DNA-Loaded Bacterial Ghosts Efficiently Mediate Reporter Gene Transfer and Expression in Macrophages

ARTICLE

doi:10.1016/j.ymthe.2004.09.024

DNA-Loaded Bacterial Ghosts Efficiently Mediate Reporter Gene Transfer and Expression in Macrophages Susanne Paukner,1,2,* Pavol Kudela,2 Gudrun Kohl,2 Tobias Schlapp,3 Sonja Friedrichs,4 and Werner Lubitz1,2 1

Institute of Microbiology and Genetics, Vienna University Biocenter, Dr. Bohrgasse 9, A-1030 Vienna, Austria 2 BIRD-C GmbH & Co KEG, Scho¨nborngasse 12, A-1080 Vienna, Austria 3 Bayer AG, Animal Health, Osterather Strasse 1a, D-50739 Cologne, Germany 4 Bayer AG, Animal Health, Alfred-Nobel-Strasse 50, D-40789 Monnheim, Germany

*To whom correspondence and reprint requests should be addressed at Sandoz GmbH, Brunner Strasse 59, A-1235 Vienna, Austria. Fax: +43 1 86659 785. E-mail: [email protected].

Available online 16 December 2004

There is a demand for efficient and safe DNA delivery vehicles mediating gene transfer and expression. We present bacterial ghosts as a novel platform technology for DNA delivery and targeting of macrophages. Bacterial ghosts are cell envelopes of gram-negative bacteria that are devoid of the cytoplasmic content. Escherichia coli ghosts were loaded with plasmid DNA and linear double-stranded DNA. Confocal laser scanning microscopy and flow cytometry confirmed that the DNA localized to the inner lumen of bacterial ghosts and was not associated with the outer surface of the bacteria. Up to ~6000 plasmids could be loaded per single ghost and the amount of loaded DNA correlated with the DNA concentration used for loading. E. coli ghosts loaded with plasmids encoding the enhanced green fluorescent protein (EGFP) targeted efficiently murine macrophages (RAW264.7) and mediated effective gene transfer. The EGFP was expressed by more than 60% of the macrophages as measured by flow cytometry detecting the green fluorescence and immunocytochemical staining with antibodies specific for EGFP. Key Words: bacterial ghosts, DNA delivery vehicle, RAW264.7, macrophages, EGFP expression, transfection

INTRODUCTION DNA as a therapeutic drug for gene therapy and vaccination is one of the most striking innovations in medical and veterinary sciences. However, the pharmaceutical application of DNA vector systems requires the transfer and the enhanced expression of the DNA-encoded protein, e.g., antigen [1,2]. Currently, viral and nonviral delivery systems are used [3]. Due to the ease of production, the lower toxicity, and the higher biological safety profile, nonviral delivery systems are likely to be favored in DNA vaccination and gene therapy [3,4]. These include polyplexes, lipoplexes, or lipopolyplexes; microparticles; cochleates; and attenuated live bacteria [4,5]. Here, we describe and characterize a novel nonviral DNA delivery and targeting vehicle, bacterial ghosts. Bacterial ghosts are nondenatured gram-negative bacterial cell envelopes devoid of cytoplasmic content and produced by the controlled expression of the plasmidencoded BX174 gene E [6]. Through the created trans-

MOLECULAR THERAPY Vol. 11, No. 2, February 2005 Copyright C The American Society of Gene Therapy 1525-0016/$30.00

membrane tunnel structure bacterial ghosts can be filled with proteins, drugs, DNA, and other water-soluble substances [7,8]. Bacterial ghosts efficiently target antigen-presenting cells [8–10] and other eukaryotic cells [11]. They are, therefore, potential carrier and targeting vehicles for DNA and drugs [12,13]. As a nonviral and nonliving delivery system with the capacity to be loaded with DNA, they are a safe alternative to liposomal and other particulate carrier systems. Macrophages are involved not only in the resolution of injuries and tissue remodeling [14] but also in the progression and onset of various diseases, including the growth and spread of, for example, malignant tumors, HIV infection, and inflammation in rheumatoid arthritic joints [5]. Macrophages have been proposed as cellular delivery vehicles for adoptive immunotherapy, as they localize to sites of inflammation and tumors, adhere to the endothelium, and transmigrate to the focus of injury [14]. Potential applications of macrophage transfection include gene-dependent enzyme

215

ARTICLE

prodrug therapy, the expression of cytokines for the stimulation of macrophage tumoricidal activity, the expression of IFN-h to mediate resistance to HIV infection, and the expression of scavenger receptors for the therapy of atherosclerosis [5]. There is, therefore, increasing emphasis on targeting of macrophages by DNA delivery vehicles and expression of transgenes in macrophages. As macrophages form a link between innate and adaptive immunity, the expression of DNA-encoded foreign transgenes might be an important approach in the treatment of inflammatory and autoimmune diseases [14]. Moreover, macrophages are professional antigen-presenting cells and therefore may play a crucial role in DNA vaccination [3]. Gene transfer to macrophages involves HIV-based vectors [5], adenoviruses [14], polylysinated mannose [15], and a variety of liposomal systems [5,16]. However, transfection of macrophages has proved to be difficult, because viral systems impose relatively low size limits (4.5–8 kb per particle) and lack specificity for macrophages [5]. Furthermore, nonviral systems are not efficient, e.g., transferrin lipoplexes, or raise safety concerns, e.g., attenuated strains of Leishmania sp., Listeria sp., and Salmonella sp. [5,16]. As macrophages efficiently internalize bacterial ghosts [8], we studied their efficacy to mediate transgene expression in the murine macrophage cell line RAW264.7. We describe a method for loading Escherichia coli ghosts with plasmid DNA encoding the marker gene enhanced green fluorescent protein (EGFP) and our investigation of the localization of the DNA associated with the bacterial ghosts and quantification of the loaded DNA by quantitative realtime PCR.

