A new DNA vehicle for nonviral gene delivery: supercoiled ... - Nature

Gene Therapy (1997) 4, 1341–1349  1997 Stockton Press All rights reserved 0969-7128/97 $12.00

A new DNA vehicle for nonviral gene delivery: supercoiled minicircle A-M Darquet1, B Cameron2, P Wils1, D Scherman1 and J Crouzet2 1

UMR 133 CNRS/Rhoˆne-Poulenc Rorer and 2Rhoˆne-Poulenc Rorer Gencell, Centre de Recherche de Vitry-Alfortville, 13, Quai Jules Guesdes, 94403 Vitry sur Seine, France

Plasmids currently used for nonviral gene transfer have the disadvantage of carrying a bacterial origin of replication and an antibiotic resistance gene. There is, therefore, a risk of uncontrolled dissemination of the therapeutic gene and the antibiotic resistance gene. Minicircles are new DNA delivery vehicles which do not have such elements and are consequently safer as they exhibit a high level of biological containment. They are obtained in E. coli by att site-specific recombination mediated by the phage l integrase. The desired eukaryotic expression cassette bounded by the l attP and attB sites was cloned on a recombinant plasmid. The expression cassette was

excised in vivo after thermoinduction of the integrase gene leading to the formation of two supercoiled molecules: the minicircle and the starting plasmid lacking the expression cassette. In various cell lines, purified minicircles exhibited a two- to 10-fold higher luciferase reporter gene activity than the unrecombined plasmid. This could be due to either the removal of unnecessary plasmid sequences, which could affect gene expression, or the smaller size of minicircle which may confer better extracellular and intracellular bioavailability and result in improved gene delivery properties.

Keywords: DNA transfection; gene therapy; site-specific recombination; l integrase; gene expression; DNA vaccine

Introduction Plasmids currently used in preclinical and clinical trials of nonviral gene transfer contain a prokaryote origin of replication and an antibiotic resistance marker, both of which are undesirable. Clinical use would lead to the dissemination of prokaryotic replicative recombinant DNA in patients. Endogenous Enterobacteriacae, in which such plasmids replicate, may acquire the recombinant DNA resulting in an uncontrolled dissemination of the therapeutic and antibiotic resistance genes. These recombinant bacteria might then have a selective advantage, particularly if the corresponding antibiotic is administered. Moreover, prokaryotic sequences can have other undesirable effects. For instance antibodies have been produced against prokaryotic d-endotoxin after injection of a Bacillus thuringiensis plasmid carrying the d-endotoxin gene, expressed from cryptic upstream eukaryotic expression signals.1 Expression of the neomycin resistance marker alters the expression of genes in mammalian cultured cells.2 Furthermore, elements in the plasmid backbone such as the ampicillin-resistance gene (AmpR) result in a lower efficiency of transgene expression.3 In addition, short immunostimulatory DNA sequences, containing CpG dinucleotides, present in the prokaryotic backbone of plasmids, function as adjuvants for the immunogenic properties of plasmids encoding antigens.4 This DNA adjuvant effect would be valuable for gene immunization, or gene immunotherapy for cancer, but is

Correspondence: J Crouzet Received 11 April 1997; accepted 31 July 1997

clearly undesirable for other applications for which an immune stimulation is not required. The aim of this work was to develop safer genetic material for nonviral gene therapy. We produced supercoiled recombinant DNA molecules, called minicircles, which contain only the therapeutic expression cassette. These minicircles are the product of in vivo excision of the desired cassette by site-specific recombination between the two attP and attB sequences driven by E. coli bacteriophage l integrase. We obtained and purified a minicircle carrying a luciferase expression cassette (lucminicircle). These minicircles have no plasmid backbone sequences and thus avoid the deleterious effects of prokaryotic sequences present in the plasmid. In addition, these recombinant molecules presented gene transfer properties superior to those of standard plasmids. Therefore, minicircles are a promising alternative to plasmid DNA for nonviral gene therapy in terms of biosafety, improved gene transfer and potential bioavailability, due to their minimal size.

Results Construction of pXL2650 Plasmid pXL2650 carries the Photinus pyralis luciferase gene (luc) expression cassette (luc-cassette). This construct is 7.4 kb and has the following components (Figure 1): the ColE1 origin of replication (ie a pBR322-derived plasmid), the E. coli galK gene, an ampicillin resistance gene, the 530 bp attP sites of l phage and the 31 bp E. coli minimal attB sequence5 in the same orientation, and the luc-cassette from pGL2 control (Promega, Madison, WI,

Supercoiled DNA molecules for nonviral gene transfer A-M Darquet et al

1342

USA). Between the l attP and attB sites the luc reporter gene is under the control of the SV40 early promoter, and upstream from the intron of the SV40 small-t antigen, the SV40 early polyadenylation signal and an enhancer from the SV40 early promoter.

