on average COS cells took up approximately 3.3 × 105 plasmids/cell and yet less than 50% of cells showed evidence of transgene expression by X-gal staining. This result contrasts with the high efficiency of adenoviral vectors. For example, we found that with 1 or 10 infectious units of adenovirus/cell we were able to transduce approximately 20% and 90% (respectively) of airway epithelial cells(
). Here we have investigated the cellular mechanisms involved in cationic lipid-mediated gene transfer using DMRIE/DOPE and have identified several important barriers to gene transfer and expression. They include the following.
Formation of the Lipid-DNA Complex
One function of the cationic lipid is to complex with the DNA via electrostatic forces to compact the otherwise extended DNA structure, thereby allowing entry into the cell. Our data suggest that even at a lipid-DNA ratio that provides maximum transfection, the population of complexes is very heterogeneous, and it is not clear which form of complex is the most efficient at transfection. In order to increase the efficiency of transfection, methods of formulating a homogeneous population of complexes that can efficiently enter the cell seem critical. Such an effort may also be of value in terms of decreasing the amount of lipid required and thereby decreasing potential toxicity.
Entry of DNA into the Cell
Our data suggest that the major mechanism of entry of lipid-DNA complexes into COS and HeLa cells (at least for DMRIE/DOPE) is through the endocytic pathway. The results reported here do not support a major role for a fusion mechanism. For these cells the entry step was not an important barrier.
Escape of DNA from the Endosomes
We found that endocytosed complexes fuse into large aggregates that assume a perinuclear localization. The appearance of the complex in these endosomes is often that of a highly ordered tubular structure. The data suggest that efforts to facilitate escape of DNA from the endosome may be of considerable value in increasing the efficiency of gene transfer.
Dissociation of the DNA from Lipid
Although formation of a complex between DNA and lipid is important for cell entry, it also appears critical that after the DNA has entered the cell that it be released from the lipid for expression in the nucleus. Our data with nuclear injection of a lipid-DNA complex support the argument that when the DNA is bound and compacted it is not transcriptionally active.
Entry of the DNA into the Nucleus
Our data suggest that this step may be one of the most important limitations for successful gene transfer.
In the presence of these barriers, each of which can provide a major limitation, how is it that cationic lipids can mediate gene transfer and expression? Our data suggest that cationic lipid-mediated transfection is a rather inefficient process that proceeds through a mass action effect(
). A large amount of DNA is delivered to the cell, a small percentage of that is released from the endosomes, and a small percentage of that makes its way from the cytoplasm to the nucleus where it is transcribed. The inefficiency of each step suggests that specific attention must be paid to developing ways to overcome each of the different barriers. Although we know of no specific toxicity associated with delivery of a large amount of DNA to a cell, delivery of large amounts of lipid does have toxicity. Focusing attention on each of the barriers to gene transfer may allow a decrease in the amount of lipid required and hence reduce toxicity. If efficiency can be improved and lipid toxicity minimized, cationic lipids could be attractive vectors for diseases in which repeated administration is required.
In considering the various barriers encountered with lipid-mediated gene transfer, it is interesting to remember that viruses, such as adenovirus, have solved many of these problems. They bind to specific receptors for cell uptake, they have mechanisms for release of viral DNA from the endosome, and they have mechanisms to target the DNA to the nucleus(
). Perhaps a better understanding of the mechanisms used by a variety of viruses will allow us to adapt some of the advantages and features of viral systems and yet avoid their disadvantages in designing better nonviral vector-mediated gene transfer techniques.
Our data suggest that there is much opportunity for improving cationic lipid-mediated gene transfer and that as the process is improved, it could be used successfully in an even larger number of applications.
We thank Pary Weber, Aurita Puga, Terri McDonnell, and Theresa Mayhew for excellent assistance. We thank the DERC DNA Core (supported by National Institutes of Health Grant DK25295) for technical support. We thank Dr. M. Daniel Lane for the gift of pMT3-SEAP, Dr. Alan Smith for the gift of pCMV-βGal, Dr. Bernard Moss for the gift of vTF7.3, and Dr. Eddy Lee (Genzyme) for the gift of isatoic ester-labeled DMRIE. We thank Dr. Phil Felgner for the gift of the DMRIE/DOPE and for helpful discussions.
Ellis R.W. Vaccines: New Approaches to Immunological Problems. Butterworth-Heinemann,