RESULTS Fluorescence Microscopy and Flow Cytometric Analysis of Bacterial Ghosts Loaded with pEGFP-N1 We loaded the E. coli ghosts with the plasmid pEGFPN1 encoding the EGFP under the control of the strong eukaryotic CMV promoter or alternatively with FITClabeled nonhomologous archaeal phage BCH1-derived linear, double-stranded DNA (dsDNA; 400 bp). This was performed by resuspending the lyophilized bacterial ghosts in the DNA-containing buffer. We stained the pEGFP-N1 loaded and washed bacterial ghosts with the DNA intercalating dye SYBR Green I. Flow cytometry revealed that the bacterial ghosts were loaded with pEGFP-N1, as the mean fluorescence intensity (MFI) of the E. coli ghosts loaded with approximately 2000 plasmids per ghosts was 28.8, whereas that of empty ghosts was 10.2 (Fig. 1a). Fluorescence microscopy of E. coli ghosts loaded with the FITC-labeled archaeal, nonhomologous, linear dsDNA confirmed that N96% of the bacterial ghosts were loaded with DNA and that

216

doi:10.1016/j.ymthe.2004.09.024

FIG. 1. (a) Flow cytometric analysis of E. coli ghosts filled with pEGFP-N1 and stained with the DNA intercalating dye SYBR Green I. In the histograms (fluorescence intensity vs. frequency) it is shown that the bacterial ghosts filled with plasmid DNA, ~2000 plasmids per single ghost, show a distinct shift in fluorescence (FL1 log) compared to the empty bacterial ghosts as also indicated by the mean fluorescence intensities (MFI). (b) E. coli ghosts filled with nonhomologous phage-derived, FITC-labeled, linear dsDNA (400 bp). Overlay of differential interference contrast and fluorescent image. Arrows indicate bacterial ghosts filled with linear dsDNA (630 original magnification).

the DNA was stably associated with the bacterial ghosts (Fig. 1b), as it was not removed by washing once. Localization of Loaded Nonhomologous Archaeal Phage BCH1-Derived Linear dsDNA in Bacterial Ghosts To investigate if the DNA was associated with the interior or with the outer surface of the bacterial ghosts, we compared the DNA loading capabilities of (i) E. coli ghosts and (ii) living intact and (iii) lyophilized intact E. coli. For this, we used nonhomologous archaeal phage BCH1-derived linear, dsDNA (400 bp) fluorescently labeled with FITC for loading of the bacterial ghosts

MOLECULAR THERAPY Vol. 11, No. 2, February 2005 Copyright C The American Society of Gene Therapy

doi:10.1016/j.ymthe.2004.09.024

ARTICLE

and the intact cells. Flow cytometry revealed that the E. coli ghosts could be loaded with the DNA (Fig. 2). This was indicated by a distinct fluorescence shift of the ghost population compared to the control ghosts (empty) and by the threefold higher MFI. In contrast, the fluorescence of lyophilized, intact and alive, intact E. coli did not increase (Fig. 2), demonstrating that they could not be loaded with DNA. This indicated an interaction of the DNA with the inside of the bacterial ghosts rather than with the outer surface, as the inside was accessible to DNA only through the lysis tunnel structure of the bacterial ghosts. To visualize the localization of the loaded DNA, we loaded sulforhodamine B-labeled bacterial E. coli ghosts with the nonhomologous FITC-labeled linear dsDNA and investigated by confocal microscopy (Fig. 3a). Several zscan sections (0.122 Am) through the bacterial ghosts showed that the FITC-labeled dsDNA (green) was located within the bacterial ghosts (red), as shown in the overlay of the fluorescence microphotographs of one middle zscan section (Fig. 3a). To exclude the possible influence of the fluorophore on the DNA’s binding affinity to the bacterial ghosts, we loaded unlabeled E. coli ghosts with pEGFP-N1 (~2000 plasmids per ghost). Subsequently we detected the loaded plasmids by fluorescence in situ hybridization (FISH) using EGFP-specific Cy3-labeled probes. The bacterial ghost membranes were stained with MitoTracker Green FM (Fig. 3b). Analysis of several z-scan sections through the bacterial ghosts by confocal laser scanning microscopy revealed that the pEGFP-N1 (red) was associated with the interior of the bacterial ghosts (green) and not bound to the outside (Fig. 3b), confirming the results obtained with the FITC-labeled, linear dsDNA. Loading of E. coli Ghosts with pDNA—Optimization of the Loading Procedure To optimize the loading procedure, we loaded the E. coli ghosts with the pEGFP-N1 under various conditions, including incubation temperature, incubation time, and the DNA concentration. The DNA loads of bacterial ghosts (~35 ng/mg) were not significantly altered by incubation at temperatures of 4, 24, or 378C. A 2-min incubation time was sufficient for the loading of the bacterial ghosts; longer incubation times (up to 120 min) did not increase the pEGFP-N1 loading, as the DNA load stayed constant. The DNA concentration used for the loading was positively correlated (R 2 = 0.981) with the

FIG. 2. DNA loading capability of (i) E. coli ghosts; (ii) intact, lyophilized E. coli; and (iii) intact, living E. coli. The fluorescence (FL1 LOG) of the bacterial ghosts and intact bacteria (solid line) measured by flow cytometry was compared to that of the unloaded controls (dashed line). Successful loading resulted in the fluorescence shift of the E. coli ghost population indicated by the MFI, whereas the fluorescence of intact cells of living or lyophilized E. coli remained unchanged.