Production of luc-minicircle The attP and attB sites are two short fragments of the phage l genome and the E. coli chromosome, respectively. Phage l integrase mediates the integration of the circular phage genome in the E. coli chromosome by recombination. This reaction needs two host proteins (FIS and IHF)5–7 in addition to the integrase. If the two att sites are in the same orientation on the same replicon, the recombination will result in the excision of a supercoiled molecule.8 Thus, Int-mediated recombination of pXL2650 (Figure 1) will lead to a minicircle with a luc-cassette of 3.4 kb and a 4 kb pXL2650-derivative containing the origin of replication, the antibiotic marker gene and the recombinant site attL termed a miniplasmid. The minicircle carries only the reporter gene, with eukaryotic expression signals, and the recombinant site attR (Figure 1). The E. coli strain D1210HP was used in this study; it is a derivative of D1210 and harbors a l thermosensitive lysogen (xis kil− cI857) defective for lethal and lytic functions of the prophage thus preventing cell lysis.9 Upon thermal shift to 42°C the induction of a lytic cycle occurs and the integrase is produced resulting in the Intmediated reaction. However, the reverse reaction requiring the Xis protein does not occur since the lysogen is xis−. Subsequently, the E. coli DNA topoisomerase IV will decatenate the two products of the reaction.10 Plasmid DNA extracts from D1210HP pXL2650 cultures shifted from 30°C to 42°C contained the expected resolution products of pXL2650 (Figure 2a), ie the lucminicircle and the miniplasmid. Restriction digests with HindIII, NsiI and SphI which cut only the luc-minicircle at one, two and three sites, respectively, gave the expected fragments and left the miniplasmid uncut (Figure 2b). AlwNI, MluI, SnaBI, ScaI and XmnI for which there are restriction sites only in the miniplasmid gave the expected patterns of digestion leaving the luc-minicircle undigested (Figure 2b). This recombination was not observed in D1210HP pXL2650 cultured at 30°C or in D1210 harboring pXL2650 indicating it was Int-mediated (Figure 2a). Thus Int can mediate in vivo excision of a therapeutic gene expression cassette. The yield of unrecombined plasmid was around 40% of the starting material indicating that the reaction has to be improved. Figure 1 pXL2650 restriction map and scheme of in vivo site-specific recombination of pXL2650. The restriction map of pXL2650 is shown at the top of the figure. (a) Pairing of the two att sites of phage l (arrows indicate their relative orientations). (b) The att sites recombine in the bacteriophage l integrative reaction, which is catalyzed by the l integrase and requires the E. coli proteins IHF and FIS. (c) Recombination resolves pXL2650 into two molecules: one is a plasmid containing the origin of replication and the ampicillin resistance gene (miniplasmid), the other is the luc-minicircle. E. coli DNA topoisomerase IV is required for the decatenation of the two resolved molecules. A, AlwNI; H, HindIII; M, MluI; Ns, NSiI; S, SnaBI; Sc, ScaI; Sp, SphI; Xb, XbaI; X, XmnI; ori, ColE1derived origin of replication; luc, luc-cassette; galK, E. coli galK gene; AmpR, ampicillin resistance gene. Horizontal striped arrow, AmpR; vertical striped arrow, luc; black square arrow, attP and attB sequences; waved arrow, galK gene; white box, SV40 small-t antigen intron; black box, SV40 early polyadenylation signal; hatched box, SV40 enhancer from early promoter; dotted box, SV0 early promoter.

Purification of the luc-minicircle Supercoiled molecules obtained after recombination in D1210HP pXL2650 were digested with XmnI and AlwNI, both of which cut only the miniplasmid and pXL2650 (Figure 1). Supercoiled molecules were then isolated by CsCl-ethidium bromide (CsCl-EtBr) density gradient centrifugation. Analysis by agarose gel electrophoresis showed that the supercoiled luc-minicircle preparation was free of pXL2650 and the miniplasmid (Figure 3). Restriction digest analysis confirmed that the preparation contained the luc-minicircle (data not shown). However, two minor bands (A and B) with lower electrophoretic mobilities were also observed in the lucminicircle preparation (Figure 3). HindIII digestion of the supercoiled preparation gave a single DNA band of