MOLECULAR THERAPY Vol. 11, No. 2, February 2005 Copyright C The American Society of Gene Therapy

217

ARTICLE

doi:10.1016/j.ymthe.2004.09.024

bacterial ghost can be loaded. As the molecular weight of the plasmid was 3.09  106 g/mol, 1 ng of pEGFP-N1 corresponded to 1.945  108 plasmids. On average, a concentration of 5 Ag pEGFP-N1 per milligram (dry wt) of bacterial ghosts corresponded to approximately 2000 plasmids per ghost. In this study the highest value was approximately 2200 plasmids per E. coli ghost (Fig. 4).

amount of pEGFP-N1 recovered from the bacterial ghosts (Fig. 4). We converted the measured pEGFP-N1 load of the bacterial ghosts (Ag/mg) to the average plasmid numberper ghost using the number of bacterial ghosts counted by flow cytometry and assuming that every

Stable Association Between E. coli Ghosts and Loaded DNA To measure the stability of the association between the loaded plasmid DNA (pDNA) and the bacterial ghosts, we pelleted the loaded E. coli ghosts and further washed them five times with Hepes-buffered saline (HBS). We quantified the pEGFP-N1 content in the bacterial ghosts and in the washing buffer by real-time PCR (Fig. 5). The initial amount of pDNA in the ghosts pellet comprised the loaded pDNA together with the residual pDNA in the interghost space. Approximately two-thirds of pDNA—mainly pDNA of the inter ghost space—was removed by one washing (not shown in Fig. 5), resulting in the absolute DNA amount of 150 F 6 Ag pDNA per 100 Al ghost suspension (corresponding to 471 F 107 Ag pEGFP-N1/mg ghosts). After the second washing ~75% of the original ghost-associated pDNA remained associated with ghosts, whereas the rest, most likely the DNA in the interghost space, was removed. In the following two washes, only 9.6 and 12.4 Ag (6.5%) of pDNA were removed. By the fifth washing only 1% of the loaded

FIG. 4. Correlation of DNA loading solution and DNA recovered from loaded ghosts. The direct correlation of the DNA concentration used for loading with the amount of pEGFP-N1 recovered from washed loaded E. coli ghosts is indicated by the linear regression line (line of the best fit) and its associated R 2 of 0.9811. Each point is the mean of quadruplicate measurements F standard deviation and each experiment was repeated at least twice.

FIG. 5. Effects of repeated washing on the DNA content of E. coli ghosts loaded with pEGFP-N1. The bars show the absolute amount of DNA (Ag) associated with the bacterial ghosts and the absolute amount of DNA (Ag) measured in the supernatant (striped bars) after the number of washing steps indicated on the x axis.

FIG. 3. Localization of loaded DNA in E. coli ghosts visualized by confocal laser scanning microscopy. (a) Sulforhodamine B-labeled E. coli ghosts (red) were loaded with nonhomologous FITC-labeled linear dsDNA (green). (b) Unlabeled E. coli ghosts loaded with pEGFP-N1 (~2000 plasmids/ghost). pEGFP-N1 was detected by in situ hybridization with Cy3-labeled probes (red) specific for the EGFP, and ghost membranes were stained with MitoTracker Green FM (green). All fluorescence photomicrographs (i, ii) were taken of one middle zscan section through the middle plane of the loaded bacterial ghosts and subsequently overlaid (iii). Direct overlays of red and green fluorescent structures are represented by the yellow color.

218

MOLECULAR THERAPY Vol. 11, No. 2, February 2005 Copyright C The American Society of Gene Therapy

ARTICLE

doi:10.1016/j.ymthe.2004.09.024

pDNA was removed from the pellet. The loaded pDNA thus remained stably associated with the bacterial ghosts, whereas the pDNA of the interghost space was removed by simple repeated washing. Delivery of Loaded pEGFP-N1 to and EGFP Expression in Macrophages We investigated the uptake of the bacterial ghosts loaded with the nonhomologous, FITC-labeled, linear, dsDNA by the murine macrophages by fluorescence microscopy (Fig. 6). We observed ingested single bacterial ghosts loaded with fluorescent DNA as well as a cluster of internalized ghosts together with released DNA in the macrophages (Fig. 6). As the loaded DNA was delivered to and released within the macrophages, we further investigated the transfection efficiency mediated by E. coli ghosts loaded with plasmid DNA pEGFP-N1 (~970 plasmids per ghost) encoding the EGFP. We allowed the macrophages to take up the loaded bacterial ghosts (at a ghost-to-cell ratio (GCR) of 500) for 2 h and 48 h later examined them by fluorescence microscopy. The macrophages exhibited green fluorescence in their cytoplasm (Fig. 7). Further, we confirmed EGFP expression by macrophages by flow cytometric measurement of the green fluorescence (FL1 log) compared to the control cells (Fig. 8, top). This resulted in a fluorescence shift of the whole cell population relative to controls (Fig. 8, top). The percentage of EGFP-expressing cells, estimated using the contour blots (side scatter vs fluorescence (FL1)), was 63.1 F 0.5% EGFP-positive cells with E. coli ghosts (Fig. 8, top). In addition, the EGFP was detected by anti-EGFP

FIG. 7. Fluorescence microphotograph of RAW264.7 macrophages expressing EGFP as a result of ingestion of pEGFP-N1-loaded E. coli ghosts. RAW264.7 cells were incubated for 2 h with E. coli ghosts loaded with approximately 970 plasmids per ghost at a cell-to-ghost ratio of 500. Subsequently the unattached and not internalized bacterial ghosts were removed and the macrophages were allowed to express the EGFP for 48 h. The photomicrograph is an overlay of the fluorescence image and the transmission light image.

primary antibodies and R-phycoerythrin (R-PE)-conjugated anti-IgG secondary antibodies (FL2 log). Compared to the control cells, 58.5 F 3.8% of the cells were EGFP positive (Fig. 8, bottom), confirming the results obtained by measurement of the green fluorescence. Fluorescence was thus due to the expression of the EGFP and not to autofluorescence.