1343

Figure 2 Production of luc-minicircle in D1210HP. (a) Analysis of the Int-mediated reaction in D1210 HP pXL2650. Plasmid pXL2650 was introduced into two E. coli strains, D1210 and D1210HP. D1210HP is D1210 carrying the l xis− kil − cI857 phage lysogen. Upon thermal induction at 42°C, l integrase is produced in D1210HP and the luciferase cassette is excised giving a luc-minicircle. Lanes 1 and 6: supercoiled DNA ladder; lane 2: plasmid DNA extracts from D1210HP pXL2650 after a thermal shift to 42°C; lane 3: uninduced D1210HP pXL2650; lane 4: D1210 pXL2650 shifted to 42°C and lane 5: uninduced D1210 pXL2650. Sizes in kb of the supercoiled DNA ladder are indicated on the left. Arrows A and B correspond to supercoiled minicircle and supercoiled miniplasmid, respectively. (b) Restriction digest of extrachromosomal molecules from induced D1210HP pXL2650. Undigested extrachromosomal molecules from D1210HP pXL26520 after induction (lane 1) and digested with AlwNI (lane 2), XmnI (lane 3), ScaI (lane 4), SnaBI (lane 5), MluI (lane 6), HindIII (lane 9), SphI (lane 10) and NsiI (lane 11). Indicated by an arrow: A, supercoiled minicircle; B, supercoiled miniplasmid; C, supercoiled pXL2650; D, linearized minicircle; E, linearized miniplasmid; F, linearized pXL2650. Supercoiled DNA ladder (lane 8) and linear DNA ladder (lane 7).

3.4 kb (Figure 3) the minor DNA species being totally digested. Similar results were obtained with other restriction enzymes cutting the luc-minicircle, whereas ScaI does not linearize this DNA preparation. The extra bands may thus correspond with dimers and nicked luc-minicircles (supercoiled plasmid DNA preparations are often contaminated by both nicked and dimer forms of the plasmid). Indeed, the mobility of band A corresponded with that of the luc-minicircle dimer as estimated by com-

parison with supercoiled DNA molecular weight standards (Figure 3). A 475 bp, 32P-labeled, AmpR gene probe hybridized with both pXL2650 and miniplasmid, but not with DNA bands A and B. However, both bands A and B hybridized with a 3 kb 32P-labeled probe for the luccassette. Luc-minicircle preparations were also analyzed after treatment with E. coli DNA topoisomerases I, II and IV. Topoisomerase II (gyrase) supercoils relaxed molecules, whereas topoisomerase I and IV relax supercoiled


1344

Figure 3 Analysis of the purified luc-minicircle. Lanes 1 and 8: supercoiled DNA ladder; lanes 2 and 7: linear DN ladder; lane 3: HindIII-digested and lane 4 undigested purified luc-minicircle; lane 5: plasmid DNA preparation from induced, and lane 6 uninduced, D1210HP pXL2650. In lane 4 bands A and B correspond to dimer and nicked luc-minicircle, respectively. Sizes in kb of the supercoiled and linear DNA ladders are indicated on the left and on the right, respectively.

molecules. DNA from band A and luc-minicircle DNA were relaxed by topoisomerase I and IV, but gyrase had no effect (data not shown). This indicates that the DNA in band A corresponds with supercoiled luc-minicircle dimer and that the luc-minicircle is supercoiled. The electrophoretic mobility of DNA in band B was not affected by the action of any topoisomerase indicating that it was nicked DNA, presumably nicked luc-minicicle generated during the purification and storage of the supercoiled preparation. Therefore, the purified luc-minicircle preparation was indeed free of miniplasmid and unrecombined plasmid.

Influence of total DNA concentration on transfection efficiency We developed a transfection protocol allowing the comparison of gene expression from DNAs of different molecular weights. Transfection experiments were performed using the DNA transfer agent RPR 120535 which consists of a dioctadecyl lipid moiety linked to a spermine head group via a primary amine to generate a linear structure. This lipopolyamine has in vitro transfection properties in the same range as those of standard cytofectins.11 Luc gene expression efficiencies for luc-minicircle (3.4 kb), pGL2 control (6 kb) and pXL2650 (7.4 kb) should ideally be compared using the same quantity of luc-cassette, and thus differing amounts of total DNA. However, in preliminary experiments the efficiency of transfection

depended not only on the molarity of the luc-cassette but also on the amount of DNA used. Similar results have been reported by Barthel et al. 12 We found that there was a linear relationship between the amount of pXL2650 used to transfect NIH 3T3 cells and the resulting luciferase activity (Figure 4) but only when the total amount of DNA added was fixed (1 mg), plasmid pBluescript II KS being used as carrier DNA. This plasmid contains only prokaryotic DNA and presumably does not interfere with the eukaryotic transcription machinery and luc-cassette expression. When no carrier DNA was added, the relationship between the pXL2650 concentration and luciferase activity was not linear: there was a lag in the response, ie low transfection efficiency at low DNA concentration (0.1 and 0.2 mg per assay). Therefore, to compare the transfection efficiencies of supercoiled molecules of different sizes, carrier DNA was used so that the total amount of DNA was the same in each assay.

Luciferase activities from luc-minicircle transfected cells The potency of luc-minicircle for luciferase gene expression was studied by transfections with 1 mg of pXL2650, 0.81 mg of pGL2 control or 0.46 mg of luc-minicircle per assay with a complement of pBluescript KS II to 1 mg as necessary. The molarity of the luc-cassette was thus the same for all tests. Three charge ratios (3, 6 and 9) of RPR120535 to DNA were assayed. Five different cell lines were studied: NIH 3T3 (murine


Influence of plasmid size on transfection efficiency To confirm that the size of the supercoiled molecule affects the efficiency of gene transfer, we constructed a large (15.3 kb) plasmid, named pXL3067, containing the luc-cassette. This plasmid has a backbone derived from pRK41515 and the same luc-cassette as pGL2 control and pXL2650. Cells were transfected with 1 mg of pXL3067, 0.49 mg of pXL2650, 0.39 mg of pGL2 control or 0.22 mg of luc-minicircle (equivalent molarity of luc-cassette) made up to 1 mg with pBluescript KS II. Transgene expression in H460, NIH 3T3 and RSM cells transfected with pXL3067 was 19-, 76- and 545-fold lower, respectively, than in the same cell lines transfected with lucminicircle (Figure 6). These results indicate that the larger the transfected supercoiled molecule, the lower the transfection efficiency.

Discussion Figure 4 Influence of total DNA concentration on gene expression. Comparison of transfection efficiencies with different amounts of pXL2650, made up (K) or not (G) to a total of 1 mg of DNA with pBluescript KS II and complexed with RPR 120535 at a charge ratio of 3. Results are expressed in RLU/mg of protein. T bars indicate the standard deviation (s.d.).

fibroblast), H460 (human nonsmall cell lung carcinoma), 3LL (mouse Lewis lung carcinoma), human aortic smooth muscle (HSM), and rabbit aortic smooth muscle (RSM) cells. Human smooth muscle cells are interesting for gene transfer purposes because of their proximity to the lumen surface and their abundance in the vessel wall.13 The strength of reporter gene expression (highest to lowest) was RSMC, NIH 3T3, 3LL, H460 to HSMC. This is not surprising as expression of reporter genes in HSMC is much lower than that obtained in cell lines commonly used for transfection, such as NIH 3T3 fibroblasts, especially with the SV40 early promoter.14 The results of the transfection experiments are shown in Figure 5, with luciferase activities normalized for total protein content. No activities above background were detected in lysates of cells transfected with DNA in the absence of cationic lipid (data not shown). Luc-minicircles expressed luciferase activity in all the cell types tested. The activities, measured as relative light units (RLU), were two- to four-fold (3LL and NIH 3T3), five- to seven-fold (H460) and 10-fold (RSM) higher in tested cells transfected with the lucminicircle, than with the unrecombined plasmid pXL2650. In HSM cells no significant luciferase activity was detected following transfection with pXL2650, whereas transfection with luc-minicircle gave 840 RLU/mg of protein, the background being lower than 6 RLU/mg of protein. Transfection with the pGL2 control plasmid led to luciferase activities similar to those obtained with pXL2650, except in H460 and RSM cells where the values were intermediate between those of lucminicircle and pXL2650. These findings were confirmed with three independent preparations of luc-minicircle, pGL2 control and pXL2650 DNA, showing that they were not a preparation effect. The experiments were reproduced between two and four times in each cell line, giving the same ratio of RLU between supercoiled molecules. Moreover, in NIH 3T3 a dose–response study showed that the observed effect was conserved at least from 1 to 0.5 mg of DNA per well.

We constructed and studied new DNA supercoiled molecules called minicircles which can be used for nonviral gene transfer. They have no bacterial origin of replication or antibiotic resistance marker and are thus safer than plasmids currently used in preclinical and clinical trials. The probability that minicircles would be disseminated following their use for gene therapy is low due to the lack of bacterial replication origin. We cannot exclude the possibility that an E. coli lysogenized by l phage could integrate the minicircle into its genome, through homologous recombination between attR sites. However, this would require two low probability events: the presence of an E. coli l lysogen and homologous recombination. The same is true for standard plasmid antibiotic resistance genes, whose homologous genes could be carried not only by E. coli, but also many other bacteria. We show that minicircles have higher transfection efficiencies than commonly used plasmids and much higher efficiencies than larger plasmids. The differences were partly cell dependent, the largest differences being observed with RSM and HSM cells. An alternative strategy to obtain minicircles is to use in vitro recombination involving either a recombinase, or restriction digestion followed by ligation. However, such approaches would be difficult to scale up for industrial use whereas in vivo recombination is simple and only needs the production of recombinase to be induced. The backbone sequences in the luc-minicircle are only 290 bp long, shorter than the 2.5–3 kb16 and 2 kb17 of the backbones in standard and improved plasmids used in the clinic for nonviral gene transfer. Overall, minicircles are about half the size of standard plasmids assuming the transgene is shorter than 1 kb (the case for growth factors and cytokines, for example) and expressed under the control of short viral expression signals such as the CMV immediate–early promoter and the SV40 polyadenylation signals. One limitation of the efficiency of in vivo nonviral gene delivery has been attributed to low extracellular and intracellular bioavailability of the delivered complexes or plasmids. As minicircles are smaller than standard plasmids, under particular conditions of complex formation, complexes between cationic lipids and minicircle could be smaller than those obtained with standard plasmids, assuming that complexes form between one DNA molecule and the transfecting agent. Small complexes should