DISCUSSION

FIG. 6. Fluorescence microphotograph of RAW264.7 macrophages with internalized E. coli ghosts loaded with FITC-labeled, nonhomologous archaeal phage-derived linear dsDNA (indicated by arrows). The macrophages were allowed to internalize the loaded bacterial ghosts at a cell-to-ghost ratio of 1:500 for 2 h until the unbound ghosts were removed, and the macrophages were examined by fluorescence microscopy. The photomicrograph is an overlay of the fluorescent image and transmission light image.

MOLECULAR THERAPY Vol. 11, No. 2, February 2005 Copyright C The American Society of Gene Therapy

We have shown that (i) E. coli ghosts can be loaded with DNA, namely pDNA and nonhomologous linear dsDNA, (ii) the DNA load correlated directly with the DNA concentration that was used for loading, and (iii) the maximal pDNA loads (5000–6000 plasmids per ghost) were obtained only with highly concentrated (~35 mg/ ml) and very pure pEGFP-N1 devoid of RNA and protein impurities. Confocal microscopic and flow cytometric studies revealed that the pDNA as well as linear dsDNA was associated with the interior of the bacterial ghosts and not bound to the outer surface. Since the net surface charge of outer bacterial membranes is negative [20,21] due to phosphate groups of the LPS core polysaccharide, binding of the negatively charged DNA is unlikely. In preliminary experiments the pDNA loads correlated with pH and Ca2+ concentration (data not shown), indicating an electrostatic interaction of the loaded DNA with positively charged moieties of the bacterial ghost interior, e.g., amine groups. According to Brown et al. [1], amine groups and positive charges are essential for the transfection competence of other delivery systems like cati-

219

ARTICLE

doi:10.1016/j.ymthe.2004.09.024

FIG. 8. EGFP expression by the RAW264.7 macrophages after ingestion of E. coli ghosts loaded with pEGFP-N1. The overlays of the histograms show a distinct shift in fluorescence of the cells treated with pEGFP-N1loaded bacterial ghosts (solid line, filled gray) compared to cells treated with the bacterial ghosts alone (dashed line, unfilled) due to either (top) the green EGFP fluorescence (FL1) or (bottom) the red fluorescence (FL2) from the immunochemical detection with polyclonal rabbit anti-EGFP primary antibodies and R-PE-conjugated anti-rabbit secondary antibodies. Numbers indicate the percentage of the EGFP-expressing cells that exhibit brighter fluorescence (higher FL1 and FL2 values) than the control cells, which was estimated by quantitative analysis of contour blots showing side scatter (SSC) vs. FL1 or FL2. (Experiments were independently repeated at least twice.)

onic lipids or polymers. Possible effects of the DNA sequence on the loading capacity are unlikely, because the bacterial ghosts could also be loaded with linear nonhomologous Archaea phage (BX174)-derived dsDNA and plasmids other than pEGFP-N1 (data not shown). While the DNA loading level was not sequence dependent, preliminary experiments indicated that there might be an upper plasmid size limit, as very large plasmid sizes (greater than three times the size of pEGFP-N1) resulted in lower DNA loads. Still the absolute size limit remains to be determined. However, it is likely that with decreasing DNA size the DNA load would increase. As shown in a previous study, the macrophages (RAW264.7) can be targeted efficiently by the bacterial ghosts, with 89% of the cells internalizing fluorescencelabeled bacterial ghosts that released a reporter substance [8]. Here, we showed that the DNA loaded in bacterial ghosts is delivered to the macrophages (Fig. 6) and that the plasmid-encoded reporter gene EGFP is expressed in about 60% of cells. Whether the transfection efficiency could be further increased by the use of a linear vector instead of plasmids remains to be determined. However, by comparison other DNA delivery systems that are used to transfect macrophages, e.g., lipid-based transfection reagents, result in poor transfection efficiencies (b5%), even when optimized with transferrin [22]. Our transfection efficiencies are comparable to those (85–95%) of monocyte-derived macrophages after infection with Salmonella typhimurium carrying pDNA [23]. As the use of live S. typhimurium as a DNA carrier is likely to raise safety concerns, bacterial ghosts are a promising alternative, with comparable transfection efficiencies in macrophages. A comparison with transfection efficiencies of other DNA delivery vehicles is difficult, as usually only the total luciferase activity of transfected cells is given rather

220

than the percentage of transfected cells. This is a common approach in transfection experiments, but makes it impossible to assess the proportion of cells transfected, as high levels of reporter gene activity could arise from a relatively small proportion of the target cells [5]. How the DNA escapes the phagoendolysosome in which the bacterial ghost is entrapped after receptormediated phagocytosis remains unknown. The DNA escape is probably mediated by the phosphatidylethanolamine present in the E. coli ghosts’ membrane [24], as was reported for DNA delivered by dioleoylphosphatidylethanolamine liposomes [25]. Regardless, we have shown that the DNA has entered the nucleus and effected transcription of the encoded reporter gene, as the EGFPspecific mRNA was detected in the cytoplasm by RNA FISH (data not shown). Bacterial ghosts represent promising DNA delivery vehicles for DNA vaccination, as Pasteurella haemolyticaghosts loaded with pEGFP-N1 mediate efficient EGFP expression in human monocyte-derived dendritic cells (52% of cells transfected [26]) and humoral immunity could be raised in a murine vaccination study in which mice were immunized with ghosts containing DNA [26]. Taking into account that, e.g., P. haemolytica ghosts are excellent vehicles to target colon carcinoma cells [27] and various other cell types, such as melanoma cells (unpublished data), transfection efficiencies might be as high. However, this remains speculation until exact data are provided since the mechanism of DNA release and transfection within the targeted cells could be completely different. The possibility of packaging up to several thousand plasmids (each ~5 kb) into bacterial ghosts is an advantage for this type of DNA vaccine delivery, compared to restrictions on gene or plasmid size reported for other