1345


1346

Figure 5 Comparison of luciferase activities following transfection of various cell types with luc-minicircle, pGL2 control, or pXL2650 complexed with RPR120535. Transfections were performed at various charge ratios (3, 6 and 9) with the same molarities of luc-cassette and the same amount of total DNA (1 mg). Results are given in RLU/mg of protein for NIH 3T3 fibroblasts (a), 3LL carcinoma cells (b), H460 carcinoma cells (c), RSM cells (d) and HSM cells (e). T bars indicate the standard deviation (s.d.).


Figure 6 Influence of the size of supercoiled DNA on gene expression from the luc-cassette. Results are presented in RLU/mg of protein for four different supercoiled molecules (luc-minicircle, pGL2 control, pXL2650 and pXL3067) at three charge ratios of RPR120535 to DNA with three cells lines: NIH 3T3 fibroblasts (a), H460 carcinoma cells (b) and RSM cells (c). T bars indicate the standard deviation (s.d.).

have better characteristics than larger complexes mainly because their diffusion coefficient is higher: the diffusion coefficient is inversely proportional to the molecular weight of the complex.18 Complex size also plays a role in endocytosis, the mechanism by which complexes enter the cell.19,20 This step is limited by the size of particles that can be taken up21 and complexes of diameter greater than 200 nm are not efficiently taken up by the endocytic pathway. Another possible advantage of minicircles is better

bioavailability, again dependent on the size of the delivered material. Indeed, DNA migration from the cytoplasm to the nucleus appears to be a crucial step for successful gene transfer.19 It has been shown that the rate of protein transport from the cytoplasm to the nucleus is influenced by the number of nuclear localization sequences per protein, and most importantly by the size of the protein.22 Similarly, the size of transfected DNA molecules may determine their diffusion and/or transport to the nuclear pore. It has been reported that the percentage of b-galactosidase-positive myotubes following injection of plasmid DNA carrying the lacZ gene increases as the site of injection is closer to the nucleus,23 indicating that the diffusion of the plasmid from a site in the cytoplasm distant to the nucleus may be limiting.23 The efficiency of gene expression from linear and supercoiled DNA is different. Buttrick et al24 showed that luciferase activity is 50- to 100-fold greater for supercoiled plasmids than for linear forms. In our experiments the topology of transfected DNA molecules was investigated by agarose gel electrophoresis. Luc-minicircle, pGL2 control, pXL2650 and pXL3067 preparations contained similar ratios of relaxed to supercoiled forms. Differences in the topology of the transfected DNA molecules are thus unlikely to be the reason for the differences of reporter gene activities observed. The enhanced gene delivery properties of minicircles may be due to absence of plasmid sequences, since plasmid backbone elements have been shown to have an effect on gene expression. The replacement of an ampicillin resistance gene by a kanamycin gene led to a twofold increase in luciferase activity. 3 In our experiments, although all the molecules tested carry the same luc-cassette, it is surrounded by different sequences that might interfere with expression. For instance, small differences in supercoiling of the molecules, the presence of cryptic sites for the binding of transcription factors or short bacterial immunostimulatory sequences which have been shown to interfere with gene expression by stimulating IFN-a4 production might be responsible for all or part of the observed effects. The luc-minicircle was purified on density gradients, a method which is not acceptable for purification of clinical grade material. Only chromatographic methods are suitable for preparing clinical grade DNA but they would not allow the separation of minicircle from miniplasmid.25 We have developed an affinity chromatographic technique for the purification of plasmid DNA based on the sequence-specific formation of a triple helix between an immobilized oligonucleotide and a specific sequence present on the plasmid.26 This method could easily be applied to separate minicircles from miniplasmids by cloning a short homopurine sequence next to the transgene cassette, thus allowing the specific triple helix interaction between minicircle and an appropriate oligonucleotide covalently coupled to a chromatographic support. The yield of recombination would have to be close to 100% to avoid contamination with unrecombined plasmid during affinity chromatography. This could be achieved by optimizing the culture conditions of the recombinant E. coli and by overexpressing int. If these approaches are unsuccessful, affinity chromatography specific for sequences present in the miniplasmid and the

1347


1348

unrecombined plasmid would allow elimination from the minicircle preparation.