MOLECULAR THERAPY Vol. 11, No. 2, February 2005 Copyright C The American Society of Gene Therapy

doi:10.1016/j.ymthe.2004.09.024

carrier systems, e.g., for VP-1 virus-like particles (1.5–1.8 kb) [5,28]. Furthermore, bacterial ghosts have been loaded with plasmids harboring unmethylated CpG motifs, which have been reported to be strong adjuvants in DNA vaccination triggering innate immunity [29]. Moreover, preformed immune responses against the bacterial ghost envelopes as a result of a preceding infection may even enhance the uptake of the bacterial ghosts by macrophages mediated by the strong Fc receptor [30]. Bacterial LPS on the bacterial ghosts’ surface is not a safety concern as was clearly shown in previous studies. No significant fever responses due to endotoxicity are induced in rabbits [31] and the stimulatory dose for similar TNF-a production is 100-fold lower for bacterial ghosts than for free LPS [10]. In addition, the bacterial ghost delivery system can be easily scaled up for animal studies, and prepared ghosts could be stored at 808C for at least 4 months without a significant loss of loaded pDNA. In conclusion, we have demonstrated that bacterial ghosts loaded with pDNA efficiently target macrophages and mediate reporter gene expression and effect transfection of these cells at much higher rates than other nonliving, nonviral systems. Bacterial ghosts thus have great potential as a delivery system for gene therapy and DNA vaccination.

MATERIAL

AND

METHODS

Production of Bacterial Ghosts Bacterial ghosts from E. coli NM522 [sup E thi-1 D(lac-proAB) D(mcrBhssdS;)5(rK-mK-) (FVproABlacI q ZDM15] were produced by the controlled expression of the phage-derived lysis protein E as described elsewhere [7,8,17]. The nonlysed bacteria (b0.05%) were inactivated with gentamycin (50 Ag/ml) and streptomycin (100 Ag/ml) [8]. Subsequently the bacterial ghosts were washed three times with PBS (phosphate-buffered saline, pH 7.4), suspended in distilled water, lyophilized, and stored at 48C. Preparation of Plasmid DNA and FITC-Labeled Linear dsDNA for Loading The plasmid pEGFP-N1 (4.7 kb; Clontech, Palo Alto, CA, USA) was prepared by large-scale (4 L) overnight culture of E. coli using the standard alkaline method essentially described by Horn [18]. LPS, proteins, and RNA were removed by ammonium acetate precipitation and RNase I digestion. After the final precipitation of the DNA with ethanol, the DNA was dissolved in HPLC water (Sigma, Vienna, Austria) at a concentration of approximately 25 mg/ml and stored frozen at 208C. The plasmid DNA was shown to be free of RNA and chromosomal bacterial DNA by agarose gel electrophoresis and its concentration was determined by measurement of absorbance at 260 and 280 nm. The purity of the pDNA preparation was calculated from the ratio OD260/280 and only preparations with a ratio of 1.9–1.93 were used for loading of the bacterial ghosts. The fluorescence labeling of the nonhomologous linear dsDNA was performed by PCR. A randomly chosen 400-bp DNA fragment of the archaebacterial phage BCH1 was amplified with the 5V-FITC-labeled primers 5V-CGGCAGGTTTCATCCAGGAG-3V and 5V-TAACAGCACGCCGGAACTGA-3V (VBC Genomics, Vienna, Austria). The 50-Al reactions, containing 1.75 nM dNTPs, 0.5 AM each oligonucleotide primer, 1 Ag BCH1 DNA, and 1 U TaqDNA polymerase in polymerase buffer, were subjected to the following conditions: 4 min predenaturation at 948C; 35 cycles of 30 s denaturation at 948C, 30 s annealing at 608C, and 1 min

MOLECULAR THERAPY Vol. 11, No. 2, February 2005 Copyright C The American Society of Gene Therapy

ARTICLE

extension at 688C; 5 min at 48C final temperature. The amplified PCR products (linear dsDNA) were stored frozen at 208C protected from light. Loading of Bacterial Ghosts with pEGFP-N1 Standard loading procedure . Bacterial ghosts were loaded with pEGFPN1 by diffusion of plasmid DNA through the lysis holes into the ghosts. The lyophilized bacterial ghosts were resuspended in HBS (100 mM NaCl, 10 mM sodium acetate, 10 mM Hepes, pH 7) containing pEGFP-N1 (6–20 mg/ml) at a constant ratio of 10 mg lyophilized bacterial ghosts per 100 Al DNA solution. Subsequently, the suspension was supplemented with CaCl2 (final concentration 25 mM) and incubated for 10 min at 248C with agitation. The bacterial ghosts were then pelleted by centrifugation (11,600 g), washed once with 1 ml HBS, and stored aliquoted (~1 mg) as pellets at 808C for more than 4 months without any significant loss of the loaded DNA. The absence of dividing cells was confirmed by determination of the colony-forming units in 1 mg bacterial ghosts on Luria broth agar plates at 288C for 24 h. Optimization of the loading procedure. Incubation temperature, DNA concentration, and incubation time were varied systematically to optimize the loading procedure. For the optimization of the incubation temperature, the bacterial ghosts were incubated with the DNA (6.2 mg/ ml in HBS, pH 7, 25 mM CaCl2) at 4, 24, and 378C for 30 min until pelleted. For optimization of DNA concentration, bacterial ghosts were incubated with DNA concentrations ranging from 23.4 to 0.047 mg/ml in HBS (pH 7, 25 mM CaCl2) at 378C for 30 min. For incubation time, the lyophilized bacterial ghosts were incubated with the DNA (6.2 mg/ml in HBS, pH 7, 25 mM CaCl2) at 248C for 2, 10, 30, 60, 120, and 180 min until pelleted. Stability of the pEGFP-N1–Ghosts Association To test the stability of the pEGFP-N1–ghosts association, the E. coli ghosts were loaded according to the standard loading procedure, pelleted by centrifugation, and washed five times with 1 ml HBS each. After each washing, an aliquot of 100 Al corresponding to 1 mg bacterial ghosts was withdrawn and centrifuged at 11,600 g for 5 min. The pEGFP-N1 in the supernatant and in the bacterial ghost pellet was isolated quantitatively using a mini preparation kit (Peqlab, Erlangen, Germany) and further quantified by real-time PCR as described below. Isolation of the Loaded Plasmids and Quantitation by Real-Time PCR The bacterial ghost-associated pEGFP-N1 was quantified by real-time PCR after plasmid isolation from the loaded ghosts (~1 mg dry wt). The isolation was done in duplicate with a mini preparation kit (Peqlab) that is based on a standard alkaline lysis method using columns for affinity purification. The plasmid DNA was eluted from the columns twice with 50 Al Tris–HCl (10 mM, pH 8) and stored at 48C. Evaluation of the mini preparation kit was done by parallel extractions of six DNA standards with known concentrations in duplicate followed by quantitation in duplicate. On average 84.6 F 12.7% (mean F SD) of the loaded DNA was extracted. For the plasmid quantitation real-time PCR was performed by the amplification of the EGFP gene (700 bp) using the Corbett Research RotorGene 2000 (Corbett Research, Mortlake, Australia). The amplification reaction was monitored with the DNA intercalating dye SYBR Green I (Roche, Mannheim, Germany). The oligonucleotides 5V-GGTGAGCAAGGGCGAGGAG-3V and 5V-TTACTTGTACAGCTCGTCCATG-3V (VBC Genomics) were used as forward and reverse primers. pEGFP-N1 purified with the mini preparation kit (Peqlab) was used for calibration. RNA and chromosomal DNA in the pEGFP-N1 preparation was not detectable by agarose gel electrophoresis. The DNA concentration of the calibration pEGFP-N1 was quantified by measurement of absorption at 260/280 nm (ratio OD260/280 1.90–1.93). Aliquots of the calibration DNA (~250 Ag/ml) were stored at 208C. To obtain a standard curve at least five serial dilutions (10 2 to 10 6) of the pEGFP-N1 were prepared. Each sample and standard was quantified in duplicate by real-time PCR. Two reactions with water instead of template served as blank. Each 25-Al reaction included DyNAzyme II DNA polymerase (Finnzymes, Finland; 1 U), the DyNAzyme reaction buffer, dNTPs (0.4 mM each), oligonucleotide primers (2 pmol),

221

ARTICLE SYBR Green I (final dilution of 1 in 105), and template DNA (25–2.5 ng). The thermal cycling conditions were as follows: an initial denaturation of 948C (3 min), followed by 26 cycles of denaturation at 948C for 45 s, annealing at 608C for 1 min, and extension at 728C for 1 min. Fluorescent signal data were acquired at the end of each extension phase. This was followed by a 3-min hold at 728C and a melt curve analysis from 72 to 998C. The final melt curve analysis and agarose gel electrophoresis showed that there were no artifacts like primer dimers present in the amplification mixture. For quantitative analysis, only standard curves with a correlation coefficient of at least 0.998 were accepted. As the pDNA isolation and the real-time PCR were each done in duplicate, the mean value and the standard deviation for each sample were calculated from four values. Loading of Bacterial Ghosts, Lyophilized Intact Bacteria, and Living Bacteria with FITC-Labeled Linear dsDNA To show that bacterial ghosts, but not intact or living bacteria, could be loaded with DNA, each was incubated with FITC-labeled linear dsDNA (in HBS, pH 7). For this, 1010 lyophilized E. coli ghosts (~10 mg), 1010 lyophilized intact E. coli cells, and 1010 living E. coli (in the logarithmic growth phase) were incubated with 100 Al FITC–dsDNA at 248C for 10 min. The bacterial ghosts and bacteria were then washed once with 1 ml HBS (pH 7), suspended in 0.5 ml HBS, and analyzed by flow cytometry using the flow cytometer Coulter EpicsXL (Coulter Corp., Miami, FL, USA). Forward scatter (FSC) and side scatter (SSC) were processed in logarithmic gain. Gating of the bacteria and bacterial ghosts was performed by measurement of FSC vs SSC and the fluorescence (FL-1) of the FITC–dsDNA-loaded bacterial ghosts was measured. Fluorescence Labeling of the Bacterial Ghosts and Loading with FITC-Labeled DNA For the localization of the pDNA, the E. coli ghosts were labeled with sulforhodamine B (10 mM in HBS; Sigma–Aldrich, Steinheim, Germany) at 378C for 1 h. The bacterial ghosts were then washed with HBS (on average five times) until no sulforhodamine B was detected photometrically in the washing solution. Subsequently the labeled bacterial ghosts were loaded with the nonhomologous FITC-labeled, linear dsDNA derived from the phage BCH1 by incubation of ~10 mg ghosts in 100 Al DNA solution at 248C for 10 min. The bacterial ghosts were then harvested, washed once with 1 ml HBS, and mounted for confocal microscopy. In Situ Hybridization of Loaded pEGFP-N1 with Cy3-Labeled Probes A FISH was performed to localize the loaded pEGFP-N1 plasmids within the bacterial ghosts. We used Cy3-labeled probes (700 bp) specific for EGFP. For this, the EGFP encoded by pEGFP-N1 was amplified using the same EGFP primers and PCR conditions as in the quantitative PCR protocol. In addition to the dCTPs, Cy3 3–dCTPs (Amersham Biosciences, Buckinghamshire, UK) (final concentration of 0.04 mM) were used. The hybridization was performed as described before [19]. Briefly, paraformaldehyde-fixed bacterial ghosts were affixed to poly-l-lysine-coated glass slides, blocked with salmon sperm DNA, transferred to ethanol baths at increasing concentrations, and incubated with the probes for 16 h. Finally, the washed bacterial ghosts were stained with 0.25 Ag/ml MitoTracker Green FM (Molecular Probes, Leiden, The Netherlands) and examined by confocal microscopy. Confocal Laser Scanning Microscopy and Fluorescence Microscopy The sulforhodamine B-labeled E. coli ghosts loaded with the FITC-labeled linear dsDNA (400 bp) and the pEGFP-N1 loaded E. coli ghosts after FISH were examined at a magnification of 1000 using a confocal laser scanning microscope (Leica DMIRES; Leica, Heidelberg, Germany) equipped with an He/Ne laser (Leica) exciting the FITC at 488 nm, the sulforhodamine B at 563 nm, and the Cy3 at 543 nm. The detection of the FITC fluorescence was performed at wavelengths in the range of 565–590 nm, that of the sulforhodamine B in the range of 580–620 nm, and that of Cy3 in the range 650–680 nm using a black/white scanner (Leica CTCMIC; Leica). The scans through the bacterial ghosts in the z axis (z-scans) had a maximal distance of 0.122 Am. The scans were recorded on video, which showed the position of the FITC–DNA within the E. coli ghosts. Fluorescence microscopy was performed using an epifluorescence microscope (Axioplan; Zeiss, Vienna, Austria) and photos were taken