Materials and methods Standard DNA manipulation All standard manipulations, including digestion by restriction enzymes, purification of DNA fragments by electroelution, electrophoresis of DNA, E. coli transformation, purification of supercoiled DNA by a CsCl-EtBr density gradient, and Southern blot analysis were performed according to previously described techniques.26 Intermediate plasmid constructions were introduced into E. coli DH5a [F-, F80lacZDM15, D(lacZYA-argF)U169, deoR, TecA1endA1, hsdR17, phoA, supE44, l- thi-1, gyrA96, relA1] (Clontech Laboratories, Palo Alto, CA, USA). DNA concentrations were measured by reverse phase HPLC analysis, using a Poros R2/H column (100 × 4.6 mm; PerSeptive Biosystems, Cambridge, MA, USA) according to described methods.26 Restriction enzymes and T4 DNA ligase were purchased from Biolabs (New-England Biolabs, Beverly, MA, USA), Gibco-BRL (Life Technologies SARL, Cergy Pontoise, France) or Amersham (Amersham, Les Ulis, France). Oligonucleotides were synthesized on an Applied Biosystems 394 DNA/RNA synthesizer28 (Perkin Elmer, Courtaboeuf, France) and purified as previously described.29 Supercoiled and 1 kb linear DNA ladders were from Gibco-BRL and Promega, respectively. Plasmid construction Plasmid pXL2648 was constructed from pNH16a 9 which contains the attP site of bacteriophage l. The attB site was introduced into pNH16a using oligonucleotides, 5476 (59-AATTGTGAAGCCTGCTTTTTTATACTAACTTGAGCGG-39) and 5477 (59-AATTCCGCTCAAGTTAGTATAAAAAAGCAGGCTTCAC-39), which reconstitute the minimal attB sequence.5 The two oligonucleotides were hybridized and ligated into the EcoRI site in pNH16a giving plasmid pXL2648 which has an EcoRI site between the attP and attB sequences which are in the same orientation. Plasmid pXL2649 was constructed by inserting the KmR gene cassette of pUC4KIXX (Pharmacia, Uppsala, Sweden) into the EcoRI site in pXL2648. The presence of BamHI restriction sites on either side of the KmR cassette allowed the insertion of the 3.15 kb BamHI–luc–BglII fragment of pGL2-control (Promega), giving plasmid pXL2650, containing the luc-cassette between the att sites. Plasmid pXL3040 corresponds to pGL2-control in which the 1.6 kb BamHI–KmR-BamHI fragment of pUC4KIXX was inserted into the BglII site. Plasmid pXL3067 was constructed in inserting the 5.1 kb BamHI– luc–KmR–NheI fragment from pXL3040 between the XbaI and BamHI sites in pRK415,15 leading to a 15.3 kb plasmid harboring the same expression cassette as pXL2650. Production of luc-minicircle E. coli was grown in LB medium supplemented with 50 mg/ml of ampicillin (LB Ap). Plasmid pXL2650 was introduced into two E. coli strains, D1210 [F-hsdS20, supE44, ara-14, galK2, proA2, leuB6, rpsL20, xyl5, mtl1, recA, mcrB, D(mcrC-mrr), lacIQ]30 and D1210HP (D1210 lysogenized with the l cI857 xis− kil− phage31 ). The transformed strains

were grown overnight at 30°C in LB Ap, diluted 1/100 in LB Ap and incubated at 30°C until the OD610 reached 2. The l phage lytic cycle was induced by a thermal shift to 42°C for 10 min. The cultures were further incubated for 30 min at 30°C and bacteria then harvested. Extra chromosomal DNA was purified from these cells by a standard plasmid DNA purification technique,27 using CsCl-EtBr density gradients. The purified DNA was then digested with AlwNI and XmnI. A second density gradient allowed the separation of the luc-minicircle from the linearized miniplasmid and pXL2650. Luc-minicircle preparations were concentrated 10-fold with an ultrafree-15, Biomax 30 centrifugal filter device (Millipore, Saint Quentin, France) according to the manufacturer’s recommendations. Luc-minicircle preparations were analyzed by the action of purified E. coli topoisomerases I, II and IV according to described techniques.10,32 The 3 kb BamHI–BglII fragment from pGL2 control containing the luc gene and the 475 bp XmnI–BglII fragment from the Amp R gene of pNH16a were 32P-labeled using the rediprime kit (Amersham).

In vitro transfection protocol Cells were maintained in appropriate medium at 37°C under a 5% CO2 humidified atmosphere. Dulbecco’s modified Eagle’s medium (DMEM) (Gibco-BRL) supplemented with 100 units/ml of penicillin, 100 units/ml of streptomycin, 20 mm l-glutamine and 10% fetal calf serum (Gibco-BRL) was used for NIH 3T3 (ATCC, Rockville, MD, USA), 3LL,25 H460 cells.33 RSM cells were obtained as previously described34 and were grown in the same medium as above except that 20% fetal calf serum was used. Smooth muscle cell basal medium (Clonetics, San Diego, CA, USA) supplemented with 5 mg/l bovine insulin, 50 mg/l gentamicin sulfate, 50 mg/l amphotericin-B, 0.5 mg/l human epidermal growth factor, 2 mg/l human fibroblast growth factor and 10% fetal calf serum (Clonetics) was used for HSM cells (AOSMC 2708; Clonetics). Confluent cells were trypsinized and seeded in 24-well microtiter plates. Cells were transfected at 60– 80% confluence, which corresponds to 18 h after seeding for NIH 3T3, 3LL, H460, RSM cells and 1 week after seeding for HSM cells. The medium was changed every day during the week before the transfection of HSM cells. Equal volumes of stock solution (10 mm solution in water obtained after heating for 30 min at 50°C) of RPR12053511 (formula: H2N (CH2)3 NH(CH2) 4NH(CH)2 )3 NHCH2 CONHCH2CON[(CH2) 17-CH3 ]2] and supercoiled DNA solution in 150 mm NaCl were mixed and incubated for 10 min at 20°C. Cells were transfected with either pXL2650, pGL2 control, luc-minicircle or pXL3067. The total amount of DNA added to each well was adjusted to 1 mg with pBluescript KS II (Stratagene Cloning System, La Jolla, CA, USA). Cells in 24-well microtiter plates were washed before transfection with 500 ml of serum free-medium, once for HSM and RSM cells and twice for NIH 3T3, 3LL and H460 cells. Transfections were performed in 500 ml of serum-free medium by adding 50 ml of lipid per DNA solution to each well. Each transfection experiment was performed in triplicate. The media were supplemented 2 h after transfection with either 10% of the appropriate serum for NIH 3T3, 3LL, H460 and HSM cells, and 20% for RSM cells. Luciferase activity was assayed 48 h later.


Luciferase activity Luciferase activity was measured using the Luciferase assay system according to the manufacturer’s recommendations (Promega) using a LUMAT LB 9501 luminometer (EG&G, Berthold, Evry, France). Cells were incubated for 30 min in a lysis buffer (25 mm Tris-phosphate, pH 7.8, 2 mm dithiothreitol, 2 mm 1,2 diaminocyclohexane-N, N, N9, N9-tetraacetic acid, 10% glycerol, and 1% Triton X100) for NIH 3T3, 3LL, H460 and RSM cells. For HSM cells the lysis buffer was supplemented with a 50-fold diluted protease inhibitor cocktail from Boehringer Mannheim (Meylan, France). The cells were detached by scraping and the lysates were centrifuged. The supernatants obtained were assayed for luciferase activity (5 ml from all transfected cells except for HSM for which 50 ml was used). Emitted light was measured for a period of 10 s for NIH 3T3, 3LL, H460 and RSM cells, and 120 s for HSM cells, and is expressed in RLU. The protein concentrations of the samples were measured using the Pierce BCA assay (Interchim. Asnie`res, France).35 Results are expressed in RLU/mg of protein, ± standard error.

Acknowledgements We thank T Ciora for oligonucleotide synthesis, C Ciolina and G Pollet for reverse phase PHPLC DNA analysis, C Dubertret and G Jaslin for their help with HSM cells culture and K Berthelot for her help with RSM cell culture. G Byk is acknowledged for the generous gift of RPR120535. We thank F Blanche for the gift of purified E. coli DNA topoisomerases and protocols for modifying topology of purified circular DNA. J-F Mayaux and J-B Le Pecq are acknowledged for their constant interest and support during the work. This work was part of the BioAvenir program supported by Rhoˆne-Poulenc, the French Ministry of Research, and the French Ministry of Industry.

References 1 Pang ASD. Production of antibodies against Bacillus thuringiensis delta-endotoxin by injecting its plasmids. Biochem Biophys Res Commun 1994; 202: 1227–1234. 2 Valera A, Perales JC, Hatzoglou M, Bosch F. Expression of neomycin-resistance (neo) gene induces alterations in gene expression and metabolism. Hum Gene Ther 1994; 5: 449–456. 3 Hartikka J et al. An improved plasmid DNA expression vector for direct injection into skeletal muscle. Hum Gene Ther 1996; 7: 1205–1217. 4 Sato Y et al. Immunostimulatory DNA sequences necessary for effective intradermal gene immunization. Science 1996; 273: 352–354. 5 Landy A. Dynamic, structural, and regulatory aspects of l sitespecific recombination. Ann Rev Biochem 1989; 58: 913–949. 6 Pollock T, Nash HA. Knotting of DNA caused by a genetic rearrangement. J Mol Biol 1983; 170: 1–18. 7 Patsey RL, Bruist MF. Characterization of the interaction between the lambda intasome and attB. J Mol Biol 1995; 252: 47–58. 8 Gellert M, Nash H. Communication between segments of DNA during site-specific recombination. Nature 1987; 325: 401–404. 9 Hasan N, Szybalski W. Control of cloned gene expression by promoter inversion in vivo: construction of improved vectors with a multiple cloning site and the Ptacpromoter. Gene 1987; 56: 145–151. 10 Peng H, Marians KJ. Escherichia coli topoisomerase IV. Purification, characterization, subunit structure, and subunit interactions. J Biol Chem 1993; 15: 24481–24490.