222

doi:10.1016/j.ymthe.2004.09.024

with a black/white LCD camera and further processed with Metaview software. The overlays of the fluorescence photomicrographs were further processed with Adobe PhotoShop 5.0 software (Adobe, San Jose, CA, USA). Cell Culture The murine macrophage cell line (RAW 264.7) was cultured in Roswell Park Memorial Institute (RPMI) medium 1640 with l-glutamine (Gibco BRL, Invitrogen, Groningen, The Netherlands) containing 50 Ag/ml gentamycin. The cells were cultured in humidified atmosphere (90%) with 5% CO2 at 378C. Twelve hours prior to the addition of the bacterial ghosts, 1  105 cells were seeded in 24-well plates. Delivery of Loaded DNA to Murine Macrophages To show that the DNA packaged into bacterial ghosts was delivered to and released within the macrophages, 1  105 macrophages/well were seeded on glass slides in 24-well plates 12 h prior to the analysis. Subsequently, the bacterial ghosts loaded with FITC-labeled, nonhomologous linear dsDNA were added to the culturing medium (RPMI) at a cell-to-ghost ratio of 1:500 and were allowed to attach to and be internalized by the macrophages for 2 h. The unbound bacterial ghosts were then removed by washing the macrophages with PBS, and the macrophages were examined by fluorescence microscopy as described above. Transfection Experiments and Immunocytochemistry For the transfection experiments, 1  105 macrophages/well were seeded in 24-well plates 12 h prior to the transfection. Transfections were performed in complete medium (200 Al per well) containing the bacterial ghosts loaded with pEGFP-N1 at a GCR of 500. The average plasmid load of the bacterial ghosts was 970 plasmids per E. coli ghost. The bacterial ghosts were allowed to adhere to and be taken up by macrophages for 2 h. Subsequently the cells were washed with PBS to remove unbound bacterial ghosts. After another 48 h incubation the transfection efficiency was determined. The cells were trypsinized, washed with PBS, and fixed with 4% paraformaldehyde and examined for EGFP expression by flow cytometry (FL-1). Cells treated with unloaded E. coli ghosts served as control. Data obtained by the measurement of the green fluorescence were confirmed by immunostaining. For this the transfected cells were suspended in 0.5 ml 1 FACS Permeabilizing Solution 2 (Becton–Dickinson Immunocytometry Systems, San Jose, CA, USA), incubated at room temperature for 10 min, and washed with FACS staining/washing buffer (0.5% FCS, 0.1% NaN3 in PBS). Immunostaining of the EGFP was performed with polyclonal rabbit anti-EGFP antibodies (1:200; Clontech) and R-PE-conjugated polyclonal goat anti-rabbit IgG (1:100; Molecular Probes) as the secondary antibody at 48C for 30 min each. The cells were then washed three times with FACS staining/washing buffer and fixed with 1% paraformaldehyde in PBS at 48C for 30 min. The R-PE fluorescence was measured by flow cytometry (FL-2) using the flow cytometer EPICS ALTRA Flow Cytometer (Beckman–Coulter, Miami, FL, USA). Cells incubated with unloaded bacterial ghosts and treated with the antibodies served as control.

ACKNOWLEDGMENTS We thank Prof. Dr. Franz Gabor (Institute of Pharmaceutic Technology and Biopharmacy, Vienna, Austria) and the Cancer Research Institute (Bratislava, Slovak Republic) for provision of the flow cytometers. Further we thank DDr. Hopmeier and Dr. Walter Krugluger at the Krankenanstalt Rudolfstiftung (Vienna, Austria) for supporting the completion of this study by providing the Corbett Research Rotor-Gene 2000 for real-time PCR. We also gratefully acknowledge Dr. Katri Jalava (Marie Curie Fellow, Ref. No. QLK3-GH-0060086-03) and Dr. Phil Cowan for reading and correction of the manuscript. This work was supported BIRD-C GmbH & Co KEG, Vienna, Austria. RECEIVED FOR PUBLICATION JUNE 10, 2004; ACCEPTED SEPTEMBER 26, 2004.

REFERENCES 1. Brown, M. D., Schatzlein, A. G., and Uchegbu, I. F. (2001). Gene delivery with synthetic (non viral) carriers. Int. J. Pharm. 229: 1 – 21.