11 Byk G, Dubertret C, Schwartz B, Scherman D. Lipopolyamines as transfection agents and pharmaceutical uses thereof. Wo Patent 97/18185. 12 Barthel F, Remy JS, Loeffler JP, Behr JP. Gene transfer optimization with lipospermine-coated DNA. DNA Cell Biol 1993; 6: 553–560. 13 Lynch CM et al. Long-term expression of human adenosine deaminase in vascular smooth cells of rats: a model for gene therapy. Proc Natl Acad Sci USA 1992; 89: 1138–1142. 14 Pickering JG et al. Liposome-mediated transfer into human vascular smooth cells. Circulation 1994; 89: 13–21. 15 Keen NT et al. Improved broad-host-range plasmids for DNA cloning in Gram-negative bacteria. Gene 1988; 70: 191–197. 16 Caplen NJ et al. Liposome mediated CFTR gene transfer to the nasal epithelium of patients with cystic fibrosis. Nature Med 1995; 1: 39–46. 17 Hobart PM, Margalith M, Parker SE, Khatibi S. Plasmids suitable for IL-2 expression. Wo Patent 96/17063. 18 Jain RK. Transport of molecules in the tumor interstitium: a review. Cancer Res 1987; 47: 3039–3051. 19 Zabner J. Cellular and molecular barriers to gene transfer by a cationic lipid. J Biol Chem 1995; 270: 18997–19007. 20 Friend DS, Papahadjopoulos D, Debs RJ. Endocytosis and intercellular processing accompanying transfection mediated by cationic liposomes. Biochim Biophys Acta 1996; 1278: 41–50. 21 Gao X, Huang L. Cationic liposome-mediated gene transfer. Gene Therapy 1995; 2: 710–722. 22 Lanford RE, Kanda P, Kennedy RC. Induction of nuclear transport with a peptide homologous to the SV40 antigen transport signal. Cell 1986; 46: 575–582. 23 Horn NA, Meek JA, Budahazi G, Marquet M. Cancer gene therapy using plasmid DNA: purification of DNA for human clinical trials. Hum Gene Ther 1995; 6: 565–573. 24 Buttrick PM, Kass A, Kitsis RN. Behavior of genes directly injected into the rat heart in vivo. Circ Res 1992; 70: 193–198. 25 Bissery M, Gue´nard D, Gue´ritte-Vogelein F, Lavelle F. Experimental antitumor activity of taxotere (RP 56976, NSC 628503), a taxol analogue. Cancer Res 1991; 51: 4845–4852. 26 Wils P et al. Efficient purification of plasmid DNA for gene transfer using triple-helix affinity chromatography. Gene Therapy 1997; 4: 323–330. 27 Sambrook J, Fritsh EF, Maniatis T. Molecular Cloning: a Laboratory Manual. Cold Spring Harbor Laboratory Press: New York, 1989. 28 Sinha ND, Biernat J, McManus J, Ko¨ ster H. Polymer supplied oligonucleotide synthesis XVIII: use of b-cyanoethyl-N,N-dialkylamino-/N-morpholino phosphoramidite of deoxynucleosides for the synthesis of DNA fragments simplifying deprotection and isolation of the final product. Nucleic Acids Res 1984; 12: 4539–4557. 29 Sawado M, Van Dyke MW. A rapid method for the purification of deprotected oligodeoxynucleotides. Nucleic Acids Res 1991; 19: 674. 30 Sadler JR, Tecklenburg M, Betz JL. Plasmids containing many tandem copies of a synthetic lactose operator. Gene 1980; 8: 279–300. 31 Podhajska AJ, Hasan N, Szybalski W. Control of cloned gene expression by promoter inversion in vivo: construction of the heat-pulse-activated att-nutL-p-att-N module. Gene 1985; 40: 163–168. 32 Lynn RM, Wang JC. Peptide sequencing and site-directed mutagenesis identify tyrosine-319 as the active site tyrosine of E. coli DNA topoisomerase I. Proteins 1989; 6: 231–239. 33 Banks-Schlegel SP, Gazdar AF, Harris CC. Intermediate filament and cross-linked envelope expression in human lung tumor cell lines. Cancer Res 1985; 45: 1187–1197. 34 Chamley H, Campbell G, McConnel J. Comparison of vascular smooth muscle cells from adult human, monkey and rabbit in primary culture and in subculture. Cell Tiss Res 1977; 177: 503–522. 35 Wiechelman K, Braun R, Fitzpatrick ES. Investigation of the bicinchoninic acid protein assay: identification of the groups responsible for color formation. J Anal Biochem 1988; 175: 231–237.

1349


1350