MOLECULAR THERAPY Vol. 11, No. 2, February 2005 Copyright C The American Society of Gene Therapy

ARTICLE

doi:10.1016/j.ymthe.2004.09.024

2. Schleef, M. (2001). Plasmids for Therapy and Vaccination. Wiley-VCH Verlag, Weinheim. 3. Gurunathan, S., Klinman, D. M., and Seder, R. A. (2000). DNA vaccines: immunology, application, and optimization. Annu. Rev. Immunol. 18: 927 – 974. 4. Scha¨tzlein, A. G., and Uchegbu, I. F. (2001). Non-viral vectors in gene delivery. Drug Des. Delivery 1: 17 – 23. 5. Burke, B., Sumner, S., Maitland, N., and Lewis, C. E. (2002). Macrophages in gene therapy: cellular delivery vehicles and in vivo targets. J. Leukocyte Biol. 72: 417 – 428. 6. Witte, A., Wanner, G., Sulzner, M., and Lubitz, W. (1992). Dynamics of PhiX174 protein E-mediated lysis of Escherichia coli. Arch. Microbiol. 157: 381 – 388. 7. Huter, V., et al. (1999). Bacterial ghosts as drug carrier and targeting vehicles. J. Controlled Release 61: 51 – 63. 8. Paukner, S., Kohl, G., Jalava, K., and Lubitz, W. (2003). Sealed bacterial ghosts—novel targeting vehicles for advanced drug delivery of water-soluble substances. J. Drug Target. 11: 151 – 161. 9. Felnerova, D., et al. (2003). T-cell specific immune response induced by bacterial ghosts. Med. Sci. Monit. 10. Haslberger, A. G., et al. (2000). Activation, stimulation and uptake of bacterial ghosts in antigen presenting cells. J. Biotechnol. 83: 57 – 66. 11. Groger, M., et al. (2000). Dermal microvascular endothelial cells express the 180-kDa macrophage mannose receptor in situ and in vitro. J. Immunol. 165: 5428 – 5434. 12. Lubitz, W. (2001). Bacterial ghosts as carrier and targeting systems. Expert Opin. Biol. Ther. 1: 765 – 771. 13. Jechlinger, W., et al. (2002). Bacterial ghosts as carrier and targeting systems for antigen delivery. In Vaccine Delivery Strategies (G. Dietrich and W. Goebel, Eds.), pp. 76 – 98. 14. Kluth, D. C., Erwig, L. P., Pearce, W. P., and Rees, A. J. (2000). Gene transfer into inflamed glomeruli using macrophages transfected with adenovirus. Gene Ther. 7: 263 – 270. 15. Opanasopit, P., et al. (2002). Serum mannan binding protein inhibits mannosylated liposome-mediated transfection to macrophages. Biochim. Biophys. Acta 1570: 203 – 209. 16. Dokka, S., et al. (2000). High-efficiency gene transfection of macrophages by lipoplexes. Int. J. Pharm. 206: 97 – 104. 17. Marchart, J., et al. (2003). Protective immunity against pasteurellosis in cattle, induced by Pasteurella haemolytica ghosts. Vaccine 21: 1415 – 1422. 18. Horn, N. A., Meek, J. A., Budahazi, G., and Marquet, M. (1995). Cancer gene

MOLECULAR THERAPY Vol. 11, No. 2, February 2005 Copyright C The American Society of Gene Therapy

19.

20. 21. 22.

23.

24. 25.

26. 27.

28. 29. 30.

31.

therapy using plasmid DNA: purification of DNA for human clinical trials. Hum. Gene Ther. 6: 565 – 573. Pogliano, J., Ho, T. Q., Zhong, Z., and Helinski, D. R. (2001). Multicopy plasmids are clustered and localized in Escherichia coli. Proc. Natl. Acad. Sci. USA 98: 4486 – 4491. Beachey, E. H. (1981). Bacterial adherence: adhesin–receptor interactions mediating the attachment of bacteria to mucosal surfaces. J. Infect. Dis. 143: 325 – 345. Harkes, G., et al. (1992). Physicochemical characterization of Escherichia coli: a comparison with gram-positive bacteria. Cell Biophys. 20: 17 – 32. Simoes, S., et al. (1999). Transfection of human macrophages by lipoplexes via the combined use of transferrin and pH-sensitive peptides. J. Leukocyte Biol. 65: 270 – 279. Montosi, G., et al. (2000). Wild-type HFE protein normalizes transferrin iron accumulation in macrophages from subjects with hereditary hemochromatosis. Blood 96: 1125 – 1129. Wilkinson, S. G. (1988). Gram-negative bacteria. In: Microbial Lipids, pp. 299–487. Academic Press, San Diego. Farhood, H., Serbina, N., and Huang, L. (1995). The role of dioleyl phosphatidylethanolamine in cationic liposomes mediated gene transfer. Biochim. Biophys. Acta 1235: 289 – 295. Ebensen, T., et al. (2004). Bacterial ghosts are an efficient delivery system for DNA vaccines. J. Immunol. 172: 6858 – 6865. Paukner, S., Kohl, G., and Lubitz, W. (2004). Bacterial ghosts as novel advanced drug delivery systems: antiproliferative activity of loaded doxorubicin in human Caco-2 cells. J. Controlled Release 94: 63 – 74. Henke, S., et al. (2000). Enhanced in vitro oligonucleotide and plasmid DNA transport by VP1 virus-like particles. Pharm. Res. 17: 1062 – 1070. Stacey, K. J., Sweet, M. J., and Hume, D. A. (1996). Macrophages ingest and are activated by bacterial DNA. J. Immunol. 157: 2116 – 2122. Ahsan, F., Rivas, I. P., Khan, M. A., and Torres Suarez, A.I. (2002). Targeting to macrophages: role of physicochemical properties of particulate carriers—liposomes and microspheres—on the phagocytosis by macrophages. J. Controlled Release 79: 29 – 40. Mader, H. J., et al. (1997). Endotoxicity does not limit the use of bacterial ghosts as candidate vaccines. Vaccine 15: 195 – 202.

223